PhotonsToPhotos does the Canon EOS R5 Mark II and it’s good

[docsmith]

My point is that there IS noise reduction being applied across the whole ISO spectrum with the R5 II.
It may be less total NR compared to what was being done for low ISOs on the R5, but there is certainly more of it.
Personally, I would hope Canon weren't having to bake in NR to get the DR curves to match?

To each their own...but, as I understand it, Canon is simply looking at the readout of adjacent pixels and if one pixel is out of whack, they are calming it down a bit. Noise reduction, yes. But, is this really the type of NR that will improve DR? I am thinking this is more about tidying up the image a bit. Making the image cleaner, which is something I have noticed going to the R5 from 5DIII and even 5DIV. I see an increase in DR, an increase in latitude while editing, and yet, the images are also a bit cleaner to begin with. Overall, an increase in IQ that does not appear artificial to me in the slightest.

 

[David_B]

Richard, this is a very useful link, horshack is doing a great job collating measurements made by users

https://horshack-dpreview.github.io/RollingShutter/

One more thing against the Sony A7RV, its readout time is 0.1s - yes 100ms for those who think it is better than the R5 with its 16.5ms.

Yes, I posted that link in a thread elsewhere. I don't know how up to date it is beause it doesn't have the Sony A9-II and I don't know how they'd represent the A9-III (0ms). The Hassleblad X2D has a readout time of 0.16s (and is twice the price of the R5-II), so I'll wait for you to tell us all how inferior that is to the Canon R5-II. Hassleblad have a BSI sensor too. But readout speed, that's what matters, right? Nothing else?

It's amusing to watch people here clutch on to some random new measurement or spec where their kit is better than others as if that one number was the be all and end all for photogrpahy.

 

[David_B]

Actually stacked sensors have been around for over a decade now. Readout speeds of 2ms or less has been possible for some time. Now what can the human vision actually see?

A quick google tells me that processing delay of the human eye is about 80ms.

 

[David_B]

Yes but in terms of noise with video it's better? That's what the video is saying right? and the highlights are better with c log 2.. so I wonder what the DR is then if it's better in video or not.

You'll have to wait for someont to do a test! I've seen videos of video camera DR being measured and they use a completely different method.

 

[neuroanatomist]

Readout speeds of 2ms or less has been possible for some time. Now what can the human vision actually see?

There are many phenomena that take place on time scales much faster than human vision can perceive. The photochemical reaction that actually drives vision, the isomerization of rhodopsin, takes only ~200 femtoseconds (that's 0.0000000002 ms to put in the units relevant for sensor readout), decades ago I worked in the lab that timed that reaction using femtosecond laser pulses and resonance Raman spectroscopy. But the cellular signaling that follows is much slower, on the order of 60 ms for high acuity vision and 30 ms for low acuity vision, plus more time for those photoreceptor signals to be received and perceived by the brain.

However, human vision can see static subjects very well. An example of a static subject is the still image generated by a camera. That still image is affected by the readout speed of the sensor. In that context, why does it matter than the readout speeds of most sensors are much faster than human vision can perceive?

If you want a relevant bit of information for the speed of human vision, consider the flicker fusion rate. A computer monitor at 60 Hz is flashing at 16.7 ms intervals and looks static.

 

[SvO2]

To each their own...but, as I understand it, Canon is simply looking at the readout of adjacent pixels and if one pixel is out of whack, they are calming it down a bit. Noise reduction, yes. But, is this really the type of NR that will improve DR? I am thinking this is more about tidying up the image a bit. Making the image cleaner, which is something I have noticed going to the R5 from 5DIII and even 5DIV. I see an increase in DR, an increase in latitude while editing, and yet, the images are also a bit cleaner to begin with. Overall, an increase in IQ that does not appear artificial to me in the slightest.

The 5DIII/IV don't have the processing being referred to here. Whatever Canon is doing seems to occur on all the Digic X cameras. I'm not really convinced it is noise reduction, but I couldn't tell you what it is.

 

[docsmith]

The 5DIII/IV don't have the processing being referred to here. Whatever Canon is doing seems to occur on all the Digic X cameras. I'm not really convinced it is noise reduction, but I couldn't tell you what it is.

Sorry if it wasn't clear, but I was referring to the statement regarding the R5 and other recent cameras. The reference to the 5DIV/5DIII was how the R5 has a different noise quality when compared which may be the result of the NR people are referencing.

 

[AlanF]

Even the Sony A1, A7, etc sensors show the same pattern that Bill is citing as proof of NR on the R5II. Bill just decided that eyeballing the magnitude of the pattern is enough to decide whether or not to mark a sensor as having NR. And there hasn't been one person to definitively prove that the spectral pattern is noise reduction vs some other form of signal processing. Only the current Nikon bodies seem to definitively have no signs of any sort of signal processing.

You are right about the lack of proof. Bill can mimic the patterns by using an image blur on RAW, but that doesn't rule out other processes. Best not to worry about it.

 

[Baywoof]

I would have preferred a comparison of the the R5 II with the Sony A7RV, not the A1, in terms of dynamic range and image quality. That would be a fairer comparison. Also, as a R5 owner shooting stills, I see no convincing argument to upgrade to a R5 II with only incremental improvements and most them in the video realm. I would rather put the savings of not upgrading into lenses, which I think is Canon's intention. The alternative is moving to a system that is more photocentric, not videocentric.

 

[SvO2]

I would have preferred a comparison of the the R5 II with the Sony A7RV, not the A1, in terms of dynamic range and image quality. That would be a fairer comparison. Also, as a R5 owner shooting stills, I see no convincing argument to upgrade to a R5 II with only incremental improvements and most them in the video realm. I would rather put the savings of not upgrading into lenses, which I think is Canon's intention. The alternative is moving to a system that is more photocentric, not videocentric.

I mean I probably would plan to save my 4k if I already owned an r5, but I don’t see how better AF, higher frame rates, less rolling shutter, and pre capture don’t benefit stills shooters. If all you do is landscape, then I’d argue it’s time to move on to MF (though if canon manages to do something like DGO in stills in the future, that could change that calculus).

Regarding A1 vs A7RV, the dynamic range is not meaningfully different from the A1. In fact I would argue all of these modern full frames have dynamic ranges close enough to each other that it shouldn’t be the primary decision maker. And the a7r5 sensor is drastically slower (100ms instead of 6).

Now of course it is worth noting that the R5 ii is brand new and the Sony cameras are not. The next iterations of those will probably surpass Canon in many respects. Likewise with the z8/z9. And then eventually the r5iii.

 

[Maximilian]

…Edit: saw a few comments regarding 5DIV vs the R5/II sensor: as someone that moved from the 5DIV to the R6 (before the II) I loved the 5DIV IQ but despite being lower MP I actually thought the R6 images looked sharper (could possibly be less aggressive AA filter?) but that's subjective of course…

As a former user of a 5D4 and now of a R6m2 I can tell you, that I have the same „subjective“ experience of the R6m2 sensor delivering more detail.

 

[neuroanatomist]

I would have preferred a comparison of the the R5 II with the Sony A7RV, not the A1, in terms of dynamic range and image quality. That would be a fairer comparison. … The alternative is moving to a system that is more photocentric, not videocentric.

With the exception of the a9III that suffers from its global shutter, there are really no meaningful differences between any current high-end FF bodies in terms of DR and IQ. None. Anyone claiming the minute differences are meaningful isn’t stating a conclusion, they’re stating a delusion.

If you want a more photo centric system with better DR and IQ, you should look at medium format.

 

[neuroanatomist]

I mean I probably would plan to save my 4k if I already owned an r5, but I don’t see how better AF, higher frame rates, less rolling shutter, and pre capture don’t benefit stills shooters.

Some people are looking for an excuse not to buy a new camera.

 

[AlanF]

As a former user of a 5D4 and now of a R6m2 I can tell you, that I have the same „subjective“ experience of the R6m2 sensor delivering more detail.

Your "subjective" experience is borne out by quantitative evidence (courtesy of optyczne.pl). There are those, of course, whose subjectivity is counter to actual measurement.

 

[SvO2]

Your "subjective" experience is borne out by quantitative evidence (courtesy of optyczne.pl). There are those, of course, whose subjectivity is counter to actual measurement.

View attachment 219150View attachment 219151

Are the tests necessarily comparable? Based on their results the r3 would have considerably less resolving ability than both, which doesn’t seem right




And actually 1dxiii vs r6 is interesting since they share the same sensor minus the AA filter (1dxiii having a symmetric 16 pt filter and r6 having a unidirectional filter). R3 is believed to have a similar filter to the 1dxiii I believe.

If I had to guess the filter is causing the difference, but whether that is “real” resolution vs artifact, I’m not confident on.

1dxiii:


r6:


Edit: after viewing these on a computer to machine translate, they do note the AA filter is likely the difference. The claim does seem to be that it is a real resolving difference.

Any optical engineers here able to comment?

 

[AlanF]

Are the tests necessarily comparable? Based on their results the r3 would have considerably less resolving ability than both, which doesn’t seem right


View attachment 219153

And actually 1dxiii vs r6 is interesting since they share the same sensor minus the AA filter (1dxiii having a symmetric 16 pt filter and r6 having a unidirectional filter). R3 is believed to have a similar filter to the 1dxiii I believe.

If I had to guess the filter is causing the difference, but whether that is “real” resolution vs artifact, I’m not confident on.

1dxiii:
View attachment 219154

r6:
View attachment 219155

It's an AA-filter thing. Look at the MTFs at the Nyqist frequency of the R3 (top) and the R6ii (below). The AA-filter for the R6ii looks to be eliminated in the vertical direction and weakened in the horizontal whereas for the R3 it is present in both. It is the vertical component that puts the 6ii ahead in the average in the previous plots.

 

[SvO2]

It's an AA-filter thing. Look at the MTFs at the Nyqist frequency of the R3 (top) and the R6ii (below). The AA-filter for the R6ii looks to be eliminated in the vertical direction and weakened in the horizontal whereas for the R3 it is present in both. It is the vertical component that puts the 6ii ahead in the average in the previous plots.

View attachment 219156View attachment 219157

So does that equate to a true detail increase in real pictures or do aliasing artifacts cause the measured resolution to increase in these kind of tests? I guess I’m skeptical that Canon would be okay with losing ~30% resolving ability to a filter.

Also of note, it does appear the r5 ii went back to a symmetric filter instead of asymmetric, so that will probably measure worse than the original in this test.

 

[docsmith]

I mean I probably would plan to save my 4k if I already owned an r5, but I don’t see how better AF, higher frame rates, less rolling shutter, and pre capture don’t benefit stills shooters. If all you do is landscape, then I’d argue it’s time to move on to MF (though if canon manages to do something like DGO in stills in the future, that could change that calculus).

Yet I've seen comparisons of real-world MF images to FF and even crop sensor images and it is amazing how few differences there really are.

As for "better"....I have no doubt the R5 II is perhaps one of the best MILCs ever released. It, arguably, is the best general purpose MILC available today. It also addressed almost every small quibble I have the R5. Well done Canon.

But I am wondering how marginal "better" is in this instance. Right or wrong, I am thinking back to the 1DX to 1DX II evolution. As hyped as the 1DX II was at the time (20.2 MP!!! 14 fps continuous shooting!!!), in hindsight, it was pretty marginal. Even the memory cards did not survive to the next iteration.

As someone considering upgrading but also thinking of standing pat with the R5:

• Better AF you say? I took this image following the osprey with a red dot gun sight mounted to the hot shoe...not looking through the viewfinder or liveview. This is how good eye-detect AF is.

How much better is the AF going to get? I have dozens of these from about 90 minutes when the tide was right. And if/when I do want better AF, would the R5 II really be the camera I want? Or the R1? By the time you add a grip, it isn't that much more.

• Higher FPS. Sure, 30 is more than 20. And I do like the control over fps in ES (that is a quibble I have the the R5). But I almost always shoot EFCS at 8 to 12 fps when I want high frame rates. I like having 20 fps, but I rarely use it. The mechanical shutter specs are unchanged.
• Less rolling shutter. Taken at face value, this affects those using ES and video. Not that critical for me. However, the underlying driver of less rolling shutter, faster readout speed, might actually help with AF between frames. That would be a benefit. A marginal benefit, but again, if this is important, then the R1 has even faster readout speed.
• Pre-capture. Yeah...that is a nice feature.

While I agree that the DR difference between the R5 and R5 II are likely meaningless, the fact that they are lower, even meaninglessly lower, when improvements I see appear fairly marginal or irrelevant for me, I completely get why some people might wonder if glass is a better investment at this time:

I would rather put the savings of not upgrading into lenses, which I think is Canon's intention.

I am also eye-balling a few different lenses at the moment. If I do not get the R5 II or....R1.
I suspect Canon will be getting some money from me in one form or another.

 

[SvO2]

Yet I've seen comparisons of real-world MF images to FF and even crop sensor images and it is amazing how few differences there really are.

Which is all the more reason to not care about these marginal dynamic range differences. The DR difference between crop and FF and FF and MF is much larger than what we’re seeing here

As for "better"....I have no doubt the R5 II is perhaps one of the best MILCs ever released. It, arguably, is the best general purpose MILC available today. It also addressed almost every small quibble I have the R5. Well done Canon.

But I am wondering how marginal "better" is in this instance. Right or wrong, I am thinking back to the 1DX to 1DX II evolution. As hyped as the 1DX II was at the time (20.2 MP!!! 14 fps continuous shooting!!!), in hindsight, it was pretty marginal. Even the memory cards did not survive to the next iteration.

As someone considering upgrading but also thinking of standing pat with the R5:

• Better AF you say? I took this image following the osprey with a red dot gun sight mounted to the hot shoe...not looking through the viewfinder or liveview. This is how good eye-detect AF is.

View attachment 219158
How much better is the AF going to get? I have dozens of these from about 90 minutes when the tide was right. And if/when I do want better AF, would the R5 II really be the camera I want? Or the R1? By the time you add a grip, it isn't that much more.

• Higher FPS. Sure, 30 is more than 20. And I do like the control over fps in ES (that is a quibble I have the the R5). But I almost always shoot EFCS at 8 to 12 fps when I want high frame rates. I like having 20 fps, but I rarely use it. The mechanical shutter specs are unchanged.
• Less rolling shutter. Taken at face value, this affects those using ES and video. Not that critical for me. However, the underlying driver of less rolling shutter, faster readout speed, might actually help with AF between frames. That would be a benefit. A marginal benefit, but again, if this is important, then the R1 has even faster readout speed.
• Pre-capture. Yeah...that is a nice feature.

While I agree that the DR difference between the R5 and R5 II are likely meaningless, the fact that they are lower, even meaninglessly lower, when improvements I see appear fairly marginal or irrelevant for me, I completely get why some people might wonder if glass is a better investment at this time:


I am also eye-balling a few different lenses at the moment. If I do not get the R5 II or....R1.
I suspect Canon will be getting some money from me in one form or another.

I mean there’s probably a photographer out there who can get that same shot with manual focus with enough tries. And certainly folks with early dslrs have also gotten similar shots. So I don’t think it’s a matter of certain shots being impossible vs possible, but rather the probability of actually getting the shot. That should be improved with the new camera. Maybe now you’ll be able to get it with a slower lens and poorer light.

Yes the R1 has faster readout speed. And probably better AF (though I’m curious as to whether or not the greater pixel density overcomes some of downsides of not having cross type AF). But it’s not 45MP. Otherwise I would have preordered that instead. And sure, currently I mostly use EFCS with my r6ii. But that’s largely because I want to avoid the rolling shutter artifact. ES is usable in more scenarios with the new camera.

If your current camera does what you need it to - that’s great, you get to save some money. But that doesn’t mean there aren’t real improvements with the new one. And it’s not like people have to upgrade every generation. Perhaps waiting for the R5iii or a potential R5s or R1s would be the more appropriate course of action for you.

 

[AlanF]

So does that equate to a true detail increase in real pictures or do aliasing artifacts cause the measured resolution to increase in these kind of tests? I guess I’m skeptical that Canon would be okay with losing ~30% resolving ability to a filter.

Also of note, it does appear the r5 ii went back to a symmetric filter instead of asymmetric, so that will probably measure worse than the original in this test.

It depends on what you are photographing and the number of Mpx on your sensor - the smaller the pixels, the fewer the aliasing artefacts. I used the Nikon D850 for a year or two. Its 45 Mpx sensor without an AA-filter gave superbly sharp images and only very, very rarely aliasing artefacts. The R5 sensor isn't quite as sharp and I have never had any Moire problems with it. I did with the 5DIV on occasion. Canon tends to be conservative. I read somewhere but I can't vouch for it as I am not into video that Canon uses the AA-filter for the video side of things.