Canon Interview: EOS R1 is the true flagship

[neuroanatomist]

Who understands, understands. Who doesn't but can and wants to, will.

Who believes they understand but doesn’t, probably never will.

 

[Omn]

As @neuroanatomist has pointed out you are confusing looking at individual pixels rather than the actual image.

No I'm talking about whole image performance.
Affected by pixel level performance.

Digital image is comprised of pixels. Which carry some properties. So those properties affect the image.

A simple illustration, for example, is that a square of 4 pixels each of 3x3 microns^2 has the same area as 1 pixel of 6x6 microns^2 and the 4 small combined have the same well depth as the single larger so images from both viewed at the same size are looking at he same well depth despite one sensor having 4x the number of pixels.

No it doesn't work that way.

1) No they don't have the same well depth.
You are missing a distinction between visible pixel image info and spatial sampling and electronic-level light intensity capture ability before there is any pixel data, manifesting in the amount of charge a pixel can hold and you are playing math the wrong way.

2) Single pixel well capacity defines quantised signal intensity range (upper part), you don't get to capture brighter values by joining 4 pixels.

3) In shadows those smaller pixels have poorer sensitivity threshold and saturation performance and having more of them won't fully help in compensating what is lacking.

Viewing size is crucial to comparing the performance of cameras.

With full size you can see actual superiority of larger pixels and with downscaling trickery (defeating the purpose of higher pixel count acquisition) you can convince folks of all kinds of nonsense.

If you viewed the FF image at 1.6x1.6 the size of the APS-C image, then both images would appear to have the same DR and noise per unit area.

If you viewed from Mars phone cameras from 2005 are like 1DXIII.
And potatoes are like rocks.

Therefore > it doesn't matter if image is recorded by quantising signal captured by larger or smaller pickup and processing tech and R1 should have actually had 40 Mp and Canon engineers are clueless.

 

[Omn]

Who believes they understand but doesn’t, probably never will.

Depends on how deep is one stuck with his beliefs.

 

[Omn]

Please provide these measurements, I would be interested in seeing them for myself as I am interested in the data to support your statements. (this is not any sort of dig at you, I am genuinely interested in seeing your measurements and the environment you collected them in)

I'd be glad to but cannot at this point and form due to related project I'm working on. Perhaps with time.

I can summarise the findings though.

With all camera makers and in stills and cine cameras it's always the same and with same generation devices larger pixel cameras have superior properties to smaller pixel cameras which manifests in superior DR, lattitude, tonal stability, tonal separation, subtle tone differentiation etc. In some cases older generation larger pixel cameras have superior properties then newer generation smaller pixel cameras.

One property finding: sensors with larger pixels have more consistent saturation response throughout exposure range. The smaller the pixel the worse it gets.

I'm assuming one of the reasons for increased chroma noise in smaller pixel/higher pixel count cameras (you can see it in dpreview raw comparison links I posted) is because camera makers try to compensate sensor saturation drop in low levels with sensor response specific camera colour profile which boosts the low level values and consequently enhances chroma noise as well.

Camera makers generally improve the performance through generations but these principles related to pixel pitch difference remain.

And 6 microns look like the best balance.

Therefore > 24Mp for 36mm for a flagship

 

[neuroanatomist]

I'd be glad to but cannot at this point and form due to related project I'm working on. Perhaps with time.

In this corner, empirical data from a trusted, knowledgeable source showing no practical difference in DR between the 45 MP sensor of the R5 and the 24 MP sensor of the R3. Granted, that’s with a limited number of parameters measured…he doesn’t quantify things like ‘subtle tone differentiation’ whatever you mean by that).

And in this corner, “I have better measurements,” but I “cannot at this point” show them to you. So just trust me, I’m right.

It’s like watching the boxing gold medal winner in the ring with a toddler.

 

[josephandrews222]

I'd be glad to but cannot at this point and form due to related project I'm working on. Perhaps with time.

....

And 6 microns look like the best balance.

Therefore > 24Mp for 36mm for a flagship

Fascinating.

Sort of.

An American president from the 1980s got a lot of mileage out of the following phrase: "Trust, but verify."

I think those three words, the words of 'The Great Communicator' (or maybe the words of his speechwriter ), apply in this case..

I await the data (real data) and thoughtful discussion that together will supply genuine verification of the claims made here.

Until that time...I think I'll continue to check my emails for the R5MkII (hopefully) headed my way.

 

[Omn]

In this corner, empirical data from a trusted, knowledgeable source showing no practical difference in DR between the 45 MP sensor of the R5 and the 24 MP sensor of the R3. Granted, that’s with a limited number of parameters measured…

Priceless.

Show me this "empirical DR data".

And in this corner, “I have better measurements,” but I “cannot at this point” show them to you. So just trust me, I’m right.

No one here cares about tackling your trust issues.

You could have clicked on dpreview hig iso raw examples and confirm a part of what I've said if you really wanted to but you didn't want to so see above about beliefs.

 

[Omn]

I await the data (real data) and thoughtful discussion that together will supply genuine verification of the claims made here.

Until that time...I think I'll continue to check my emails for the R5MkII (hopefully) headed my way.

Anyone else also biased to purchase decision and expecting being convinced in truthfullness, save your time and ignore my posts from now on.

 

[nickditoro]

Ohhhhh don’t sell the 5dsr. Man. That to me is one of canons best cameras. I have two and I love them. Might not have the best iso performance but every time I use that camera I am blown away.

I hear you about the 5DSR, which I use only for architecture with a tilt-shift lens. Since settings are controlled, I don't worry about ISO. It's just that unless I trade in my Canon and Fuji bodies, I won't be able to afford the SL3, which, at 60MP, should be able to handle architecture along with whatever else I throw at it except for sports. And even there, I don't shoot professionally and generally spray and pray with my 1DX3.

 

[nickditoro]

LOL. Is it demosaiced? Is it downsampled? Then it’s not RAW. I can export a JPG file as a DNG, but that doesn’t reconstitute the raw image data.

Leica haters can can throw crap terms around to try to prove a point, but Leica claims that what it calls its "Triple Resolution Technology" produces raw image files in DNG format and JPEGs that can be recorded at 60, 36 or 18 megapixels, always using the full sensor area.

 

[vikingar]

Omn does have a point, even if badly argued without concrete data.

(EDIT: any statisticians, please let me know if the probability distribution of a large pixel is actually really the same as the probability distribution of the sum / mean of 4 smaller pixels. Theoretically I mean, aside from noise and all technical difficulties.)

The R5 does seems to strike a balance between quality and pixel count, and in most cases I would agree that you cannot see the difference. But I would also not trust any empirical measurements blindly that boil quality down to a single number without verifying them as well. Notwithstanding mistakes, it's very easy for a number (or multiple numbers) to turn out the same even if the pictures look different. Anyone who's implemented image processing algorithms knows that there are a zillion pitfalls and difficulties mapping math to what people actually perceive.

I am going to post a DPReview comparison of R5, R6, R3, 1DxIII at 102400 ISO. Yes I am aware that no one shoots at that ISO and that it's an edge case at the extremes, but I couldn't find any pushed RAWs for online comparison. You can still see a difference at 51200, but below that it's harder.

I am also aware that many parts of the image look fine on the R5, and that the other cameras don't have exactly the same sensor tech. But if you look at the linked area yourselves you will see that certain darker or colored patches look distinctly worse on the R5.

Take a look: https://www.dpreview.com/reviews/im...=1&x=-0.1645782001043483&y=0.6793647520894319

 

[neuroanatomist]

Show me this "empirical DR data".

It was posted already. Were you unable to see it, or did you just not comprehend it?

Unlike you, Bill Claff publishes his measurements and his methods of obtaining them.

No one here cares about tackling your trust issues.

Trust is earned. Making statements that are counter to established information and not backing those claims up with data does not earn trust.

You could have clicked on dpreview hig iso raw examples and confirm a part of what I've said if you really wanted to but you didn't want to so see above about beliefs.

I downloaded the low ISO RAWs for the R5II and R5, the exposures from their studio comparison tool are substantially different. It’s cute that you are suggesting use of an unreliable tool to support the conclusions you’ve drawn from ‘data not shown’.

Unless you are capable of replying with something informative that can be substantiated, you need to bother replying at all.

 

[Anon]

Maybe to add to the "Whats next" question:

Here is a survey from a german tech website, asking people about their habits with cameras / smartphones. A few interesting insights.

Q: When buying a new camera it will probably be:
1. a new smartphone
2. a small compact camera
3. a proper, but fixed lens camera
4. a ILC




Q: Do you have a dedicated camera? what kind is it?
1. digital camera with integrated lens
2. fixed focal length digital camera
3. DSLR
4. MILC
5. Analog camera
6. Instant camera
7. something else
8. No, but I would like to have one
9. No, and I dont want one
10. No, but I had one in the past




Q: What sensor size does your camera have?
1. Smaller than 1 inch
2. 1 inch
3. MFT
4. APS-C
5. Full Frame
6. Medium Format
7. Does not have a sensor aka analog camera




Q: What brand is your camera?




Q: How many lenses do you own?



Source: https://www.computerbase.de/2024-08/fotografie-wie-nehmt-ihr-fotos-auf-mit-smartphone-oder-dslm/

I hope this is interesting to some of you It might be only a small sample within a "tech bubble" community, but nice to see nonetheless. Cheers

 

[neuroanatomist]

I am going to post a DPReview comparison of R5, R6, R3, 1DxIII at 102400 ISO. Yes I am aware that no one shoots at that ISO and that it's an edge case at the extremes, but I couldn't find any pushed RAWs for online comparison. You can still see a difference at 51200, but below that it's harder.

ISO 102400 is already a pushed RAW for the R5. The other three cameras have a top native ISO of 102400, the R5 tops out at 51200 so ISO 102400 is a 1-stop in-camera push.

 

[Omn]

It was posted already. Were you unable to see it, or did you just not comprehend it?

Priceless.

Let's try to help with the ignorance and hopefully arrogance reduces as well.

1) Those are not camera DR data charts.

2) Best stills cameras are 13+ stops of captured dynamic range, which is not visible with a simple test and with a display rendering OETF. This is why you can "bring back" highlights in RAW which are there but you don't see them because of how the image data is rendered for display.

3) When you throw this DR into a testing methodology aiming at some concept of "photographic range" you are:
a) not evaluating full dynamic range of the camera
b) reprocessing imagery altering captured data and altering the basis of evaluation
c) stuffing larger range into smaller range

So...the results will be limited by the method and context of evaluation. And the actual differences between full DR will not be visible if a larger range is evaluated within a smaller range context, will they ?

Also, chart shows about 7 stops of DR at ISO 3200. Anyone thinking their camera lost almost half od dynamic range on ISO 3200 setting and captured less DR then 20 y.o. 1/3" sensor consumer camera ?

Hello logic.

I downloaded the low ISO RAWs for the R5II and R5,

Last attempt.

Larger pixel to smaller pixel camera comparison.

No need to download anything. Open that 2x2 RAW comparison.

- 1DXIII on upper left
- Nikon ZF on the lower left
- Any new ultraMp FF on the right
- Put ISO 6400, move about and compare

This will show just one part of what I'm talking about. There is much more.

Like highlight performance. You can see this...in neither of those tests. Where are the highlights ? Oh well...they are somewhere. Where are the overexposure tests ? Oh well...we don't do these. What happens with bright saturated values and rolloff to clip ? Oh well, that doesn't matter.
Must be coincidence.

Option #2
You can just get R3 and R5 and shoot side by side high contrast scenes, do over and under exposures, count the stops and compare highlight and shadow performance.

High contrast and real dark instead of evenly lit paper and objects in that dpreview test are not the same and will reveal more.

No need for someone else to convince you. If you don't see it then, no babble online is going to help. If you do see it, beliefs from megapixel race propaganda crumble.

 

[neuroanatomist]

Priceless.

Yes, a priceless failure in your understanding. Your logic applies to a single pixel, it does not apply to a sensor full of them. If it did, the 5DsR and 7DII would have similar noise and DR in their resulting images, because they are equivalent sensor generations and the pixels are essentially the same size.

It's ok, you can keep on measurebating your single pixel until it falls off. Have a nice day.

 

[Omn]

Yes, a priceless failure in your understanding.

Wishful thinking may feel nice but won't compenstate for the lack of understanding.

Try starting with learning the difference between camera dynamic range and whatever is plotted on those charts.

Your logic applies to a single pixel, it does not apply to a sensor full of them.

Zero sense.

There is enough of useful stuff in previous posts, I'm not wasting time on fantasy land strawman nonsense.

 

[vikingar]

Also, chart shows about 7 stops of DR at ISO 3200.

That is because Photographic Dynamic Range (PDR) uses a different definition, more specifically a SNR (signal to noise ratio) of 20 as the lower acceptable bound instead of 1 (also an arbitrary limit BTW):

Sensor Analysis Primer – Engineering and Photographic Dynamic Range

www.photonstophotos.net

My definition of Photographic Dynamic Range is a low endpoint with an SNR of 20 when adjusted for the appropriate Circle Of Confusion (COC) for the sensor.

So any PDR values will be lower than DR and are not comparable. But both are sort of arbitrary definitions.

With PDR there's also an adjustment for sensor size / circle of confusion (COC), which makes the value useful for comparisons.

From: https://www.photonstophotos.net/Gen...Primer/Photographic_Dynamic_Range_Summary.htm

Recall that this value is normalized for the Circle Of Confusion (COC) that is appropriate for the sensor size.

PDR is the dynamic range you would expect in an 8x10" print viewed at a distance of about arms length.

As I understand this is essentially a slight "blur" applied before the SNR calculation designed to simulate how the image would look when viewed from a set distance, i.e. it averages out small noise that you wouldn't notice and allows comparing FF and APS-C cameras.

It is also mentioned that the values for the R, G, B channels are calculated separately and then averaged.

I'm not sure I fully understand everything, but I think this essentially tries to do a "print-out" comparison with some threshold for acceptable noise, which makes sense. It does not seem to fully account for the qualititative color differences between images or the types of noise. Also the ISO values are nominal only, so exposures may actually be different between cameras I think. And due to the COC adjustment I think the results might only be valid up to a certain magnification.

 

[dougsv]

Thanks for the heads up - they appeared today. The points for the R5ii are all downward pointing triangles, which means that noise reduction is being applied to raw throughout the whole iso range.

FYI at least in ISO 100 the NR applied to the R5ii is very very weak (source: Bill himself in DPReview forums). According to him, the closer to a horizontal line the energy spectrum is, the less NR it has applied.

This is for the R5ii:

This is for the R5:



And the Z9 (a camera who has a circle at base ISO):

 

[neuroanatomist]

...I'm not wasting time on fantasy land strawman nonsense.

That's all you've been doing here so far. I'm glad you're going to stop.