Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Sensor dynamic range and native ISO
#1
Just came across this article 
https://petapixel.com/2020/06/24/this-is...n-dxomark/

1Dx markiii scored badly in DXO mark because it has higher native ISO and measurements are made at native ISO, but at equivalent ISO speeds it does pretty well.   A sports photographer  won't mind this if it gives him any advantage at high ISO setting.
 
Here's what DXO themselves have to say about it 

https://www.dxomark.com/canon-eos-1d-x-m...or-review/

So again the dynamic range values are measured at native ISO (which is basically the ISO speed used on tripod for landscapes and where we need the most of dynamic range)
#2
DXO is writing nonsense again, what else is new?


You seem to misunderstand what DXO is saying. The native ISO setting on both cameras is the same: ISO 100. That is the same for both cameras. DXO claims that they "measured" ISO 63 at ISO 100 for the mkII and that they "measured" ISO 136 at ISO 100 for the mkIII.
They claim that they "measure" a different ISO than what the manufacturer does. This means that they do not understand ISO very well, and they should reread how ISO settings are defined.
And they mess up things even more. The mkIII has an ISO range that starts at ISO 100. It expands down to ISO 50. Yet, somehow, DXO labels manufacturer ISO 100 as ISO 50 and manufacturer ISO 200 as ISO 100, and so on. In their graphs, they put manufacturer ISO 50 (according to them) next to ISO 200 from the mk II and for instance the Nikon D5. 
Things are not right in their measurements and I do not know how exactly they got them messed up, but they are not correct. And as always, they do not really look at what is wrong, but come with some nonsense explanation.

If you are interested in RAW DR of the Canon EOS 1D X mkIII compared to the Canon EOS 1D X mkII, Nikon D6 and Sony A9 II:
https://www.photonstophotos.net/Charts/P...20ILCE-9M2
#3
(06-25-2020, 06:31 AM)toni-a Wrote: Just came across this article 
https://petapixel.com/2020/06/24/this-is...n-dxomark/

1Dx markiii scored badly in DXO mark because it has higher native ISO and measurements are made at native ISO, but at equivalent ISO speeds it does pretty well.   A sports photographer  won't mind this if it gives him any advantage at high ISO setting.
 
Here's what DXO themselves have to say about it 

https://www.dxomark.com/canon-eos-1d-x-m...or-review/

So again the dynamic range values are measured at native ISO (which is basically the ISO speed used on tripod for landscapes and where we need the most of dynamic range)

Hi Toni,
I have dig into the sensor technology in the past. There is something fundament that you have to understand. The sensor design begins with definition of sensor parameters- mostly based on use cases. Just to name few of them
-         Sensor size
-         Number of pixels.
-         Light sensitivity – so called base ISO value
-         Dynamic range
-         Color reproduction
 
As anything in the nature parameters play agains ach other. I have electrical engineer background and  understand it. This fundamental apply in many science area.  I’ll try to use biological example – close to your background.  You cannot have animal that fly as an bird, swim as an fish and run as cheetah. There are such all rounder animals like ducks, they do everything but nothing is best.
Same principe applies in sensor design too. In general sensor with higher DR have lower ISO performance. Canon designers made a great job with balancing those parameters. What need the average 1Dx shooter?
-         Higher frame rate
-         Fast AF
-         Excellent low light performance even in it scarify other IQ parameters – e.g. Megappixel count, DR, color reproduction
DXo try to use quotative numbers to represent The problem is that the average person doesn’t  know the meaning of this numbers.
 
 
#4
(06-25-2020, 08:22 AM)miro Wrote:
(06-25-2020, 06:31 AM)toni-a Wrote: Just came across this article 
https://petapixel.com/2020/06/24/this-is...n-dxomark/

1Dx markiii scored badly in DXO mark because it has higher native ISO and measurements are made at native ISO, but at equivalent ISO speeds it does pretty well.   A sports photographer  won't mind this if it gives him any advantage at high ISO setting.
 
Here's what DXO themselves have to say about it 

https://www.dxomark.com/canon-eos-1d-x-m...or-review/

So again the dynamic range values are measured at native ISO (which is basically the ISO speed used on tripod for landscapes and where we need the most of dynamic range)

Hi Toni,
I have dig into the sensor technology in the past. There is something fundament that you have to understand. The sensor design begins with definition of sensor parameters- mostly based on use cases. Just to name few of them
-         Sensor size
-         Number of pixels.
-         Light sensitivity – so called base ISO value
-         Dynamic range
-         Color reproduction
 
As anything in the nature parameters play agains ach other. I have electrical engineer background and  understand it. This fundamental apply in many science area.  I’ll try to use biological example – close to your background.  You cannot have animal that fly as an bird, swim as an fish and run as cheetah. There are such all rounder animals like ducks, they do everything but nothing is best.
Same principe applies in sensor design too. In general sensor with higher DR have lower ISO performance. Canon designers made a great job with balancing those parameters. What need the average 1Dx shooter?
-         Higher frame rate
-         Fast AF
-         Excellent low light performance even in it scarify other IQ parameters – e.g. Megappixel count, DR, color reproduction
DXo try to use quotative numbers to represent The problem is that the average person doesn’t  know the meaning of this numbers.
 
 
Hi Miro,

I'd like to make an addition here:
there also is a choice in dealing with noise in in-camera-processed RAWs. We do know that Nikon and Sony filter out noise in some clever way before even storing the RAW files, thus, in a way, reducing noise artificially in the RAW files these cameras create.

Canon has chosen not to do this.

The net result of this is that Sony and Nikon have cleaner RAW files, and based on DxO criteria then have higher DR numbers. However, they also have less real detail towards the lower end of the DR range, f.e., the disappearance of stars in night sky shots, whereas Canon files retain more real detail, even though the noise levels are higher. DR numbers unfortunately are based on noise levels ....

BTW, it is possible to process Canon RAW files in such a way they appear to have similar DR levels as Sony and Nikon RAW files. There are a few (engineering-like) articles on the subject on the internet, although I do not have any links to those at the moment. I did read a few over the years.

Having said that, A DR of over 10 for me is not restrictive at all. I developed a Zone System for use with film many years ago, wiyh B&W in ideal circumstances approximately having a DR of (8 to) 10, and still adhere to the same principles for digital, be it adjusted for digital. That works fine for me, and that means that a DR of 8 to 10 stops is plenty for me. If I need more, I just create an HDR file composed of several shots with different exposures.  

Kind regards, Wim
Gear: Canon EOS R with 3 primes and 2 zooms, 4 EF-R adapters, Canon EOS 5 (analog), 9 Canon EF primes, a lone Canon EF zoom, 2 extenders, 2 converters, tubes; Olympus OM-D 1 Mk II & Pen F with 12 primes, 6 zooms, and 3 Metabones EF-MFT adapters ....
#5
(06-25-2020, 09:22 AM)wim Wrote:
(06-25-2020, 08:22 AM)miro Wrote:
(06-25-2020, 06:31 AM)toni-a Wrote: Just came across this article 
https://petapixel.com/2020/06/24/this-is...n-dxomark/

1Dx markiii scored badly in DXO mark because it has higher native ISO and measurements are made at native ISO, but at equivalent ISO speeds it does pretty well.   A sports photographer  won't mind this if it gives him any advantage at high ISO setting.
 
Here's what DXO themselves have to say about it 

https://www.dxomark.com/canon-eos-1d-x-m...or-review/

So again the dynamic range values are measured at native ISO (which is basically the ISO speed used on tripod for landscapes and where we need the most of dynamic range)

Hi Toni,
I have dig into the sensor technology in the past. There is something fundament that you have to understand. The sensor design begins with definition of sensor parameters- mostly based on use cases. Just to name few of them
-         Sensor size
-         Number of pixels.
-         Light sensitivity – so called base ISO value
-         Dynamic range
-         Color reproduction
 
As anything in the nature parameters play agains ach other. I have electrical engineer background and  understand it. This fundamental apply in many science area.  I’ll try to use biological example – close to your background.  You cannot have animal that fly as an bird, swim as an fish and run as cheetah. There are such all rounder animals like ducks, they do everything but nothing is best.
Same principe applies in sensor design too. In general sensor with higher DR have lower ISO performance. Canon designers made a great job with balancing those parameters. What need the average 1Dx shooter?
-         Higher frame rate
-         Fast AF
-         Excellent low light performance even in it scarify other IQ parameters – e.g. Megappixel count, DR, color reproduction
DXo try to use quotative numbers to represent The problem is that the average person doesn’t  know the meaning of this numbers.
 
 
Hi Miro,

I'd like to make an addition here:
there also is a choice in dealing with noise in in-camera-processed RAWs. We do know that Nikon and Sony filter out noise in some clever way before even storing the RAW files, thus, in a way, reducing noise artificially in the RAW files these cameras create.

Canon has chosen not to do this.

The net result of this is that Sony and Nikon have cleaner RAW files, and based on DxO criteria then have higher DR numbers. However, they also have less real detail towards the lower end of the DR range, f.e., the disappearance of stars in night sky shots, whereas Canon files retain more real detail, even though the noise levels are higher. DR numbers unfortunately are based on noise levels ....

BTW, it is possible to process Canon RAW files in such a way they appear to have similar DR levels as Sony and Nikon RAW files. There are a few (engineering-like) articles on the subject on the internet, although I do not have any links to those at the moment. I did read a few over the years.

Having said that, A DR of over 10 for me is not restrictive at all. I developed a Zone System for use with film many years ago, wiyh B&W in ideal circumstances approximately having a DR of (8 to) 10, and still adhere to the same principles for digital, be it adjusted for digital. That works fine for me, and that means that a DR of 8 to 10 stops is plenty for me. If I need more, I just create an HDR file composed of several shots with different exposures.  

Kind regards, Wim

I have seen you write that before, but it remains (mostly) incorrect.
The main reason Sony has higher DR in RAW is not because they filter noise, it is because of how/where the signals get amplified. They simply have lower read out noise.

Canon has caught up with that somewhat (see 5D mk IV, EOS R for instance). And the Canon 1D X mk III does better (DR wise at base ISO) than the Nikon D6 and Sony A9 II.

The high DR is in low ISO, not in high ISO with NR.

That with some cameras/some camera settings noise gets filtered out at low light/long exposure/high ISO settings has nothing at all to do with high DR in low ISO setting RAWS.
#6
(06-25-2020, 09:31 AM)Brightcolours Wrote: I have seen you write that before, but it remains (mostly) incorrect.
The main reason Sony has higher DR in RAW is not because they filter noise, it is because of how/where the signals get amplified. They simply have lower read out noise.

Canon has caught up with that somewhat (see 5D mk IV, EOS R for instance). And the Canon 1D X mk III does better (DR wise at base ISO) than the Nikon D6 and Sony A9 II.

The high DR is in low ISO, not in high ISO with NR.

That with some cameras/some camera settings noise gets filtered out at low light/long exposure/high ISO settings has nothing at all to do with high DR in low ISO setting RAWS.
"(mostly) incorrect" is debatable obviously.

Yes, Sony and Nikon read out is structured differently at the hardware level, and therefore generates less readout noise, I am aware of that. However, the processing in hardware is different too, resulting in files that look cleaner and appear to have lower noise levels, but lose some detail caused mostly by the in-hardware algorithms used.

The articles I mentioned were scientific papers, in which the sensors and electronics were taken apart and measurements were done at different output levels of the hardware itself. As to the starry sky shots: that was just an example (and done properly, they should be done at base iso). Star shots should be done with NR off as well.

Regards, Wim
Gear: Canon EOS R with 3 primes and 2 zooms, 4 EF-R adapters, Canon EOS 5 (analog), 9 Canon EF primes, a lone Canon EF zoom, 2 extenders, 2 converters, tubes; Olympus OM-D 1 Mk II & Pen F with 12 primes, 6 zooms, and 3 Metabones EF-MFT adapters ....
#7
About noise and Canon sensors I totally agree with WIM, Canon files might have more noise but they have more details and that's easily recognizable. from ISO 6400 and beyound my A6000 photos look watery as if airbrush was applied, I prefer by far the results from 7D2 and 750D despite being more noisy, especially the noise reduction applied to RAW can't be removed.
#8
(06-25-2020, 10:49 AM)wim Wrote:
(06-25-2020, 09:31 AM)Brightcolours Wrote: I have seen you write that before, but it remains (mostly) incorrect.
The main reason Sony has higher DR in RAW is not because they filter noise, it is because of how/where the signals get amplified. They simply have lower read out noise.

Canon has caught up with that somewhat (see 5D mk IV, EOS R for instance). And the Canon 1D X mk III does better (DR wise at base ISO) than the Nikon D6 and Sony A9 II.

The high DR is in low ISO, not in high ISO with NR.

That with some cameras/some camera settings noise gets filtered out at low light/long exposure/high ISO settings has nothing at all to do with high DR in low ISO setting RAWS.
"(mostly) incorrect" is debatable obviously.

Yes, Sony and Nikon read out is structured differently at the hardware level, and therefore generates less readout noise, I am aware of that. However, the processing in hardware is different too, resulting in files that look cleaner and appear to have lower noise levels, but lose some detail caused mostly by the in-hardware algorithms used.

The articles I mentioned were scientific papers, in which the sensors and electronics were taken apart and measurements were done at different output levels of the hardware itself. As to the starry sky shots: that was just an example (and done properly, they should be done at base iso). Star shots should be done with NR off as well.

Regards, Wim

You still appear to be conflating high ISO noise reduction (seen mostly years ago in the D3 era, concerning those scientific papers you are referring to) with higher DR at low ISO (where no NR is being applied). 
The high ISO NR also depends/ed on exposure time and/or high ISO NR being enabled or not. Making blanket statements about Sony/Nikon based on some vague memories of something you read in the past does not make for the most accurate posts.

So, mostly incorrect. 

Again: just because some manufacturer applied some NR at some high ISO setting at certain conditions, does NOT mean that the high DR at low/base ISO settings is a function of NR. It is not.

Photonstophotos analises  if NR has been applied, and shows that in its charts with a triangle pointing down.
https://www.photonstophotos.net/Charts/PDR.htm

As you can see, the Pentax K1 II does a lot of "cheating" in today's field of cameras.

(06-25-2020, 10:57 AM)toni-a Wrote: About noise and Canon sensors I totally agree with WIM, Canon files might have more noise but they have more details and that's easily recognizable. from ISO 6400 and beyound my A6000 photos look watery as if airbrush was applied, I prefer by far the results from 7D2 and 750D despite being more noisy, especially the noise reduction applied to RAW can't be removed.

I'd look at what in your RAW processing causes the watery results, as it is not NR in RAW that does that.
Easily verifiable by looking at the dpreview sample image (RAW ISO 6400). If you compare it with the 7D mk II, you don't see watery results, just "sharper" more harsh noise. If you convert with software that does its own heavy NR you are not aware about, that would be an explanation of the "watery" results you see.
https://www.dpreview.com/reviews/sony-alpha-a6000/12
#9
I think that we all start mixing terminologies.  To meke the mess complete I’ll introduce my point of view.
The modern sensor end camera technology becomes very complex. I think that no one can make scientific analyzes on current sensor technology.
1.      At the begin was simple – CCD sensor sends analog values – there was simple analog antialiasing filter -> ADC -> processing of ADC data to RAW file was easily reverse ingenered.
2.      Then come Canon with C-MOS sensor technology in dSLRs =
3.      Then Canon start playing with color filters in Bayer pattern – It was also reverse engineered.
4.      BIS aaapears on the market – It allow manufacture tu put more intligence on photosensors. Analog frontend become more complex,  Journalist start to use over-simpkified terms that nowone knows the meing – e.g. sensor readout.
5.      Complexity races once again – The traditional SLR manufacture start to struggle keep in pace with software & hardware  development. New start apper in the horizont - “Computational photography ” – The smartphone small sensors and lenses start to deliver IQ similar to tohose of big cameras.
 
 
The hardcore guys still use the word cheating but no one understand what they main. It is even worse, they don’t know what they main with word “cheating.” Is Apple cheating, Is Addobe cheating, Is canon JPEG cheating or is RAW file cheating.


My point is that I don’t care on this intermediate stages. I look at the final result. One picture is good or not. You can use whatever you camera likes, whatever you post-processing software/ Apps use. Whatever you media likes – still photo,video, Paper, web, metal , plastic etc. The final result is what counts.
#10
(06-25-2020, 11:16 AM)Brightcolours Wrote: You still appear to be conflating high ISO noise reduction (seen mostly years ago in the D3 era, concerning those scientific papers you are referring to) with higher DR at low ISO (where no NR is being applied). 
The high ISO NR also depends/ed on exposure time and/or high ISO NR being enabled or not. Making blanket statements about Sony/Nikon based on some vague memories of something you read in the past does not make for the most accurate posts.

So, mostly incorrect. 

You are making unsubstantiated assumptions here. I only gave the starry sky images as an example, basically because everybody knows those. 

I used to have a 1D Mk III at some point, which had similar complaints. A small piece of software I acquired that used similar algorihtms as to what Nikon did at the time, gave similar results DR-wise at base iso. It was nice to experiment with this, but it was another step in the processing flow, and I didn't need it anyway.

I haven't kept track of any of this, other than occasionally browsing new articles on the subject, if I happened to come acoss them, but basically Canon chooses to not process noise prior to storing "RAW" results as much as Nikon and Sony do. That they have a slight disadvantage in their current line-up of processors with off-sensor processing of signals is clear, but it isn't really 2 stops of DR.

Besides, as mentioned, I do not really care about this.

Regards, Wim
Gear: Canon EOS R with 3 primes and 2 zooms, 4 EF-R adapters, Canon EOS 5 (analog), 9 Canon EF primes, a lone Canon EF zoom, 2 extenders, 2 converters, tubes; Olympus OM-D 1 Mk II & Pen F with 12 primes, 6 zooms, and 3 Metabones EF-MFT adapters ....
  


Forum Jump:


Users browsing this thread:
1 Guest(s)