Real iso's?

[Valvebounce]

Hi guys,
I have read before about intermediate iso's being interpolated from real iso's, but always thought that the last "real" iso was the one before the H setting.

I have no knowledge of the 5D2 in the quote below, is 3200 the last iso before the H or are there more before the H?

Where does one find out about the real iso's for a particular camera? Is it just a case of my 7D has 6400, so that is the last real iso, or is that also lifted from 3200?

I hope this isn't a hanging offence, but this is a quote of dilbert from the post that raised the question.

Topic was "Why is my 5D III so noisy?"

[/quote]

It doesn't matter whether it is ISO 3200 or ISO 6400. Once the ISO is past the point where IQ drops more than 1 stop per ISO stop, increasing the ISO and then overexposing does not result in a better picture because you lose more than you gain by moving the ISO higher.

On the 5D2 the last real ISO is 3200 and everything over that is software underexposing and pulling up.

Whether it is the same on others...
[/quote]

 

[Loswr]

There are basically three 'types' of ISOs:

• Base ISO - this is the 'real' ISO for the sensor before any amplificaition. For most sensors, this is actually in the ISO 60 to ISO 80 range, not ISO 100 as many people assume.
• Native ISOs - analog amplification applied to the base signal, prior to analog to digital conversion (ADC). These values have numbers for selection (e.g. ISO 3200)
• Expanded ISOs - digital amplification applied to the signal after the maximum analog amplification, occurring after the ADC. These valuse have letters for selection (L, H1, etc.).

You're talking about 'tweener' ISOs - those are seen in some cameras but not others. Some analog amplifiers (most of them, for Canon sensors, actually) are only capable of full stop incremental amplification, so digital amplification (pushing/pulling occurring after the ADC) is applied to the signal after the 'native' full-stop analog amplification. If you look at Bill Claff's data for Canon sensors, you can see the zig-zag curves that result from these 'tweener' ISO settings, with all their cameras except the 1D X.

So, dilbert is wrong about the 5DII's ISO settings - ISO 6400 is the highest 'real' (aka native) ISO. He may be confused because ISO 3200 is the highest available ISO when ISO is set to Auto. I have no idea what he means by stating, " Once the ISO is past the point where IQ drops more than 1 stop per ISO stop" - what the heck is '1 stop of IQ'?!? Last time I checked, IQ was not measured in stops anywhere outside of dilbert-land (a fanciful place where lenses are sometimes cameras). But the reality is, noise and DR scale linearly in the 5DII from ISO 1600 all the way through H2, so the 'IQ difference' between ISO 1600 and ISO 3200 is the same as the difference between ISO 3200 and ISO 6400.

Having said that, the highest ISO one will use is a personal judgement call. With the 5DII, ISO 3200 was my highest setting for non-emergency use (but that doesn't make ISO 6400 'not real' - it's still a native ISO setting).

 

[Rienzphotoz]

interesting

 

[privatebydesign]

I turned off 1/3rd stops for iso on my 1Ds MkIII's after reading this a few years ago. http://www.northlight-images.co.uk/article_pages/cameras/canon_1ds3_noise.html

It is easy, if a little involved, to test for yourself using Keith's methodology. Some would argue this is all getting a bit too anal, in general I would agree and many people will never notice the difference, but there are occasions where you are pushing and pulling your file to the absolute limits and that is when little things like this can make a difference. This kind of thing only ever makes a difference if you are shooting RAW and you are comfortable with post processing.

 

[RGF]

Neuro

Great stuff. I learn something from your posts. I had heard about to avoid 1/3 stop ISO , but know know why. Do you have a reference for Nikon? Just wondering if they have similar behavior. Also what about point and shots?

Thanks

 

[Pi]

I turned off 1/3rd stops for iso on my 1Ds MkIII's after reading this a few years ago. http://www.northlight-images.co.uk/article_pages/cameras/canon_1ds3_noise.html

I find this site to be misleading, at least. He tests for read noise (visible mainly in the deep shadows). It is true that "software ISO" gives you higher read noise than hardware one. The shot noise however depends on the total light, and it is not much affected by the ISO. In other words, if you are concerned with the deep shadows, ISO 200 is better than ISO 160. Shooting at ISO 200, in, say Av mode, would lower the total exposure by choosing a faster speed. As a result, the noise away from the deep shadows will increase. If you are going to keep the black level at moderate levels, you will not notice the increased shadow noise anyway.

The best thing in that scenario is to shoot at ISO 200 and overexpose a little (keep the speed the same) assuming you do not blow the highlights.

 

[privatebydesign]

I turned off 1/3rd stops for iso on my 1Ds MkIII's after reading this a few years ago. http://www.northlight-images.co.uk/article_pages/cameras/canon_1ds3_noise.html

I find this site to be misleading, at least. He tests for read noise (visible mainly in the deep shadows). It is true that "software ISO" gives you higher read noise than hardware one. The shot noise however depends on the total light, and it is not much affected by the ISO. In other words, if you are concerned with the deep shadows, ISO 200 is better than ISO 160. Shooting at ISO 200, in, say Av mode, would lower the total exposure by choosing a faster speed. As a result, the noise away from the deep shadows will increase. If you are going to keep the black level at moderate levels, you will not notice the increased shadow noise anyway.

The best thing in that scenario is to shoot at ISO 200 and overexpose a little (keep the speed the same) assuming you do not blow the highlights.

Anybody pushing the limits of Canon RAW files is interested in deep shadows, everything else is easy! So your recommendation is full iso stops and expose to the right, isn't that basic knowledge and in agreement with the article?

 

[sanj]

VERY interesting indeed.
So Mr. Scientist sir (Neuro!), how do I find out the 'base' ISO for 1dx and 5d3. It would be great to know. You are the only one I know who can help.
Regards!
Sanjay

 

[privatebydesign]

VERY interesting indeed.
So Mr. Scientist sir (Neuro!), how do I find out the 'base' ISO for 1dx and 5d3. It would be great to know. You are the only one I know who can help.
Regards!
Sanjay

1DX iso 251
5D MkIII iso 159
both as per the Bill Claff link.

 

[Pi]

Anybody pushing the limits of Canon RAW files is interested in deep shadows, everything else is easy! So your recommendation is full iso stops and expose to the right, isn't that basic knowledge and in agreement with the article?

Not so simple. If you go to ISO 400, instead of ISO 260 (which is ISO 200 pushed), you can blow the highlights with the same exposure to light. So you have to decrease the total light, and this will increase the noise everywhere instead in the shadow area.

 

[privatebydesign]

Anybody pushing the limits of Canon RAW files is interested in deep shadows, everything else is easy! So your recommendation is full iso stops and expose to the right, isn't that basic knowledge and in agreement with the article?

Not so simple. If you go to ISO 400, instead of ISO 260 (which is ISO 200 pushed), you can blow the highlights with the same exposure to light. So you have to decrease the total light, and this will increase the noise everywhere instead in the shadow area.

Using full iso stops does not negate the need to not blow highlights! Show me two images files where your scenario makes an appreciable difference to the overall image quality. You might be making a valid technical point, but I believe I am making a valid user point, shadows are far and away the biggest problem for corrections with Canon RAW files, if there is a setting that elevates a problem from the shadows to anywhere else in the image, as an actual photographer processing RAW files everyday, I'll take it.

It is a shame that the Bill Claff site doesn't have data for my cameras.

 

[Pi]

Anybody pushing the limits of Canon RAW files is interested in deep shadows, everything else is easy! So your recommendation is full iso stops and expose to the right, isn't that basic knowledge and in agreement with the article?

Not so simple. If you go to ISO 400, instead of ISO 260 (which is ISO 200 pushed), you can blow the highlights with the same exposure to light. So you have to decrease the total light, and this will increase the noise everywhere instead in the shadow area.

Using full iso stops does not negate the need to not blow highlights! Show me two images files where your scenario makes an appreciable difference to the overall image quality. You might be making a valid technical point, but I believe I am making a valid user point, shadows are far and away the biggest problem for corrections with Canon RAW files, if there is a setting that elevates a problem from the shadows to anywhere else in the image, as an actual photographer processing RAW files everyday, I'll take it.

I just pointed out the misleading nature of that page (the guy has much worse articles, BTW, like the one on diffraction). What is important to you is something I cannot comment on.

It is a shame that the Bill Claff site doesn't have data for my cameras.

Why do you trust that site so much? I prefer sensorgen. I have communicated with some of the authors, and I am fairly sure that they know what they are doing.

 

[Pi]

Anybody pushing the limits of Canon RAW files is interested in deep shadows, everything else is easy! So your recommendation is full iso stops and expose to the right, isn't that basic knowledge and in agreement with the article?

Not so simple. If you go to ISO 400, instead of ISO 260 (which is ISO 200 pushed), you can blow the highlights with the same exposure to light. So you have to decrease the total light, and this will increase the noise everywhere instead in the shadow area.

Using full iso stops does not negate the need to not blow highlights! Show me two images files where your scenario makes an appreciable difference to the overall image quality. You might be making a valid technical point, but I believe I am making a valid user point, shadows are far and away the biggest problem for corrections with Canon RAW files, if there is a setting that elevates a problem from the shadows to anywhere else in the image, as an actual photographer processing RAW files everyday, I'll take it.

I just pointed out the misleading nature of that page (the guy has much worse articles, BTW, like the one on diffraction). What is important to you is something I cannot comment on.

BTW, presenting black frames as a photographic evidence of real life shooting is of a highly questionable value.

It is a shame that the Bill Claff site doesn't have data for my cameras.

Why do you trust that site so much? I prefer sensorgen. I have communicated with some of the authors, and I am fairly sure that they know what they are doing.

 

[privatebydesign]

It is a shame that the Bill Claff site doesn't have data for my cameras.

Why do you trust that site so much? I prefer sensorgen. I have communicated with some of the authors, and I am fairly sure that they know what they are doing.

I don't trust any site to any substantive degree, but I like user interface direct comparisons, I can choose what information is displayed but must assume that his measurements, even if they are not "accurate", are consistent.

As for Kieth's site, I like it (and he is a member here) because he is an actual working pro with an above average interest in the technical side which he not only demonstrates, but he always lays out his methodology so you can repeat his results of you want to. He also addresses technical ideas and solutions from the perspective of a working pro, he is happy to say things like 'this might be a better way but you'll never see the difference', and that, to me, makes a lot more sense than the Lloyd Chambers style of analyze to death regardless of actual real world differences.

YMMV.

P.S. With regards sensorgen, anybody that measures an output 14 bit file as having 14.0 stops of DR is showing signs of method failure, add in the fact that they take most of their actual measurements off the DxO site and they have lost me.

 

[bdunbar79]

Can I get some credit here?? In the post that the OP is speaking of, I DID tell dilbert he was wrong!

 

[privatebydesign]

Can I get some credit here?? In the post that the OP is speaking of, I DID tell dilbert he was wrong!

We give you credit

Now go and fix that 400 mount

 

[MxM]

I did an ISO test with my 7D.

Put the camera in a dark room (basement) leave the lenscap on, close/block the viewfinder, use RAW or disable ISO long exposure/noise reduction. Choose the settings that you want (Manual mode). Take every picture for at least 10 seconds and you will come up with al list like me.

The 7D is an 18MP camera which is equal to approx. 18MB file size. (the smaller the file size, the cleaner the image is)

As you can see is ISO 160, 320, 640 the camera's sweet-spot... The golden rule is that the limit is MP/MB+1(file size) so 19MB is in the safe/clean zone and 20MB is pushing your camera. It's no math, just hard numbers.

MxM

 

[Rocky]

Based on Bill Claff's data, The canon 30D has quite low readind noise. Any camera after that the read noise has been at least doubled up, includind the 1D-X. Why??? Is Canon is going backward on the read noise???

 

[privatebydesign]

Based on Bill Claff's data, The canon 30D has quite low readind noise. Any camera after that the read noise has been at least doubled up, includind the 1D-X. Why??? Is Canon is going backward on the read noise???

It is because you are comparing 12 bit to 14 bit files.

 

[Sporgon]

Based on Bill Claff's data, The canon 30D has quite low readind noise. Any camera after that the read noise has been at least doubled up, includind the 1D-X. Why??? Is Canon is going backward on the read noise???

It is because you are comparing 12 bit to 14 bit files.

At Building Panoramics we shoot a lot of skys for our library and of all the Canon cameras we've used the 30D was by far the worst for noise, in fact worse than the 20D for some unknown reason.

With regard to Neuro's link, this confirms what I have found in practice. I could never see this '160 ISO is less noisy than 100' statement that's bandied about, but I do find 50 gives noiseless data due to it's 'overexposure pulled back' value.

Actually it was Privatebydesign's link. I can't find the practical link with Claff's data. So the 5D mk2 has the same read noise at ISO 640 as it does at 100 ? Yet shoot a sky at ISO 640a and it's full of noise whereas 100 is ( reasonably) clean. Likewise ISO 125 is similar to 100.

If you're exposing in total darkness with the lens cap on surely you are under exposing never mind how long you leave the shutter open. What about heat generated at these long 'exposures-that-aren't' ?

?