5D III Native ISO

I have been trying to find out once and for all what the true native ISO is for my 5DIII.

I have read some info online stating that the native ISO is 100, 200, 400, however i have just watched a video on Adorama website that says 160, 320, 640 and so on.

Can someone clear this up for me please.

8)

Canon engineers have been asked this outright and there has been no clear answer. It is a matter of speculation. Out of curiosity why the intense curiosity about it?

nqjudo wrote in post #17472233
Canon engineers have been asked this outright and there has been no clear answer. It is a matter of speculation. Out of curiosity why the intense curiosity about it?

Thanks for the reply mate.

I have started an online course in digital photography and one of the elements they discuss is native ISO, this is something that i had never heard of before, so i was really curious.

Perhaps that's why the submission form for my question to Canon would not submit LOL

When I first heard about this, I spent a weekend testing my 60D at various ISOs. I came to the conclusion that the 160, 320, 640, etc. settings were just a hair better than the 200, 400, 800, etc settings. Then I tried to kick myself in the butt. If I am happily shooting at ISOs up to 6400, what difference do miniscule variations at 320 or 400 matter. It is purely an academic exercise. In real shooting, ISO is far from the largest factor affecting outcome.

Don't worry about it. This has no real, practical implications to your photography. -alex

Thanks for the reply guys, i really appreciate it.

Norry Rodgers wrote in post #17472228
I have been trying to find out once and for all what the true native ISO is for my 5DIII.

I have read some info online stating that the native ISO is 100, 200, 400, however i have just watched a video on Adorama website that says 160, 320, 640 and so on.

Can someone clear this up for me please.

8)

For still images, 100, 200, 400, and so on are the "native ISOs." Multiples of 160 actually expose at 200, 400, 800, and so on, and then pull the exposure 1/3 stop. This gives you a slightly cleaner file with slightly less dynamic range. You can use this to your advantage when shooting. If you need all the dynamic range, shoot at multiples of 100. If dynamic range isn't essential and you need more exposure, go with multiples of 160. For video, it seems the 160, 320, 640 ISOs are preferable. Avoid using ISOs that are pushed, that is multiples of 125. These will give you the most noise. For example, ISO 250 actually exposes at ISO 200, and then is digitally pushed 1/3 stop.

And if you are shooting RAW then just shoot ISO 100 and its multiples, Expose To The Right and pull it your self in post. This will guarantee the image to have the maximum DR and minimum noise possible. If you cannot lower the shutter speed, or open the aperture anymore then up the ISO, while still keeping the wanted highlight up against the RHS of the histogram. Do this until you reach the maximum non expanded ISO value. There is little point in using expanded ISO values, as they again are digitally produced, so at that point you can just push the exposure in post with the same results. The minuet that you allow the histogram drop away from the right hand side, you start to lose dynamic range. Remember in the linear space of the sensor, and RAW file the top stop holds full half of all the possible brightness values. Thats a very big penalty for not using ETTR.

Alan

Just to confuse the issue a bit, although 100,200, 400 etc.. are considered the 'native' ISO, in reality the actual true 'native' ISOs are going to be a bit different and change depending on the sensor. It seems that 100, 200 etc are the closest to the true native ISOs.

Canon has quite a bit more going on in terms of amplification after reading the voltage off the sensor which makes things a little more difficult to asses.

I'm not going to even get into what ISO's to shoot except to say that the multiples of 160 do look better for dark frames (i.e. where there is no signal). I don't usually take pictures with no signal - or very little signal so I'm not too interested in that and test in my usual operating conditions.

BigAl007 wrote in post #17472323
And if you are shooting RAW then just shoot ISO 100 and its multiples, Expose To The Right and pull it your self in post.

OK, I am going to tell 'you' (OP, not Alan) what ISO to shoot with an explanation I thought of recently so let's see if it flies.

Although ISO160 looks cleaner than ISO100 on a noise test (i.e. a shot with the lens cap on), instead of shooting at ISO160 and say 1/60s you can shoot at ISO100 and 1/30s. Then although the read noise is higher at ISO100 (canon read noise decreases with ISO, although other noise does increase), your signal is now 2x stronger. So your signal/noise is actually better at ISO100. Same goes for other multiples of 160.

One test I did is that on the dark frame noise tests ISO320 actually looks cleaner than ISO100. However, if you take a shot of a scene at ISO320 (with some dark areas so you can see the signal/noise) and the same scene at ISO100 with an exposure time 4 times as long, it should be pretty clear looking at the dark areas which one is better. Also since 320 is essentially ISO400 pulled, your highlights should look very similar in both cases in terms of detail.

This is why I don't think those dark frame noise tests are very useful, or more specifically I think people have reached incorrect conclusions from them.

BigAl007 wrote in post #17472323
And if you are shooting RAW then just shoot ISO 100 and its multiples, Expose To The Right and pull it your self in post. This will guarantee the image to have the maximum DR and minimum noise possible. If you cannot lower the shutter speed, or open the aperture anymore then up the ISO, while still keeping the wanted highlight up against the RHS of the histogram. Do this until you reach the maximum non expanded ISO value. There is little point in using expanded ISO values, as they again are digitally produced, so at that point you can just push the exposure in post with the same results. The minuet that you allow the histogram drop away from the right hand side, you start to lose dynamic range. Remember in the linear space of the sensor, and RAW file the top stop holds full half of all the possible brightness values. Thats a very big penalty for not using ETTR.

Alan

Some very in depth explanations guys, many thanks again.

I might try the lens cap shoot, never thought of that to be honest then compare the RAW files in Lightroom.

thanks again.

ejenner wrote in post #17472362
Just to confuse the issue a bit, although 100,200, 400 etc.. are considered the 'native' ISO, in reality the actual true 'native' ISOs are going to be a bit different and change depending on the sensor. It seems that 100, 200 etc are the closest to the true native ISOs.

Canon has quite a bit more going on in terms of amplification after reading the voltage off the sensor which makes things a little more difficult to asses.

I'm not going to even get into what ISO's to shoot except to say that the multiples of 160 do look better for dark frames (i.e. where there is no signal). I don't usually take pictures with no signal - or very little signal so I'm not too interested in that and test in my usual operating conditions.

OK, I am going to tell 'you' (OP, not Alan) what ISO to shoot with an explanation I thought of recently so let's see if it flies.

Although ISO160 looks cleaner than ISO100 on a noise test (i.e. a shot with the lens cap on), instead of shooting at ISO160 and say 1/60s you can shoot at ISO100 and 1/30s. Then although the read noise is higher at ISO100 (canon read noise decreases with ISO, although other noise does increase), your signal is now 2x stronger. So your signal/noise is actually better at ISO100. Same goes for other multiples of 160.

One test I did is that on the dark frame noise tests ISO320 actually looks cleaner than ISO100. However, if you take a shot of a scene at ISO320 (with some dark areas so you can see the signal/noise) and the same scene at ISO100 with an exposure time 4 times as long, it should be pretty clear looking at the dark areas which one is better. Also since 320 is essentially ISO400 pulled, your highlights should look very similar in both cases in terms of detail.

This is why I don't think those dark frame noise tests are very useful, or more specifically I think people have reached incorrect conclusions from them.

Could I suggest that you try the dark frame test with the camera set to ISO 200 and then pulled 1/3 stop in your RAW converter. Then compare it with the ISO 160 shot. Both of course with the same shutter time. I think you will find the ISO 200 shot will now be the same or possibly even better than the ISO 160 shot. If you then do the same for a relatively normal range of tones the shadow noise will generally always look better with the ISO 100 multiples used with ETTR to maximise the DR. The other advantage of doing this yourself in a modern RAW converter is control. Usually you will be able to control the highlights midtones and shadow detail along with white and black point independently. This actually gives you a big advantage over the in camera processing which simply pulls everything by 1/3 stop.

When it comes to the analogue amplification stages between the sensor and the ADC I'm only making a guess based on having a background in Electronics, and I don't know if Canon have actually published anything specific about their designs. This actually all depends on the actual signal voltage levels at the sensor. If they are such that the ADC is capable of the necessary discrimination to measure the sensor voltage to 14 bit accuracy, then the base ISO level will not require amplification of the voltage level. This is the ideal situation, and is even better if the ADC is really close to the sensor, say on the same bit of silicon (Sony Exmoor). If the ADC is a long way away on another chip then even if the signal voltage is sufficient for the ADC then you may still need to use a driver amplifier, with voltage unity gain (i.e the voltage remains the same). This will usually be the last stage of the amplifier, running the signal through the first voltage amplifier stage(s) with the gain at 1× should be no worse than running it along the conductor. For each full stop of ISO you add you double the voltage gain of the amplifier. You then will still need to run the signal through the driver stage to build the current necessary to get the voltage to where it needs to be, the ADC.

Now there is actually nothing that says you cannot change the analogue amplification stages gain by any amount you like, it is after all a simple analogue system. I believe the main reason that they do not do this is calibration. You would need to be able to accurately know just how much gain you were going to get. So that you can set other two corners of your exposure accurately enough to not to blow wanted highlight detail. It is therefore necessary to limit the number of steps of analogue gain that you are able to use, to the number of accurate bias voltage levels you have available. Limitations of the sensor physics defines your upper analogue amplification level. All of this is of course then calibrated against the ISO sensitivity standard. As I said this is based purely on speculation, and experience of combined analogue, pulse and digital electronic receivers and signal processing systems, that are not cameras.

Alan

Brilliant thread- first thread I have read every posting on.

If the processing of 'non-iso100 2x multiples' is essentially post processed in camera, then why are they present? Surely if it is cleaner to pull down the exposure from a 'native' ISO then that would be best to do in all circumstances.

On the Sensor iso1600 will not need twice the current/power/voltage(​which ever it is) of iso800, it just needs to be twice as sensitive-
as an analogy -take a slightly open curtain- to let in twice as much light the gap just needs to be twice as wide, The size of the curtain is not relevant and does not need to be known.

The base electrical figures at the sensor we do not know, it is the fractional increase/ decrease that changes the sensitivity of the sensor. ie 1% is 1/16 of 16%,
but 84% is only , well 84% of 99%
So why not have all the increments inbetween.

BigAl007 wrote in post #17472880
Could I suggest that you try the dark frame test with the camera set to ISO 200 and then pulled 1/3 stop in your RAW converter. Then compare it with the ISO 160 shot. Both of course with the same shutter time. I think you will find the ISO 200 shot will now be the same or possibly even better than the ISO 160 shot. If you then do the same for a relatively normal range of tones the shadow noise will generally always look better with the ISO 100 multiples used with ETTR to maximise the DR. The other advantage of doing this yourself in a modern RAW converter is control. Usually you will be able to control the highlights midtones and shadow detail along with white and black point independently. This actually gives you a big advantage over the in camera processing which simply pulls everything by 1/3 stop.

Alan

I think we agree and have done similar tests, I just don't like to go against the grain too much (still a lot of people swear by ISO160 multiples). I only use multiples of 100 except for video.

john5189 wrote in post #17472929
If the processing of 'non-iso100 2x multiples' is essentially post processed in camera, then why are they present? Surely if it is cleaner to pull down the exposure from a 'native' ISO then that would be best to do in all circumstances.

A few reasons:

For interesting and sometimes heated discussions on internet forums. Many people are still convinced that multiples of 160 are the way to go.
People expect to be able to change everything by 1/3 stop. It would just be weird not to be able to change ISO by 1/3 stop.
For video, where you are not shooting raw, multiples of 160 above ISO100 can be better + your shutter speed is usually fixed and you can't ETTR so you need more control.
Some people want to specify precise exposure time and aperture and are not too bothered about whether one shot is slightly noisier than another
Convenience. Again most people probably don't care about the difference in noise at ISO200 and 250.
Auto ISO - if you fix exposure time and aperture, you want to have ISO vary by 1/3 stop.
Shooting .jpeg and semi-ETTR (similar to video, multiples of 160 are best above ISO100).

I'm sure there are other reasons. Also 1D series cameras do not 'push' and 'pull' for intermediate ISOs, so they have something closer to native ISOs for all values.

I don't think it is a coincidence that Canon allows you to use ISO in 1/3 stops or whole stops and not whole stop multiples of 160 though.

BigAl007 wrote in post #17472880
Could I suggest that you try the dark frame test with the camera set to ISO 200 and then pulled 1/3 stop in your RAW converter. Then compare it with the ISO 160 shot. Both of course with the same shutter time. I think you will find the ISO 200 shot will now be the same or possibly even better than the ISO 160 shot. If you then do the same for a relatively normal range of tones the shadow noise will generally always look better with the ISO 100 multiples used with ETTR to maximise the DR. The other advantage of doing this yourself in a modern RAW converter is control. Usually you will be able to control the highlights midtones and shadow detail along with white and black point independently. This actually gives you a big advantage over the in camera processing which simply pulls everything by 1/3 stop.

When it comes to the analogue amplification stages between the sensor and the ADC I'm only making a guess based on having a background in Electronics, and I don't know if Canon have actually published anything specific about their designs. This actually all depends on the actual signal voltage levels at the sensor. If they are such that the ADC is capable of the necessary discrimination to measure the sensor voltage to 14 bit accuracy, then the base ISO level will not require amplification of the voltage level. This is the ideal situation, and is even better if the ADC is really close to the sensor, say on the same bit of silicon (Sony Exmoor). If the ADC is a long way away on another chip then even if the signal voltage is sufficient for the ADC then you may still need to use a driver amplifier, with voltage unity gain (i.e the voltage remains the same). This will usually be the last stage of the amplifier, running the signal through the first voltage amplifier stage(s) with the gain at 1× should be no worse than running it along the conductor. For each full stop of ISO you add you double the voltage gain of the amplifier. You then will still need to run the signal through the driver stage to build the current necessary to get the voltage to where it needs to be, the ADC.

Now there is actually nothing that says you cannot change the analogue amplification stages gain by any amount you like, it is after all a simple analogue system. I believe the main reason that they do not do this is calibration. You would need to be able to accurately know just how much gain you were going to get. So that you can set other two corners of your exposure accurately enough to not to blow wanted highlight detail. It is therefore necessary to limit the number of steps of analogue gain that you are able to use, to the number of accurate bias voltage levels you have available. Limitations of the sensor physics defines your upper analogue amplification level. All of this is of course then calibrated against the ISO sensitivity standard. As I said this is based purely on speculation, and experience of combined analogue, pulse and digital electronic receivers and signal processing systems, that are not cameras.

Alan

This is some great reading and very informative too.

Thanks Alan.