Another controversial topic : High ISO = less noise? [Archive]

Xavier

12-13-2009, 09:18 PM

The problem with this experiment is that you have basically exposed the two shots at different EVs. The ISO 100 shot needed a longer shutter shutter speed to compensate for the constant Aperture.
I knew this would come up. Well, EV stands for Exposure Value, correct? Exposure has nothing to do with ISO. This is a common misconception. The basic definition of Exposure as per any Photography textbook is the total amount of light falling on the photographic medium. Thus, aperture and shutter speed are the only two factors involved in 'exposure'.

When you take such an under exposed shot into post and simply boost the lightness, you are bound to get noise and you will.
The same experiment can be done on an ISO 1600 whose 'lightness' value is decreased by 4 stops. This will also produce the same results, i.e less noise @ ISO 1600. It won't be visible to the human eye, however the standard deviation figures can easily prove it.

This is why cameras have a setting called "Long exposure noise reduction", which is basically dark frame subtraction. If you shoot the ISO 100 shot with a long exposure and do a dark frame subtraction, the results would be significantly different.
Agreed.

Next problem: You are confusing two different types of noise. What you are demonstrating is read noise, which is and should be corrected.
This is read noise and it is both chroma and luminance read noise. This cannot be 'corrected', it can only be reduced at the cost of sharpness. You're confusing this noise with thermal noise which can be corrected by the dark frame subtraction method.

On the other hand, the dynamic range and per-pixel resolution of any camera drop as one goes up the ISO scale.
Agreed. Higher ISO = less highlight headroom = low DR.

Where there should be fine detail, the camera adds noise. This detail cannot be recovered no matter what. This is the "High ISO noise" that people talk about in general.
Nope, people usually associate chroma and luminance noise with high ISO noise, which is actually caused due to underexposure and not high ISO.

synn

12-13-2009, 09:41 PM

Exposure has nothing to do with ISO.

Eh, Exposure is determined by the triangle of Aperture, Shutterspeed and ISO. You change one, the other two get affected. That's pretty much the basic principle behind photography.

My point is that you have correctly exposed one shot and under exposed the other. This is hardly the scientific way to do it. Any experiment done to demonstrate objective data should be done on the level grounds and this wasn't.

You can underexpose the high ISO shot as you claim, but the experiment would still be skewed.

The correct way to use base ISO for lowlight photography is to meter properly and use dark frame subtraction. When you follow and incorrect procedure and claim that you've come across noise, the claim doesn't hold much value.

As I understand, read noise is a characteristic of the sensor and remains constant no matter what image you're shooting. Chroma noise on the other hand occurs because the photosites fail to capture the light rays properly. On film, this is less of an issue as lightrays coming in at any angle can be captured by it. On a digital sensor, there is a limitation to the angle at which a lightray can fall and be captured. When shooting at a higher ISO, the photosite simply does not have time enough for all the frequencies of visible light to fall on it at the required angle. Hence the noise.

I just shot 300 odd images at a gig yesterday, properly metered and all have a significant amount of chroma and luminance noise in them.

...and all were at 1600 ISO.

Deltaone

12-13-2009, 09:52 PM

Your experiment is wrong. To start with your experiment is actually used as an example of how not to compare noise. Pushing a 100 iso shot by 4 shots does not equate to 1600 iso.

Ok, now you just said, only shutter speed and aperture matter for exposure. This does beg the question, why is iso even there in the first place. For fun? No (just in case :P ). IT is a triangle as synn mentioned. This is something that does not need textbooks to explain. Take a couple of photos and its easy enough to see.

Now why is a 4 stop post overexposed picture showing noise? Not exactly rocket science there. Its in post. The sofware is trying to do the impossible. Raw has at most 2 stops of headroom in it. Not four. You can notice this when trying to lighten any shot especially the ones involving blacks.

You want to compare two isos, fine. Keep the exposure the same. i.e. For the settings mentioned above, either the aperture or the shutterspeed has to go down to compensate for the 100 ISO.

Ok, even simpler. Put your cam in A mode or Av for canons. Point at anything, dont matter what. In A, you control the aperture. So fix it at some value. Now meter it at iso 100, then meter the same shot at iso 1600. The shutter speed goes up. Why? Because the camera is compensating for the increased iso by reducing exposure time to give an equivalent exposure.

The basic triangle of photography has remain unchanged for many many years now. Noise is confusing. The more you read, the more you're likely to get confused between the various types and kinds of noise. So do it the simple way. Take photos and experiment. But when you do that, make sure you're doing the right experiment, not hte wrong one.

PS: This was discussed previously somewhere and i thought it to be at an end. Apparently not :P.

Xavier

12-13-2009, 09:58 PM

Eh, Exposure is determined by the triangle of Aperture, Shutterspeed and ISO. You change one, the other two get affected. That's pretty much the basic principle behind photography.

My point is that you have correctly exposed one shot and under exposed the other. This is hardly the scientific way to do it. Any experiment done to demonstrate objective data should be done on the level grounds and this wasn't.

You can underexpose the high ISO shot as you claim, but the experiment would still be skewed.

The correct way to use base ISO for lowlight photography is to meter properly and use dark frame subtraction. When you follow and incorrect procedure and claim that you've come across noise, the claim doesn't hold much value.

As I understand, read noise is a characteristic of the sensor and remains constant no matter what image you're shooting. Chroma noise on the other hand occurs because the photosites fail to capture the light rays properly. On film, this is less of an issue as lightrays coming in at any angle can be captured by it. On a digital sensor, there is a limitation to the angle at which a lightray can fall and be captured. When shooting at a higher ISO, the photosite simply does not have time enough for all the frequencies of visible light to fall on it at the required angle. Hence the noise.

I just shot 300 odd images at a gig yesterday, properly metered and all have a significant amount of chroma and luminance noise in them.

...and all were at 1600 ISO.

Nope, ISO doesn't come in the equation. For apparent brightness, it does. For 'exposure', it doesn't
Check any Photography textbook for the definition of exposure. The sensitivity of the medium plays no role in it. The amount of light is the only factor.

And read noise does not remain constant, but depends upon pixel location, pixel colour and exposure time.

Xavier

12-13-2009, 10:03 PM

synn

12-13-2009, 10:11 PM

To be honest, I'd rather take pictures rather than pander over books over academic data that contributes nothing to my creative flow. That said, you're still wrong and possibly confused by a bit too much of reading:

You fail to understand that There are three factors involved in a properly exposed photograph. Exposure indeed is the total amount of light received, but to receive the total amount of light that is required for a certain shot, the medium would have to be sensitive eonough for the selected SS+Aperture combo.

"The exposure for a photograph is determined by the sensitivity of the medium used. For photographic film, sensitivity is referred to as film speed (http://en.wikipedia.org/wiki/Film_speed) and is measured on a scale published by the International Organization for Standardization (http://en.wikipedia.org/wiki/International_Organization_for_Standardization) (ISO). Faster film requires less exposure and has a higher ISO rating.

http://en.wikipedia.org/wiki/Exposure_%28photography%29

Meter a shot at say, 800 ISO in manual mode. So now the camera is at a setting that's chose totally by you. Now change the ISO to 100. Do you notice that the Meter has in fact, changed the reading? One could even say, the shot is *gasp* UNDEREXPOSED?

...anyway, this side discussion doesn't really have much to do with your original experiment. Please do repeat the experiment with different ISO, but this time; meter them properly. Let's discuss it then.

Incidentally, some contributor to Wikipedia has already done this. Here' are two shots of the same scene by a Canon camera. Both metered properly.

http://upload.wikimedia.org/wikipedia/commons/a/a5/ISO_comparison_150px.jpg

Top image: ISO 100, f/5.6, 1/350 s

Bottom image: ISO 1600, f/5.6, 1/4000 s

http://en.wikipedia.org/wiki/Image_noise
Don't make me say "I told you so". :D

Deltaone

12-13-2009, 10:17 PM

As I mentioned above, the same experiment can be conducted by pulling down an ISO 1600 shot too. And even in that case the ISO 1600 shot will have less noise than an ISO 100 shot. This nullifies your 'pushing causes noise' The noise is already there, its just not visible to the human eye because of the lack of brightness. The standard deviation figures too indicate the higher noise values of the ISO 100 shot.

Selective reading is not good for you :P. ISO is part of the photography triangle. Film speed or ISO or whatever you want to call it is part of the triangle.

Again, A mode, two shots, two metering, why is it different. I see no explanation for that most basic of questions.

The sensitivity of the medium matters. It does. This is not something that is up for debate. This is something that has been proven in over 100 years of photography.

Again, ive distilled it down to the bare essence so that there is no confusion.

A. Why is there an ISO setting if it is irrelavant and why is it so important.

B. Why does the camera give two differing shutter speeds for the exact same shot for the very same ISO. (same applies to aperture for Shutter priority).

I cannot possibly make it simpler at the moment.

Xavier

12-13-2009, 10:23 PM

Deltaone

12-13-2009, 10:33 PM

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html

From the above link.

Bottom line: Read noise at high ISO is much smaller than read noise at low ISO, in terms of the error in photon counting that it represents. Thus, better image quality is obtained for using the highest ISO for which the signal is not clipped.

Oh and btw, you should also check this link

http://photography-on-the.net/forum/showthread.php?t=730218

Read from post 7 onwards.

Why am i not surprised. Everyone goes and reads something exactly similar to that and jumps to the very same conclusion. I am getting old these days ><.

Again, my questions remain unanswered. Yes he is right. The part you have quoted is right. But noise is complicated. Very very complicated. I cannot really explain this in simple sentences so i shall admit for the moment that read noise is higher at 100 iso. There are too many types and varieties and kinds of noise to go into at this time.

Your fundamental assumption which you have somehow gotten from this that ISO does not matter is WRONG. I cannot emphasize this enough. Photography triangle -> three sides -> sensitivity is one of the sides. No argument, no question, no debate. It is a FACT. ->"a statement or assertion of verified information about something that is the case or has happened."

Now the noise thing. Post #8 in the forum is what has you confused and its not exactly surprising. It happens to most people.

Heres the simple version. Noise is confusing. There are many types of noise. Basic facts regarding one type of thing can be interpreted rather wrongly in another way.

Again, i reiterate. Where is the answer to my two simple questions of the post above? I have marked them rather clearly as A and B. That should provide half your answer. Or else do i have to go into a detailed discussions involving photos, gain, electronics and what not? Seriously, i dont have the strength for that, it takes too long and it has been done too many a time before.

I reiterate again. Selective reading is bad for you.

Please answer questions A and B above :P.

synn

12-13-2009, 10:33 PM

Xavier

12-13-2009, 10:52 PM

You quoted something in bold, but you failed to notice the most important part of the sentence.

"for which the signal is not clipped."

In the real world, this means that the signal may be clipped from anything above the base ISO, in which case the optimal ISO IS the base ISO. Go back a few posts and read up on how the photosites don't get enough time to capture all frequencies. This correlates with that.

I just provided you with a real world example of how two shots, exposed the same have differing levels of noise. Just so happens that the higher ISO shot has more noise. How in the world are you going to defend it?

Nevermind reading up/ asking me to read something. Go out there in the real world, repeat the same experiment and come back to me with two photographs.

I'll be right here.

I did mention that high ISO does result in a loss of highlight headroom and thus is more prone to clipping. However, high ISO != noise and this is the basic crux of my experiment.

The example that you provided has two different exposures. The underexposed (ISO 1600) obviously will have higher noise. Unless we agree on the definition of exposure, this is going nowhere.

synn

12-13-2009, 10:59 PM

No they aren't.

Both shots are exposed exactly the same and your complete lack of understanding of the exposure triangle is making you beat the dead horse for the umpteenth time.

I could explain the whole shebang one more time for you, but those would once again, fall on deaf ears. So yep, this is going nowhere. We can agree on that bit.

Do keep taking your shots at high ISOs no matter what. Me on the other hand, would choose it as per requirements of the scene. I'll deal with noise when I come across it and unless someone changes the basic rules of photography overnight, I'll have to deal with it with my high ISO shots than base-ISO, long exposure shots.

Cheerio.

Deltaone

12-13-2009, 11:02 PM

I did mention that high ISO does result in a loss of highlight headroom and thus is more prone to clipping. However, high ISO != noise and this is the basic crux of my experiment.

The example that you provided has two different exposures. The underexposed (ISO 1600) obviously will have higher noise. Unless we agree on the definition of exposure, this is going nowhere.

Are my posts invisible or something?? I can see them on my monitor. CAN ANYONE READ THIS???

Good lord, man. Your definition of exposure is wrong. I tried stating it, explaining it, then stating it again. Simplified it as much as i could, distilled it into two very simple questions. And for some reason, you don't seem to be able to read those questions. So seriously, are my posts invisible or what?

Answer those questions and you will realize your definition of exposure is wrong. ISO is part of the equation. This is not up for debate. This is beating a dead horse way above and beyond its grave.

Again, questions A and B above. Visible i hope. Answer will lead to the answer you are missing.

Bibudesh

12-14-2009, 10:58 AM

Are my posts invisible or something?? I can see them on my monitor. CAN ANYONE READ THIS???

Good lord, man. Your definition of exposure is wrong. I tried stating it, explaining it, then stating it again. Simplified it as much as i could, distilled it into two very simple questions. And for some reason, you don't seem to be able to read those questions. So seriously, are my posts invisible or what?

Answer those questions and you will realize your definition of exposure is wrong. ISO is part of the equation. This is not up for debate. This is beating a dead horse way above and beyond its grave.

Again, questions A and B above. Visible i hope. Answer will lead to the answer you are missing.

Your posts can't be read, quoted and replied bcoz you are right !..
This thread took 1 hour of my billable time to go thru wiki and read the same thing again.

synn

12-14-2009, 02:50 PM

Interestingly, something on these lines came up as I was comparing the 7D and the D300s at DxO mark right now (Both cameras were just added to the database).

Tonal Range:

Tonal range indicates how many gray levels are distinguishable up to noise in an image. The tonal range corresponds to an integration of the signal-to-noise ratio (SNR) over the dynamic range of the sensor, and is therefore representative of the average noise level. The larger the tonal range, the better the image: noise is lower, and subtle nuances can be distinguished. It is expressed on a logarithmic scale as the bit depth necessary to encode that number of gray levels: an increase of 1 bit of tonal range corresponds to doubling the number of distinguishable gray levels.

Color Sensitivity

Color sensitivity is an extension of tonal range that takes into account the color rendering involved in raw conversion: it basically indicates how many colors are distinguishable in an image, up to noise. Typically, if the sensor is slightly color blind, color rendering implies a strong digital color amplification and therefore a strong increase of noise. The greater the color sensitivity, the better the image: noise will be lower and with lower coloration. Color sensitivity is expressed on a logarithmic scale as the bit depth needed to encode a given number of colors: an increase of 1 bit of color sensitivity corresponds to doubling the number of distinguishable colors.

http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/(appareil1)/329|0/(appareil2)/336|0/(onglet)/0/(brand)/Canon/(brand2)/Nikon (http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/%28appareil1%29/329%7C0/%28appareil2%29/336%7C0/%28onglet%29/0/%28brand%29/Canon/%28brand2%29/Nikon)

Click the tonal range and color sensitivity tabs to see how noise and ISO correlate with each other. Keep in mind, this is irrespective of the Aperture and Shutterspeed. This is purely sensor performance and noise levels through the ISO scale.

(If Xavier remembers, this was pretty much exactly what I said the last time this debate came up. That the higher the ISO, the greater the gain and thus, more noise than signal in the S/N ratio.)

Vicky

12-14-2009, 07:35 PM

Exposure is determined by the triangle of Aperture, Shutterspeed and ISO. You change one, the other two get affected. That's pretty much the basic principle behind photography.

My point is that you have correctly exposed one shot and under exposed the other. This is hardly the scientific way to do it.

+1

YPut your cam in A mode or Av for canons. Point at anything, dont matter what. In A, you control the aperture. So fix it at some value. Now meter it at iso 100, then meter the same shot at iso 1600. The shutter speed goes up. Why? Because the camera is compensating for the increased iso by reducing exposure time to give an equivalent exposure.

The basic triangle of photography has remain unchanged for many many years now.

+1 again!

Don't really feel the need to waste any more words on this:)

Daniel Browning

01-18-2010, 10:41 AM

I would really like to weigh in on this thread, but first I think we need to clear up the meaning of the word exposure. Synn and Deltaone: I'm sorry to say that you guys are misinformed. I kindly suggest that we take the discussion to the following thread:

ISO is not a part of exposure.

After we discuss that, then we can move onto the meat of this thread.

Daniel Browning

01-24-2010, 10:13 PM

The problem with this experiment is that you have basically exposed the two shots at different EVs.

That is incorrect. Both shots have the exact same EV according to every authoritative definition of the term (textbooks, photography references, the dictionary, etc.).

When you take such an under exposed shot

Either both shots are underexposed or neither is, because they both have the exact same exposure. Furthermore, they were both developed for the same brightness, which makes the effective ISO to be 1600 on both.

When you take such a [shot] into post and simply boost the lightness, you are bound to get noise and you will.

The point of this post is to explain how and why ISO selection affects noise. When exposure is limited, increasing ISO reduces noise for a given output raw conversion (i.e. developer).

This is why cameras have a setting called "Long exposure noise reduction", which is basically dark frame subtraction.

You are mistaken:

If you shoot the ISO 100 shot with a long exposure and do a dark frame subtraction, the results would be significantly different.

No, they would not. Even if he did slow the shutter speed by a factor of sixteen (four stops), it still wouldn't be slow enough to enable dark frame subtraction. Again, that has nothing to do with this thread whatsoever.

Next problem: You are confusing two different types of noise.

No. Xavier has it right.

What you are demonstrating is read noise, which is and should be corrected.

Yes. I agree.

On the other hand, the dynamic range and per-pixel resolution of any camera drop as one goes up the ISO scale.

You are right about dynamic range dropping, but not per-pixel resolution. In the case of a fixed exposure, which is the subject of the thread, resolution can only stay the same or increase with higher ISO settings (due to less read noise).

Where there should be fine detail, the camera adds noise.

You have this exactly backwards. Relative to a fixed exposure (any given tonal level), the cameras adds the most amount of noise at low ISO. As the analog gain (ISO) is increased, the amount of noise added by the camera stays the same, but the tonal level in the image affected by it becomes more distant, resulting in less noise overall.

This detail cannot be recovered no matter what. This is the "High ISO noise" that people talk about in general.

No. The "high ISO noise" that people talk about is caused by reduced exposure.

Eh, Exposure is determined by the triangle of Aperture, Shutterspeed and ISO.

No. Exposure is determined by light intensity and time. There are many factors that affect light intensity, including scene luminance, flash settings, ND settings, and f-number.

You change one, the other two get affected. That's pretty much the basic principle behind photography.

No. Of the three you listed, only f-number and shutter speed are a change in exposure. (You didn't mention flash settings or ND settings, which are also changes in exposure.) ISO is not a change in exposure. Changing the ISO may or may not affect the way you want to expose for an ideal photo, but it certainly doesn't change exposure directly itself.

My point is that you have correctly exposed one shot and under exposed the other. This is hardly the scientific way to do it. Any experiment done to demonstrate objective data should be done on the level grounds and this wasn't.

You are mistaken. Both photos have the exact same exposure, and the test is on level ground. It proves that when you are limited by exposure (i.e. not in ample light), it's best to use the highest ISO that doesn't clip highlights you care about.

The correct way to use base ISO for lowlight photography is to meter properly and use dark frame subtraction.

You don't understand the point of the thread, which is to discuss the situation when exposure is limited by factors outside your control, such as motion blur or depth of field, for example. In that case, the correct way to use lower ISO settings is when you need more highlight headroom at the cost of noise. The correct way to use high ISO settings is for less noise at the cost of highlight headroom.

When you follow and incorrect procedure and claim that you've come across noise, the claim doesn't hold much value.

There is nothing incorrect about choosing exposure first and ISO second -- that is the absolute best method in low light.

As I understand, read noise is a characteristic of the sensor and remains constant no matter what image you're shooting.

Correct.

Chroma noise on the other hand occurs because the photosites fail to capture the light rays properly.

That is wrong. Chroma noise is a product of the raw conversion and spectral sensitivity of the color filter array.

On film, this is less of an issue as lightrays coming in at any angle can be captured by it. On a digital sensor, there is a limitation to the angle at which a lightray can fall and be captured.

It's true that digital sensors have angle of response limitations, but that has absolutely nothing to do with chroma noise whatsoever.

When shooting at a higher ISO, the photosite simply does not have time enough for all the frequencies of visible light to fall on it at the required angle. Hence the noise.

That's wrong. Are you just making this up as you go along?

synn

01-24-2010, 10:23 PM

Lets do this one more time, shall we?

Just because you say you're right a thousand times make you right. I've given enough samples, links and quotes in the various threads that seem to be popping up on this matter every other day to prove my view on this matter and frankly, I can't be bothered to do it again. No offense, but I'd rather be taking pictures with that time. I'm a photographer first and shall always remain to be so.

Interpret that in any way you want. Keep saying I (And every one who has a different take on the matter than you) is wrong. Some like Coke, some prefer Pepsi. And the world goes round 'n round.

Daniel Browning

01-24-2010, 10:39 PM

Keep saying I (And every one who has a different take on the matter than you) is wrong.

This isn't about individual taste or choice. These are verifiable facts. The corrections I've given you are not my personal opinion, they are statements of fact that are backed up by real references such as photography textbooks, the dictionary, international standards organizations, etc. You have only posted your opinion, and it is wrong, so I am correcting it.

Some like Coke, some prefer Pepsi. And the world goes round 'n round.

Again, this isn't a matter of taste like the choice between Coke and Pepsi. It's a matter of fact, like whether a long prime lens is called a "long lens" or a "zoom lens". The fact is that calling it a zoom lens is wrong, even if a few novices are not aware of it. In the same way, your understanding of exposure is simply wrong, not one of your artistic choices.

synn

01-24-2010, 10:44 PM

You have only posted your opinion

So you HAVEN'T read my posts.
Thanks for the clarification. I now read your posts in a new light (Pun unintended).

Daniel Browning

01-24-2010, 11:30 PM

Your experiment is wrong.

No, he got it right.

To start with your experiment is actually used as an example of how not to compare noise. Pushing a 100 iso shot by 4 [stops] does not equate to 1600 iso.

Yes it does. Both shots are the exact same ISO 1600 according to the three most popular methods allowed for the definition by the International Organization for Standardization (ISO).

The only difference is exactly the point of the thread: one is pushed four stops with analog gain, and the other is pushed four stops with digital gain.

Ok, now you just said, only shutter speed and aperture matter for exposure. This does beg the question, why is iso even there in the first place. For fun? No (just in case :P ).

ISO changes brightness -- not exposure. You think that exposure and brightness are the same, but they are not.

IT is a triangle as synn mentioned. This is something that does not need textbooks to explain. Take a couple of photos and its easy enough to see.

It's obvious from photos that ISO is not a part of exposure. The photon shot noise is the exact same from photos that have a six-stop difference in ISO setting, which proves it.

Now why is a 4 stop post overexposed picture showing noise? Not exactly rocket science there. Its in post. The sofware is trying to do the impossible. Raw has at most 2 stops of headroom in it. Not four.

These statements demonstrate that you do not understand raw photography. Raw files have whatever amount of headroom the photographer wants them to have. If you want 0 stops headroom, you expose for that. If you want 6 stops headroom, you can expose for that. What you call "raw headroom" is just the difference between the default conversion and one that takes advantage of the unused headroom that already existed. AE meters tend to pick between 1.5 and 4.5 stops of headroom, depending on settings and model, but that does not affect what is possible in the raw file.

You want to compare two isos, fine. Keep the exposure the same.

He did. What you mean is "if you want to compare two ISOs, use a lower exposure for the high ISO". That would only prove that reducing exposure increases noise for a given brightness. It would invalidate the whole point of the thread, which is to analyze the effect of ISO when exposure is the same.

Ok, even simpler. Put your cam in A mode or Av for canons. Point at anything, dont matter what. In A, you control the aperture. So fix it at some value. Now meter it at iso 100, then meter the same shot at iso 1600. The shutter speed goes up. Why? Because the camera is compensating for the increased iso by reducing exposure time to give an equivalent exposure.

No, it is not an equivalent exposure. It is equivalent brightness. The difference is vital.

Take photos and experiment. But when you do that, make sure you're doing the right experiment, not hte wrong one.

Xavier did the right experiment to prove that increasing ISO decreases noise for a fixed exposure.

Deltaone

01-25-2010, 12:34 AM

Ok, im going to go about this in another way. There is a reason why i refer to this as semantics. The point xavier makes about high iso and noise, is correct, although partial.

The real statement is not high iso causes less noise. But for a given shutter speed and aperture (exposure if you want :P) use the highest iso you can to avoid clipped highlights.

Now, here's the real problem. For most of us here, there are very few situations where you have to control both SS and aperture. Usually it is concern about motion for shutter speed or DOF for aperture that we are concerned about. Given a situation where you have no option but to change both, using the lowest iso possible is well kinda dumb. This was something i had assumed was obvious to anyone who had taken a few photographs and played around with any professional camera. The whole point of all these settings is to get an image with an acceptable lighting anyways.

Now, again, the fact that given if the first two are fixed, you have to change the third to get a result is why its called a triangle. As i mentioned in the other thread, it is to simplify this that it is called an exposure triangle, less confusion is always good, i've always thought. So given if you're under a constraint with the first two, and the base iso gives you an underexposed shot, you will increase the iso till you get the correct brightness (happy now? :P) you want. This i had assumed to be common sense. I had forgotten the oft repeated statement that common sense is not so common....

In a given sensor, there is an acceptable iso after which the signal gain of the camera produces an unacceptable amount of noise. Modern camera are pushing that number higher and higher. Real world scenario? The d70/80 struggles with iso 800 at anything like low light, the d90 does 1600 in the same situation with ease. That is a sensor issue.

That said, more light = less noise. Low iso means the other two point of the triangle get affected letting more light in, so less noise. Combine the above point together and you get the common advise, dont go too high in the iso scale. This is simple enough for most to understand.

However, if someone were to fix an aperture and ss for some reason (one would assume you have some decent knowledge of photography to want one specific or particular set of those), one would also assume, the person in question would be intelligent enough to use an iso appropriate to the situation, which makes the whole argument fascile in the first place.

The problem comes from two places.
a) people who assume low iso automatically means low noise. Wrong? yes. But doesn't really harm anyone and doesn't affect the photo either, unless theyre dumb enough to actually get an underexposed shot and then try to increase the brightness in post and then end up with noise and then complain. For such people, well that is why there are many an article and many a lolcat even :P.

b)People who assume the converse is true and simply think high iso means less noise. Which is again wrong. And actually leads to some rather funny and useless photographs which do nothing in the first place and just confuses people even more.

There is a reason why most sites and people for that matter prefer to go with A. Less confusion, more photography. One would assume the latter is actually beneficial. However, if the point is to just argue for the sake or argument, well hey, thats all well and good if you have the time and the patience for it. Which is why i prefer route A, take the simpler route and let people actually go out, take photos and learn by themselves.

A day out in the field photographing things is worth many many books of photography. -> Delta (thats me btw). :P.

Daniel Browning

01-25-2010, 04:53 AM

The real statement is not high iso causes less noise. But for a given shutter speed and aperture (exposure if you want :P) use the highest iso you can to avoid clipped highlights.

Yes, but why do you say to "use the highest iso you can"? The reason why is because it has less noise. So when exposure is limited, high ISO causes less noise.

Now, here's the real problem. For most of us here, there are very few situations where you have to control both SS and aperture.

The only time that is true is when you don't shoot in low light, so you never have a reason to leave ISO 100 anyway, and there is no point in joining a thread about high ISO.

using the lowest iso possible is well kinda dumb.

Not at all. There are many times where it is vitally important to choose a lower ISO setting in order to trade noise for additional highlight headroom. I routinely shoot ISO 400 and push 3 or 4 stops on my 5D2 in order to get additional highlight headroom. But even more often I shoot ISO 1600 and pull 2 stops in order to get less noise at the cost of reduced highlight headroom. Neither choice is dumb.

This was something i had assumed was obvious to anyone who had taken a few photographs and played around with any professional camera.

Yes, I agree that it should be obvious.

Now, again, the fact that given if the first two are fixed, you have to change the third to get a result is why its called a triangle. As i mentioned in the other thread, it is to simplify this that it is called an exposure triangle, less confusion is always good, i've always thought.

I responded to this in the other thread already.

So given if you're under a constraint with the first two, and the base iso gives you an underexposed shot, you will increase the iso till you get the correct brightness (happy now? :P) you want.

That only works for a jpeg photographer; it has nothing to do with raw photography. You assume that there is a fixed raw conversion and post processing that will occur, but in raw photography there need be nothing of the kind, so it is not the whole story.

In raw photography (low light), what matters most for selection of ISO is the highlight clipping and SNR curve (i.e. noise). The "correct brightness" can be anything the photographer wants, as long as the highlights aren't clipped and the rest of the image isn't too noisy. So if you choose the ISO setting that will give you the "correct brightness" (e.g. ISO 100), then you are going to have more noise than you would if you had chosen the one that will give you the best SNR for the highlights you care about (e.g. ISO 400) -- because you can always decrease brightness back to the "correct brightness" in post.

In a given sensor, there is an acceptable iso after which the signal gain of the camera produces an unacceptable amount of noise.

No. Again, the noise produced by the camera only goes *down* with increasing ISO. If it didn't, there would never be a point to using high ISO. And in fact, on many cameras that is precisely how it works. The only reason why people think high ISO settings cause more noise is because they reduce exposure when they use high ISO. They don't realize that high ISO with the same exposure has less noise.

That said, more light = less noise.

Agreed.

Combine the above point together and you get the common advise, dont go too high in the iso scale. This is simple enough for most to understand.

There is nothing wrong with that advice when light is ample. But the whole point of this thread is for situations when exposure is fixed and/or there is low light.

However, if someone were to fix an aperture and ss for some reason (one would assume you have some decent knowledge of photography to want one specific or particular set of those),

It doesn't require any decent knowledge of photography to need to shoot in low light. Perhaps there many photographers who never shoot in low light, but personally I do it a lot.

one would also assume, the person in question would be intelligent enough to use an iso appropriate to the situation, which makes the whole argument fascile in the first place.

No, because you are assuming that the "appropriate" ISO is the one that achieves "correct" brightness (what the photographer desires). But one point of this thread is that for raw photographers that is often suboptimal. A different ISO is needed if you want less noise (high ISO) or more highlight headroom (low ISO).

However, if the point is to just argue for the sake or argument, well hey, thats all well and good if you have the time and the patience for it.

That is not it. The purpose is to increase knowledge of photography.

Daniel Browning

01-25-2010, 05:12 AM

You fail to understand that There are three factors involved in a properly exposed photograph.

No, Xavier demonstrated his understanding of the factors just fine. It is you who mistakenly thinks that ISO is exposure.

[...] to receive the total amount of light that is required for a certain shot, the medium would have to be sensitive eonough for the selected SS+Aperture combo.

No. Exposure has nothing to do with the amount of light "received" by the medium, nor does is it changed by how sensitive the medium is. You are confusing exposure with brightness. Brightness is indeed affected by how much light is received, and it is affected by sensitivity. But exposure is not.

"The exposure for a photograph is determined by the sensitivity of the medium used.

You misunderstand what is meant on that wikipedia page. It is not saying that exposure is defined to include ISO. It is saying that in order to get a desired brightness in the developed image (i.e. converted JPEG), the choice of exposure is affected by sensitivity.

Meter a shot at say, 800 ISO in manual mode. So now the camera is at a setting that's chose totally by you. Now change the ISO to 100. Do you notice that the Meter has in fact, changed the reading? One could even say, the shot is *gasp* UNDEREXPOSED?

No. They are both underexposed. Both will have the exact same photon shot noise at every matching tonal level. The difference is that the ISO 800 shot will have less read noise, and will be multiplied by a factor of 8. Whereas the ISO 100 setting will have more read noise and must be multiplied in raw conversion to achieve the same brightness. (Or perhaps the ISO 100 shot has the brightness desired by the photographer, in which case the ISO 800 shot must be reduced in brightness to bring it down to the desired level.)

...anyway, this side discussion doesn't really have much to do with your original experiment. Please do repeat the experiment with different ISO, but this time; meter them properly. Let's discuss it then.

In real life, it's not always possible to increase the exposure. For example, it may be because you don't want motion blur. Or you want sufficient DOF. Or you're already using the most expensive lens you can afford and you can't afford one with a faster aperture. This thread addresses the optimal choice of ISO in those common circumstances.

Incidentally, some contributor to Wikipedia has already done this. Here' are two shots of the same scene by a Canon camera. Both metered properly.

http://upload.wikimedia.org/wikipedia/commons/a/a5/ISO_comparison_150px.jpg

Top image: ISO 100, f/5.6, 1/350 s

Bottom image: ISO 1600, f/5.6, 1/4000 s

http://en.wikipedia.org/wiki/Image_noise
Don't make me say "I told you so". :D

All that proves is that decreasing exposure increases noise. The point of this thread is that increasing ISO reduces noise at any given tonal level.

Daniel Browning

01-25-2010, 06:02 AM

You quoted something in bold, but you failed to notice the most important part of the sentence.

"for which the signal is not clipped."

No, Xavier did indeed notice and understand it. It is you who failed to do the latter.

In the real world, this means that the signal may be clipped from anything above the base ISO, in which case the optimal ISO IS the base ISO.

The only time that is the case is in ample light. This thread is important for low light, not ample light.

Go back a few posts and read up on how the photosites don't get enough time to capture all frequencies. This correlates with that.

No, it has nothing to do with your completely incorrect understanding of chroma noise and sensitivity over angle of incidence.

I just provided you with a real world example of how two shots, exposed the same have differing levels of noise. Just so happens that the higher ISO shot has more noise. How in the world are you going to defend it?

No defense is required because your example is not exposed the same at all, according to any authoritative definition of the word exposure:

Photography, 9th Edition, by London and Upton
The Manual of Photography by Jacobson et al.
Other photography textbooks
The dictionary
Even wikipedia (which is not authoritative)

Nevermind reading up/ asking me to read something. Go out there in the real world, repeat the same experiment and come back to me with two photographs.

I'll be right here.

Until you understand the meaning of the word exposure, there is no point.

synn

01-25-2010, 06:43 AM

Erm, the example I provided IS from Wikipedia, which quotes an authoritative source for the experiment. There are things called footnotes, you know?

Until you understand the meaning of the word exposure, there is no point.

http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=7692

Defines the exposure at the focal plane of cameras for values of two exposure parameters, i.e. field luminance and film speed.

Yeah, I guess ISO.org got it all wrong too, then. Those poor souls.

Daniel Browning

01-25-2010, 07:24 AM

Erm, the example I provided IS from Wikipedia, which quotes an authoritative source for the experiment.

First of all, the noise examples from wikipedia only prove that increasing exposure reduces noise. That has no bearing on this thread, which is about low light, when you have already increased exposure as much as you can, and now have only the choice of a variety of iso settings.

Second, that does nothing whatsoever to prove your point that ISO is a part of exposure. The example makes no such claim. In fact, Wikipedia is quite clear:

In photography, exposure is the total amount of light allowed to fall on the photographic medium (photographic film or image sensor) during the process of taking a photograph.

Third, Wikipedia is not an authoritative source, even though it happens to be correct in this case. The Image_noise page does have links to authoritative references that state that ISO is not a part of exposure.

http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=7692

Defines the exposure at the focal plane of cameras for values of two exposure parameters, i.e. field luminance and film speed.

Clearly you lack the reading comprehension necessary to decipher this simple statement, because it refutes the very point you are trying to make. Hint: "exposure parameter" is not the same as "exposure".

There are things called footnotes, you know?
[...]
Yeah, I guess ISO.org got it all wrong too, then. Those poor souls.

I find it entertaining that you resort to rude and juvenile behavior when you are proven wrong.

synn

01-25-2010, 07:29 AM

Exposure indeed is the total amount of light received, but to receive the total amount of light that is required for a certain shot, the medium would have to be sensitive eonough for the selected SS+Aperture combo.

This is the original point I made. Now where in that have i said ISO IS Exposure? I said ISO is a factor in properly exposing a shot and stand by it. Hint: They are two different things.

I find it entertaining that you have an uncanny knack of putting words in other people's mouths.

Daniel Browning

01-25-2010, 07:40 AM

This is the original point I made. Now where in that have i said ISO IS Exposure?

In the very first sentence of the very first post of this thread:

The problem with this experiment is that you have basically exposed the two shots at different EVs.

You thought they were different EV because you thought ISO affected Exposure Value, but it doesn't.

I said ISO is a factor in properly exposing a shot and stand by it. Hint: They are two different things.

I agree with you there.

I find it entertaining that you have an uncanny knack of putting words in other people's mouths.

If I'm wrong above, and you really did not say "exposed the two shots at different EVs", then I will apologize for implying otherwise. But as far as I know, you really did say that, and therefore you really did believe that the change in ISO was a change in exposure.

synn

01-25-2010, 08:14 AM

In hindsight, i shouldn't have said EV there. Yes, I agree with you on that. But my point there is exactly what I quoted above and I see you agree with it.

The thing I'm trying to put across here is that there is no SINGLE exposure for a scene. You can't meter a scene for ISO 100 and then argue that it's applicable for every film speed available. It simply isn't. Optimal exposure for EVERY scene is dependent on your film speed and it's fairly dumb if anyone says it isn't.

Daniel Browning

01-25-2010, 09:10 AM

The thing I'm trying to put across here is that there is no SINGLE exposure for a scene. You can't meter a scene for ISO 100 and then argue that it's applicable for every film speed available. It simply isn't. Optimal exposure for EVERY scene is dependent on your film speed and it's fairly dumb if anyone says it isn't.

I agree with all that as well. That's not what this thread is about.

When you have ample light, the optimal procedure is to stick to ISO 100 and then increase the exposure until it's just right. But when you do not have ample light, you have already exhausted all the methods of increasing exposure. So the only question left is what is the optimal ISO to use. The answer is this (which I'm sure you already know): the ISO with the lowest read noise that doesn't clip highlights you care about.

For example, ISO 400 has four times less read noise than ISO 100 in my 5D2. So let's say I'm shooting a low contrast scene in low light, and I've already increased exposure as much as possible. At ISO 100, the image is just perfect -- exactly how I want it: very dark and moody, with very little midtones and no highlights. I could take the picture at ISO 100 -- but that's not optimal, and that's where this thread comes in. As demonstrated in this thread, I could shoot the picture at ISO 400 with the same exposure, then reduce the brightness in post with -2 EC. Since it has low contrast, there are no additional blown highlights, and the brightness is the same. But one thing is different, the ISO 400 shot has far less noise, which means more dynamic range, less shadow noise, improved color depth, and more. For a fixed exposure, ISO selection is a trade-off between highlight headroom and noise. High ISO for less noise and low ISO for more headroom.

synn

01-25-2010, 09:20 AM

Shooting something at 100ISO and boosting it 4 stops in post is not the right way to demonstrate it.

You're right about the headroom part, which is what I need for my lowlight HDRs. I do some cityscaping where blown highlights are something I cannot afford to have. In a real world situation, my 200 ISO shots give me enough headroom and minimal noise.

http://farm3.static.flickr.com/2529/3799854913_7c93336175.jpg

Trust me on this, me and Delta tried every ISo possible there and this was the cleanest output we got. Pushing gain adds noise. There are no two ways about it. The problem, as we repeatedly say is that we're dealing with different types of noise. Arising from different things.

The difference maker is that while I can do something about noise later on, there's not a damn thing I can do about the lack of headroom.

And so, base ISO makes the most sense for me in a lowlight situation. As a photographer.

Daniel Browning

01-25-2010, 09:36 AM

Shooting something at 100ISO and boosting it 4 stops in post is not the right way to demonstrate it.

Let's set that aside for a moment and consider the situation in my last post: *reducing* brightness in post instead of boosting it. Then would you feel it is a valid way to demonstrate it?

Trust me on this, me and Delta tried every ISo possible there and this was the cleanest output we got.

I've tested the noise of each ISO setting on my camera as well.

Pushing gain adds noise. There are no two ways about it.

No, gain reduces noise relative to any given tone in the image. The reason why it seems like gain adds noise is because you do it with reduced exposure and/or you never equalize brightness.

For example, on most Canon DLSRs, when you go from ISO 100 to ISO 200 and reduce exposure by 1 stop, the reduced exposure causes noise to increase. But the read noise at ISO 200 is improved so much, that it almost (but not quite) cancels out the increase in noise caused by reduced exposure. That is why many Canon photographers shoot ISO 200 so often -- the noise is almost identical to ISO 100, but it allows a full stop less exposure.

The difference maker is that while I can do something about noise later on, there's not a damn thing I can do about the lack of headroom.

I agree.

And so, base ISO makes the most sense for me in a lowlight situation. As a photographer.

In the case of cityscapes, I stick to low ISO as well, because I can usually just increase the exposure duration. But many other low light situations I'm in do not have that luxury, and it's in those cases where a choice between noise and headroom has to be made. Most of the time I choose to reduce noise by shooting ISO 1600 and reducing brightness by 2 stops in post.

synn

01-25-2010, 09:44 AM

No, gain reduces noise relative to any given tone in the image. The reason why it seems like gain adds noise is because you do it with reduced exposure and/or you never equalize brightness.
Basic logic and electronics suggest that the more you push gain, the S/N ratio gets altered in the favor of noise. This is why I gave Xavier the example of the electric guitar amplifier. When you reduce the brightness in your high gain image, you're merely hiding the noise that arose from it.

I refer you again to this post: http://www.thephotographer.in/darkroom/showpost.php?p=11061&postcount=19

Which demonstrates sensor noise at all gain levels IRRESPECTIVE OF the aperture and shutterspeed chosen. This is measured sensor performance.

Daniel Browning

01-25-2010, 11:00 AM

Basic logic and electronics suggest that the more you push gain, the S/N ratio gets altered in the favor of noise.

Not so in this case; I'll explain below.

This is why I gave Xavier the example of the electric guitar amplifier.

Sorry, I could not see anything about an electric guitar amplifier in the thread. In any case, audio equipment makes for an *excellent* demonstration of what is happening here.

When you record a low volume sound (-40db) with a microphone and feed it into a noisy ADC (-40 db noise floor), you get a poor SNR of just 1:1. Record a loud sound (-3 dB) with the same noisy ADC, and you get a very good SNR (almost 100 times better SNR). In either case, you can normalize both recordings to the same audio level (say -12) for comparison.

Louder sound is equivalent to more exposure. Sometimes it's not possible and you just have to record whatever sound (or exposure) you can get. In those cases, it can be very beneficial to use analog gain (amplifier). Now, instead of feeding the noisy ADC with a low volume sound (-40 db), you gain up the sound to -3 db. The ADC will still add its noise, but the SNR will be improved greatly. The trade-off is just like image sensors: increased analog gain results in less headroom, which is why it has to be balanced with the benefit of less noise.

This is exactly what happens in Canon cameras. The late stage electronics (probably the ADC and other electronics) add a fixed (input-referred) amount of read noise to the signal (especially temporal FPN). When you increase gain, it increases the distance between that noise and any given tonal level in the image, just as in the audio example above.

When you reduce the brightness in your high gain image, you're merely hiding the noise that arose from it.

Huh? I can only guess that I did not make myself clear. I hope it will help if I explain it again:

The first shot is f/2.8 1/60 ISO 100 and has the desired brightness with the default conversion settings. The second shot is f/2.8 1/60 ISO 400, and it doesn't clip any detail (scene requires only 1.5 stops headroom), but it is two stops too bright with the default conversion settings. Change the conversion settings to reduce brightness by two stops. Now both shots have the exact same brightness and no clipping in either. But the second shot has much less read noise.

Does that make it clear? Or do you still feel that reducing the brightness of the high ISO shot so that it matches the desired brightness is merely hiding the noise?

I refer you again to this post: http://www.thephotographer.in/darkroom/showpost.php?p=11061&postcount=19

Which demonstrates sensor noise at all gain levels IRRESPECTIVE OF the aperture and shutterspeed chosen. This is measured sensor performance.

It does not demonstrate that, because it assumes a different exposure for each ISO setting.

(If Xavier remembers, this was pretty much exactly what I said the last time this debate came up. That the higher the ISO, the greater the gain and thus, more noise than signal in the S/N ratio.)

The chart only shows that reduced exposure makes SNR worse, it does not isolate the effects of ISO on a fixed exposure - as in a low light situation, which is the point of this thread.

synn

01-25-2010, 11:25 AM

It does not demonstrate that, because it assumes a different exposure for each ISO setting.Did I not mention that they are measuring pure sensor performance, irrespective of Aperture and SS?

I really do not have time to extract quote by quote, but please go through: http://www.dxomark.com/index.php/eng/Technologies

As for the gain part, The original reply was in one of the dozen threads on this matter.

Anyhoo, get yourself a 10W Marshall amp. They are cheap as anything. Plug a six string in. turn the volume down to 1. Turn gain down to 1. (Low light, low ISO). Give it a strum.

Now do the same with volume at 1 and gain at 10.

Once again, do the experiment with volume at 10, gain at 1 and then 10.

Get back to me if you found volume 1, gain 10 a cleaner signal than volume 10, gain 1.

Lastly, please. Do post some of your works demonstrating your first hand experiences with all these. For all I see from you are essays on these matter with minimal clicking involved. All this supposed mastery of technique should theoretically make you a master photographer beyond comparison, but you seem to present very little evidence for the same.

No offense.

Daniel Browning

01-25-2010, 11:50 AM

Did I not mention that they are measuring pure sensor performance, irrespective of Aperture and SS?

Yes, you mentioned that. And as I stated, you are wrong. I know far more about the dxomark measurements than you do; I have taken plot points from their chart and reverse engineered their data to extrapolate input referred quantities such as read noise and full well capacity in electrons using a least squares fit. For example, I extrapolated 4200 e- FWC and 3.4 e- read noise for the 5D2 in order to compare DXOMark's measurements of the 5D2 with my own camera.

I really do not have time to extract quote by quote, but please go through: http://www.dxomark.com/index.php/eng/Technologies

Nothing in there supports your position.

Anyhoo, get yourself a 10W Marshall amp.

As I said, that is not the case with image sensors. Input-referred read noise goes down with higher analog gain in every single measurement out there, including dxomark.com, clarkvision.com, my own, etc. I gave you an audio analogy that actually reflects the reality of the situation, which is not the case with your guitar amp analogy.

All this supposed mastery of technique should theoretically make you a master photographer beyond comparison, but you seem to present very little evidence for the same.

No offense.

None taken, as even a fool knows that photography is far more than knowing the optimal choice of ISO. Perhaps you'll learn that some day and then understand why your childish provocations are void of effect.

No offense.

synn

01-25-2010, 12:12 PM

Yet another essay, without the requested evidence. Why are you so afraid of doing the amplifier experiment? Because it'd bring your theory crumbling down? signal strength and gain are the same for any sort of electronic equipment that receives an analog signal, whatever form of energy it may be in. Hint: This test has been done within the context of photography before. Google is your friend.

"The SNR decreases with the gain, or equivalently, with the ISO settings."
http://www.dxomark.com/index.php/eng/Technologies/Noise-characterization/Noise-in-mid-tones

"

Shadows (http://www.dxomark.com/itext/tech_ess_char/image024.gif): the SNR increases 6dB for every EV and loses 6dB for each doubling of the ISO setting.
Midtones (http://www.dxomark.com/itext/tech_ess_char/image025.gif): the SNR increases 3dB for every EV and decreases by 3dB for each doubling of the ISO setting.
Highlights (http://www.dxomark.com/itext/tech_ess_char/image026.gif): the SNR is constant and does not depend on the ISO."

http://www.dxomark.com/index.php/eng/Technologies/Noise-characterization/Summary

But it's ok, we get it. DxO has it all wrong, you're the correct one.

Anyway, I just wasted 20 mins of my time looking for pictures that you have clicked. Nothing on Flickr, but I did come across your account in DPReview. After a LOT of clicking, I found ONE picture. Well, two if you count the out of focus one that you posted initially. (Kinda shocking how all this number crunching makes a photographer's eyes blind towards the obvious loss of focus).

What I DID find there though, are umpteen neverending debates in the same vein. (To be fair, MOST people on DPReview do the exact same thing). Going through those reminded me of some self proclaimed experts in Harmony Central who go on and on about how Eddie Van Halen's technique isn't right or how George Lynch would fail a basic exam in music theory, but can't cut a semi-decent solo if their lives depended on it.

Speaking of which,

I know far more about the dxomark measurements than you do; I have taken plot points from their chart and reverse engineered their data to extrapolate input referred quantities such as read noise and full well capacity in electrons using a least squares fit. For example, I extrapolated 4200 e- FWC and 3.4 e- read noise for the 5D2 in order to compare DXOMark's measurements of the 5D2 with my own camera.But it STILL haven't made you a good photographer. I'm not even remotely joking when I say I've seen Delta throw away flower shots which looked way better than the one you posted the other day. Compositionally AND technically. With bottom of the barrel equipment.

The likes of you thrive on places like DPReview where people can spend a lifetime arguing over charts and numbers without ever bothering to go outside to shoot something. Places like this however, are filled with PHOTOGRAPHERS. There are kids here who know jack squat about theory and have churned out amazing pics, way better than the 4 you have posted in your entire tenure. Now I don't proclaim to be an expert artistically or mathematically when it comes to photography, but I HAVE learned a thing or two in my humble time spent with the artform and I'm only glad to help them out with whatever I know. Not bombard them with essays that reek of narcissism. (Half of which aren't accurate, incidentally. I could post another dozen links that disprove your theory that noise doesn't increase with gain, but I doubt if it'd make a difference).

Because it's all about becoming a better PHOTOGRAPHER.

as even a fool knows that photography is far more than knowing the optimal choice of ISO. Perhaps you'll learn that some day ...and then what? Become as good a PHOTOGRAPHER as you? Don't wanna brag or anything, but I think I've got that covered.

People like Vicky, Ken etc. teach others new things every now and then, but ALWAYS back it up with real world examples. Not you. Never you. you could probably read a textbook on photography backwards doing a handstand, but I seriously doubt if you could take a real world picture demonstrating the same.

As I said in my first post yesterday, I had zero interest in pursuing this debate with you because I had every feeling that you'll get the last word in come what may, refuting all evidence presented otherwise. Well, I don't have anywhere near the patience that the DPReview guys, so congrats! You're the first guy ever to be on my ignore list from this moment on.

p.s. Someone please let me know if this dude actually posts a picture one of these days. I most definitely would want to take a look at that.

Daniel Browning

01-25-2010, 12:36 PM

Because it's all about becoming a better PHOTOGRAPHER.

If you really believed that, you wouldn't have to resort to these sad little tirades every time you are proven wrong. You choose to ignore the facts instead of learning something new about photography. I share my photographs with colleagues and clients in the form of prints. In real life. So I have no need to share them online; certainly not to prove anything to the likes of you. I participate online for the more advanced discussions that I cannot have with colleagues in real life. If I do post my photographs online, it will certainly not be to satisfy your adolescent preconceptions about the validity of experimentally verifiable information.

Deltaone

01-25-2010, 04:08 PM

OMFG. My poor head, it aches. Daniel, im sure you think, taking apart a post and replying in sentences seems rather intelligent and logical to you. But here again, i have to say this. You are missing the goddamn forest for the trees. Try to understand some basic linguistics. A sentence is not merely a collection of the meaning of its word and a paragraph is not merely a collection of the sentences. Which is why replying on a sentence to sentence basis is quite frankly rather dumb. I cant be bothered with quoting that whole thing i mentioned earlier.

What you would have realized if you had actually taken the time to read it, is that you have actually just gone and said the same thing ive said as though you were refuting something. Good lord man, take the time to read and understand a post first, we can do the arguing after. Do you realize you just restated exactly pretty much what i was trying to say in a different way?

Now, i hope i dont have to repeat all that i said before. There is a reason why it is called select an iso as high as you can.

Now, past a certain iso a sensor will push up the gain in favor of noise. THIS IS NOT DEBATABLE. IT IS A FACT. Try to get this through your head. You talk of reverse engineering dxo results and what not, but i do wonder how this rather simple and basic aspect of physics got right past you. You talk about highlight room on your 5dm2. Let me ask you this, have you used any of the older gen cameras? IF you have, you would quite easily know what im talking about.

Light reduces noise, which is quite obvious. High iso will work in a situation where there is plenty of light but you still need a high shutter speed for example. I've used 1600 iso oh the d70s in good lighting conditions and gotten good photographs. Try 400 iso in a low light situation with dark color and blacks in it, and the photo turns out unusable. That is a sensor limitation. In that particular case, the sensor has to push the gain to a point where the noise overwhelms the image and there just isn't enough light.

Now, in low light situations, HIGH ISO IS DETRIMENTAL DEPENDANT ON THE SENSOR. Forget dxo, forget every other thing. Why the hell do you think we upgrade to better cameras and use the higher iso settings? WHy is the D3 or the D700 considered so much better in high iso settings than the other low end bodies? Why does the 1000D for a good canon example, do so much better than the 400D at 1600 ISO? Advancements in sensor technology.

There are many kinds of noise. However, dependent on sensor there are limits to how much you can push your damn iso dependant on the light available.

Here are some facts for you since you seem to love them. There is no correct lighting for any given scene. It is whatever the photographer want to get out of the scene, hence the term creative control. The whole point of a dslr is creative control, to get something out of a scene that i want to. Now, that said, unfortunately do have to live with certain laws of physics. As technology improves, sensors improve as well, which is why modern gen dslrs can do high iso a hell of a lot better than older ones.

NO, simply saying high iso reduces noise is just FLAT OUT WRONG. (Note i dont like using capitals to make a point, but it just seems you are hell bent on selective reading, maybe this will help.)

More light = less noise.
High iso past a certain point (Dependant on sensor and light available to sensor) introduces more noise.

For a given scene, assuming the light is low, and you have to use a specific aperture and shutter speed combo (again rare, as usually you want to control either DOF or motion blur, and rarely both, unless you're shooting sports or fast moving objects), you have to ramp up the iso. Now, there is a reason why camera makers charge quite a bit of money for their high end models and why only those high end models are capable of high iso performance the low end models can only dream about. You want low light sports shooting? you get a damn D3 or the 1dm3 for canon. You cant do that with a 1000D or a d60, they just dont do the kind of high iso that you want.

So far, i had thought that your beef was with the definition of teh term exposure. Which is fine by me. THe whole point of this thread is people reading some articles and not really understanding what it is about. Which is why in my previous post i had clarified. I do interestingly note that you kind of left out the last part where i explain WHY such explanations are simplified adn the harm in doing them in the two ways outlined.

However, it seems now from your posts, that you are now arguing that high iso = less noise period and that is not dependant on anything else. In which case that is just flat out wrong and i dont even need to point to any textbook or authoritative source anywhere. A camera and a few minutes is all thats needed.

Which is why my original statement, for a given aperture and SS combination, use the highest iso to avoid clipping highlights. Dont just use the highest iso your camera has, that is just dumb. It wont work. Unless you are now stating every single camera tester and testing place in the world is flat out wrong, as are the camera manufacturers, and of course a few trifling laws of physics?

Heres the thing, arguing can be fun, i love arguing, however, there is no point in arguing if you're going to be a selective reader and just going to say "im right because im right". So this time, try to read through, understand, and then you can post something as a reply.

As i kind of made a point in the post above, it is dependant. This is something every single photographer realizes after you play around with it after a while.

synn

01-25-2010, 04:25 PM

A camera and a few minutes is all thats needed.

You see, that's only applicable to people who use cameras for real world applications. I'm not sure if our friend here is into all that.

"However, just as very sensitive films are known to be very grainy, parallels can be drawn for digital cameras, since high sensitivities are related to high gain and noise amplification."

"While it is a common practice for camera vendors to emphasize high ISO settings on their cameras, it must be said that high ISO does not mean good image quality. Any serious photographer knows that the lowest ISO should be used to shoot a scene with a longer exposure time."

http://www.dxomark.com/index.php/eng/Technologies/Measurement-definitions

But of course, we've all been conned by DxO into believing in something that's clearly not true!

http://farm4.static.flickr.com/3439/3846856909_4b79c0b9bb.jpg

200 ISO. But hey, what do I know? Should'a used 1600 instead!

Daniel Browning

01-25-2010, 11:41 PM

Do you realize you just restated exactly pretty much what i was trying to say in a different way?

That is not what happened.

Now, past a certain iso a sensor will push up the gain in favor of noise. THIS IS NOT DEBATABLE. IT IS A FACT.

You are mistaken. This is very easy to prove experimentally:

http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8201_rawnalyze_0ev.png

http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8202_rawnalyze_-4ev.png

The noise produced by the camera only goes *down* with increasing ISO. The only reason why you think high ISO settings cause more noise is because you reduce exposure when you use high ISO or don't equalize brightness. You don't realize that high ISO with the same exposure has less noise.

You talk of reverse engineering dxo results and what not, but i do wonder how this rather simple and basic aspect of physics got right past you.

You are still wrong, as proven in the experiment above.

You talk about highlight room on your 5dm2.

No I didn't. I already explained that highlight headroom is a choice of the photographer - it does not vary with camera unless you blindly follow the autoexposure meter.

Let me ask you this, have you used any of the older gen cameras?

Of course. Nothing older than a 10D (not that it has any relevance).

Light reduces noise, which is quite obvious. High iso will work in a situation where there is plenty of light but you still need a high shutter speed for example. I've used 1600 iso oh the d70s in good lighting conditions and gotten good photographs. Try 400 iso in a low light situation with dark color and blacks in it, and the photo turns out unusable. That is a sensor limitation. In that particular case, the sensor has to push the gain to a point where the noise overwhelms the image and there just isn't enough light.

No. The reason why your ISO 400 shot looks worse has nothing to do with what you mentioned -- it's simply due to different white balance and analysis of lower values that received less light.

Now, in low light situations, HIGH ISO IS DETRIMENTAL DEPENDANT ON THE SENSOR.

That is very wrong and terrible advice.

Forget dxo, forget every other thing. Why the hell do you think we upgrade to better cameras and use the higher iso settings? WHy is the D3 or the D700 considered so much better in high iso settings than the other low end bodies? Why does the 1000D for a good canon example, do so much better than the 400D at 1600 ISO? Advancements in sensor technology.

That's correct, but none of it does anything to support your position.

There are many kinds of noise. However, dependent on sensor there are limits to how much you can push your damn iso dependant on the light available.

It's true that there are many kinds of noise, but your representation of how to push the ISO is badly mischaracterized. The truth is that in low light, high ISO has less noise and low ISO has more headroom. That is all.

Here are some facts for you since you seem to love them. There is no correct lighting for any given scene. It is whatever the photographer want to get out of the scene, hence the term creative control. The whole point of a dslr is creative control, to get something out of a scene that i want to.

I agree, and there is nothing I've said that contradicts that.

Now, that said, unfortunately do have to live with certain laws of physics.

I find it entertaining that you throw around the phrase "laws of physics" yet lack even a basic understanding of the elements involved.

As technology improves, sensors improve as well, which is why modern gen dslrs can do high iso a hell of a lot better than older ones.

I agree, but again, irrelevant.

NO, simply saying high iso reduces noise is just FLAT OUT WRONG.

It takes only 5 minutes to do an experiment that proves you are wrong, as I did above. The reason you are misled is because you don't understand the implications of fixed exposure and brightness.

For a given scene, assuming the light is low, and you have to use a specific aperture and shutter speed combo (again rare, as usually you want to control either DOF or motion blur, and rarely both, unless you're shooting sports or fast moving objects),

I find it entertaining that you still don't understand the situation where this thread applies. It's not "rare" -- it's every low light situation. Every circumstance where the photographer cannot increase the exposure as much as desired.

you have to ramp up the iso. Now, there is a reason why camera makers charge quite a bit of money for their high end models and why only those high end models are capable of high iso performance the low end models can only dream about. You want low light sports shooting? you get a damn D3 or the 1dm3 for canon. You cant do that with a 1000D or a d60, they just dont do the kind of high iso that you want.

True, but again irrelevant.

So far, i had thought that your beef was with the definition of teh term exposure. Which is fine by me.

The definition of the word exposure is a very simple topic. It doesn't even require a 5 minute experiment to prove: all you have to do is look it up in the dictionary. But I still had to correct you on it before I even start to address your misconceptions about ISO.

THe whole point of this thread is people reading some articles and not really understanding what it is about.

No, Xavier understood everything just perfectly. It is you who cannot see past your own dearly-held misconceptions in the face of incontrovertible evidence.

Which is why in my previous post i had clarified. I do interestingly note that you kind of left out the last part where i explain WHY such explanations are simplified adn the harm in doing them in the two ways outlined.

Your explanations are wrong.

However, it seems now from your posts, that you are now arguing that high iso = less noise period and that is not dependant on anything else.

No. I have made my point very clear that it depends on a fixed exposure (i.e. it only applies to every low light situation).

In which case that is just flat out wrong and i dont even need to point to any textbook or authoritative source anywhere. A camera and a few minutes is all thats needed.

Funny that you have not taken the few minutes to post any raw files to prove it. If you did try to, you would find you are wrong.

Which is why my original statement, for a given aperture and SS combination, use the highest iso to avoid clipping highlights.

I agree with that.

Dont just use the highest iso your camera has, that is just dumb.

Of course that would be dumb. Xavier explained this in the first post.

Heres the thing, arguing can be fun, i love arguing, however, there is no point in arguing if you're going to be a selective reader and just going to say "im right because im right". So this time, try to read through, understand, and then you can post something as a reply.

If only you lived up to your own words.

Daniel Browning

01-25-2010, 11:59 PM

I see that synn has again edited his post to add new comments. I'll go back to it and reply to the additional comments.

Why are you so afraid of doing the amplifier experiment? Because it'd bring your theory crumbling down?

As I said, your amplifier "experiment" (ha!) bears no resemblance at all to real life because the read noise is caused by late stage camera electronics such as the ADC.

signal strength and gain are the same for any sort of electronic equipment that receives an analog signal, whatever form of energy it may be in. Hint: This test has been done within the context of photography before. Google is your friend.

I feel sorry for you.

"The SNR decreases with the gain, or equivalently, with the ISO settings."
http://www.dxomark.com/index.php/eng/Technologies/Noise-characterization/Noise-in-mid-tones"

Shadows (http://www.dxomark.com/itext/tech_ess_char/image024.gif): the SNR increases 6dB for every EV and loses 6dB for each doubling of the ISO setting.
Midtones (http://www.dxomark.com/itext/tech_ess_char/image025.gif): the SNR increases 3dB for every EV and decreases by 3dB for each doubling of the ISO setting.
Highlights (http://www.dxomark.com/itext/tech_ess_char/image026.gif): the SNR is constant and does not depend on the ISO."

http://www.dxomark.com/index.php/eng/Technologies/Noise-characterization/Summary

Again, all of that refers only to the effects of reduced/increased exposure. It does not address how read noise varies with analog gain (ISO setting).

But it's ok, we get it. DxO has it all wrong, you're the correct one.

DxO has everything correct; it is you who lacks the basic reading comprehension necessary to understand the DxO web site.

Now I'll respond to synn's new post:

"However, just as very sensitive films are known to be very grainy, parallels can be drawn for digital cameras, since high sensitivities are related to high gain and noise amplification."

Again, it's funny that you are quoting DxO without understanding it. The noise here refers to floating exposure.

"While it is a common practice for camera vendors to emphasize high ISO settings on their cameras, it must be said that high ISO does not mean good image quality. Any serious photographer knows that the lowest ISO should be used to shoot a scene with a longer exposure time."

http://www.dxomark.com/index.php/eng/Technologies/Measurement-definitions

Wow, you still don't get it? They added a clarification that should make it simple enough even for you: "with a longer exposure time". Thus, the noise is caused by lower exposure times. This thread is about low light situations when exposure time cannot be increased.

http://farm4.static.flickr.com/3439/3846856909_4b79c0b9bb.jpg

200 ISO. But hey, what do I know? Should'a used 1600 instead!

This is funny for several reasons. First that you think such a tiny web image offers any input one way or the other. I can post an ISO 12,800 shot at that size that has no visible noise.

Second, because you confuse the difference between low ley and limited exposure. You had the luxury of choosing a long exposure duration, but in many low light situations that is not possible due to motion blur.

Third, because ISO 1600 would indeed have reduced noise at the cost of highlight headroom. Not that the difference would have been visible in your thumbnail-sized web image, but the fact remains. That is why high ISO settings offer the benefit of reduced noise when highlight headroom is worth the trade - clearly not the case in your example of a long exposure duration shot.

It would have been better for you to remain silent and be thought a fool; now that you opened your mouth you have removed all doubt.

Xavier

01-26-2010, 12:00 PM

Wow! A lot has happened here since I last posted. Anyway, In a nutshell, most of us do the right thing, but almost everyone does it under a wrong assumption. What's wrong in correcting wrong notions that people have? A photographer might be pushing his ISO in low light conditions because he knows that shooting at low ISO and pushing it in post will create a very noisy image, but does he/she know why this low ISO produces so much noise? One can call brightness and exposure the same thing, but it doesn't change the fact that they are as different as zooms and primes.

P.S : A lot of personal attacks on this thread was totally uncalled for. We're talking about something technical here, not the aesthetic value of a person's photographs.

synn

01-26-2010, 12:23 PM

Xavier: Since you're a person who walks the walk and talks the talk, I'll tell you why there's a lot of disagreements.

The very fact that some people argue that a higher ISO shot is "Underexposed" because it doesn't get as much light as a lower ISO shot is the bone of contention here. Is it really "Underexposed"?

I point you to a little book called "Understanding Exposure" written by a person who has forgotten more about photography than anyone here. Mr. Bryan Peterson. He states that

"The choice of Aperture and shutterspeed is directly influenced by the ISO selected".
(p14)

If you read on, you will understand what he is saying. Whatever combination of Aperture and SS that you arrive at is the RIGHT one (At least 6 such combos exist btw. but that's besides the point) for THAT ISO. So if you're shooting in Aperture Priority at f1.8 and get 1/125 for ISO 100 and say, 1/500 for ISO 400, they are both CORRECT exposures for THAT ISO. To say either one isn't the right exposure is simply incorrect.

With that in context, a higher ISO shot IS more noisy. It's not that complex a concept and overthinking is just gonna make life more difficult for most.

...and before a certain somebody interjects with his very own interpretation of clear, concise text YET AGAIN; let me re-iterate that I have less than zero interest in holding a conversation with a textbook photographer. Unless you have your own photos to back yourself up with, get a new hobby. Such people are no better than the fat guy down in the pub who goes "Rossi shouldn't have downshifted and should've taken the inside line" and so on.

Xavier, I highly recommend that you read this book for answers to a lot of your questions.

Daniel Browning

01-26-2010, 12:58 PM

"The choice of Aperture and shutterspeed is directly influenced by the ISO selected".
(p14)

That's true. And it doesn't conflict with my position. If some one said that Bryan Peterson is saying that ISO is a part of exposure just because it affects the choice of exposure, they would be wrong.

Furthermore, choosing ISO first and then exposure is the optimal procedure for beginners, JPEG photographers, ample light, film, etc. But it is not the optimal procedure for low light and advanced raw photography, which is best served by first maximizing exposure as much as possible, then letting that affect the choice of ISO.

With that in context, a higher ISO shot IS more noisy.

True. But that's not the context of this this thread, which is to optimize the best choice of ISO to reduce noise in low light.

Xavier, I highly recommend that you read this book for answers to a lot of your questions.

Xavier is way past beginner-level books like "Understanding Exposure"; he's not afraid to learn new information, even if it conflicts with his previous understanding.

...and before a certain somebody interjects with his very own interpretation of clear, concise text YET AGAIN; let me re-iterate that I have less than zero interest in holding a conversation with a textbook photographer. Unless you have your own photos to back yourself up with, get a new hobby. Such people are no better than the fat guy down in the pub who goes "Rossi shouldn't have downshifted and should've taken the inside line" and so on.

When new evidence is presented that contradicts the prior understanding, there are two responses. The wise man accepts that his understanding was wrong and gains knowledge. The fool clings to his misconceptions and resorts to petty talent contests.

synn

01-26-2010, 01:00 PM

***you are ignoring this user***

best.post.ever.

Xavier

01-26-2010, 01:42 PM

Synn, I have read Understanding Exposure and a couple of other Peterson books. His books good no doubt, and what you quoted is true too. But in no way does it say that ISO is a 'part' of the exposure equation. Again, as Daniel has mentioned on this thread and the other one, Scene brightness is very important to most photographers since this determines how good the final output really is. I agree with this, but that does not warrant for a change in the definition of Exposure which in no way has ISO in the picture.

As I said, most of us are doing the right thing (Increasing ISO only in low light situations since doing the same in well lit scenes would cost valuable headlight room), but most of us do not know why we're doing it.

synn

01-26-2010, 02:02 PM

But in no way does it say that ISO is a 'part' of the exposure equation. That is not what I said right now. I said what ever exposure your meter suggests is RIGHT for THAT particular ISO. So the meter readings I get for 100 ISO and 1600 ISO are both RIGHT. Neither is underexposed.

With this in mind and unless we are shooting moving subjects, there is absolutely no need to use a high ISO setting. The example you have posted in the first post is of still life, which isn't going anywhere while you shoot at 100 ISO. If you're shooting moving object, well then; even a novice knows a higher ISO setting is desired.

Two shots, taken with only a sliver of light. one metered at 200 ISO (I accidentally named the file 100ISO. Doesn't matter) and the other at 1600 ISO. Both CORRECT exposures for THAT particular ISO setting.

ISO 200:

http://i226.photobucket.com/albums/dd9/first_synn/Test_2/ISO100.jpg

ISO 1600:

http://i226.photobucket.com/albums/dd9/first_synn/Test_2/ISO1600.jpg

ISO 200 crop:

http://i226.photobucket.com/albums/dd9/first_synn/Test_2/ISO100crop.jpg

ISO 1600 crop:

http://i226.photobucket.com/albums/dd9/first_synn/Test_2/ISO1600crop.jpg

Scene brightness is very important to most photographers since this determines how good the final output really is.

I see the same amount of scene brightness in both images. And a whole lot more noise in the ISO1600 shot. And like I said, the subjects weren't going anywhere, so why exactly would I want to use ISO 1600?

Deltaone

01-26-2010, 02:36 PM

synn

01-27-2010, 09:45 AM

Delta nails it in one go.

...and now, here's evidence presented in YOUR way. Same Aperture and SS, two different ISOs (Lord knows why this is even relevant in the real world, but let's do it nonetheless).

Nikon D70, Sigma 50 f2.8 Macro, A = f2.8, SS = 3 secs. EXTREME low light conditions. (No tripod, but I rested it on level ground. Hence the slight difference in framing).

Here's ISO 200: (Image A)

http://i226.photobucket.com/albums/dd9/first_synn/iso_200.jpg

Here's ISO 1600: (Image B)

http://i226.photobucket.com/albums/dd9/first_synn/iso1600.jpg

(Both images exported as RAW from Lightroom with zero noise reduction, same WB and color profile. Maximum quality).

Now here's the ISO 1600 shot with 3 stops cut from the "Exposure" slider: (Image C)

http://i226.photobucket.com/albums/dd9/first_synn/iso1600-3stop.jpg

Ok, now LR automatically boosts the brightness etc. based on your camera model, so I zeroed down the ISO 200 shot (The one I metered for). (Image D)

http://i226.photobucket.com/albums/dd9/first_synn/iso_200_equalized.jpg

...and here's Image C equalized to match image D:

http://i226.photobucket.com/albums/dd9/first_synn/iso_1600_equalized.jpg

----

Now let's compare crops, shall we?

ISO 200
http://i226.photobucket.com/albums/dd9/first_synn/iso_200_equalized_crop.jpg

ISO 1600

http://i226.photobucket.com/albums/dd9/first_synn/iso_1600_equalized_crop.jpg

ISO 200

http://i226.photobucket.com/albums/dd9/first_synn/iso_200_equalized_crop2.jpg

ISO 1600

http://i226.photobucket.com/albums/dd9/first_synn/iso_1600_equalized_crop2.jpg

ISO 200

http://i226.photobucket.com/albums/dd9/first_synn/iso_200_equalized_crop3.jpg

ISO 1600

http://i226.photobucket.com/albums/dd9/first_synn/iso_1600_equalized_crop3.jpg

Now what was that about higher ISOs being cleaner again?

------

This is EXACTLY what Delta was talking about. Not all sensors are alike. No way in hell. Sensor technology advances as companies find new ways to get CLEANER HIGHER ISO OUTPUT. This same experiment may show no apparent difference at all on say, a 1D Mk IV or a D3S coz their ISO1600 is so goddamn clean, but try again with ISO 200 vs ISO 102400. I guarantee ya, you'll get the same results. Even worse.

Higher gain adds noise. It's a FACT and no amount of equalizing and whatnot can hide it beyond a limit. That's simple physics. My camera is limited to ISO 1600 and not say, ISO 102400 because IT WOULD GIVE PISS POOR RESULTS IN ANYTHING ABOVE ISO 1600. (Sorry for the capitilization, but it's for the benefit of those who practice selective reading).

That's all we've got. I don't think we need to or would say anything more.

Daniel Browning

01-27-2010, 01:00 PM

And you have done it again. You agree that newer gen cameras do better high iso than older ones, you agree newer gen sensors do better high iso than older ones, you agree higher end cameras do better iso than lower end ones, and yet all that is irrelevant? Seriously?

How on earth is that irrelevant. Doesn't that prove the point that each sensor has a limit beyond which the pictures become useless? Or has even that gotten past you? How much simpler does it have to get?

Everything you're saying hinges entirely on changing the exposure and/or brightness at the same time as ISO. My position is about circumstances where that is not possible, such as in low light, where you cannot increase the exposure any more. If you were to consider things from that angle you would realize that every camera is the same in this regard, new and old: high ISO has the same or less noise as low ISO.

Now, you post two screens from a software comparing two shots. I note the shots in question have the same A and SS values but differing iso. I further assume that its a 1:1 crop from what i can see there. Now i have no idea of what that software is and what many of the other numbers mean (i am ignorant in a great many things :P ). That said, did you just take two shots of the lens cover or something black and then go 1:1 and prove your point?

No, the lens cap was not on. The software is Rawnalyze. The selected portion of the image is dominated by photon shot noise, so according to my position, both images should have the same noise level when shown at the same brightness. In order to show them at the same brightness, I darkened the ISO 1600 shot by four stops. It proves to things: that high ISO has the same or less noise as low ISO (when exposure/brightness is fixed), and that there is a trade off between clipped highlights and noise (these images showing extremely bad clipped highlights to emphasize the point).

Here are the raw files:

http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8201.CR2

http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8202.CR2

Let me know if you have any other questions about the comparison.

This is why i say selective reading is bad for you.

For someone who says that so frequently, I find it interesting that you have still failed to grasp the very premise of this entire thread: low light photographer where exposure is fixed. You are still making statements that are fully and completely predicated on the idea of using less exposure for higher ISO. It is the reduced exposure and/or changed brightness, not the high ISO, that is the cause of the effects you see and talk about.

Now, i dont do much low light shooting unless im doing long exposure shots because the D70 just cannot do what i want it to do. That said, usually i shoot in Aperture priority, i select the DOF i want, let the camera choose the shutterspeed. In case i have no tripod on hand (which is most of the time) and my hand jsut cant do justice to the shutterspeed indicated (which is also most of the time), i increase my iso to get the shutterspeed up to a reasonable amount.

That is suboptimal technique. Increasing the ISO only enough to get the shutter speed up to where you want it can very often result in an ISO that is lower than it could be. The optimal technique is to increase exposure as much as possible (same as you did), but then increase ISO as much as possible.

For example, if the strongest exposure you can manage is f/2.8 1/60 (for whatever reason, e.g. motion blur), and at ISO 200 you get the desired brightness, then the optimal technique would be to *increase* the ISO even further, and leave exposure the same. The more you increase the ISO, the less noise you'll have in the raw file. The cost, of course, is highlight headroom. If you can only increase 1 stop before clipping highlights, that's something. Sometimes you can increase 2 or 3 stops, and the noise benefit will be great. Other times you'll have to reduce ISO to get more headroom.

Now, if i were to show some samples of how badly the sensor on the D70s does with higher iso, you would undoubtedly say that is because the shot is underexposed and not because of the high iso. Now you also mention having used cameras as old as the 10d? IF so you would know how higher iso in the 5dm2 is much more usable than in say the 350D.

Yes, but it is again irrelevant. All cameras, new and old, have less noise (or same) at higher ISO for a fixed exposure/brightness.

Heres another experiment for you then. Take two cameras, in the case of nikon, a D300 and a D70. For canon i guess 350D and 1000D maybe? (not very sure on the iso ranges of especially the older canons).

Identical settings for both. Same scene. Fix whatever A and SS you want. Put them both at 1600 ISO. Take the shot. Now, everythign else is the same. Only the camera itself differs. According to what you say, the noise should be the same? Because there is no other difference right?

Nothing I said implies that the noise should be the same between different cameras. I have no idea where you come up with this stuff.

Now, heres a simple law of physics for you. There is no such thing as a perfect amplifier. Not so far anyways. Which is why there is a limit to which you can push the gain on any signal before you get noise. Which is why as technology advances the amount you can push increases.

I already explained this to you. The late state electronics (e.g. ADC) add a fixed amount of noise relative to signal. They are the primary source of read noise in most DSLRs. When you amplify the signal (high ISO / high analog gain), their fixed contribution is now relatively less to any given tonal level.

Why is this so hard to grasp? Better sensor = better noise control. Better and newer sensor = better high iso performance. This is something noone argues about, which is the whole point of all these higher end new gen cameras (other than the build quality) which is why they keep pushing the highest iso they can go to in each new gen.

I can't believe you spent so many words arguing with a straw man. I never said anything of the kind.

Why do you think camera makers restrict each sensor to a certain iso? Because theyre bored?

You've knocked another straw man down, but you still haven't address my position at all.

And no camera so far can actually go to the highest iso it can go to, without degradation. This is so blatantly obvious that i don't know how you can just skip part it.

Again, your statement here is predicately wholely on changing exposure and brightness at the same time as you change ISO. My entire position is based on keeping those fixed. If you did that, you would see that you're wrong.

See, this is why i said, don't do selective reading. You just oppose selected sentences and say irrelevant or something else to the parts that don't fit your argument? Seriously?

I disagree.

Using an iso appropriate to a scene is kind of a given factor. But for reasons i still don't understand, you seem to labor under the misapprehension that regardless of camera or sensor, higher iso will always do better in noise than lower iso? How on earth can you get to that particular conclusion is the part which completely escapes me.

I get to that particular conclusion because of evidence. When exposure/brightness is fixed, high ISO always has the same or less noise.

After yet another rather long post, do you now understand why i quoted the difference between various sensors, prices and a few other things in my previous posts? Is that still considered "irrelevant and immaterial" or would you clarify on why it is irrelevant and immaterial?

It is irrelevant and immaterial because my position applies equally to all sensors and prices. For a fixed exposure/brightness, even cheap and old sensors have less noise at high ISO. Similarly, expensive new sensors have less noise at high ISO.

PS: Synn has just gone and posted a couple of pics which well actually do the same thing im trying to explain here.

In his first comparison, he changed the exposure by three stops, so it has nothing to do with my position or this thread, which is the times when you don't have the luxury of increasing exposure (i.e. low light).

PPS: I don't post in paragraphs so that it is easier for you to quote and rebut each. That is pointless unless i was making a separate and distinct point in each paragraph. I do paragraphs to make reading easier. So please, at least this time, read the whole thing, and then reply, i do note rather interestingly that you have done the same thing in my last post, read the parts you want, interpreted it in your own way, and then replied with the assurance no doubt that you are right. (which is nice, but since this is supposed to be an argument, i kinda like it when people actually understand what im saying and then refute said argument, which makes arguing all the more fun, dont you think?)

This time I was very careful to read your entire post several times before I wrote my response. I cannot see where I misinterpreted your posts, though I'm sure I made some mistakes somewhere. Sorry for that. If you point them out, I will apologize in specific.

Daniel Browning

01-27-2010, 01:23 PM

...and now, here's evidence presented in YOUR way. Same Aperture and SS, two different ISOs (Lord knows why this is even relevant in the real world, but let's do it nonetheless).

The reason why it is relevant in the real world is this. One technique results in the less noise. The other technique results in more noise. By learning the technique, you have the information necessary to make the optimal trade between noise and highlights.

You did not post any raw files. All you've proved is that Adobe renders files differently depending on ISO and settings. Even your brightness and color are not equal between the shots. You need to use a raw converter that is capable of converting both shots the exact same way except for one minor difference: one single mathematically multiplication of the pixel values. Rawnalyze is the best tool for this purpose, but other converters like RawTherapee can do it too.

You can use yousendit.com to upload raw files for free. Or, you can open the raw files yourself in Rawnalayze, and when you realize you were

Higher gain adds noise. It's a FACT and no amount of equalizing and whatnot can hide it beyond a limit. That's simple physics.

That is incorrect. Higher gain settings have the same or less read noise when exposure/brightness is fixed, like in your experiment.

Deltaone

01-27-2010, 04:06 PM

OMFG. Seriously? I was the one who said the iso you can go up to in any given camera depends upon the sensor in it, you were the one who said that no, it was quite wrong. Which made no sense whatsoever so i spent quite a bit of time and effort simplifying things so that that was clear to you. Now you suddenly claim that you never said anything of that sort? All that you said in your previous posts about the sensor being irrelevant is now something that i just misinterpreted?

Heres the thing, you're just flat out wrong. This is something that has been proved to you time and again. When given sources like DXO you claim that you know better than others and that we are misinterpreting it. When given logical reasons, you just bluster and say its flat out wrong. Heres a simple fact for you. You are wrong. This is not open to question or doubt. The whole reason why i went as far as i did to simplify things was because i thought a logical reasoned argument would actually get through to you. And you accuse me of building a straw man? You read selectively, take sentences out of context and reply to something which you assume i said. When i say that you are basically saying what i said back to me by doing so, you assure me that i have misunderstood.

Heres something else. There is more than one kind of noise. Every single reply of yours however is only contending with one of them, what the others dont count? Or maybe you just dont plain like them? With all your knowledge of physics you do know of such things as leakage currents and and various kinds of noise which gain brings into the equation. Here's some ultrabasic physics to you. Gain is done by an amplifier. There is no such thing, i repeat, NO SUCH THING, as a perfect amplifier. And since we have such a problem, there is also no such thing as a signal that is perfectly amplified to any level with no distortion or noise if you will. That is why improvements in technology make better sensors which can do better high iso.

When i point out that to you by means of two different cameras, you argue that you never said the noise would be same between two different sensors? SEriously? You were the one who was arguing that the sensor made no difference?

Also, you refer to my technique as suboptimal. I have not read many photography books or articles, this i will freely admit. I will also freely admit im not even a superb or exceptional photographer by any means. At best, i would consider myself slightly above average. Even in this forum there are many, many people who take pictures far better than i do.

That said, i have been taking pictures in low light for quite some time. So what technique i use for what purpose is basically up to me. So far, i have very rarely if at all come to a low light situation where i need to fix both values. Fine i'll even give you that, which is why synn did that experiment in the first place. An experiment, mind you, which was totally unnecessary. But i knew that you would claim that the exposure was varied and the noise in teh first set was produced by the lesser exposure. Fine, so its done your way.

So finally you have proof, and now its adobe's fault? Seriously?

So if you are given sites like DXO, you claim we are all misinterpreting it and since you know "more" you are the only one who can actually understand it.

I give you simplified arguments and you change the whole thing, understand what you want out of it, take it out of context and tell me the same thing im saying, or more concisely such pearls of wisdom as "you are wrong" and we are of course to accept it.

You're given photographic proof and now it's adobe's fault?

And you accuse me of doing a strawman argument. Your temerity, me likes .... :P.

Now, again, your last part in the post above this. Higher gain will not have the same amount of noise. It only works to a point. A POINT.

Which is why, newer sensors can do better and cleaner high iso than older ones. This is something i've been repeatedly saying, and you answer it by saying of course it will be different between cameras?......

Newer technology, better sensors, = better and cleaner high iso. Which is why older cameras had horrible 1600 iso, latest ones go up to 102500 iso. Why is this so hard to grasp?

Higher gain settings will increase the basic amplifier noise. As amplifier technology improves, the amount will go down. Until such a date as a perfect amplifier can be built, there will always be a limit as to how high an iso you can go to.

This is something every photographer knows. Something borne out by the review sites (which we dont understand obviously), the manufacturers (who are all out to fool us and steal our money, i assume), and personal experience (which is irrelevant because we are just not good enough...).

See, anyone can keep arguing by finding something which makes no sense and sticking adamantly to a point. You did claim that you joined these online forums to have discussions. Well? If so, a discussion involves more than one side going "im right and you're wrong because i'm right and you dont understand". It involves a bit more give and take.

The experiment which synn did was precisely done simulating conditions which are pretty much irrelevant to most real world scenario's simply to prove to you in actual proof that you are mistaken. But no, now its the way adobe processes its files.......

Here's the thing. If you admit that modern sensors handle high iso noise better than older ones, you also have to admit the sensor comes into the equation. Now that means the signal gain of the sensors is much better in the newer ones than the older ones. Which means gain adds noise which is better controlled in newer sensors than older ones.

See, you claim to have used every camera since the 10D or so, so why is this concept so hard to grasp? Especially for a low-light shooter, the whole point of upgrading to a better camera is especially for the better high iso performance the new sensors can deliver. Which in turn is because technology improves. Maybe someday we will have such a thing as a perfect amplifier. In which case, yes then you can turn up the gain as much as you want without any detrimental effects.

But till then, there will always be a limit, a limit that is even now being extended, but a limit nevertheless.

synn

01-27-2010, 04:40 PM

Ok, I gotta temporarily un-ignore our resident expert to give a reality check:

All you've proved is that Adobe renders files differently depending on ISO and settings.

L O Effin' L.

So after Me, Delta, DxoMark, Bryan Peterson and Marvin the Martian, Adobe is now wrong too!

O HAI GUYS! BREAKING NEWS! DANIEL BROWNING IS RIGHT. ENTIRE WORLD PROVEN WRONG. MOAR AT 11

If your so-called theory cannot be proven to the naked eye using tools that PHOTOGRAPHERS use (In fact, it was proven WRONG), you probably need to look at another hobby to kill time. Might I suggest ice fishing?

This so-called theory is supposedly there to help you take better pictures in the dark. Well Newsflash, photographers who take low light pics (Or any pics, for that matter) use Lightroom and Photoshop to process their pictures. Not sit in front of Rawnalizequaglimificator for hours and pixel peep. And as seen by the NAKED EYE, the higher ISO image was proven to be more noisier.

this is why I pity you for being a textbook photographer. Coz for all your reading prowess, you simply have no idea of what to do or what can be done in a real world shooting situation.

You can use yousendit.com to upload raw files for free.

Oh golly gee, really? For free you say? Surely, there must be some catch!

You did not post any raw files.

I'll be glad to. Just as soon as you post the godawesome pictures that you've taken in your illustrious career.

Higher gain settings have the same or less read noise when exposure/brightness is fixed, like in your experiment.

Let me just use your favorite sentence as a reply. You know, the one that magically nullifies everyone else's posts without ever having to prove yourself.

"That is incorrect."

Or better yet,

"You are wrong"

Or the best one yet,

"You are mistaken".

Cheerio. Back to ignoreville you go.

Daniel Browning

01-28-2010, 01:39 AM

OMFG. Seriously? I was the one who said the iso you can go up to in any given camera depends upon the sensor in it, you were the one who said that no, it was quite wrong.

It is wrong. For a fixed exposure and brightness, you can increase ISO as much as you want and noise will only stay the same or go down.

Which made no sense whatsoever so i spent quite a bit of time and effort simplifying things so that that was clear to you. Now you suddenly claim that you never said anything of that sort? All that you said in your previous posts about the sensor being irrelevant is now something that i just misinterpreted?

I've only ever been talking about a fixed exposure/brightness. I don't know why you kept making statements based on variable exposure or variable sensors.

When given sources like DXO you claim that you know better than others and that we are misinterpreting it.

That's because you are misinterpreting it. In the case of DXO, you fail to grasp the difference between the ISO setting on the camera and the ISO of the final image. In the context of the discussion on the DXO page, both of the demonstration images I posted in this thread have the same ISO -- even though the ISO setting on the camera is different.

When given logical reasons, you just bluster and say its flat out wrong.

No. I gave proof (raw files) to verify my position. You have given no proof of your position whatsoever.

You read selectively, take sentences out of context and reply to something which you assume i said.

Again, if you would like to point out anything in specific where I have misinterpreted you, please let me know and I will apologize. The only misinterpretation I can see is when you thought I said that all sensors are the same, regardless of technology, which prompted you to go into a long tirade on a ridiculous topic.

There is more than one kind of noise. Every single reply of yours however is only contending with one of them, what the others dont count? Or maybe you just dont plain like them? With all your knowledge of physics you do know of such things as leakage currents and and various kinds of noise which gain brings into the equation.

There are two reasons for that. First, my position is based on fixed exposure/brightness. In that case, photon shot noise always remains the same, no matter the ISO setting. Second, this information is most relevant to low light. And in low light, read noise predominates total noise power, and it changes with ISO setting (for a fixed exposure/brightness). The reason it predominates is because total noise power is the square root of the sum of all noises squared. Because of this, when one noise is sufficiently higher than the others, it doesn't matter what the other noises are -- the one predominating noise determines the total noise power. In the case of low light, read noise is the noise that predominates. Leakage currents and other various kinds of noise do not come into the equation.

There is no such thing, i repeat, NO SUCH THING, as a perfect amplifier. [...] That is why improvements in technology make better sensors which can do better high iso.

The fact that there is no such thing as a perfect amplifier does not mean that the self-noise of the amplifier is the predominant noise source in the camera. In fact it isn't -- the late stage electronics are dominant, and the increased gain of the amplifier reduces their significance. Of course, I explained all this already.

When i point out that to you by means of two different cameras, you argue that you never said the noise would be same between two different sensors? SEriously? You were the one who was arguing that the sensor made no difference?

I've never said anything that implies two different sensors can be compared under the position I described. All I said is that for a fixed exposure and brightness, the high ISO setting results in less (or same) noise than low ISO. How you got to two different sensors from that is a mystery.

So far, i have very rarely if at all come to a low light situation where i need to fix both values.

If you do not need to fix both values, then you do not need to ever worry about ISO at all! Just set your camera to ISO 100, fix one of the values, and let the other fall where it may for the optimal exposure. On the other hand, if you ever shoot in a circumstance where you *do* have to have a certain minimum amount of motion blur or depth of field, then you will have to fix both values. After increasing them as much as possible, the next question is how much to increase ISO. The answer to that second part is the point of this thread: the higher the ISO, the less the noise (and headroom).

Fine i'll even give you that, which is why synn did that experiment in the first place. An experiment, mind you, which was totally unnecessary. But i knew that you would claim that the exposure was varied and the noise in teh first set was produced by the lesser exposure. Fine, so its done your way.

I said nothing of the kind. I believe synn when he said that they have the exact same exposure. I did not say that exposure was varied.

So finally you have proof, and now its adobe's fault? Seriously?

I did not say it's adobe's fault. You are misinterpreting the Adobe raw conversion to mean something that it does not. If you posted raw files or used a different raw converter, you could see for yourself. More on this below in my response to synn.

So if you are given sites like DXO, you claim we are all misinterpreting it and since you know "more" you are the only one who can actually understand it.

Addressed above.

I give you simplified arguments and you change the whole thing, understand what you want out of it, take it out of context and tell me the same thing im saying, or more concisely such pearls of wisdom as "you are wrong" and we are of course to accept it.

You're given photographic proof and now it's adobe's fault?

Not at all. JPEG files are not proof at all, so you've given none. Any raw files taken on the conditions described are proof. I gave some examples.

Newer technology, better sensors, = better and cleaner high iso. Which is why older cameras had horrible 1600 iso, latest ones go up to 102500 iso. Why is this so hard to grasp?

Straw man again. You must like that guy because it's a lot easier for you to argue with him.

Higher gain settings will increase the basic amplifier noise. As amplifier technology improves, the amount will go down. Until such a date as a perfect amplifier can be built, there will always be a limit as to how high an iso you can go to.

This is something every photographer knows. Something borne out by the review sites (which we dont understand obviously), the manufacturers (who are all out to fool us and steal our money, i assume), and personal experience (which is irrelevant because we are just not good enough...).

Again, all of your statements are predicated on varying exposure/brightness, which has nothing to do with my position.

See, anyone can keep arguing by finding something which makes no sense and sticking adamantly to a point. You did claim that you joined these online forums to have discussions. Well? If so, a discussion involves more than one side going "im right and you're wrong because i'm right and you dont understand". It involves a bit more give and take.

I posted evidence.

The experiment which synn did was precisely done simulating conditions which are pretty much irrelevant to most real world scenario's simply to prove to you in actual proof that you are mistaken.

First, the conditions are not irrelevant in the real world -- they apply any time that exposure is limited, such as in low light. Second, JPEGs from one flawed raw converter are not proof for generally applicable truth.

Here's the thing. If you admit that modern sensors handle high iso noise better than older ones, you also have to admit the sensor comes into the equation. Now that means the signal gain of the sensors is much better in the newer ones than the older ones.

Non sequitor. Noise is improved in a variety of ways with new technology, including the read noise before level before the variable gain amplifier even comes into play -- and even on cameras that lack any gain amplifier whatsoever, such as digicams and Medium Format. For example, read noise may have been improved by optimizing the Si/SiO gate width to reduce trapped carriers that cause read noise.

See, you claim to have used every camera since the 10D or so,

I never said anything of the sort.

Maybe someday we will have such a thing as a perfect amplifier. In which case, yes then you can turn up the gain as much as you want without any detrimental effects.

For fixed exposure/brightness, gain results in the same or less noise.

Daniel Browning

01-28-2010, 03:45 AM

So after Me, Delta, DxoMark, Bryan Peterson and Marvin the Martian, Adobe is now wrong too!

I've only ever said that you and Delta are wrong; never that DXO, Bryan Peterson, or Adobe is wrong. Specifically, you two cite sources that do not state what you think they do. First when you misunderstood the word exposure, and now the effect of ISO setting on noise in a fixed exposure/brightness. The entire world is doing fine. It is only you two who have been proven wrong by the experiments here.

If your so-called theory cannot be proven to the naked eye using tools that PHOTOGRAPHERS use

There are other raw converters besides Adobe.

This so-called theory is supposedly there to help you take better pictures in the dark. Well Newsflash, photographers who take low light pics (Or any pics, for that matter) use Lightroom and Photoshop to process their pictures.

That is true, but it doesn't change the fact that Lightroom 2.x does very poor with high ISO files, turning out poor resolution and blotchy tones. Many other raw converters do a much better job, including RPP, RT, Aperture, and others. Just because Adobe does a poor job at one specific aspect of their software does not affect whether or not a raw file has more noise or not.

O HAI GUYS! BREAKING NEWS! DANIEL BROWNING IS RIGHT. ENTIRE WORLD PROVEN WRONG. MOAR AT 11
Oh golly gee, really? For free you say? Surely, there must be some catch!
I'll be glad to. Just as soon as you post the godawesome pictures that you've taken in your illustrious career.
Cheerio. Back to ignoreville you go.

I thought that the baboons on the discovery channel were entertaining, but your adolescent chest beating is even funnier. The only part that I dislike about your neanderthal talent contest is the cowardice. It's unfortunate that you are so afraid of being proven wrong that you wont even post your raw files.

Anyway, thanks for the laughs.

Daniel Browning

01-28-2010, 03:47 AM

Here is some more proof at the same exposure and the same brightness. First, using the same raw files that I posted above, but with a different raw converter. Unlike Adobe, it has a linear exposure compensation slider. It's called RawTherapee.

ISO 100:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8201_rt_plus0_crop.png

ISO 1600:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8202_rt_minus4_crop.png

As you can see, the noise level is the same or less in the ISO 1600 file. If the brightness is increased (but still kept the same), the difference becomes more obvious:

ISO 100:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8201_rt_plus2_crop.png

ISO 1600:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_8202_rt_minus2_crop.png

Now it's clear that the ISO 1600 shot has much less noise than the ISO 100 file.

Here is a different set of raw files:

http://thebrownings.name/images/2009-07-29-iso-compare/IMG_7281_500.CR2

http://thebrownings.name/images/2009-07-29-iso-compare/IMG_7278_500.CR2

And this time with Canon DPP:

ISO 100:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_7278_500_canon_plus2_crop-0.png

ISO 1600:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_7281_500_canon_minus2_crop-0.png

Again it shows higher gain has less noise for a fixed exposure/brightness. It also shows that the cost is highlight headroom.

Here they are in rawtherapee:

ISO 100:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_7278_500_raw_therapee_crop.png

ISO 1600:
http://thebrownings.name/images/2009-07-29-iso-compare/IMG_7281_500_raw_therapee_crop.png

Again it proves that higher gain has less noise for a fixed exposure/brightness.

Deltaone

01-28-2010, 06:52 PM

You know, im not omniscient or prescient, but here we go again. Why do i have to keep repeating the same things over and over again?

First point, the reason why i went on and on about the sensors is because you claimed, im not sure in this thread or the other, than the sensor din't matter. I was hoping to correct that notion of yours.

Because i seemed to me that if you understood the sensor actually mattered, and that newer sensors handle high iso better, you would also understand why i said that how high an iso you can go to depends on the sensor. At which point, you said, that the reason to go high iso is because of the noise and it has nothing to do with the sensor.

Which is the one part i don't get. You do admit that newer generation sensors can handle high iso better than older ones? Yes? No? I need an answer to this one first. Mind you, this is not really a question as the answer is rather obviously yes.

Now, it is precisely because of that reason why we invest more money in newer cameras to get the additional capabilities that they bring, pushing iso is the newest game with 102500 being the highest at the moment if im not mistaken. Which is also why a camera like the D90 can do a cleaner 1600 ISO than a D70 WITH THE SAME SETTINGS.

I told you to do an experiment with two cameras of two different generations with the same A and SS. That will ensure that the light hitting the sensor remains the same. The fact that one will handle noise better than the other will at least prove that one sensor is better than the other at handling noise. I fail to see why this is irrelevant because it is something that i've been tryign to get through to you that the sensor matters when you're fixing your values.

Allow me to elaborate. In the other thread you were adamant on you definition of exposure being limited to just two things. I also took the time to explain why iso or sensitivity of the medium is also included in the "exposure" triangle. Regardless or not of whether the "definition" you so seem to love includes it or not, it is an unavoidably important part of the equation when you're taking a photograph.

A photograph depends on 3 things ( basic, there are a hell of a lot more, but im restricting myself to the bare basics at the moment). A, SS, and time.

Now, in a low-light situation, if you have the luxury of increasing the third factor, i.e. time, you can use a low iso well enough. But unfortunately, such is not always possible, hence we increase the sensitivity in teh case of film and gain in the case of digital sensors so as to reduce the time.

Which is why in a given scenario you cannot ignore the time. If time is a restriction, then ASA/ISO comes into play. Which is why the exposure triangle includes ISO even if the technical definition does not. I thought i had made this clear earlier, but let it be.

Now, in this triangle if you have to fix two values, then you have to modify the third to get the desired result, which in this case is a photograph with whatever effect you desire.

Let us take a low light situation where the camera meters and shows you that a shutter speed longer than you want is required. So you keep the other two the same (for whatever reason, btw i did mention that motion was a category maybe you din't see it) and increase the iso till you get the effect you want out of the scene. This is basic common sense.

Noone, would be dumb enough to fix both of them and then just shoot at the lowest iso possible, because that wouldn't give you the effect you want, all you would get is an underexposed shot. you could push it in post, but there are limits as the noise would probably be unbearable. So you use an iso that is high enough to get the scene brightness you can. So far so good?

Now, herein lies the essential problem. Lets say i need an iso of 1600 to get the effect that i want out of it. Your 5dm2 handles 1600 easily, and exceptionally well. The D70s that i have does it exceptionally poorly especially in low light with fixed values. This is where the iso capability of the sensor/camera come into play. A camera that can do 1600 iso well like the modern generation ones will give me a cleaner output than one which can't handle it well. Which is why i have been trying time and again to state that the sensor/camera does actually matter. It is the most vitally important part in determining how high an ISO you can go to, in a low-light situation where your A and SS are fixed.

This is why i kept stating that the sensor is important because how the sensor handles high iso is just as important to the selection of iso as anything else. Now, yes lightroom is not the best way to handle raws, this has been proven before. However, changing color profiles can change a few things, but are you actually suggesting that its actually introducing noise that wasn't there in the first place?

See, unfortunately, in real-world scenarios, since most of us at least are not professional photographers, we prefer to use tools like lightroom because it gives a good combination of simplicity and flexibility. See, there are a lot of other factors which come into play in the real world which don't really come into play in a theoretical discussion, which is why most of my arguments are heavily slanted in favor of practical applications rather than theoretical and based on my experience in the same.

And well that is why i said, i am rarely if even in a scenario which requires my fixing both, since i don't really shoot low light sports. And my tastes in low light photography lean more into the long-exposure section anyways. Now in real life situations another reason why i like the D90 and the D300 are their capabilities in reducing chromatic aberration which means i can actually make do with crappier lenses. There are many many factors which come into play in a real life scenario including the color and kind of lighting, what particular effect you want to get out of it etc.

All of these factors affect my choice of settings for a scene i wish to shoot. My practical experience so far has shown that using 1600 iso in a low light situation even with fixed settings of both A and SS give me demonstrably more noise than i want out of it, and i have had to throw away a hell of a lot of shots because of that. At the same time i have observed other cameras doing a lot better in the same scenarios with a lot better noise control as well.

So yes, to me, what camera i use, and the capabilities of it, are a vital factor when selecting what iso i shoot at, and most people i have talked to, seen and met with seem to share the exact same opinion. How high an iso you can go to depends on the capability of your sensor. Which is why the statement is, for a fixed exposure, use an iso that gives you a result you want, dependant on the sensor involved.

And yes, btw i realize that its not just the sensor per se, but various other components including the ADC, the amplifier and a few others which affect how the camera ultimately handles the noise in the image. I just felt saying sensor was easier. However, that hardly negates my point that the highest iso you can use is dependant on the sensor.

Oh, and yes, we use lightroom, it is what we use for postprocessing. The point of our photography is the end effect, to get something we want out of a scene. And as such, that is the benchmark i use because it is what enables me to get what i want out of it. Yes, more learned persons than me have pointed out that lightroom does not actually give pure raw and i've read those very same articles and understood the implications as well. But again, to me, at present this is a hobby, and as such i prefer to spend more time taking the photographs i like with a minimum amount of time spend on the computer afterwards. I know i can control even the noise that the D70s gives at iso 1600 but my purely personal preference is not to spend a couple of hours on each photo i take. Id rather go out and take more photos. That is an individual choice.

So, as ive said before, i say so again, what settings you use are dependant on a great many things and whichever camera/sensor you are using and its capabilites at any given iso do come into the picture in a large way.

Which is why i keep maintaining that simply using a high iso is no guarantee, it depends on the scene and imo, far more importantly the capabilities of the sensor/camera you are using.

I've always beleived the best way for anyone to gain an interest in something is by actually doing it. You gain far more knowledge, experience, and interest in photography by actually going out and shooting. Which is why many of the things in life are greatly simplified so as to make it easier to understand and increase the interest. Which is why basic physics taught in schools does not include quantum physics, even though some of the basics we are taught are essentially "wrong" by today's understandings.

Deltaone

01-28-2010, 06:53 PM

So yea, use the appropriate iso for the appropriate situation is something which photographers learn rather quickly when they're out shooting. Yes, there are such notions as the lowest iso is always the cleanest (the reason for that is because cam manufacturers chose the base iso dependant on the SNR and it is the cleanest), and yes, in low light you cant just use the lowest iso if you have to fix two of the other values. Which is why i kept saying the exposure triangle is so important. As long as you understand these three are interrelated you also understand the importance of changing the iso. That said, i'd say simply saying higher iso is cleaner is actually more harmfull than that as visual results in the majority of real world scenarios would bear that to be rather wrong.

So again, its no straw man, it is your insistence at first that the sensor din't matter, your later insistence that high iso will definitely have less noise (which btw is the topic name) which is rather glaringly wrong as an absolute statement, and now your slightly modified statement that higher iso will have lesser than or equal noise.

See, why i've been going on about this now? I haven't changed my stance in either thread. I have been at pains to try and explain why things are done the way they are done, while you have been adamant on technical definitions (ironically i even explained why its explained the way it is in the other thread) and how you are absolutely right to the exclusion of anything else.

And also the most important fact of all, your tests are done with a 5DM2 which handles iso 1600 exceptionally well. The D70s doesnt. Color profiles and converters not withstanding, the D70s does horrible at that iso. Hence the different results.

Btw, a question about your shots. You have equalized the brightness of the two shots yes? Which is the base? i.e. did you increase brightness for the iso 100 or decrease for the 1600?.

Daniel Browning

01-29-2010, 04:38 AM

First point, the reason why i went on and on about the sensors is because you claimed, im not sure in this thread or the other, than the sensor din't matter. I was hoping to correct that notion of yours.

I did not say that. Not in this thread or any other. Either you misunderstood me or I did not make myself clear.

At which point, you said, that the reason to go high iso is because of the noise and it has nothing to do with the sensor.

What I said is correct for its context. When you say "how high an iso you can go to depends on the sensor", that is only true in the context of floating exposure/brightness, such as in ample light -- it is not true in the context of fixed exposure/brightness, such as in low light. In the fixed exposure/brightness context, you can choose any ISO setting you want and it's simply a trade off between noise and highlight headroom; it applies equally to all sensors, old and new alike.

Here is an example. Here is what I think you are saying:

Let's say that the Canon 10D with a certain exposure (f/4 1/500) has a certain level of noise at ISO 100 that you find acceptable. You're saying that newer cameras can use less exposure and higher ISO to achieve the same brightness and similarly acceptable noise levels. Let's say the 7D with f/11 1/500 ISO 800 gives you the same acceptable level of noise that you had with the 10D above. Therefore, high ISO is limited by the camera.

What you're saying has nothing to do with what I've said. I'm talking about low light where you don't have the luxury of using stronger exposures. I'm talking about changing the ISO when exposure/brightness is fixed. Here is an example of what I'm talking about:

Let's say that you have a Canon 10D with a certain exposure/brightness that you need. You can set it to ISO 100 and get lots of noise, but preserve a lot of highlights. Or you can set it to ISO 800 and get much less noise, and clip 3 stops of highlights.
Now go to the 7D with the same exposure. You can set it to ISO 100 and get lots of noise, but preserve a lot of highlights. Or you can set it to ISO 800 and get much less noise, and clip 3 stops of highlights.

Do you see the difference? Either way the 7D will have less noise than the 10D, but that's obvious and has nothing to do with my point: that both sensors benefit in the same way from the selection of ISO.

A photograph depends on 3 things ( basic, there are a hell of a lot more, but im restricting myself to the bare basics at the moment). A, SS, and time.

Are you saying that "Shutter Speed (SS)" and "time" are different, or was that just a minor mistake?

Now, in this triangle if you have to fix two values, then you have to modify the third to get the desired result, which in this case is a photograph with whatever effect you desire.

That is only true for a JPEG photographer. In raw photography, you can leave the ISO alone and still get the desired result. For example, on Canon cameras, HTP is a metadata feature that recommends a 1 stop brightness boost to the raw converter. If you are at f/4 1/500 ISO 100 and the default raw conversion is one stop too dark, you can make the default increase by one stop if you enable HTP. Of course, that's only the default. You can boost it by more stops if you want. Or maybe you decide that the noise is too high at ISO 100+1 stop push, then you can set ISO 200, but with the knowledge that you're losing a full stop of highlights. If the noise is still too high, you could set ISO 400 and do a 1 stop pull, but now you're losing two stops of highlights. It might be well worth it if you need to get noise down.

So you keep the other two the same [...] and increase the iso till you get the effect you want out of the scene. This is basic common sense.

That is a good method for a beginner who is using the camera JPEG and trying to learn how to get brightness correct. But for an intermediate level photographer who knows about raw and wants to get the least amount of noise possible, it is suboptimal. If you need more highlight headroom than is offered by the default raw conversion in your camera, then following your procedure will give you an ISO that is too high and clips too many highlights. If you need less highlight headroom than the default offered by your camera's raw converter, then that method will give you an ISO that is too low and has more noise than the optimal ISO. The optimal procedure is to increase exposure as much as possible, then increase ISO as much as possible without clipping highlights you care about.

you could push it in post, but there are limits as the noise would probably be unbearable.

The only time when pushing in post has more noise than in-camera gain is when the gain does not reduce noise for a fixed exposure/brightness. For example, on the 20D, the analog gain goes up to ISO 1600, but the difference between 400 and 1600 is relatively minor, so many times it's better to leave the gain at 400 and push the rest of the way in post to preserve highlights. Other times, the decrease in read noise is worth the loss of highlights. What I've been saying all along is that the choice of ISO is a simple tradeoff between noise and highlights.

Now, yes lightroom is not the best way to handle raws, this has been proven before.

I'm glad you agree. It is a matter of taste, though; I've run into people who actually prefer LR 2.x for high ISO stuff.

However, changing color profiles can change a few things, but are you actually suggesting that its actually introducing noise that wasn't there in the first place?

No, I'm not suggesting that. I'm just saying it does a bad job with high ISO shots compared to most other converters. I don't know why.

....we prefer to use tools like lightroom because it gives a good combination of simplicity and flexibility.

That's fine. I heard that Adobe finally fixed their high ISO output in LR 3.x, but I haven't tested it for myself.

And yes, btw i realize that its not just the sensor per se, but various other components including the ADC, the amplifier and a few others which affect how the camera ultimately handles the noise in the image. I just felt saying sensor was easier.

Me too.

So again, its no straw man, it is your insistence at first that the sensor din't matter,

I see now that it's not a straw man, just a misunderstanding. I stated an idea that applies equally to all sensors; you thought I was saying that all sensors are the same. I guess I can see now how those two might easily be confused.

your later insistence that high iso will definitely have less noise (which btw is the topic name) which is rather glaringly wrong as an absolute statement, and now your slightly modified statement that higher iso will have lesser than or equal noise.

You're right. I should have stated from the beginning that it's the same or less, not just less. Thank you for the correction.

And also the most important fact of all, your tests are done with a 5DM2 which handles iso 1600 exceptionally well. The D70s doesnt. Color profiles and converters not withstanding, the D70s does horrible at that iso. Hence the different results.

It's not just the 5D2, it's every DSLR. If you think your D70 is an exception, then prove it. Take 5 minutes to do the simple and easy experiment and your raw files. I've seen measurements of dozens of cameras, so I can already guess what you will find: that ISO 1600 has the same or less noise as ISO 100 in fixed exposure/brightness. (BTW, minor clarification: one set of images I posted are from the 5D; the others are 5D2.)

Btw, a question about your shots. You have equalized the brightness of the two shots yes? Which is the base? i.e. did you increase brightness for the iso 100 or decrease for the 1600?.

I did both. Either way you do it, the noise is the same or less in the high ISO shot. The first is ISO 1600 -4 and ISO 100 +0. The rest of them are ISO 1600 -2, ISO 100 +2 IIRC.

Deltaone

01-29-2010, 04:24 PM

What i've been trying to say is how high an iso you can go changes from camera to camera. For example the nikon D80 actually has a hi extended setting above 1600, which is pretty much useless in any kind of light to be honest. Even with a fixed exposure, i have found that older gen cameras (i must admit to not having used much of canon, although i do know their sensors have handles noise a hell of a lot better than older gen ccd nikons), like the d70, 60, and 80 to quote three examples produce demonstrably more noise at higher iso in low light conditions especially at the top of their range. How much varies from camera to camera, but my point was that in a given low light situation i would not go 1600 iso on the d70 where i might chance it on the d80 because of the way the sensors handle noise.

That apart there is a reason why i asked you that last question. And i must admit that this still confuses me. Now it might be my lack of knowledge, but how exactly do you meter the scene without fixing the iso? Aperture yes, you fix yourself. Now, normal situations i can want a shutterspeed which is faster than x but to arrive at a certain fixed shutterspeed you have to meter it against something right? So how do you meter your shot without deciding the iso you want? Which is why my question on which is the base.

Reason being, something ive seen so far, at least with said cameras is that, you really cant get more than 2 stops of headroom out of the raw in these nikons. Especially if youre trying to push it up. The noise in the dark areas becomes unbearable at anything over 2 stops. This is what i've seen so far. I need to try something else, let me see if i can get something done tomorrow.

Daniel Browning

01-30-2010, 12:57 AM

how exactly do you meter the scene without fixing the iso?

When you are in low light, and your f-number is already as fast as you can make it and your exposure duration is as long as you can bear, then it really doesn't matter what the meter says, because there's nothing you can do to increase exposure. The only choice now is whether you increase gain in the camera (throwing away highlights to reduce noise) or gain in post (preserving highlights at the cost of increased noise). Neither of these affect exposure, but they do affect which ISO setting will be used. If you need the meter to help you chose the ISO setting that will give you the desired brightness, then that is fine. Ideally, cameras would let us have the choice of metadata ISO or analog ISO. Canon offers one stop of metadata ISO (they call it "HTP"), for example, but it would be better if we had full control. Then we would be able to set the analog ISO to 100 and the metadata ISO to 1600 (for example) and the meter would tell us that the shot is four stops underexposed for ISO 100 (the true sensitivity of the camera) and normally exposed for ISO 1600 (the brightened image in post).

Reason being, something ive seen so far, at least with said cameras is that, you really cant get more than 2 stops of headroom out of the raw in these nikons. Especially if youre trying to push it up.

Just to make sure we're talking about the same thing, let me point out that I'm not referring to the the difference between the default the white point of the JPEG conversion in the camera and the white point of the raw conversion that is used after highlight recover operations. I am referring to the number of stops between middle gray and the white point (AKA saturation or dmax) in the output image. As I think you know, it's also possible for the headroom to be whatever you want it to be. Even the Nikon cameras you are talking about can be used with 5 stops of headroom if you adjust your exposure and raw conversion accordingly.

The noise in the dark areas becomes unbearable at anything over 2 stops.

It's true that noise gets worse as you increase highlight headroom, as long as the total dynamic range that is used goes up by the same amount as the headroom.

Deltaone

01-31-2010, 03:38 PM

Er, that first part doesnt really make any sense. Aperture yes, its dependant on the lens, so yea theres a cap on it. The shutterspeed? Yes, if you dont have a remote, you're limited to 30 seconds. But that said, short of metering how do you find out what shutterspeed to use?

Lets say you're using the 50 1.4. So you've got your aperture at 1.4, and then what exactly do you do to your shutterspeed? Any camera will go as high as 30 seconds without a remote and well thats longer than required unless you're going for some extreme long exposures in an astro situation. However only one or at most one set of shutterspeeds will give you the scene brightness you want. You make the shutterspeed too fast, its underexposed, too slow and its overexposed. Which is why you meter right?

Yes, there are situations like the ones i mentioned before. I've got my lens at 1.4, and i need a handheld shot. Now my hands dont really work at anything slower than 1/6 and thats pushing it. So now i have a limit. However, its not an exposure limit. because the camera can actually get more light, its a limitation of my hands to hold it steady for long enough/lack of a tripod.

Either way, my question is this. Obviously you fix your aperture first. Now dependant on what iso you've chosen the camera is going to give you a shutterspeed which it calculates as optimum for the scene, hence the metering part. You choose a higher iso, you get a faster SS, lower iso, slower SS. I can understand changing iso to get the SS you want. But how on earth do you fix the SS without chosing the iso? THe only other way that i know how to would be to use a light meter and calculate which i guess is an option but one which most prefer not to do as modern dlsr meters can do it accurately enough.

So, my question is, how exactly do you determine a fixed shutterspeed because you can't unless you fix the iso beforehand on the camera (assuming you're using the camera's meter). So how exactly do you fix it? This is the part i don't get.

Now, the headroom part. Im talking in extremely simplistic terms. In a given shot in raw in lightroom, before conversion, i can get a maximum of two stops of headroom to play around with in the exposure or brightness. (it also depends on the scene to a certain extent). BUt if i increase any more than that the noise values change to unacceptable levels. Same thing works in converse, if i reduce it any more than that noise that wasn't evident before creeps in. This is all before any jpeg conversion and while still in lightroom.

Vicky

01-31-2010, 07:05 PM

High ISO = Lower Noise

/\ I wanted to give this theory a try...

Say, I was in a situation wherein I wanted to take a photo and had with me a lens with a max aperture of f/2. Lets say I was without a tripod and the longest shutter speed I could hand hold without camera shake was 1/30s.

So, in this situation I was 'limited' to a max 'exposure' of 1/30, f/2. Now, I took three shots of the exact same scene with the exact same 'exposure' but different ISO settings.

The 1st shot was with the camera's base ISO 100. Incidentally, the camera meter showed this as 2stop underexposed.
The 2nd shot was with the camera's ISO set to 400. Incidentally, the camera meter showed this as properly exposed.
The 3rd shot was with the camera's Max ISO of 1600. Incidentally, the camera meter showed this as 2stop overexposed.

Here are the shots SOOC:

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5400_sooc.jpg

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5401sooc.jpg

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5402_sooc.jpg

Then to equalize the 'brightness' of all three, I adjusted the 1st to +2 stops and the 3rd to -2 stops in the RAW converter keeping all other parameters unchanged:
http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5400_plus2.jpg

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5401sooc.jpg

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5402_minus2.jpg

Here are the 100% crops from the above:
http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5400_a.jpg

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5401_b.jpg

http://i247.photobucket.com/albums/gg143/Vikash2007/IMG_5402_c.jpg

The 1st shot which was originally underexposed and later given a +2stop exposure boost in the RAW converter has the maximum visible noise. Well, this was kinda of expected.

However, the real points of interest are:

#1 The 3rd shot is definitely not showing lower noise than shot no 2:(
#2 The 3rd shot is clearly having that 'washed out' look whereas no 2 has richer colors and better contrast.

Moreover, had this been a high contrast scene, the shot no 3 would have a lot of washed out highlights which may be impossible to recover even in RAW:(

Would love to hear the inputs of others on this..

Daniel Browning

01-31-2010, 11:20 PM

But that said, short of metering how do you find out what shutterspeed to use?

The longest duration possible that has the amount of motion blur that you want.

Lets say you're using the 50 1.4. So you've got your aperture at 1.4, and then what exactly do you do to your shutterspeed?

Personally, if I'm shooting stationary subjects in low light with 50mm on the 5D2, I tend to go for around 1/50 and bracket liberally; I find that at least 1/3 will be critically sharp. But if I am shooting indoor sports, I might use 1/1000 instead, to stop motion blur. Both are the longest duration possible for the amount of motion blur that I want ("none" in this case).

However only one or at most one set of shutterspeeds will give you the scene brightness you want. You make the shutterspeed too fast, its underexposed, too slow and its overexposed. Which is why you meter right?

I have no problem with metering, but I find that I never need to use it for choosing an exposure duration in low light, because the duration is always dictated precisely by motion blur. I use the meter more often when there is ample light to choose from a variety of shutter speeds.

Now my hands dont really work at anything slower than 1/6 and thats pushing it. So now i have a limit. However, its not an exposure limit. because the camera can actually get more light, its a limitation of my hands to hold it steady for long enough/lack of a tripod.

Well, I think the heat death of the universe in 48 billion years does set the ultimate exposure limit. :) Personally, I prefer a definition where "limited exposure" includes factors such as how steady your hands are and how much motion blur you want in order to achieve your goal for the photo.

Either way, my question is this. Obviously you fix your aperture first. Now dependant on what iso you've chosen the camera is going to give you a shutterspeed which it calculates as optimum for the scene, hence the metering part. You choose a higher iso, you get a faster SS, lower iso, slower SS. I can understand changing iso to get the SS you want. But how on earth do you fix the SS without chosing the iso?

Again, by choosing the longest one possible before you even touch the ISO. If you choose your ISO first, then let exposure duration float, you're going to end up with some pictures that have far more noise than the others. When exposure is limited, it is far better to set the exposure duration at the maximum possible, then adjust ISO to match the desired brightness. For example, it's far better to go from ISO 1600 1/60 to ISO 400 1/60 than it is to go from ISO 1600 1/60 to ISO 1600 1/250. It's important to always have the longest exposure possible.

So, my question is, how exactly do you determine a fixed shutterspeed because you can't unless you fix the iso beforehand on the camera (assuming you're using the camera's meter). So how exactly do you fix it? This is the part i don't get.

When exposure is limited, I fix the exposure duration by choosing the longest one possible. If that clips too many highlights at the lowest gain, then I'm not in low light and I can pick from a variety of exposure durations.

This is all before any jpeg conversion and while still in lightroom.

When I say "conversion", I'm not talking about converting uncompressed images into JPEG; I'm talking about raw conversions. Lightroom does not have any method for viewing the unconverted raw files: the only thing you can ever possibly see in LR is a processed conversion, just as the camera JPEG is a processed conversion. If you want to see the image with less modification, you have to use a different raw converter.

Daniel Browning

01-31-2010, 11:43 PM

#1 The 3rd shot is definitely not showing lower noise than shot no 2:(

As mentioned earlier in the thread, that is normal and expected. For example, when the tonal values examined are dominated by photon shot noise.

You might think that this means the four-word summary from the thread title is incorrect ("high ISO = less noise"), but I think it just means that the four words are not sufficient to explain the full conditions, i.e. that the four-word summary does not embody all the conditions described in the thread. For example, it would be clearer if it said:

"High ISO settings on the camera = less noise"

But that's only true when the raw conversion results in equal brightness. So to add clarity it should actually say:

"High ISO settings on the camera combined with the same ISO rating in the final converted image = less noise".

But actually, that can be misleading too, because it only has less noise in the parts of the raw file that are affected by read noise. So it should really say:

"High ISO settings on the camera combined with the same ISO rating in the final converted image = less noise in the parts of the image that are affected by read noise, which varies by your taste in conversion".

But then again, just because there is less noise doesn't mean the difference will be very noticeable, so there are other caveats:

"High ISO settings on the camera combined with the same ISO rating in the final converted image = less noise in the parts of the image that are affected by read noise, which varies by your taste in conversion and may be small enough to appear only the same, depending on your perception".

By that time, the summary is far too long to put in the subject line. It's better to use a short subject line and explain all the conditions at length in the thread.

#2 The 3rd shot is clearly having that 'washed out' look whereas no 2 has richer colors and better contrast.

That's purely and simply a limitation of your raw converter. As I demonstrated above, many raw converters are capable of delivering the exact same color and contrast. (And really, it's not hard at all for raw developers to achieve: it's simple multiplication and division.)

If you post your raw files (e.g. yousendit.com), I can convert them in several other converters that do show the same contrast and color richness between all the ISO settings.

Moreover, had this been a high contrast scene, the shot no 3 would have a lot of washed out highlights which may be impossible to recover even in RAW:(

Agreed. That's why the balance is between highlights and noise.

Deltaone

02-01-2010, 01:04 AM

Er, so let me get this straight. The shutterspeed is dictated by motion or lack thereof? Which is what i said before? And the same thing i said in my previous post with an example?

Again an example. A low light scene, lets say its a bridge, nothing particularly important vis a vis motion blur in the scene itself. Im shooting at F4 lets say. At iso 100 the camera shows me a shutterspeed of 1/6. Now, thats the limit of my handheld ability at best, but i don't want to risk it. I need a slightly greater shutterspeed. IF i were to just arbitrarily increase it, i get an underexposed shot. So i do the next thing, i increase my iso to 800 and end up with a shutterspeed of 1/25 which is what i need. I take my shot and get a shot with the scene brightness i desire and the shutterspeed i desire.

The reason i brought up the question of metering is simple enough. If your aperture has hit its limit, then there is only one possible shutterspeed for you assuming you want a properly exposed image. Now, you say the shutterspeed is dictated by motion, that makes sense (mind you this is what i've been saying about the method i've been using but you said it was only for jpeg photographers?).

In low light, if you're aperture has hit its limit, and you need a faster shutterspeed, the only thing you can do is increase your iso unless you want to get an underexposed shot. And even in those cases there are limits. Because for example, the d70s only goes to 1600 and i have been to many cases where that was just not enough. Even 1600 wont give me a shutterspeed fast enough to do the job unless i want a severely underexposed shot.

Now-----

---"Again, by choosing the longest one possible before you even touch the ISO. If you choose your ISO first, then let exposure duration float, you're going to end up with some pictures that have far more noise than the others. When exposure is limited, it is far better to set the exposure duration at the maximum possible, then adjust ISO to match the desired brightness. For example, it's far better to go from ISO 1600 1/60 to ISO 400 1/60 than it is to go from ISO 1600 1/60 to ISO 1600 1/250. It's important to always have the longest exposure possible."-----

For the first part, what i've been saying all along is what you;'ve just said in your last post. You fix your aperture, see if the SS is fast enough and if not increase the ISO till you get an SS fast enough (unless you want a low iso underexposed shot). So the floating part only refers to having the SS go above whatever speed i need it to be. So basically what i;ve been saying all along is the correct method?

Your second part confuses me. If yu admit its better to go from 1600 iso 1/60 to 400 iso 1/60 (im assuming here scene brightness is good enough at 400 iso) you've basically just admitted lower iso is better. OR am i misinterpreting something here.

And for the second part, 1600 iso 1/60-1/250, the only reason id do that if i actually needed 1/250 to get the stop motion that i needed in that particular scene. Why on earth would i just randomly cut off 2 stops unless i needed it?

And yes, you always go for the longest exposure possible, which is also the lowest iso. For a fixed aperture, the lowest iso will give you the longest exposure for a properly metered shot. The only real reason to go to the higher iso is when you need a faster shutter speed and the light isn't strong enough to support that faster ss.

So after this many pages were back to what i originally said?

Vicky

02-01-2010, 09:09 AM

As mentioned earlier in the thread, that is normal and expected.

That's why the balance is between highlights and noise.

That's exactly how I concluded after my little test.

- Using this technique didn't get me an image with lower visible noise using my regular conversion tools.
- Using this technique, there would be higher possibility of clipped highlights in high contrast scenes.

Hence, using this technique wouldn't really be worthwhile in real world shooting scenarios.

Again, by choosing the longest one possible before you even touch the ISO. If you choose your ISO first, then let exposure duration float, you're going to end up with some pictures that have far more noise than the others. When exposure is limited, it is far better to set the exposure duration at the maximum possible, then adjust ISO to match the desired brightness.

In my everyday shooting, I determine the final 'exposure' or 'scene brightness' as you may call it, by changing the three parameters namely Aperture, Shutter speed and ISO. Of these, I try to keep the ISO to the lowest position possible and only increase it given the situation wherein my choice of optimal Aperture and SS are giving me an underexposed shot.

Nothing new here. As this, I believe, is the standard practice followed by most photographers. I've yet to see a photographer who first selects an ISO and then adjusts the AV and SS to get the desired exposure:(

Daniel Browning

02-01-2010, 09:11 AM

Er, so let me get this straight. The shutterspeed is dictated by motion or lack thereof?

Yes.

Which is what i said before? And the same thing i said in my previous post with an example?

It still isn't clear to me that what you meant is the same as what I meant, but it's no big deal. I'll clarify below.

I need a slightly greater shutterspeed. IF i were to just arbitrarily increase it, i get an underexposed shot.

Agreed.

So i do the next thing, i increase my iso to 800 and end up with a shutterspeed of 1/25 which is what i need. I take my shot and get a shot with the scene brightness i desire and the shutterspeed i desire.

It seems to me that your method is focused on achieving a certain brightness with default raw conversions, whereas to me that is completely unimportant. All I care about is achieving the optimal noise and saturation level. If the scene has very high dynamic range, I may set it to ISO 100 and 1/25, giving me an extra 3 stops of highlight headroom at the cost of some noise in the shadows. Or if I don't care about the highlights and I just want the least noise possible, I would set it to ISO 1600 1/25 instead of 800, to eek out slightly less noise. Either way, I'll get the brightness in post; all I care about for selection of ISO (after I've maximized exposure) is the tradeoff between noise and headroom.

Now, you say the shutterspeed is dictated by motion, that makes sense (mind you this is what i've been saying about the method i've been using but you said it was only for jpeg photographers?).

No. You said "Now, in this triangle if you have to fix two values, then you have to modify the third to get the desired result, which in this case is a photograph with whatever effect you desire." What you said is only true for JPEG photographers. Raw photographers can get the desired brightness no matter what the parameters are -- all that changes is the noise and headroom. So raw photographers should not care about brightness at the time of the photo -- they care about noise and headroom. Let me elaborate using your example above:

JPEG Scenario: Setting ISO 800 1/25 gives the desired brightness. It also has the benefit of a low noise level. But unfortunately it clips 3 stops of important highlights. Since this camera has no HTP setting or custom JPEG curve, it cannot be fixed. We have to live with the 3 stops of clipped highlights. (The low noise is nice, but we would have been happy to trade some extra noise to get those 3 stops back.)

Raw Scenario: Setting ISO 100 1/25 does not give the desired brightness. Bring the file into post and further work is necessary to get the desired brightness. Use compressed exposure compensation to increase brightness 3 stops without clipping any highlights. Downside: more noise than the 800 shot -- but you knew how much extra noise it would be and you felt it was an acceptable compromise. Upside: three extra stops of highlight headroom means the image has a nice, film-like rolloff in the highlights.

I hope that makes it clear what I meant by the difference between JPEG/raw. Only one of them has to care about brightness at the time of the photo. The other one has to care about noise and highlights.

In low light, if you're aperture has hit its limit, and you need a faster shutterspeed, the only thing you can do is increase your iso unless you want to get an underexposed shot.

Let me clarify that what you mean by "underexposed" is "too dark with a default raw conversion". I would point out that the raw photographer has the luxury of considering non-default raw conversions. (And with the right camera tools, such as metadata ISO, it wouldn't even require any extra work.)

And even in those cases there are limits. Because for example, the d70s only goes to 1600 and i have been to many cases where that was just not enough. Even 1600 wont give me a shutterspeed fast enough to do the job unless i want a severely underexposed shot.

The ISO setting on the camera does not limit the effective ISO of the final image. My 5D2 has ISO 3200 and 25,600; but I never use them because read noise is only the same (not less). So when I want to shoot ISO 51,200 (and I do, sometimes), I just set the camera to ISO 1600 and underexpose by five stops, which gives me the same noise as shooting ISO 25,600 underexposed 1 stop, except it has 4 stops more headroom. Which ties things back to my point: selection of ISO is about highlights and noise.

Your second part confuses me. If yu admit its better to go from 1600 iso 1/60 to 400 iso 1/60 (im assuming here scene brightness is good enough at 400 iso) you've basically just admitted lower iso is better. OR am i misinterpreting something here.

The sentence you quoted says "it's better to X than to Y". You only quoted "it's better to X". Let me try putting it another way. Given the following choices:

Go from ISO 1600 1/60 to ISO 400 1/60
Go from ISO 1600 1/60 to ISO 1600 1/250.

What I said is that the first is better than the second.

And for the second part, 1600 iso 1/60-1/250, the only reason id do that if i actually needed 1/250 to get the stop motion that i needed in that particular scene. Why on earth would i just randomly cut off 2 stops unless i needed it?
[/qutoe]

I'm glad that you would not do that.

[QUOTE=Deltaone;13009]
And yes, you always go for the longest exposure possible, which is also the lowest iso.

It's only the lowest ISO in the final developed image. The actual ISO setting on the camera may not be the lowest.

For a fixed aperture, the lowest iso will give you the longest exposure for a properly metered shot. The only real reason to go to the higher iso is when you need a faster shutter speed and the light isn't strong enough to support that faster ss.

No, there are other reasons to go to a higher ISO, such as reducing read noise if you don't need the highlight headroom.

synn

02-01-2010, 12:14 PM

That's exactly how I concluded after my little test.

- Using this technique didn't get me an image with lower visible noise using my regular conversion tools.
- Using this technique, there would be higher possibility of clipped highlights in high contrast scenes.

Hence, using this technique wouldn't really be worthwhile in real world shooting scenarios.

Bingo!
I'm glad I'm not the only one who feels this way. ;)

In my everyday shooting, I determine the final 'exposure' or 'scene brightness' as you may call it, by changing the three parameters namely Aperture, Shutter speed and ISO. Of these, I try to keep the ISO to the lowest position possible and only increase it given the situation wherein my choice of optimal Aperture and SS are giving me an underexposed shot.

Nothing new here. As this, I believe, is the standard practice followed by most photographers. I've yet to see a photographer who first selects an ISO and then adjusts the AV and SS to get the desired exposure:(

Ditto.
My nightclub shoots are done in mostly in Aperture priority (And sometimes Manual) and I almost always go for f1.4 for artistic reasons as well as to allow maximum light in. The SS used has a limit because i'm shooting handheld without a stabilized lens and because the DJs are moving around a lot. I have no OPTION but to shoot with 1600 ISO then. And with all this, I still end up with a ton of noise.

http://farm5.static.flickr.com/4017/4315421067_d1db91a46c.jpg

f1.4, SS 1/30, ISO 1600.

As I was telling Delta the other day, I wouldn't increase exposure even if he was standing still like a rock. Even One thirds more exposure clips the highlights badly and still give me a noisy image (In fact, I did exactly that for a few shots just to see how it goes. It went as expected).

I'd rather let Noise Ninja or something take care of the noise (That I get NO MATTER WHAT) than throw a shot away coz of severely blown highlights.

So I'm with you Vicky. There are places where theory helps. But in the real world, we go with what works.

Deltaone

02-01-2010, 02:25 PM

Yes, my efforts are directed towards getting a certain brightness for a very simple reason. The camera has both a meter and a display screen. I use them both to see what kind of image i'm shooting and what kind of effect the settings ive chosen have given me. And i use those to determine if i will get the effect i want. Unfortunately i have no way of deciding without using either of them what the correct settings for that particular scene would be. I cannot arbitrarily chose a SS and then go to post and see if i can get the scene brightness i want. Which is why you use both the meter and the display behind. Once you use them, you can then decide whether to under or overexpose as needed.

Now, the clipped highlights part. Wasn;t my statement far previously that you should only raise your iso till your highlights dont get clipped? If i get clipped highlights why would i raise my iso?

That said, heres the real problem. Yes, i can do a lot with raw in post. But unless i have a baseline to start off with, a proper metering for the exposure if you will, how on earth am i supposed to know how much im supposed to modify it in post? Just shoot, then go post and try raising the brightness to see if it works? The problem with that method is there is a limit to how underexposed you can go, raising the brightness just won't work at times.

Which is why i use the meter to give me a baseline reading of what the ideal brightness would be for that scene, what the camera can do in that particular scene and then go on from there.

I cannot look at a scene and say, hmm ill use 1.4 and 1/60 here and then go to post and try to bring up the brightness that way.

Now you mention default raw conversions. Are you stating that by using other tools other than lightroom, i can actually increase the exposure or brightness of an image by more than 2 stops without increasing noise? Because i've been using lightroom for a while now and especially in low light situations, no you just can't do that. There just isnt that much room in the raw.

The whole reason i've been going on about metering is because most people shoot in A or Av mode. You set the aperture to get your desired DOF. The camera selects the SS based on whatever iso you've chosen. Normally, its kept at the lowest or base iso of the camera. If the SS isn't fast enough for the particular scene, you can try increasing the iso to see if you can obtain a SS fast enough. Why do i do this? Because then i have a baseline to work off of. I can decide and know what is the proper metering for that shot and then underexpose or overexpose or whatever by a stop or two if the scene demands it. Now i need that baseline to know how much i need to do in post and if its even possible to do something about it in post.

Now the clipped highlights part. I would have thought that one was obvious. If i were to use a setting that would clip my highlights and i needed said highlights, er why on earth would i use it? In high contrast scenes, sometimes one shot cannot just give you all that you need. You can use exposure blending or HDR to get the full range of lighting that the shot needs. THat is something else alltogether.

And the part that really amazes me. You have a software that allows you to increase exposure in post by 5 stops without increasing noise? Er what is it and can i see samples ?

Vicky

02-01-2010, 05:23 PM

You have a software that allows you to increase exposure in post by 5 stops without increasing noise?

Wow!:eek::eek:

Which RAW converter is that? Now, I'm excited:)

Daniel Browning

02-01-2010, 10:49 PM

Now, the clipped highlights part. Wasn;t my statement far previously that you should only raise your iso till your highlights dont get clipped? If i get clipped highlights why would i raise my iso?

What if getting the correct brightness (in a default conversion) means clipping four stops of highlights? What if preserving the highlights you care about means having a scene brightness that is four stops too dark? What if you care about noise more than highlights? From what you said, I was under the impression that you chose ISO based on default brightness, not just noise and highlights.

But unless i have a baseline to start off with, a proper metering for the exposure if you will, how on earth am i supposed to know how much im supposed to modify it in post?

Personally, I have no problem with this at all. I just use the UniWB histogram. In any case, the ideal camera would have enough metadata to describe the desired brightness.

Just shoot, then go post and try raising the brightness to see if it works? The problem with that method is there is a limit to how underexposed you can go, raising the brightness just won't work at times.

The only limit is noise, which will vary by camera model. Some cameras, like Medium Format digital backs, have no analog gain whatsoever -- and all the ISO settings are just metadata. In the future, I hope they can get rid of the late-stage read noise that makes analog gain useful now. At that point, *all* cameras will use only metadata ISO, and everyone will be raising brightness in post (from ISO 100 to even ISO 12800) instead of using analog ISO settings.

Now you mention default raw conversions. Are you stating that by using other tools other than lightroom, i can actually increase the exposure or brightness of an image by more than 2 stops without increasing noise?

No. I'm simply referring to the brightness you get without raising settings. For example, ISO 400 1/60 and ISO 1600 1/250 both have the same brightness in a default raw conversion. ISO 400 1/250, on the other hand, does not. You would have to manually increase brightness (or set the camera metadata to instruct the raw converter to do so) before it would equal brightness.

Because i've been using lightroom for a while now and especially in low light situations, no you just can't do that. There just isnt that much room in the raw.

It's not "room" in the raw file, it's the SNR over the all the tonal levels.

The whole reason i've been going on about metering is because most people shoot in A or Av mode.

When exposure is limited, that mode is very suboptimal. It will result in some images having an exposure duration that is shorter than the maximum. The Manual Auto-ISO setting is far better if your camera has one that works. That way you always ensure the least amount of noise by using the maximum duration. Then the least important setting, ISO, moves around according to the scene luminance.

Daniel Browning

02-01-2010, 10:51 PM

And the part that really amazes me. You have a software that allows you to increase exposure in post by 5 stops without increasing noise? Er what is it and can i see samples ?

First of all, I don't know if the reason why you rephrased what I said in a grossly inaccurate way is by intention or miscomprehension, but let me quote myself to clear it up.

"I just set the camera to ISO 1600 and underexpose by five stops, which gives me the same noise as shooting ISO 25,600 underexposed 1 stop, except it has 4 stops more headroom".

As for the software that is capable of that, it's almost anything and everything out there except Adobe. Raw Photo Processor, Raw Therapee, Digital Photo Processor, any of the dozens of dcraw derivitives, libraw, ... and that's just the free ones. The commerical raw converters that are capable of this (that I'm aware of) are Phase 1 Capture 1 Pro, Apple Aperture, DxO Optics Pro, and Bibble Pro 5. It's no surprise that so many converters have this simple feature, to get the same noise, all it requires is one multiplication operation in linear light space -- nothing more.

Deltaone

02-02-2010, 02:45 AM

See its the ambiguity in your statements that i just dont understand.

I will take this point by point. Yes i use the default brightness shown on the lcd back, it is what i use to figure out whether the settings ive used have clipped any highlights in the image. Because i'd rather find out then than go back home and find out too late. I had thought that to be obvious enough not to merit statement.

Now i dont know about ideal cameras. But the way i shoot, and the way i understand this, you have to have a baseline to work off of. I cant just arbitrarily decide settings for any one scene. Short of using the zone method or the less accurate sunny 16 rule, the only other way to figure out the settings that i know off is the meter. Now the meter runs based off what iso you've selected, it works the same way in film. That aside, unless you have a baseline, saying a shot is one stop under or overexposed holds no meaning whatsoever. When i say i reduce one from 6 i get 5 it makes sense. But without knowing the 6 how on earth do i deduce the 5. Just knowing the one less part doesn't help does it? Which is why i use the meter in the first place. I can then look at the output and decide which way to go.

I know the limit is noise and it depends on sensor, which is what i've been saying all along btw.

Er i know how to equalize brightness, which is why ive been mentioning the triangle all along. It is by equating the three sides of the triangle that i know a shot is equalized. Again a rather obvious point which is why i prefer to use the word exposure in teh sense to include all 3 to reduce confusion.

Now the last part is ambiguous as hell. How on earth does using A or Av limit exposure in any way? There is no limit to exposure. The only limits to exposure are set by the photographer himself dependent on scene needs. So how does using A limit it any more than using M is the part i dont get. The reason i use A is so i know what the baseline metering is and i have a good idea of how much i can move around. I unfortunately don't have a skill of looking at a scene and arbitrarily deciding which shutterspeed to use. Which is what i mentioned before.

Again with the example. I was at a club recently. The same one synn posted hte last pic of btw. I was using the D70. Now, with the aperture at 1.4 which is wide open for the 50, i was getting shutterspeeds in the range of 1/6 for a scene i was trying to shoot. Now this was at iso 800. I had one option left, raise the iso to 1600 get 1/15 and shoot. I know how the d70 works, and with lightroom the noise is way too much in that kind of scene and so i left it at that.

Now let us assume i needed a shutterspeed of 1/500 in that situation. Which means at iso 1600 which btw is the maximum that the d70s goes to, id have a shot that is underexposed by 5 stops. If i just used the method you just said, put it in manual, auto iso all the way to the top and shoot it, id get a black screen basically that no amount of post could recover.

So why is M and auto iso (btw i dont like it because the cameras i've used so far dont do well at high iso in low light) better than A mode in any way?

Now for the post 2. Yes i know 5 stops underexposure on 1600 is one stop on 25600, that much is basic math, kinda obvious. Moreover i dont beleive the noise will be the same. That said, my question is simple enough. What software allows you to increase your exposure by a margin of 5 stops without increasing your noise.

See, this is my problem. On the 5dm2, you could theoritically underexpose a 1600 shot by four stops and push it in post. I mention this because it actually has an iso 25600 which means the sensor is good enough to push the gain up four stops. But as mentioned above, the d70s doesnt have that much room. Thats sensor dependant. I still dont know if its possible on the 5dm2, im just guessing here.

Again, i ask my question, you have stated quite a few times that raw has as much room as you want. So which software allows you to underexpose a shot by 5 stops and then push it up in post without increasing noise?

From my experience, especially in low light where you need as long an exposure as you can get, if you underexpose a shot by 4-5 stops all you're getting is a blank screen or nearly blank screen. Because the sensor just doesn't get enough time to record the scene since the shutterspeed is too fast.

One final note about auto-iso. I know ken rockwell loves and recommends this, and i tried it a couple of times as well. The reason i gave up on this is simple enough, i want control over what iso my pictures are taken in because i know the characterestics of my sensor and what kind of noise it produces. Which is why i gave it up soon enough. Then again, this is the same man who states program auto is all you need, so i guess its up to you.

I should probably state this just to be clear. I know about highlight clipping. I've mentioned this before and more importantly id thought it would be rather obvious. Why on earth would i shoot something which clips highlights if im after the highlights in the first place? Let us move past the obvious shall we?

Daniel Browning

02-02-2010, 12:55 PM

Short of using the zone method or the less accurate sunny 16 rule, the only other way to figure out the settings that i know off is the meter.

I use the UniWB histogram.

Er i know how to equalize brightness, which is why ive been mentioning the triangle all along. It is by equating the three sides of the triangle that i know a shot is equalized.

Yes, but the brightness may be equalized with "ISO" setting in the camera (analog gain) or with digital gain in the raw converter (correctly called ISO).

Again a rather obvious point which is why i prefer to use the word exposure in teh sense to include all 3 to reduce confusion.

Personally, I find that using the correct definition reduces confusion.

Now the last part is ambiguous as hell. How on earth does using A or Av limit exposure in any way?

Because there is only one exposure duration that is the maximum length possible for any given shot. By using aperture priority mode, you are not fixing the camera to use that one value, but instead allowing the duration to move around as the meter dictates, sometimes using the maximum value and other times using shorter values, all the while leaving the ISO the same. It is far better to always use the maximum value and let ISO move around with the meter (i.e. M+AutoISO).

So why is M and auto iso better than A mode in any way?

Because it always ensures that the maximum possible exposure is used.

Moreover i dont beleive the noise will be the same.

It's a very simple matter to prove. The experiment will only take me 15 minutes. But before I bother going through with it, I want to make sure that you understand what I said. For example, it's possible that someone might have read my post and thought that I was saying "ISO 1600 pushed 0 stops has the same noise as ISO 1600 pushed 5 stops", which is silly of course. So before I bother taking pains to prove it to you, would you be so kind as to rephrase the following quote from my post into your own words?

"I just set the camera to ISO 1600 and underexpose by five stops, which gives me the same noise as shooting ISO 25,600 underexposed 1 stop, except it has 4 stops more headroom"

That would help assure me that we're on the same page and it wont be a wasted effort.

On the 5dm2, you could theoritically underexpose a 1600 shot by four stops and push it in post. I mention this because it actually has an iso 25600 which means the sensor is good enough to push the gain up four stops. But as mentioned above, the d70s doesnt have that much room. Thats sensor dependant. I still dont know if its possible on the 5dm2, im just guessing here.

Of course sensors have different performance, I never said they were all the same or that your D70s could match the 5D2. Rather, the reason I use specific examples is to illustrate a point that is generally applicable. My point was not that all cameras are capable of a nice looking ISO 51,200. It was that there is a certain general principle: the balance between highlights and noise for analog gain is the same among all of them. In fixed exposure/brightness, high ISO has the same or less noise, and low ISO has more headroom.

That said, my question is simple enough.

My answer was simple enough.

What software allows you to increase your exposure by a margin of 5 stops without increasing your noise.

I gave you a list of nine software programs that have the feature required for the conditions in my quote above.

One final note about auto-iso. I know ken rockwell loves and recommends this,

Even a broken clock is right twice a day. KR is wrong on just about everything, but I agree with him on AutoISO.