Initial testing shows Canon’s new 32.5mp APS-C sensor improves dynamic range over predecessor

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,318
7,055
Good news on the high ISO as the 80D was pretty average, same as the 7DII.
Yes, the 80D is pretty average: the same as the 7DII, Nikon D7200 and D500 (despite their having a smaller crop factor) and all the top-rated APS-Cs, because they are all just about the same at high iso! It does help to read the earlier posts in a thread - the 90D is unlikely to be significantly better than the 80D.
 
  • Like
Reactions: dtaylor

Architect1776

Defining the poetics of space through Architecture
Aug 18, 2017
597
579
118
Williamsport, PA
This discussion was started in the Download D90 thread. I'll copy my posting from there to move the discussion here.

I am very suspicious about the analysis on the FM site. The DR at high iso with efficient modern sensors, which are of very similar efficiency, is limited mainly by the photon flux and at any given high iso is nearly the same for all the leading makes and for their different models for the same size sensor. And FF are about an ev higher than APS-C. The FM analysis has nearly an ev increase of the 90D over the 80D, putting it close to that of the 5DIV and 1DXII and leaping ahead of Nikon and Sony APS-Cs by more than you thinks the physics would allow.
Perhaps canon is innovating and beginning to lead the way again.
Are you a sensor designer?
 

Durf

Picture Taker - Image Maker
In a similar vein, what I am looking for is a high density sensor so I can fit a 100-400mm to it and have similar reach to an 800mm on a 20 Mpx FF. A 32.5 Mpx crop or an 83 Mpx FF will do that, and that is as much as I can reasonably carry nowadays. I would like FF with crop facility so I can choose smaller RAW files rather than crop in post. But, at my age, you have to live in the present and photograph now rather than wait for an unknown future. The 90D specs fit the bill for me, not perfectly but good enough.
Personally I think in a year or so from now we'll see many 7D2 users shooting with the 90D, perhaps even a couple pro wildlife shooters.

I basically only shoot the 70-300mm L or the Sigma 150-600mm on my 80D, so if the 90D is "somewhat" better than the 80D (which I'm sure it is) I'll pick one up in the next year or so when I can catch it on sale. By next summer or so this 90D will be used by many and we'll know more about how it actually performs and the results that are possible with it. By the sounds of it thus far it actually may just be one of Canons best sellers in 2020.
 

Architect1776

Defining the poetics of space through Architecture
Aug 18, 2017
597
579
118
Williamsport, PA
I'm bored so I'll post an essentially irrelevant comment. I could care less about 24p. I could care less about slight differences in DR, or even more than slight differences. I'd rather have an AA filter than not. I have two Canon DSLRs (AA filters, of course) and one Fujifilm camera (no AA filter). I really don't see any advantage to no AA filter. Any slight differences in sharpness are irrelevant as far as I'm concerned. I might even say that pictures can be too sharp. Buy what you want and argue about the technicalities if you want.
30p is actually shown to be superior to 24p anyway. 24p was just a way to use less film when the film makers would have preferred 30p. Sort of a historical item ignored by the 24p snobs.
 

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,318
7,055
Perhaps canon is innovating and beginning to lead the way again.
Are you a sensor designer?
As far as the sensor is concerned, its S/N at high iso and the DR are limited by the shot noise of the photon flux and the quantum efficiency of the CMOS sensor (qe), and the electronic noise from the circuits is relatively low at high iso. According to a Poisson distribution for the photon flux, the S/N for a flux of n photons will be sqrt(n). The signal is given by n*qe, and the qe of modern CMOS sensors is already very good see, for example, https://www.vision-systems.com/home...mizing-manufacturing-for-a-sensitivity-payoff
I do not know what Canon is doing with the electronic circuits and amplifiers.
 
Last edited:

melgross

EOS RP
Nov 2, 2016
621
410
I'm bored so I'll post an essentially irrelevant comment. I could care less about 24p. I could care less about slight differences in DR, or even more than slight differences. I'd rather have an AA filter than not. I have two Canon DSLRs (AA filters, of course) and one Fujifilm camera (no AA filter). I really don't see any advantage to no AA filter. Any slight differences in sharpness are irrelevant as far as I'm concerned. I might even say that pictures can be too sharp. Buy what you want and argue about the technicalities if you want.
I agree. 24p is a lot of crap. It has very little to do with movie-like quality. That’s mostly due to color, dynamic range and shooting style. It’s amazing that even writers for the photo sites know so little about it that they automatically say how bad it is when 24p isn’t available. I’d bet they never use it when it is.

in addition, 24 was used for simple reasons, none of them having to do with movie-like quality, which no one thought about back then. It was all about how much film a camera could hold in the magazine, and how much shooting time it could give. That had to do with size, clumsiness and cost, as film and processing was very expensive. They shot at the slowest speed that would give, upon playback, with frame doubling, a smooth look.

thats it! It’s of little value today. In fact, it’s worse, requiring software techniques to allow it to work with most playback equipment.

as far as AA goes, the higher the resolution, the weaker the filter can be. Leica didn’t use one with their 24mp cameras, and moiré was abundant wherever there was fine detail.

medium format doesn’t use the filter either, and for years, many fashion photographers give up on medium format because of fatal moiré from clothing. I used to beta test backs for Leaf, with a colleague, and it was a major problem until resolution was high enough, and even then, you could find it if you looked.
 

mb66energy

EOS R
Dec 18, 2011
1,452
328
Germany
www.MichaelBockhorst.de
This discussion was started in the Download D90 thread. I'll copy my posting from there to move the discussion here.

I am very suspicious about the analysis on the FM site. The DR at high iso with efficient modern sensors, which are of very similar efficiency, is limited mainly by the photon flux and at any given high iso is nearly the same for all the leading makes and for their different models for the same size sensor. And FF are about an ev higher than APS-C. The FM analysis has nearly an ev increase of the 90D over the 80D, putting it close to that of the 5DIV and 1DXII and leaping ahead of Nikon and Sony APS-Cs by more than you thinks the physics would allow.
I am physicist but no sensor specialist: If you reduce the amount of light by a factor of 4 you increase the statistical noise (determined by number of photons) by a factor of sqrt(4) = 2 so increasing ISO by two full steps (e.g. 200 to 800) you will have reduced DR by a factor of two. The D90 curve are close to that theoretical limit, the D80 is worse.
Maybe it is a matter of cleaner read out / better ADCs (on Sensor?) / better amplifiers ...
 

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,318
7,055
I am physicist but no sensor specialist: If you reduce the amount of light by a factor of 4 you increase the statistical noise (determined by number of photons) by a factor of sqrt(4) = 2 so increasing ISO by two full steps (e.g. 200 to 800) you will have reduced DR by a factor of two. The D90 curve are close to that theoretical limit, the D80 is worse.
Maybe it is a matter of cleaner read out / better ADCs (on Sensor?) / better amplifiers ...
I did show the formula for S/N just two posts before yours and mentioned these points.
 

Architect1776

Defining the poetics of space through Architecture
Aug 18, 2017
597
579
118
Williamsport, PA
As far as the sensor is concerned, its S/N at high iso and the DR are limited by the shot noise of the photon flux and the quantum efficiency of the CMOS sensor (qe), and the electronic noise from the circuits is relatively low at high iso. According to a Poisson distribution for the photon flux, the S/N for a flux of n photons will be sqrt(n). The signal is given by n*qe, and the qe of modern CMOS sensors is already very good see, for example, https://www.vision-systems.com/home...mizing-manufacturing-for-a-sensitivity-payoff
I do not know what Canon is doing with the electronic circuits and amplifiers.
Again,Are you some sort of sensor designer?
Or are you just Googling this stuff.
Most of us do not care except that it works and apparently defies physics which is wonderful.
Canon leading the way into territory others still think impossible.
 

Timedog

EOS R
Aug 31, 2018
58
41
I've looked at some RAW files and am pretty impressed but I can't help think that the images would be just a little better if Canon had decided to drop the AA filter. Canon's attitude towards these filters seems to be like the Foreman in the first chapter of the Hitchikers Guide to the Galaxy who says "It's a free way. You have to BUILD free ways."
If you want fake detail (aliasing), why not just add sharpening in post?
 

Architect1776

Defining the poetics of space through Architecture
Aug 18, 2017
597
579
118
Williamsport, PA
80d came out 3 years ago. guess no one here thinks about the future. canon puts out new cameras every 3 -4 years. will it hold up to the needs of users and competition in 3-4 years in the 2020s.
it is NOT needs by any means. Most all users can not utilize what is in their cameras today. In fact they do not even know what the camera can do without reading the instructions, over and over. Barring some major tech advancement the minuscule incremental changes year to year are virtually negligible. Marketing hype is what collectors fall prey to and as we know Sony leads the way as they have to change their products, TVs et all on a 6-12 month cycle and they seem to do the same philosophy with cameras.
 

Quirkz

EOS RP
CR Pro
Oct 30, 2014
272
195
Well.. Canon raws are always compressed, the craw is a lossy compression. And depending the way the measurement is done lossy compressed file could actually seem to have more DR if the compression acts like a crude noise reduction. This obviously doesn't mean lossy is better, it just happens to score higher in a single metric when nothing else is taken into account.
This is only sometimes true :). Canon raws are compressed with a lossless compression by default. Recent cameras introduced v3 (cr3) flies which can support a lossy compression for somewhat smaller file sizes, but the user needs to explicitly enable this. (And canon seems to have strongly favored quality over compression even here, with generally imperceptible difference to the human eye.)

For your reference:

 

AlanF

Hands. Face. Space.
CR Pro
Aug 16, 2012
7,318
7,055
Again,Are you some sort of sensor designer?
Or are you just Googling this stuff.
Most of us do not care except that it works and apparently defies physics which is wonderful.
Canon leading the way into territory others still think impossible.
I am an experimental scientist who analyses data such as these and I have designed rapid reaction spectrophotometers and spectrofluorimeters, in the distant past using PM tubes and subsequently digital sensors. The principles involved are simply basic physics and statistics that are routine to many scientists and engineers.
 

Joules

EOS R
CR Pro
Jul 16, 2017
1,214
1,357
Hamburg, Germany
Again,Are you some sort of sensor designer?
[...]
Most of us do not care except that it works and apparently defies physics which is wonderful.
Did you miss this part:

I've been talking to Bill Claff today about all this, and he's not agreeing with the results. He has a slightly different model that he's using for all this, but he's not seeing much difference between this new sensor and the 80D sensor.
It seems very likely that there is some error in the measurement process of the original post in the FM Forum.

If the results were defying physics that would be the case without a doubt. You don't have to be a sensor designer to understand how to measure and interpret the signals they deliver.

Edit: I spoke too soon, looks like there was no measurement mistake. Apologies to the person that posted the measurements on FM. However, it still doesn't look like like there's any reason to get hyped about the DR improvement:

 
Last edited:

Proscribo

EOS RP
Jan 21, 2015
262
135
This is only sometimes true :). Canon raws are compressed with a lossless compression by default. Recent cameras introduced v3 (cr3) flies which can support a lossy compression for somewhat smaller file sizes, but the user needs to explicitly enable this. (And canon seems to have strongly favored quality over compression even here, with generally imperceptible difference to the human eye.)

For your reference:

That's what I said, craw is the lossy option. ;)
 
  • Like
Reactions: Quirkz

Act444

EOS R
May 4, 2011
1,120
191
I'm bored so I'll post an essentially irrelevant comment. I could care less about 24p. I could care less about slight differences in DR, or even more than slight differences. I'd rather have an AA filter than not. I have two Canon DSLRs (AA filters, of course) and one Fujifilm camera (no AA filter). I really don't see any advantage to no AA filter. Any slight differences in sharpness are irrelevant as far as I'm concerned. I might even say that pictures can be too sharp. Buy what you want and argue about the technicalities if you want.
For me, it depends on the application, but I find that generally, it seems to get in my way more often than it doesn’t. I’m basing this on the average amount of sharpening I find myself applying on average in post. This is more of an issue with 5D4 images - the AA filter on the previous 5D3 was much weaker and I found that one to provide a good balance.
 

Dragon

EF 800L
May 29, 2019
309
252
That argument will have actual relevance once 16-bit readouts become commonly available in APS-C and FF sensors. I’m not sure we’ll be there in 3-4 years, but maybe.
We may see 16 bit readouts, but fast 16 bit readouts are not likely anytime soon Every bit added basically doubles the settling time of a slope A/D. This is the reason why the very fast modes of the Sony sensors are only 12 bit. There are other A/D approaches that can get to 16 bits fast, but they don't easily fit on an imaging chip (and they are much more likely to add digital noise).
 
  • Like
Reactions: Sharlin