Canon R5 read noise

[DanK]

When I upgraded my Canon 5D Mark III not long ago, the primary reason I chose the 5D Mark IV over the R5 was price. Net of selling my old camera, the R5 would have cost me 2.5 times as much, because it's a new model and the Mark IV was on sale for a month. Nonetheless, I had a list of pros and cons. One of the (hypothetical, as it turns out) cons for the R5 was the possibility of greater noise from smaller photosites (45 vs. 30 mpx). Turns out not to be the case: improvements in sensor and amplification technology apparently offset the difference. I don't understand the reason for the big non montonicity between ISO 320 and 400 on the R5, but the notes at the bottom have a few clues.

It's hard to read at this size, so I'll note that blue is the R5.

[lunaticitizen]

When I upgraded my Canon 5D Mark III not long ago, the primary reason I chose the 5D Mark IV over the R5 was price. Net of selling my old camera, the R5 would have cost me 2.5 times as much, because it's a new model and the Mark IV was on sale for a month. Nonetheless, I had a list of pros and cons. One of the (hypothetical, as it turns out) cons for the R5 was the possibility of greater noise from smaller photosites (45 vs. 30 mpx). Turns out not to be the case: improvements in sensor and amplification technology apparently offset the difference. I don't understand the reason for the big non montonicity between ISO 320 and 400 on the R5, but the notes at the bottom have a few clues.

It's hard to read at this size, so I'll note that blue is the R5.

Emphasis mine.

Isn't that the famous 'dual conversion gain' from Aptina?^[1][2] It's like the camera has two base ISOs. Apparently the R5 is the first Canon tested by Bill Claff^[3] that utilizes the technology.

References:

1. Aptina, 2010, Leveraging Dynamic Response Pixel Technology to Optimize Inter-scene Dynamic Range, https://www.photonstophotos.net/Apti...WhitePaper.pdf
2. Bill Claff, 2015, Sony A7S DR-Pix Read Noise, https://www.photonstophotos.net/Gene...Read_Noise.htm
3. Bill Claff, 2020, Canon EOS R5 Sensor Measurements at PhotonsToPhotos, https://www.dpreview.com/forums/post/64211915

[dje]

Emphasis mine.

Isn't that the famous 'dual conversion gain' from Aptina?^[1][2] It's like the camera has two base ISOs. Apparently the R5 is the first Canon tested by Bill Claff^[3] that utilizes the technology.

References:

1. Aptina, 2010, Leveraging Dynamic Response Pixel Technology to Optimize Inter-scene Dynamic Range, https://www.photonstophotos.net/Apti...WhitePaper.pdf
2. Bill Claff, 2015, Sony A7S DR-Pix Read Noise, https://www.photonstophotos.net/Gene...Read_Noise.htm
3. Bill Claff, 2020, Canon EOS R5 Sensor Measurements at PhotonsToPhotos, https://www.dpreview.com/forums/post/64211915

Yes I believe that’s the case. There is an extra capacitance in each pixel that is switched in for low ISO and switched out above a certain ISO setting. Sony and Nikon have been using it for a while now. The Aptina article is a good read.

Dave

[DanK]

Thanks for the explanation and link.


Sent from my iPad using Tapatalk

[lunaticitizen]

For the R5, the practical consequence of having two base ISO is like this when deciding the optimal exposure (and if you shoot raw):

1. Decide on the f-stop based on the DOF you need
2. Decide on the shutter speed based on motion blur
3. If the metered ISO is below 400, then set the ISO to 100. If the metered ISO is 400 or above, then set the ISO to 400.

I'm mainly an aperture-priority-auto-ISO-JPEG-only shooter so the above strategy doesn't really matter to me

[dje]

For the R5, the practical consequence of having two base ISO is like this when deciding the optimal exposure (and if you shoot raw):

1. Decide on the f-stop based on the DOF you need
2. Decide on the shutter speed based on motion blur
3. If the metered ISO is below 400, then set the ISO to 100. If the metered ISO is 400 or above, then set the ISO to 400.

I'm mainly an aperture-priority-auto-ISO-JPEG-only shooter so the above strategy doesn't really matter to me

Hi Leo

Setting ISO to either 100 or 400 could be done but this would generally require adjusting the lightness in post if you are setting aperture and shutter speed manually. This should be OK because the camera seems to be ISO-less. However I think it would be easier for many to simply use the ISO setting predicted to give suitable lightness straight off (whether it be 100, 200 500 etc), even for raw shooters.

Dave

[DanK]

This should be OK because the camera seems to be ISO-less.

Given the nonmonotonicity of the curve below 400, it can't be ISO-less at low ISOs, no?

However I think it would be easier for many to simply use the ISO setting predicted to give suitable lightness straight off (whether it be 100, 200 500 etc), even for raw shooters.

I agree, particularly if you set the camera for 1-stop ISO increments, as I do. That would get rid of the two aberrant high values between 200 and 400.

For me, the bonus would be the considerably lower read noise at 400 compared to that of the 5D IV I use now. 400 is a setting I use a great deal for shots with TTL flash because of the balance it often provides between ambient flight and flash. However, the ISO 400 shots I get now are generally quite clean if well exposed.

[royent]

Is 'nonmonotonicity' a word?

[DanK]

Is 'nonmonotonicity' a word?

Yes: https://en.wiktionary.org/wiki/nonmonotonicity

[dje]

Given the nonmonotonicity of the curve below 400, it can't be ISO-less at low ISOs, no?

Dan I see where you are coming from. Perhaps the camera could be described as having two sets of ISO-less regions, one below and one above ISO400? If you limit your lightening in post to one of those regions, there should be no problem with noise. However if you used an ISO setting of 100 and tried to lighten the image in post by say 4 stops (to an effective ISO1600), you would lose the advantage of the dual conversion gain feature. It wouldn't be ISO-less in that case.

Dave

PS: I don't really like the term ISO-less as it is a very confusing term IMO.

[lunaticitizen]

Hi Leo

Setting ISO to either 100 or 400 could be done but this would generally require adjusting the lightness in post if you are setting aperture and shutter speed manually. This should be OK because the camera seems to be ISO-less. However I think it would be easier for many to simply use the ISO setting predicted to give suitable lightness straight off (whether it be 100, 200 500 etc), even for raw shooters.

Dave

Hi Dave,

Yeah, the strategy is mainly for technical users who don't mind if their pictures are very, very dark when viewed through the back monitor of the camera. I'm not one of them

In case you push shadows a lot in post though, I can see why the strategy is beneficial for the shooter. See this.

His unabridged exposure strategy is much more complicated than the one I posted above.

[lunaticitizen]

Given the nonmonotonicity of the curve below 400, it can't be ISO-less at low ISOs, no?



I agree, particularly if you set the camera for 1-stop ISO increments, as I do. That would get rid of the two aberrant high values between 200 and 400.

For me, the bonus would be the considerably lower read noise at 400 compared to that of the 5D IV I use now. 400 is a setting I use a great deal for shots with TTL flash because of the balance it often provides between ambient flight and flash. However, the ISO 400 shots I get now are generally quite clean if well exposed.

If you're sensitive about apparent noise, according to DXOMARK the cameras are virtually indistinguishable.



The price difference is huge, though

[pnodrog]

The first graph shows the read noise against the ISO. Although there is a strange looking kick in it but the advantage of the low level region is the extension of the increase in dynamic range. From a photography point of view the dynamic range vs ISO at a set noise level is far more indicative graph of the overall performance.

If it is a similar design to the one described by Aptina it is essentially what could be called a dual ISO invariant sensor.

Just thought I should add to the confusion.....

P.S. Read noise is just one factor in overall noise. The benefit of allowing the higher charge to be collected (low ISO portion) is because the shot noise becomes less and probably more than compensates for the increased read noise.

[DanK]

If you're sensitive about apparent noise, according to DXOMARK the cameras are virtually indistinguishable.

Details aside, that's actually why I started the thread. I had assumed that the extra pixels would harm that aspect of image quality, but that turns out not to be the case.

In any case, for me, it's a moot point: I'm not going to replace my 5D IV unless someone steals it or it gets destroyed. The bottleneck for me usually me, not my camera.

[pnodrog]

Details aside, that's actually why I started the thread. I had assumed that the extra pixels would harm that aspect of image quality, but that turns out not to be the case.

In any case, for me, it's a moot point: I'm not going to replace my 5D IV unless someone steals it or it gets destroyed. The bottleneck for me usually me, not my camera.

At very high ISO there is a slight advantage in fewer pixels. The 5D IV offers good bang for bucks.

[lunaticitizen]

Details aside, that's actually why I started the thread. I had assumed that the extra pixels would harm that aspect of image quality, but that turns out not to be the case.

In any case, for me, it's a moot point: I'm not going to replace my 5D IV unless someone steals it or it gets destroyed. The bottleneck for me usually me, not my camera.

Next time when it's time to change your camera perhaps you needn't worry about pixel count since their SNRs look so similar when normalized. More pixels means less aliasing, though it's not really a problem with Canon since their cameras generally have an AA filter.

[DanK]

Indeed. That was my conclusion. Now if someone will just send me a check for $4000, I'm set to go. Although in fact, I'm still often struck by how amazing the 5D IV I now have actually is, so basent something like a theft, that next time is probably a long time away.