Petapixel: 6D vs 6DII, high ISO

[Don Haines]

Don, sorry. With all due respect, I took the liberty of pointing out that the following your statement is incorrect. DPAF tech does not results in slightly poorer high ISO performance. That's all.

Personally, I've noted that the dual pixel technology does seem to result in a slightly poorer high ISO noise.

Mt. Spokane said that, not me......

Personally, I don't know how you could tell the difference. The lithography has changed, the design has changed, the pixel size has changed, and the processing has changed.... there is no example of an identical camera, one with and the other without dual pixel, to compare between....

 

[SecureGSM]

Don, I am so sorry. I posted from my smartphone and replied quoting wrong person. Really sorry.

P.S. high ISO performance of 5D iv higher resolution and higher pixel density dpaf enabled sensor is no worse than the of 6D II.

DPAF enabled 1Dx II high ISO performance is no worse than the one of 1dx

80d vs 70D vs 7D II - same story.

 

[Don Haines]

Don, I am so sorry. I posted from my smartphone and replied quoting wrong person. Really sorry.

P.S. high ISO performance of 5D iv higher resolution and higher pixel density dpaf enabled sensor is no worse than the of 6D II.

DPAF enabled 1Dx II high ISO performance is no worse than the one of 1dx

80d vs 70D vs 7D II - same story.

No problem! It's the web and we all make mistakes, and I regularly mis-read stuff on my phone

 

[Joules]

there is no example of an identical camera, one with and the other without dual pixel, to compare between....

But we have the EOS 750D (T6i) and the EOS 800D (T7i) which both have 24.2 effective megapixels. Those might be the closest thing we have to see the difference in High ISO performance, has somebody looked at both and done a comparison? Because it has been stated multiple times that Dual-Pixel Technology has some disadvantages on IQ, but I haven't seen any proof.

 

[SecureGSM]

1dx 18Mp , no DPAF vs 1dx II 20Mp /w DPAF. 1dx II wins.

there is no example of an identical camera, one with and the other without dual pixel, to compare between....

But we have the EOS 750D (T6i) and the EOS 800D (T7i) which both have 24.2 effective megapixels. Those might be the closest thing we have to see the difference in High ISO performance, has somebody looked at both and done a comparison? Because it has been stated multiple times that Dual-Pixel Technology has some disadvantages on IQ, but I haven't seen any proof.

 

[Joules]

1dx 18Mp , no DPAF vs 1dx II 20Mp /w DPAF. 1dx II wins.

there is no example of an identical camera, one with and the other without dual pixel, to compare between....

But we have the EOS 750D (T6i) and the EOS 800D (T7i) which both have 24.2 effective megapixels. Those might be the closest thing we have to see the difference in High ISO performance, has somebody looked at both and done a comparison? Because it has been stated multiple times that Dual-Pixel Technology has some disadvantages on IQ, but I haven't seen any proof.

Oh, right ;D

 

[Mt Spokane Photography]

Personally, I've noted that the dual pixel technology does seem to result in a slightly poorer high ISO noise. Its difficult to asses because the pixel count keeps rising as well. However, I'm not one to compare images at a low resolution, if I wanted a lower MP camera, I'd get one and pay less.

However, the advantages of Dual Pixel in accurate live autofocus, and even autofocus at f/11 or f/16 in my tests of my 5D MK IV are worth the tiny high ISO difference. I think its amazing to see my 100-400mm L with solar filter attached to autofocus at f/16 and even f/22 when its pre-focused to be close.

I don't think it's fair to compare noise on the pixel level coming different resolution sensors. No question a file from the 5dsr will look like colorful mess, but it'll be comparable when downsampled.

Down sampling means reducing the resolution. Why buy a 50 mp camera to print 3 X 5 images? If I wanted low resolution, I'd buy a low resolution camera. Some don't see it the same way, and say that if you downsampled the image to 8mb or whatever, they would look great. Thats true, but what if you want to crop severely. Then you can't do that because of the noise.

Cropping images to 8mb sized frames would be a better comparison for me.

 

[rfdesigner]

I'm not quite sure why people are still chasing this one, the physics of the current technology is more or less maxed out.

Read noise is now so low that you could only tell the difference between a real world camera and perfection (zero noise) in the last 1/4 of a stop of so of shadow of an image. (read noise ~ 1.5e (6D @ high ISO), signal of 4e results in noise of 2e)

The technology that will help isn't anything that lowers noise, but tech that increases Quantum Efficiency.

Currently we waste 2/3rds of the signal in the bayer matrix, and maybe 30~50% of the remaining signal in the front side illumination.

Foveon tech or tripple mono sensors with >90% efficient colour splitting prisms are the only tech I'm aware of that would allow the bayer matrix to be beaten in terms of quantum efficiency, maybe a stop and a bit at most.

Back side illumination would buy 2/3rds of a stop if optimised for quantum efficiency.

If we do get >90% QE in a colour sensor with the same very low readout noise then we'll have perfection and the only way to upgrade from there will be to go for bigger cameras, or perhaps someone can get God to change the laws of physics.

Anyway back to the 6D vs 6DII article.. without accurate AF, you can't really compare except in absolute black noise level, which they didn't do, and even if they had all they'd end up saying is.. it's about the same, which is exactly what everyone should expect. The 6D was the best of it's generation in terms of noise, finally getting close to the limits of the technology, we should expect all cameras to be roughly on this level now.

 

[Don Haines]

I'm not quite sure why people are still chasing this one, the physics of the current technology is more or less maxed out.

Read noise is now so low that you could only tell the difference between a real world camera and perfection (zero noise) in the last 1/4 of a stop of so of shadow of an image. (read noise ~ 1.5e (6D @ high ISO), signal of 4e results in noise of 2e)

The technology that will help isn't anything that lowers noise, but tech that increases Quantum Efficiency.

Currently we waste 2/3rds of the signal in the bayer matrix, and maybe 30~50% of the remaining signal in the front side illumination.

Foveon tech or tripple mono sensors with >90% efficient colour splitting prisms are the only tech I'm aware of that would allow the bayer matrix to be beaten in terms of quantum efficiency, maybe a stop and a bit at most.

Back side illumination would buy 2/3rds of a stop if optimised for quantum efficiency.

If we do get >90% QE in a colour sensor with the same very low readout noise then we'll have perfection and the only way to upgrade from there will be to go for bigger cameras, or perhaps someone can get God to change the laws of physics.

Anyway back to the 6D vs 6DII article.. without accurate AF, you can't really compare except in absolute black noise level, which they didn't do, and even if they had all they'd end up saying is.. it's about the same, which is exactly what everyone should expect. The 6D was the best of it's generation in terms of noise, finally getting close to the limits of the technology, we should expect all cameras to be roughly on this level now.

+1

We are getting down to differences that only matter in a lab. You need more light for any significant improvement, and that means bigger glass (presumably with a bigger sensor)..... It is not going to be cheap nor will it be as portable......

 

[rfdesigner]

I'm not quite sure why people are still chasing this one, the physics of the current technology is more or less maxed out.

Read noise is now so low that you could only tell the difference between a real world camera and perfection (zero noise) in the last 1/4 of a stop of so of shadow of an image. (read noise ~ 1.5e (6D @ high ISO), signal of 4e results in noise of 2e)

The technology that will help isn't anything that lowers noise, but tech that increases Quantum Efficiency.

Currently we waste 2/3rds of the signal in the bayer matrix, and maybe 30~50% of the remaining signal in the front side illumination.

Foveon tech or tripple mono sensors with >90% efficient colour splitting prisms are the only tech I'm aware of that would allow the bayer matrix to be beaten in terms of quantum efficiency, maybe a stop and a bit at most.

Back side illumination would buy 2/3rds of a stop if optimised for quantum efficiency.

If we do get >90% QE in a colour sensor with the same very low readout noise then we'll have perfection and the only way to upgrade from there will be to go for bigger cameras, or perhaps someone can get God to change the laws of physics.

Anyway back to the 6D vs 6DII article.. without accurate AF, you can't really compare except in absolute black noise level, which they didn't do, and even if they had all they'd end up saying is.. it's about the same, which is exactly what everyone should expect. The 6D was the best of it's generation in terms of noise, finally getting close to the limits of the technology, we should expect all cameras to be roughly on this level now.

+1

We are getting down to differences that only matter in a lab. You need more light for any significant improvement, and that means bigger glass (presumably with a bigger sensor)..... It is not going to be cheap nor will it be as portable......

Don

There is one aspect that no one anywhere seems to have measured that you might be able to help out on. (as you have a 6DII.. and a 6d?)

Dark Current.

The 7DII is a monster in this regard, while the 6D mkI is more ho-hum, would be interesting to see where the MkII stands.

So.. any chance of a really long exposure comparison? 6D vs 6DII

Perhaps something at ~ISO3200.. (almost lowest readout noise but still decent dynamic range)

I'm thinking 5 minutes not 30 seconds, clearly of something really dark, perhaps a star trail image with silloette trees?

 

[Joules]

Dark Current.

The 7DII is a monster in this regard, while the 6D mkI is more ho-hum, would be interesting to see where the MkII stands.

Sorry to go off topic here, but I have been wondering how dark current noise becomes visible in an image and searching on Google i haven't found any great examples. Is it the same as thermal noise that appears in long exposures? So, do different sensors differ visibly in how much of it they produce? In that case, I'd be interested in this too.

Here's a 100% crop of a long exposure I've taken at a beach, where the only light comes from the sky in the background and a 30% moon to the left. It was 1 am and pretty damn dark to the eye, there's 0 direct light on the front of that building. 210 seconds, f/5.6, ISO 400, Canon T3i, pushed by 1 stop in LR with a hint of clarity to show the dots better. Is that nasty intense color noise there also dark current or is it something else? Because i really dislike it, and wondered if moving to a different camera would get reduce it in scenarios like this.

 

[Don Haines]

I'm not quite sure why people are still chasing this one, the physics of the current technology is more or less maxed out.

Read noise is now so low that you could only tell the difference between a real world camera and perfection (zero noise) in the last 1/4 of a stop of so of shadow of an image. (read noise ~ 1.5e (6D @ high ISO), signal of 4e results in noise of 2e)

The technology that will help isn't anything that lowers noise, but tech that increases Quantum Efficiency.

Currently we waste 2/3rds of the signal in the bayer matrix, and maybe 30~50% of the remaining signal in the front side illumination.

Foveon tech or tripple mono sensors with >90% efficient colour splitting prisms are the only tech I'm aware of that would allow the bayer matrix to be beaten in terms of quantum efficiency, maybe a stop and a bit at most.

Back side illumination would buy 2/3rds of a stop if optimised for quantum efficiency.

If we do get >90% QE in a colour sensor with the same very low readout noise then we'll have perfection and the only way to upgrade from there will be to go for bigger cameras, or perhaps someone can get God to change the laws of physics.

Anyway back to the 6D vs 6DII article.. without accurate AF, you can't really compare except in absolute black noise level, which they didn't do, and even if they had all they'd end up saying is.. it's about the same, which is exactly what everyone should expect. The 6D was the best of it's generation in terms of noise, finally getting close to the limits of the technology, we should expect all cameras to be roughly on this level now.

+1

We are getting down to differences that only matter in a lab. You need more light for any significant improvement, and that means bigger glass (presumably with a bigger sensor)..... It is not going to be cheap nor will it be as portable......

Don

There is one aspect that no one anywhere seems to have measured that you might be able to help out on. (as you have a 6DII.. and a 6d?)

Dark Current.

The 7DII is a monster in this regard, while the 6D mkI is more ho-hum, would be interesting to see where the MkII stands.

So.. any chance of a really long exposure comparison? 6D vs 6DII

Perhaps something at ~ISO3200.. (almost lowest readout noise but still decent dynamic range)

I'm thinking 5 minutes not 30 seconds, clearly of something really dark, perhaps a star trail image with silloette trees?

I have a 6D at work, and a 6D2 at home. The site is "secure", which means that I can't bring my 6D2 in there, and it also means that the 6D is only for work and I can't take it out.... That makes it real hard to find a dark scene to use for comparisons...

I can try a comparison of the 7D2 and the 6D2 on the same scene....

 

[rfdesigner]

I have a 6D at work, and a 6D2 at home. The site is "secure", which means that I can't bring my 6D2 in there, and it also means that the 6D is only for work and I can't take it out.... That makes it real hard to find a dark scene to use for comparisons...

I can try a comparison of the 7D2 and the 6D2 on the same scene....

great.. I'll try and replicate with my 6D once I see what you've managed to produce.

 

[SecureGSM]

Would 5 min long exposure with lens cap on do the trick? I guess, it should. If so, then perhaps it makes it easier for Don to find that dark scene to use for comparison with cameras at work and at home.

I have a 6D at work, and a 6D2 at home. The site is "secure", which means that I can't bring my 6D2 in there, and it also means that the 6D is only for work and I can't take it out.... That makes it real hard to find a dark scene to use for comparisons...

I can try a comparison of the 7D2 and the 6D2 on the same scene....

great.. I'll try and replicate with my 6D once I see what you've managed to produce.

 

[rfdesigner]

Would 5 min long exposure with lens cap on do the trick? I guess, it should. If so, then perhaps it makes it easier for Don to find that dark scene to use for comparison with cameras at work and at home.

Yes this can be good. clearly I'm looking for RAW unprocessed data.

Dark Current.

The 7DII is a monster in this regard, while the 6D mkI is more ho-hum, would be interesting to see where the MkII stands.

Sorry to go off topic here, but I have been wondering how dark current noise becomes visible in an image and searching on Google i haven't found any great examples. Is it the same as thermal noise that appears in long exposures? So, do different sensors differ visibly in how much of it they produce? In that case, I'd be interested in this too.

Here's a 100% crop of a long exposure I've taken at a beach, where the only light comes from the sky in the background and a 30% moon to the left. It was 1 am and pretty damn dark to the eye, there's 0 direct light on the front of that building. 210 seconds, f/5.6, ISO 400, Canon T3i, pushed by 1 stop in LR with a hint of clarity to show the dots better. Is that nasty intense color noise there also dark current or is it something else? Because i really dislike it, and wondered if moving to a different camera would get reduce it in scenarios like this.

What happens is the apparent read-noise increases, it's a key number for astro-imaging, as poor dark current can swamp sky glow limiting ultimate sensitivity, it becomes a real problem for narrowband imaging where you're excluding 99% of the light. If you wanted a single star trail image with great long arcs you might want to do this by shooting with a stopped down lens but you'd probably choose ISO100 first, using ISO3200 just makes it easier to measure.

here's a nice graph of different canon cameras from clark vision: http://www.clarkvision.com/reviews/evaluation-canon-7dii/dark-current-compared_2-6c-v1.gif

One thought... a long and a short image of the lens cap would be a good idea. I'll try and do a star trail image the next clear night I get to show the kind of impact with the 6D... could be a while, we're getting a lot of rain/cloud at the moment.


EDIT: ok.. I've checked elsewhere and determined the 6D gain @ ISO3200 is 5ADU/e. I've taken a blank ISO3200 300s image and the noise (as measured in IRIS, in un-debayered form) is 57ADU.. this translates back to approx 2e/s, which correlates well with the above link (@20C). For reference a sub 1s image with otherwise identical settings produces only 12ADU of noise.

Interestingly there is no DC-offset in the 300s blank image.. which means Canon must be processing the RAWs, probably averaging the unlit pixels and taking this value from the rest of the frame before saving.