fredmiranda.com Login
This topic is in the 2013 archive. You can read it and like its posts here. It is read only.

Archive 2013Sony A7R sensor gets high scores from DxOMark

  
 
Tariq Gibran
Offline
Details
p.3 #1Sony A7R sensor gets high scores from DxOMark
douglasf13 wrote:
Mescalamba wrote
douglasf13 wrote
sebboh wrote
Lee Saxon wroteWhy post the meaningless, totally made up DXO "Scores" instead of the charts that are actually useful?

Also seems the D800 sensor is an (imperceptible in photos) smidge better. I would've expected the opposite from a newer camera even if it used the same or nearly the same sensor.
nikon always seems to score a smidge better than sony when they use the same sensor. same with pentax.
Iliah Borg, who does the color profiling for RPP, used to talk quite a bit about the differences in the CFAs that the various camera companies use, and how Sony sacrifices a bit of high ISO performance for better color with a less transparent CFA. I'm not sure how one gleans that from DXO Mark, though.
D3x vs A900 for example. D600 vs A99 (but thats a bit unfair due A99 SLT design). Well, they dont do that anymore, A7r is proof. Same colors as D800E, same results as D800E. I guess they were tired of making something that only few appreciate. Im sure it will work for them tho. It does for Canon for quite some time..

5 points difference is visible, 10 points is more or less obvious, 15 points is striking. Dont believe? Unless you are colorblind, compare pics from 1Ds MK3 vs 6D under daylight. Or if someone wants more striking result, A900 or A99 vs 6D.
I'm not sure that it is certain, yet. The D600 and RX1 score very similarly in terms of DR and SNR, too, despite the higher SMI of the RX1. I don't think SMI is necessarily a good indicator. Like theSuede has mentioned, the Leica M9 also handles color more like MFDB (Iliah says the same,) and it only has an SMI of 76.

I doubt that Sony has stopped using their usual CFA design. The new sensor is probably just a little more efficient than the older D800 sensor, so that could be equalizing the SNR and DR.


...and in that link I provided to the previous discussion (which many here took part in), theSuede did mention something about Sony being a special case (in a good way with regard to color accuracy).
Oct 31, 2013 at 04:18 PM
zhangyue
Offline
Details
p.3 #2Sony A7R sensor gets high scores from DxOMark
No, I won't take SMI score as a bible. I sure will take this a grain of salt. but seems the quote you referred imply that score is not irrelevant.

But a camera with SMI80 will almost certainly be a lot better than the camera with SMI60."

It is a good indication. and coincidentally, the camera having high SMI score also happen to have a good reputation of color reproduction: A900, 1DsIII etc.. and camera has low SMI score also happen to have many complain, ignore the list here

Conceptually, this isn't that hard to understand as once the signal is captured in raw, whatever profile you choose just have to boost different color channel to target color, the more gain you have to applied, the harder to achieve target color. Especially for something involve very subtle tonality change such as skin tone....

The same terminology can also be used in audio world, that mixer (boost/attenuate different frequency signal) can never achieve the sound as good as the time of recording it right.

If I have to choose RX1 now, I will choose regular version just based on this 'huge' SMI score difference than little sharpness gain.
Oct 31, 2013 at 04:48 PM
chez
Offline
Details
p.3 #3Sony A7R sensor gets high scores from DxOMark
naturephoto1 wrote:
I handled the camera at PhotoPlus and I found that the camera with the grip handled and balanced quite well.

Rich


Who cares how it balanced and handled...did it match your shoes?
Oct 31, 2013 at 04:54 PM
douglasf13
Offline
Details
p.3 #4Sony A7R sensor gets high scores from DxOMark
zhangyue wrote:
No, I won't take SMI score as a bible. I sure will take this a grain of salt. but seems the quote you referred imply that score is not irrelevant.

But a camera with SMI80 will almost certainly be a lot better than the camera with SMI60."

It is a good indication. and coincidentally, the camera having high SMI score also happen to have a good reputation of color reproduction: A900, 1DsIII etc.. and camera has low SMI score also happen to have many complain, ignore the list here

Conceptually, this isn't that hard to understand as once the signal is captured
...Show more →

The SMI difference between the RX1 and RX1R is only 5 points, so I'm not sure it matters much. As quoted on the DXO website, "In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating. It is different for low-end cameras (such as camera phones), which typically have a SMI of about 40. For this reason, we give this measurement as an indication but do not integrate it in DxO Mark."

Since the A7R scored close to the RX1R, with a 78, I wonder if the A7 will be closer to 84, like the RX1? Is it possible that some kind of artifacts from removing the AA filter are causing this discrepancy?
Oct 31, 2013 at 04:56 PM
RustyBug
Offline
Details
p.3 #5Sony A7R sensor gets high scores from DxOMark
chez wrote:
naturephoto1 wroteI handled the camera at PhotoPlus and I found that the camera with the grip handled and balanced quite well.

Rich
Who cares how it balanced and handled...did it match your shoes?



I'm waiting for the gold-plated version ... it'll go with everything.

To see or to be seen ... that is the question.
Oct 31, 2013 at 05:00 PM
Tariq Gibran
Offline
Details
p.3 #6Sony A7R sensor gets high scores from DxOMark
zhangyue wrote:
No, I won't take SMI score as a bible. I sure will take this a grain of salt. but seems the quote you referred imply that score is not irrelevant.

But a camera with SMI80 will almost certainly be a lot better than the camera with SMI60."

It is a good indication. and coincidentally, the camera having high SMI score also happen to have a good reputation of color reproduction: A900, 1DsIII etc.. and camera has low SMI score also happen to have many complain, ignore the list here

Conceptually, this isn't that hard to understand as once the signal is captured
...Show more →

Yes, but which cameras are we discussing which have a SMI difference of 20 points? That is where we know the SMI matters according to theSuede. If we are talking only 5 points, he specifically makes the case that we don't know if the camera with the higher score actually has more accurate color than the one with the lower score or vice versa. Thus, SMI is practically irrelevant because we are not comparing cameras with a huge SMI delta..at least that's what I take from what theSuede wrote - SMI can't reliably be used as an accurate indicator or color accuracy when the delta between the cameras in question is low (which seems to be the case with almost all current cameras).


Edited on Oct 31, 2013 at 05:10 PM · View previous versions
Oct 31, 2013 at 05:08 PM
skibum5
Offline
Details
p.3 #7Sony A7R sensor gets high scores from DxOMark
Tariq Gibran wrote:
Mescalamba wrote
Tariq Gibran wrote
Mescalamba wroteYea and CFA almost exactly like D800E. No A7R for me then..

I thought it has slightly worse colors than A99, but said to myself, well its JPGs. Apparently it wasnt JPGs. Sigh. :/

Well, f*ck that.. back to dSLRs.
I take it you are referring to the individual "color response" measurement and not the overall "Portrait (Color Depth)" score? Interestingly, it appears that it might be the case that Sony sensors with AA filters have a higher color response measurement than those without an AA filter. That is the case with the RX1 vs RX1r anyway. I don't know ultimately how important that measurement is - that it actually makes as much difference as it might seem...at least I believe theSuede said something to that extent. It was a while back when discussing the A900 (which does have amazing color and color response #s?) so I'm not 100% sure.
http://www.dxomark.com/About/In-depth-measurements/Measurements/Color-sensitivity

Simply called SMI. A7R result of 78/75 is a "meh". D800E has 77/74 (I guess these sensors are really very similar). And for example 1D-X has 73/74, which is a bit.. tragic. Especially compared to 1D/s MK3 .

Bottom is currently 6D with 69/70. Thats somewhere around "tragicomedy" level. Quite visible in all pics taken with it.

Obviously people dont have same vision of colors and some are more or less colorblind. Since most photographers are still men, which are even more colorblind, it doesnt matter that much as it could.

That said, SMI is directly tied to CFA used. Better you get, better SMI you will have as results. Downside is that thicker CFA means also less light = more noise (tho in case of something like D4 sensor, it wouldnt have exactly big impact on SNR).

Issue with SMI is that it doesnt tell "where those color errors are". It doesnt tell if camera cant find difference between violet and blue flowers, or pink vs red skin tone. Even then its rather good tool for getting overall idea how much good or bad colors will be.

In case of A7R, if you are fine with current Nikon and Canon offerings, you will be fine with this. Actually you will probably gain in colors too.

Me? I will pass and wait if Nikon came to senses with Nikon "Pure" DF or whatever that will be and see then.
So, if we take the RX1 (84/81) and RX1r (79/75) example I cited, your conclusion is that these two cameras are using a completely different CFA which is why they differ? I suspect there is more to it, perhaps including the AA filter as well as software tuning. Anyway, I bet the A7 with it's AA filter will likely have a better measurement. Incidentally, the Leica M 240 only scores 75/71 but I really don't hear folks talking about how bad Leica color is. I suspect the delta has to be pretty large to really matter. Per DxO:
"In practice,
...Show more →


Maybe it is due to the AA filter removal, considering these are bayer sensors I could imagine ways that might be damaging to color reproduction. Personally I think this should have had a least a weak AA filter in it, 36MP for FF isn't dense enough to get rid of the filter IMO, especially not if you have lenses that can handle it well. The lack of the AA filter is the one downside to how they made the sensor IMO. A weak AA filter in it would've been better I think and just in general too.
Oct 31, 2013 at 05:08 PM
Mescalamba
Offline
Details
p.3 #8Sony A7R sensor gets high scores from DxOMark
1D/s MK3 has markedly better and "alive" colors than most of things from Canon. Almost all "past" Canon cameras (including even simple old 5Dc) had better looking colors than today ones. And almost all had better SMI.

50D, despite its rather poor SNR characteristic is very good when it comes to colors (bit harder to take advantage of that exactly due that poorer SNR).

Nikon D300, D2Xs, D700 .. all not only "in theory" but in practice provide better colors than newer "better" models. All higher SMIs.

Yea I think that there is relation between good colors (or simply colors that have good enough separation) and CFA and SMI.


Its supposedly "non-important" part of DxOmark measuring. Well, when its so non-important, why they bother with testing every camera under two different illuminants and do rather thorough color tests. Including how much is each color channel boosted (rather important thing too, for example if SNR is close to 38dB - level of relative "noise-free" image, but blue color channel is under daylight somewhere like 2.5x boost, then you have almost guaranteed problems with blue sky when it comes to noise).


I have no doubt Sony changed CFA. I had doubts when I saw pics but thought, well, crappy JPG colors, could happen (except those JPGs are not exactly crappy in other things). They did that before, A99 doesnt have colors like A900. Its closer to A850 and has slightly strange shift (which as result produce very nice skin colors, but bit less of others. Still better than Canikon today).


M9 colors are not accurate in anything. But they look nice and if you PP it a bit, it looks a bit like digital Velvia. CFA like that is both cheap and doesnt hurt SNR (which is something that Leica most likely really didnt want to worsen even further ). Why it looks bit like MFDB might be cause of CCD and it might be (thats just a guess) 16-bit ADC inside (M8 had one) with output to 14-bits. Thats one way how to make camera with cheap-ish CFA and still colors that most will like (me included, even tho they are not 100% accurate .. or more then 76% .


6D is another way to do the same. And sorry, that I offer my opinion, but those colors are bl**dy awful. Yet, people buy it and are happy with it and its results. Which sorta proves my point about colorblindness. That and/or they simply prefer different colors. I dont take it from them, if they like it.


Anyway, its not end of the world. A7r is pretty much D800E minus Nikon AF and Nikon lens mount. Plus sensor really without AA filter. Which without doubt will please a lot of their users.

For me its without A900 colors that I hoped for. Which means end of story for me. I wont post here any further.
Oct 31, 2013 at 05:10 PM
skibum5
Offline
Details
p.3 #9Sony A7R sensor gets high scores from DxOMark
Tariq Gibran wrote:
sebboh wrote
Tariq Gibran wroteSo, if we take the RX1 (84/81) and RX1r (79/75) example I cited, your conclusion is that these two cameras are using a completely different CFA which is why they differ? I suspect there is more to it, perhaps including the AA filter as well as software tuning. Anyway, I bet the A7 with it's AA filter will likely have a better measurement. Incidentally, the Leica M 240 only scores 75/71 but I really don't hear folks talking about how bad Leica color is. I suspect the delta has to be pretty large to really matter. Per DxO:
"In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating." Given the somewhat arbitrary nature of these scores you mentioned (and theSuede previously also mentioned), I don't think these scores really mean that much in practice.
i actually i hear a lot of people whining about the colors on the m240 compared to the m9. though i think maybe the m9 scores even worse?
The measurements are really close between the M9 (76/72) and m240 (75/71) and the strict difference is inconsequential just going by this measurement alone. I think it's the case that sometimes cameras have high numbers and great color and low numbers and not so great color BUT there is not always a correlation between the two (so one can't always go by this measurement). This is probably why DXO says the measurements are not very discriminating and that they don't even use them in their overall scoring.


It's really complex, since who knows in what shades the differences occur and which shades matter more to you, etc. And one overall worse might still be better in some other shades, etc. I'd certainly bet on one with scores like 86 vs one with scores like 72 though.
Oct 31, 2013 at 05:11 PM
Tariq Gibran
Offline
Details
p.3 #10Sony A7R sensor gets high scores from DxOMark
skibum5 wrote:
Tariq Gibran wrote
Mescalamba wrote
Tariq Gibran wrote
Mescalamba wroteYea and CFA almost exactly like D800E. No A7R for me then..

I thought it has slightly worse colors than A99, but said to myself, well its JPGs. Apparently it wasnt JPGs. Sigh. :/

Well, f*ck that.. back to dSLRs.
I take it you are referring to the individual "color response" measurement and not the overall "Portrait (Color Depth)" score? Interestingly, it appears that it might be the case that Sony sensors with AA filters have a higher color response measurement than those without an AA filter. That is the case with the RX1 vs RX1r anyway. I don't know ultimately how important that measurement is - that it actually makes as much difference as it might seem...at least I believe theSuede said something to that extent. It was a while back when discussing the A900 (which does have amazing color and color response #s?) so I'm not 100% sure.
http://www.dxomark.com/About/In-depth-measurements/Measurements/Color-sensitivity

Simply called SMI. A7R result of 78/75 is a "meh". D800E has 77/74 (I guess these sensors are really very similar). And for example 1D-X has 73/74, which is a bit.. tragic. Especially compared to 1D/s MK3 .

Bottom is currently 6D with 69/70. Thats somewhere around "tragicomedy" level. Quite visible in all pics taken with it.

Obviously people dont have same vision of colors and some are more or less colorblind. Since most photographers are still men, which are even more colorblind, it doesnt matter that much as it could.

That said, SMI is directly tied to CFA used. Better you get, better SMI you will have as results. Downside is that thicker CFA means also less light = more noise (tho in case of something like D4 sensor, it wouldnt have exactly big impact on SNR).

Issue with SMI is that it doesnt tell "where those color errors are". It doesnt tell if camera cant find difference between violet and blue flowers, or pink vs red skin tone. Even then its rather good tool for getting overall idea how much good or bad colors will be.

In case of A7R, if you are fine with current Nikon and Canon offerings, you will be fine with this. Actually you will probably gain in colors too.

Me? I will pass and wait if Nikon came to senses with Nikon "Pure" DF or whatever that will be and see then.
So, if we take the RX1 (84/81) and RX1r (79/75) example I cited, your conclusion is that these two cameras are using a completely different CFA which is why they differ? I suspect there is more to it, perhaps including the AA filter as well as software tuning. Anyway, I bet the A7 with it's AA filter will likely have a better measurement. Incidentally, the Leica M 240 only scores 75/71 but I really don't hear folks talking about how bad Leica color is. I suspect the delta has to be pretty large to really matter. Per DxO:
"In practice, the SMI for DSLRs ranges between 75 and 85, and is not very discriminating." Given the somewhat arbitrary nature of these scores you mentioned (and theSuede previously also mentioned), I don't think these scores really mean that much in practice.
Maybe it is due to the AA filter removal, considering these are bayer sensors I could imagine ways that might be damaging to color reproduction. Personally I think this should have had a least a weak AA filter in it, 36MP for FF isn't dense enough to get rid of the filter IMO, especially not if you have lenses that can handle it well. The lack of the AA filter is the one downside to how they made the sensor IMO. A weak AA filter in it would've been better I think and just in general too.


Agree about the AA filter in general. I would prefer to have one. That said, it's pretty odd that in all the samples out there, moire has not really shown up yet. I'm wondering why that is.
Oct 31, 2013 at 05:13 PM
skibum5
Offline
Details
p.3 #11Sony A7R sensor gets high scores from DxOMark
Tariq Gibran wrote:
Mescalamba wrote
douglasf13 wrote
sebboh wrote
Lee Saxon wroteWhy post the meaningless, totally made up DXO "Scores" instead of the charts that are actually useful?

Also seems the D800 sensor is an (imperceptible in photos) smidge better. I would've expected the opposite from a newer camera even if it used the same or nearly the same sensor.
nikon always seems to score a smidge better than sony when they use the same sensor. same with pentax.
Iliah Borg, who does the color profiling for RPP, used to talk quite a bit about the differences in the CFAs that the various camera companies use, and how Sony sacrifices a bit of high ISO performance for better color with a less transparent CFA. I'm not sure how one gleans that from DXO Mark, though.
D3x vs A900 for example. D600 vs A99 (but thats a bit unfair due A99 SLT design). Well, they dont do that anymore, A7r is proof. Same colors as D800E, same results as D800E. I guess they were tired of making something that only few appreciate. Im sure it will work for them tho. It does for Canon for quite some time..

5 points difference is visible, 10 points is more or less obvious, 15 points is striking. Dont believe? Unless you are colorblind, compare pics from 1Ds MK3 vs 6D under daylight. Or if someone wants more striking result, A900 or A99 vs 6D.
So, you disagree with what theSuede writes with regard to this topic which I just quoted? A coincidence doesn't automatically lead to a correlation (your examples).


I think the 6D vs 1Ds3 is like 15 points or something though.


(and as he said a 5 point difference may mean nothing and the lower score may even do better in practice BUT it could also be meaningful and potentially even more than the small difference indicated)


Edited on Oct 31, 2013 at 05:19 PM · View previous versions
Oct 31, 2013 at 05:14 PM
Mescalamba
Offline
Details
p.3 #12Sony A7R sensor gets high scores from DxOMark
skibum5 wrote:
Maybe it is due to the AA filter removal, considering these are bayer sensors I could imagine ways that might be damaging to color reproduction. Personally I think this should have had a least a weak AA filter in it, 36MP for FF isn't dense enough to get rid of the filter IMO, especially not if you have lenses that can handle it well. The lack of the AA filter is the one downside to how they made the sensor IMO. A weak AA filter in it would've been better I think and just in general too.


If you test colors by simply measuring color patches, then being Bayer or not or having AA or not, shouldnt make a difference. In real life, it might. And I agree that 36 mpix FF without AA filter isnt best idea.. (pixel density under 16 mpix APS-C isnt enough).
Oct 31, 2013 at 05:15 PM
zhangyue
Offline
Details
p.3 #13Sony A7R sensor gets high scores from DxOMark

Tariq Gibran wrote:
zhangyue wroteNo, I won't take SMI score as a bible. I sure will take this a grain of salt. but seems the quote you referred imply that score is not irrelevant.

But a camera with SMI80 will almost certainly be a lot better than the camera with SMI60."

It is a good indication. and coincidentally, the camera having high SMI score also happen to have a good reputation of color reproduction: A900, 1DsIII etc.. and camera has low SMI score also happen to have many complain, ignore the list here

Conceptually, this isn't that hard to understand as once the signal is captured in raw, whatever profile you choose just have to boost different color channel to target color, the more gain you have to applied, the harder to achieve target color. Especially for something involve very subtle tonality change such as skin tone....

The same terminology can also be used in audio world, that mixer (boost/attenuate different frequency signal) can never achieve the sound as good as the time of recording it right.

If I have to choose RX1 now, I will choose regular version just based on this 'huge' SMI score difference than little sharpness gain.
Yes, but which cameras are we discussing which have a SMI difference of 20 points? That is where we know the SMI matters according to theSuede. If we are talking only 5 points, he specifically makes the case that we don't know if the camera with the higher score actually has more accurate color than the one with the lower score or vice versa. Thus, SMI is practically irrelevant because we are not comparing cameras with a huge SMI delta..at least that's what I take from what theSuede wrote - SMI can't reliably be used as an accurate indicator
...Show more →

What make 20 a magic number than 5 or 10 in score.
My 6d have grand total score of 69, which 20 below a900, so it is bad, and 5diii is 79, 10 point less than a900, so it is not bad, but now my 6d is only 10 point below 5diii, then my 6d must not bad as well

Basically, you were saying all current camera having same performance is this regard. I disagree, but I agree we should not take this number as sole indication but a good reference.
I believe 5 point between RX1 and RX1R is real and difference will be detectable for sensitive eyes. If it is profile correctable? I don't know.


Oct 31, 2013 at 05:17 PM
Tariq Gibran
Offline
Details
p.3 #14Sony A7R sensor gets high scores from DxOMark
Mescalamba wrote:
1D/s MK3 has markedly better and "alive" colors than most of things from Canon. Almost all "past" Canon cameras (including even simple old 5Dc) had better looking colors than today ones. And almost all had better SMI.

50D, despite its rather poor SNR characteristic is very good when it comes to colors (bit harder to take advantage of that exactly due that poorer SNR).

Nikon D300, D2Xs, D700 .. all not only "in theory" but in practice provide better colors than newer "better" models. All higher SMIs.

Yea I think that there is relation between good colors (or simply colors that have good
...Show more →

I agree the A900 has great color - still the best imo that I have seen out of a FF sensor.

BTW, sorry if I came off as antagonistic. That was not my intention at all and I think the discussion is both relevant and healthy.
Oct 31, 2013 at 05:18 PM
skibum5
Offline
Details
p.3 #15Sony A7R sensor gets high scores from DxOMark
Mescalamba wrote:
skibum5 wroteMaybe it is due to the AA filter removal, considering these are bayer sensors I could imagine ways that might be damaging to color reproduction. Personally I think this should have had a least a weak AA filter in it, 36MP for FF isn't dense enough to get rid of the filter IMO, especially not if you have lenses that can handle it well. The lack of the AA filter is the one downside to how they made the sensor IMO. A weak AA filter in it would've been better I think and just in general too.
If you test colors by simply measuring color patches, then being Bayer or not or having AA or not, shouldnt make a difference. In real life, it might. And I agree that 36 mpix FF without AA filter isnt best idea.. (pixel density under 16 mpix APS-C isnt enough).


yeah i guess testing large patches it would easily all average out

Edited on Nov 03, 2013 at 10:44 PM · View previous versions
Oct 31, 2013 at 05:20 PM
Tariq Gibran
Offline
Details
p.3 #16Sony A7R sensor gets high scores from DxOMark
zhangyue wrote:
Tariq Gibran wrote
zhangyue wroteNo, I won't take SMI score as a bible. I sure will take this a grain of salt. but seems the quote you referred imply that score is not irrelevant.

But a camera with SMI80 will almost certainly be a lot better than the camera with SMI60."

It is a good indication. and coincidentally, the camera having high SMI score also happen to have a good reputation of color reproduction: A900, 1DsIII etc.. and camera has low SMI score also happen to have many complain, ignore the list here

Conceptually, this isn't that hard to understand as once the signal is captured in raw, whatever profile you choose just have to boost different color channel to target color, the more gain you have to applied, the harder to achieve target color. Especially for something involve very subtle tonality change such as skin tone....

The same terminology can also be used in audio world, that mixer (boost/attenuate different frequency signal) can never achieve the sound as good as the time of recording it right.

If I have to choose RX1 now, I will choose regular version just based on this 'huge' SMI score difference than little sharpness gain.
Yes, but which cameras are we discussing which have a SMI difference of 20 points? That is where we know the SMI matters according to theSuede. If we are talking only 5 points, he specifically makes the case that we don't know if the camera with the higher score actually has more accurate color than the one with the lower score or vice versa. Thus, SMI is practically irrelevant because we are not comparing cameras with a huge SMI delta..at least that's what I take from what theSuede wrote - SMI can't reliably be used as an accurate indicator or color accuracy when the delta between the cameras in question is low (which seems to be the case with almost all current cameras).
What make 20 a magic number than 5 or 10 in score.
My 6d have grand total score of 69, which 20 below a900, so it is bad, and 5diii is 79, 10 point less than a900, so it is not bad, but now my 6d is only 10 point below 5diii, then my 6d must not bad as well

Basically, you were saying all current camera having same performance is this regard. I disagree, but I agree we should not take this number as sole indication but a good reference.
I believe 5 point between RX1 and RX1R is
...Show more →

I had no idea a 6D had such a low score. I don't know what's "magic" about 20, that was just used by theSuede as an example (probably to make the point that a huge delta is required for any sort of correlation to exist between SMI and color accuracy). So, in those examples between a 6D and A900, it probably would matter given the huge delta which is close to 20. I would still question that the 5 points alone between the RX1 and RX1r is meaningful without other information.
Oct 31, 2013 at 05:27 PM
douglasf13
Offline
Details
p.3 #17Sony A7R sensor gets high scores from DxOMark
Tariq Gibran wrote:
zhangyue wrote
Tariq Gibran wrote
zhangyue wroteNo, I won't take SMI score as a bible. I sure will take this a grain of salt. but seems the quote you referred imply that score is not irrelevant.

But a camera with SMI80 will almost certainly be a lot better than the camera with SMI60."

It is a good indication. and coincidentally, the camera having high SMI score also happen to have a good reputation of color reproduction: A900, 1DsIII etc.. and camera has low SMI score also happen to have many complain, ignore the list here

Conceptually, this isn't that hard to understand as once the signal is captured in raw, whatever profile you choose just have to boost different color channel to target color, the more gain you have to applied, the harder to achieve target color. Especially for something involve very subtle tonality change such as skin tone....

The same terminology can also be used in audio world, that mixer (boost/attenuate different frequency signal) can never achieve the sound as good as the time of recording it right.

If I have to choose RX1 now, I will choose regular version just based on this 'huge' SMI score difference than little sharpness gain.
Yes, but which cameras are we discussing which have a SMI difference of 20 points? That is where we know the SMI matters according to theSuede. If we are talking only 5 points, he specifically makes the case that we don't know if the camera with the higher score actually has more accurate color than the one with the lower score or vice versa. Thus, SMI is practically irrelevant because we are not comparing cameras with a huge SMI delta..at least that's what I take from what theSuede wrote - SMI can't reliably be used as an accurate indicator or color accuracy when the delta between the cameras in question is low (which seems to be the case with almost all current cameras).
What make 20 a magic number than 5 or 10 in score.
My 6d have grand total score of 69, which 20 below a900, so it is bad, and 5diii is 79, 10 point less than a900, so it is not bad, but now my 6d is only 10 point below 5diii, then my 6d must not bad as well

Basically, you were saying all current camera having same performance is this regard. I disagree, but I agree we should not take this number as sole indication but a good reference.
I believe 5 point between RX1 and RX1R is real and difference will be detectable for sensitive eyes. If it is profile correctable? I don't know.
I had no idea a 6D had such a low score. I don't know what's "magic" about 20, that was just used by theSuede as an example (probably to make the point that a huge delta is required for any sort of correlation to exist between SMI and color accuracy). So, in those examples between a 6D and A900, it probably would matter given the huge delta which is close to 20. I would still question that the 5 points alone between the RX1 and RX1r is meaningful without other information.


Agreed. I doubt anyone would be able to tell a 5 point difference in SMI. Things like raw converter profile differences would likely be much more obvious. That being said, 20 points could be a different story.
Oct 31, 2013 at 05:32 PM
theSuede
Offline
Details
p.3 #18Sony A7R sensor gets high scores from DxOMark
zhangyue wrote:
What make 20 a magic number than 5 or 10 in score.
My 6d have grand total score of 69, which 20 below a900, so it is bad, and 5diii is 79, 10 point less than a900, so it is not bad, but now my 6d is only 10 point below 5diii, then my 6d must not bad as well

Basically, you were saying all current camera having same performance is this regard. I disagree, but I agree we should not take this number as sole indication but a good reference.
I believe 5 point between RX1 and RX1R is
...Show more →

The magic is in "noticeable difference". I'll short that to JND from here on to abbreviate.

Your sample is a regression sample - if 90 isn't a JND from 80, and 80 isn't a JND from 70, and 70 isn't a JND from 60, and 60 isn't a JND from 50 - does that mean that 50 isn't a JND from 90 No, not really...

The thing with color is that it isn't something you can put "one number" on. That's almost impossible in theory, and definitely impossible in practice.

Two cameras can have the exact SAME score, and behave quite differently.
One may be correctable by a good profile, the other maybe not.
One may have all errors in gree-blue, the other in green-red - that'll give very different results in an image, but not in the score.

But in the metric that DxO uses, 20 is a definite "this camera will not be able to give as accurate color, even with a very good camera profile" limit.

Second that about the Sony A900. It has the most correctable [by profiling] and most pleasing [just my personal opinion] color rendition of all cameras I've tried/owned (I had the 850).
Color isn't just about hue and saturation! A color rendition is also what brightness value a color is assigned, and the balance between those three are wildly different even with almost otherwise identical sensors - as users of many similar cameras know. Even with an almost perfect camera profile, there's still some small differences left. Interactions between CFA and raw converter, interactions between noise factors before/after a profile has been applied and so on.

But a good camera profile, in good quality light (not cheap office fluorescents) will make most cameras almost unseparable by "only" doing visual inspection, for an untrained eye.
Oct 31, 2013 at 05:46 PM
philip_pj
Offline
Details
p.3 #19Sony A7R sensor gets high scores from DxOMark
For nature, buildings and general subject matter I think the a99 and RX1 colour is significantly 'better' - my criteria are cleaner, more free from artifacts/casts, colour should change smoothly in concert with tone changes, and very clean neutral tones - than a900 output. Also they drop in much easier without much WB adjustment.

I went back and tried some a900 files with a later vsn of ACR and no, it is the cameras that improved, esp in DR. It could be an artifact of my unusual subject matter, or simply preference. I do know what people mean and a900 at base ISO is excellent, but overall in diverse use and through the ISO range it's not even close re file cleanness and reliability for colour handling.

Remember all those 200/320 ISO chats, the independent developer community seemed convinced you needed ISO 320 to get rid of red blotches? I drank that koolaid one time and really paid the price as the a900 falls away in DR and colour integrity fast with increasing ISO, several stops in fact, with a corresponding noise ramp up. The noise character was very poor also, the latest Sonys retain very nice colour within the onset of noise very nicely. And why were the shadow blotches there in the first place? No idea. Nikon did better with D3X by all accounts of others.

I also like the a7r output so far - one jpeg of a woman with a camera at a window front in particular, with very special control of shadows and colour transitioning, from the western US release. I think you will find few people prefer the older Sony FF cameras overall in these respects, and the colour will be regarded as a highlight of the a7/a7r.
Oct 31, 2013 at 05:58 PM
douglasf13
Offline
Details
p.3 #20Sony A7R sensor gets high scores from DxOMark
philip_pj wrote:
Remember all those 200/320 ISO chats, the independent developer community seemed convinced you needed ISO 320 to get rid of red blotches? I drank that koolaid one time and really paid the price as the a900 falls away in DR and colour integrity fast with increasing ISO, several stops in fact, with a corresponding noise ramp up. The noise character was very poor also, the latest Sonys retain very nice colour within the onset of noise very nicely. And why were the shadow blotches there in the first place? No idea. Nikon did better with D3X by all accounts of
...Show more →

It was always the contention by Iliah and Andrey that ISO 320 improved shadow blotches while slightly reducing DR. Like many others, I did tons of tests with different converters and discovered what they were talking about. The trade off decision should have been dependent on the scene, though.
Oct 31, 2013 at 06:04 PM
FM ForumsSony ForumJoin Upload & Sell
  
 

You are not logged inLog inRegister