Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Leica & Alternative Gear | Join Upload & Sell

1       2       3      
4
       5       6       end
  

Archive 2009 · a850 vs a900, 1 stop better noise performance?

  
 
theSuede
Offline
• • • •
Upload & Sell: Off
p.4 #1 · p.4 #1 · a850 vs a900, 1 stop better noise performance?


A "good" CFA is a combination of many things, as far as I've noticed, four things that I look for: Smooth response curves, with no kinks, spikes or narrow valleys. Reasonable primary interdistances. A balance between the blue-red c-c distance and the individual filter response width. Reasonably equal sided slopes for the intersecting channels seem to make the camera more resistant to shifts when changing WB-temperature.

Munsell Labs are one of the few places having open discussions about these problems right now - the consumer companies are understandably tight-lipped about their research and compromise choices.

Otherwise I agree with mr Borg on the D2x (yes, that's D"2"x) and the A900 being the performance leaders in colour. There's a small difference with the A850 that I haven't really had the time to look closer at yet - but I seriously consider buying one privately, unless a good s/h 900 turns up nearby soon. The M9 is quite good, at least in the center 10mm of the image-circle... :-)

I'm still waiting for a reasonable performance EVIL camera with at least a 24x16mm sensor to surface. M9 doesn't count, as I know the limitations of that camera all too well... The base sensor construction is the same as the Olympus E-1 anno 2002. Sure, there's some seriously good glass to use with it, but all too costly. And the angle sensitivity is just horrendously bad. I wouldn't even consider using a symmetrical WA (like almost all Leica WA's are) on it - and neither would Leica themselves. That's why the new Summarits are a lot more retrofocal than the older lenses...



Dec 20, 2009 at 12:30 AM
kosmoskatten
Offline
• • • •
Upload & Sell: Off
p.4 #2 · p.4 #2 · a850 vs a900, 1 stop better noise performance?


Wayne; if Iceland FM tour is happening I think I will make it there. Then you should pop over to Stockholm while you are in the "neighborhood", we'll take good care of you. Regardless of what camera you are swinging.

theSuede: very interesting reading. Thank you.

Lawrence: will PM you about Leitax.



Dec 20, 2009 at 03:17 AM
wayne seltzer
Offline
• • • •
Upload & Sell: Off
p.4 #3 · p.4 #3 · a850 vs a900, 1 stop better noise performance?


Thanks Henrik!

Thanks theSuede for your technical insight/knowledge. I have also enjoyed your posts in the 7D thread. I have a few questions:

1) Do you believe the color response charts and info in DXO Mark database are accurate for the different camera makes?
2) The things you list for a good CFA above like smooth response curves can not be deduced from the DXO data, right? How important are the color sensitivity per red, green and blue channel bar graphs in DXO. How come the P45+ red channel response has more green than red but still noone complains about a P45+'s colors?
3) Even within a camera makers line the color resonse charts vary, the metamerism numbers change, color matrix multiplier numbers change but the camera will maintain the same image color look of that camera maker?
4)When you say the D2x and the A900 are the color performance leaders, is what ways do you mean, better color accuracy, less metamerism, more different tones of each color?
5) What is responisble for the difference between Sony and Nkon/Canonon?
Is it purely CFA design differences?Or just different color profile?
6)



Dec 20, 2009 at 04:39 AM
you2
Offline
• • •
Upload & Sell: Off
p.4 #4 · p.4 #4 · a850 vs a900, 1 stop better noise performance?


Thanks.I think I mostly understand what you said but a couple of questions:
What exactly does the CFA do and does it in reality provide more gradiant of tones in the green channel for the Sony camera (a900/a850) hence improving skin tones ?

From the comments it sounds like you favor the Sony over the canon; but at the same time suggest the cameras are more or less equal under good light. Am I miss-interpreting your comments or is it more along the lines that the canon (for example) will produce rich full colours but you have to work harder to obtain them ?

Also I've noticed a large uniformly of blue shift (when I talk about blue shift I'm focusing on skin tones) in many pictures displayed on the web from canon camera after the 20d (i.e, 40d,5dmk2, ...). It seems to my eyes that there is a bit of a blue overtone on the images. Now obviously this could be due to my monitor; due to these images being processed by a raw converter that shifts the colour (ala lightroom which is commonly used); my personal taste in images or something native to the canon camera. This does not seem to be the case with pentax or olympus pictures hence the question (you mentioned several times that Canon has problems with the shift from blue to red (ala purples); and so I wonder if that is related to what I am seeing or if I'm taking something totally unrelated and applying it to poor post processing or poorly calibrated monitor.

Also you point out that olympus and canon have very dense sensors but for full frame isn't Sony and Nikon denser than Canon ?


theSuede wrote:
Well, the digital communication lines out from the sensor in the D3x are only twelve lanes wide, just as the 850/900... And I seriously doubt they "hid" the missing two lanes somewhere else. There's a definite correlation between the slower read-out times 12>14bits and this. This is not 100% certain proof, but a very good indication.

The CFA's in the A850 and the A900 seem to be slightly different, but I've no hard data there yet. I have to take the time to go through the 850. This will be soon, as I'm personally interested in that camera. In fact the
...Show more



Dec 20, 2009 at 08:35 AM
zoomo
Offline
• •
Upload & Sell: Off
p.4 #5 · p.4 #5 · a850 vs a900, 1 stop better noise performance?


douglasf13 wrote:
Huh Nothing here seems trollish to me.



Oops so sorry, makes no sense at all, I was reading several threads on several forums simultaneously, and this was meant for elsewhere, I had no intention of saying anything in this thread, please accept my apologies. I can't even find the other thread again now!



Dec 20, 2009 at 03:32 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.4 #6 · p.4 #6 · a850 vs a900, 1 stop better noise performance?


Ouch, that's a lot... :-) Well, from the top:

Yes, I do believe that the DxO findings are quite accurate. I do sometimes find quite small deviations from my own measurements, or measurements made by other people I trust to "get it right", but there's nothing strange about that - sample variation and measurement methodology variation often cause even larger deltas than the ones I've seen. Trying to get something useful out of the numbers is though quite hard. Unless you're sentient in the field, you'll likely end up with a headache and still know no more than when you started reading.

You can only guess at what the CFA filter spectrum looks like from the DxO figures - but they do plot complete spectrals at some places not directly connected to the "DxO-mark" part of the site. I don't know if all of them are open to the general public. The P45+ (and some others, like the Leica sensors) have chosen a slightly different approach with a slightly wider separation of the primaries, and slightly wider filters. The red in the P45+ has its main peak at ~630nm which is a bit further out than the "normal" 610nm. To keep metamerisms as low as possible green has to follow red, and is placed at just over 550nm. This shifts the "best hue-resolution" point by just a bit, makes purples easier to get "just right", and lowers sensitivity when you use very cool or warm WB's - which isn't really that important in MDFB's - they're almost exclusively used with studio strobes... sRGB primaries are centered (best average) at R:615nm G:550nm B:465nm if you assume D65 whitepoint.
And, actually, the green isn't really "stronger" in red, you have to factor the relative sensitivities, in this case that means dividing the green value by 0.4 (x2.5).

Even if spot colour differences may be quite large between different models from the same maker, the general colour "fingerprint" of the brand is quite easy to achieve. It's just a matter of getting the profiles right - which they hopefully are capable of doing. Raw-data is NOT the same thing as colour. The transform and interpolation makes the final colour.

D2x and A900 share two very important characteristics; they have very good (small) metamerism areas, and they have very good hue resolution. Hue resolution is (can be...) measured by how small the hue differences that the camera can distinguish between are, if you place two (say) 10x10 pixel patches of each hue next to eachother.

Sony chose to not sacrifice anything on the "high-ISO" altar, but concentrated on just getting the relationships between the CFA colours "just right" in stead. Both Nikon and Canon have started to sacrifice some accuracy for better transmittivity > lower noise. Lower overall noise does give better, more accurate colour in the mid-ISO ranges though, but you loose some at base/low ISOs. The companies' "fingerprint" colours are something I have no idea why/how they've arrived at from the beginning. But the fingerprint models are more of a large, sweeping, general "tone" of the camera, the different models still exhibit very clear "personalities" when you start to look a little closer at the results.



Dec 20, 2009 at 05:04 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.4 #7 · p.4 #7 · a850 vs a900, 1 stop better noise performance?


you2:
What the CFA really does is quite complex, as it's related to two main things: trying to match the metamerism behaviour of the human vision, while trying to get as good hue-resolution as possible. The balancing between those two is not something that can be explained in short words... :-)
Humans can (due to the way that the pigments in our cones in the works in conjuction with the brain) see two or more combinations of different wavelengths of light as "the same colour". This is a large part of the CIE "standard observer" studies and statistics.
So, if we are to make a RGB-sensor as "conformant" to human vision as possible, this has to be taken into consideration. The same combinations of different wavelengts that we as humans consider "white" should be considered "white" by the camera RGB value combination.
Unfortunately, this makes the other part of the task - hue resolution - harder. It's a lot easier to get good hue resolution if you separate the filter center wavelengths ("peaks") a bit further from each other - but this also wreaks havoc on the human-observer related metamerisms...

The way manufacturers make their general "trade-off choices" is a large part of the manufacturer colour fingerprint also, and those chosen trade-offs also affect the way you can post-process your pictures quite strongly. It's said that MDFB's can be PP'd a lot harder than "normal" cameras, and in part this is very much true - due to the choices an MDFB maker is likely to do when choosing a sensor package.

Most of the Canon "problem" stems from the area that covers the skin-tones. By setting WB to get skintones "just right", you usually screw up the rest of the hues present in the picture. But in controlled lighting, building a profile to make this right is not a problem.

I did not mean that Canon/Oly have got "high density" sensors in an absolute meaning, my intent was to try to show that both companies generally try to compete by resolution - having a bit higher MB/cm2 than their nearest competitors. They have to do something to increase light-efficiency per mm2 if they don't want more noise to go with the higher resolution.



Dec 20, 2009 at 05:30 PM
you2
Offline
• • •
Upload & Sell: Off
p.4 #8 · p.4 #8 · a850 vs a900, 1 stop better noise performance?


Ok thanks. Not sure if what you said explains why canon skins are bluish but there is still an issue (for me) of camera+lens. I'll see what happens over the next month or two; maybe canon will produce a better sensor or sony will announce some lenses.


Dec 20, 2009 at 05:44 PM
douglasf13
Offline
• • • • •
Upload & Sell: Off
p.4 #9 · p.4 #9 · a850 vs a900, 1 stop better noise performance?


Wow, theSuede, thanks for taking the time to type these interesting posts. Very helpful.


Dec 20, 2009 at 08:09 PM
Tariq Gibran
Offline
• • • • • •
Upload & Sell: Off
p.4 #10 · p.4 #10 · a850 vs a900, 1 stop better noise performance?


douglasf13 wrote:
Wow, theSuede, thanks for taking the time to type these interesting posts. Very helpful.


...and somewhat confusing regarding noise comparisons between the a850 and a900. For instance, theSuede states that he finds DXO quite accurate yet DXO has the a900 practically equal to the a850 at all ISO's regarding noise yet theSuede states that "One of the major reasons that the A850 is better at higer ISO's than the A900 is that they reinforced the botched and underdimensioned powersupply lines going into the sensor. The support structure around the chip is totally different from the D3x implementation."

These observations don't seem to jive with one another. Again, technical jargon aside (and no offense to theSuede), but does anyone have any real life experience with both cameras regarding noise?




Dec 20, 2009 at 08:35 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.4 #11 · p.4 #11 · a850 vs a900, 1 stop better noise performance?


No offense taken - for all and any purpose I doubt that you can see any difference at all between the 850/900. I didn't say the difference was big, or even noticeable - just that it was there, and why it was there... The thread title is way off - the difference is probably less than 1/10Ev at the highest ISOs - but I still find it worthwhile to mention that it is an improvement, and not the other way around.

Now IF you see a difference, it's probably there because your raw-converter of choice use slightly different "base" settings and colour profiles for the different models. Or, if you're shooting jpg - better noise-suppression and better interpolation algorithms in the newer 850 firmware. Nothing bigger than this.

I'll know more from firsthand experience after X-mas week, hopefully... :-) But as far as every indication would show, I doubt I will be able to see any difference in a normal shot.



Dec 20, 2009 at 11:48 PM
wayne seltzer
Offline
• • • •
Upload & Sell: Off
p.4 #12 · p.4 #12 · a850 vs a900, 1 stop better noise performance?


ISO 100 200 400 800 1600 3200 6400
A900 1.2 1.5 1.7 1.9 2.0 2.4 3.5

A850 0.9 1.2 1.5 1.8 1.9 2.6 3.1

The ISO noise test numbers above are from Pop-Photo.
I think their statement of a full stop better noise improvement is
misleading and you can see that it is just a slight improvement like theSuede says.



Dec 21, 2009 at 01:03 AM
wayne seltzer
Offline
• • • •
Upload & Sell: Off
p.4 #13 · p.4 #13 · a850 vs a900, 1 stop better noise performance?


Thanks again theSuede for your technical insights.
I understand your point about colors near the metamerism areas of a camera may become inaccurate and are affected more by the WB setting but when I think of this preferred color of the Sony over Canon and Nikon I feel it is based more on the difference of the camera maker's color fingerprint When I see a shot of some green foliage taken by a Sony or MFDB back, the green is more vivid than a Canon Nikon rendition.I think people who like more vivid colors like the color from cameras like Sony A900, Kodak 14N and MFDB backs.
And also the reason is because the colors are more vivid rather than there are more shades of fine hue changes captured. It would be interesting if their are any other Sony cameras whose CFA had less hue resolution than the A900 and like a Canon or Nikon camera and see if that camera's colors give the same simialar feeling of more vivid color still because of the camera maker's color fingerprint.

theSuede, do you believe that it is possible to have a preset or way to tweak the colors in post to make a Canon camera's colors come close to the Sony fingerprint colors, except maybe the color near metamerism areas?



Dec 21, 2009 at 02:14 AM
Tariq Gibran
Offline
• • • • • •
Upload & Sell: Off
p.4 #14 · p.4 #14 · a850 vs a900, 1 stop better noise performance?


theSuede wrote:
No offense taken - for all and any purpose I doubt that you can see any difference at all between the 850/900. I didn't say the difference was big, or even noticeable - just that it was there, and why it was there... The thread title is way off - the difference is probably less than 1/10Ev at the highest ISOs - but I still find it worthwhile to mention that it is an improvement, and not the other way around.


Thanks for the clarification. The tread title is a question based on Pop Photo's findings that the a850 showed one stop less noise at low ISO's below 400. They claim to use DXO software for their testing but fail to go into much detail otherwise. I suspect this result can't be substantiated which is why I had asked the question. I also postulated that perhaps if a difference was there, it was a result of firmware only. Others postulated that there was a physical difference in the CFA between the two cameras that might be responsible which I believe you say there is not. If we are really only talking 1/10Ev, that's an insignificant difference in real world shooting. It's also interesting that DXOmark finds the color sensitivity between these two cameras practically identical whereas others have said there is a difference.



Dec 21, 2009 at 07:07 AM
Tariq Gibran
Offline
• • • • • •
Upload & Sell: Off
p.4 #15 · p.4 #15 · a850 vs a900, 1 stop better noise performance?


wayne seltzer wrote:
ISO 100 200 400 800 1600 3200 6400
A900 1.2 1.5 1.7 1.9 2.0 2.4 3.5

A850 0.9 1.2 1.5 1.8 1.9 2.6 3.1

The ISO noise test numbers above are from Pop-Photo.
I think their statement of a full stop better noise improvement is
misleading and you can see that it is just a slight improvement like theSuede says.


These are standard deviation numbers, not stops (which means little to me since I forgot what these actually mean via statistics!) . According to Pop Photo:
"Using DxO Analyzer software, we measure the standard deviation of grayscale patches across a full range from dark to light."

I have also learned that they do use RAW files and convert them into tiffs using the manufacturers raw converter before analyzing. So, it's nice to know they are at lest not using out of camera jpegs.



Dec 21, 2009 at 07:52 AM
douglasf13
Offline
• • • • •
Upload & Sell: Off
p.4 #16 · p.4 #16 · a850 vs a900, 1 stop better noise performance?


Good to know that they use the manufactures RAW converter. I wonder if they just leave everything on "standard," and, if so, is it just the converters profile that is off, color-wise?

As far as the cfa, theSuede can correct me if I'm wrong, but it seems there might be small differences, at least, judging by DxO's color response. Or is that just programming?



Dec 21, 2009 at 11:26 AM
you2
Offline
• • •
Upload & Sell: Off
p.4 #17 · p.4 #17 · a850 vs a900, 1 stop better noise performance?


I was looking at dx0 mark (specifically colour response) for various cameras and while the canon score relatively low; what is amusing is how high olympus scores. All of the recent cameras (e-620 not tested) score above 80; most of them in the 82-84 range. While the olympus small sensors have other issues; it seems that they have made colour a bit of a priority where as the nikon d300s scores no higher than the canon 7d; despite being 12mp and the 1dmk3 is around the same level as the sony a900 (85.75). Not sure it is fair to compare the 1dmk3 with the sony given price point; but still since there is a bit of talk about colours thought I would mention it.

It is also interesting that the 10d has a colour score of 86.22. I checked this specific body because when I was noticing blue skin tones; several of the sites I had looked at with 5dmk2 images had older images with 10d; which seemed to have better skin tones. Naturally this could just be a change in post processing or maybe not.

Edited on Dec 21, 2009 at 12:19 PM · View previous versions



Dec 21, 2009 at 11:48 AM
mawz
Offline
• • • • • •
Upload & Sell: Off
p.4 #18 · p.4 #18 · a850 vs a900, 1 stop better noise performance?


you2 wrote:
I was looking at dx0 mark (specifically colour response) for various cameras and while the canon score relatively low; what is amusing is how high olympus scores. All of the recent cameras (e-620 not tested) score above 80; most of them in the 82-84 range. While the olympus small sensors have other issues; it seems that they have made colour a bit of a priority where as the nikon d300s scores no higher than the canon 7d; despite being 12mp and the 1dmk3 is around the same level as the sony a900 (85.75). Not sure it is fair to compare
...Show more

Colour is the one thing Oly got right from the get-go with 4/3rds bodies.

One of the ironies of the high scores for DXOMark is that many Oly users complain about the poor colour in the newer bodies with the Panasonic sensors (vs. the older Kodak sensors).



Dec 21, 2009 at 11:58 AM
Tariq Gibran
Offline
• • • • • •
Upload & Sell: Off
p.4 #19 · p.4 #19 · a850 vs a900, 1 stop better noise performance?


I agree. I have always been impressed by Olympus digital color.


Dec 21, 2009 at 12:11 PM
you2
Offline
• • •
Upload & Sell: Off
p.4 #20 · p.4 #20 · a850 vs a900, 1 stop better noise performance?


Well I was comparing panasonic models (not so many of the @ dx0) to olympus and noticed that olympus seemed to score a bit better with regards to colour response (ep1 and e-510 to g1 for example; which might not be fair). I was really trying to check if move to video ready sensors influence colour response (50d to 7d and d300 to d300s); alas the gh1 did not have data.

Ho hum. Quite amazed at how low the d300s scored.

mawz wrote:
Colour is the one thing Oly got right from the get-go with 4/3rds bodies.

One of the ironies of the high scores for DXOMark is that many Oly users complain about the poor colour in the newer bodies with the Panasonic sensors (vs. the older Kodak sensors).




Dec 21, 2009 at 12:23 PM
1       2       3      
4
       5       6       end




FM Forums | Leica & Alternative Gear | Join Upload & Sell

1       2       3      
4
       5       6       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.