I saw this on the sony forum. But it is broad enough for all forums. I am reposting the link for discussion here.
The test redefines dynamic range to useable dynamic range rather than 1:1 and concludes the a7rii sony sensor is still superior in dynamic range to the 5dsr. A7r is closer above iso 400.
I like the fact that the test does not use the unusable 1:1 range of dxomark.
-------------------
Based on these measures
it looks like the dynamic range advantage of canon over iso 800 has gone away and the a7rii has 1/2 stop advantage over 5dsr.
it also looks like a7rii is better at > iso 800 than the a7r. And I have always wondered whether the comparison is linear and reflected in a more reasonable test than dxomark.
--------------------
Some interesting questions that arise to me:
1) My view is that the a7rii (best Iso landscape camera in low ISO) is still too low and needs HDR. The redefining of dynamic range to 20% snr is a better test and supports that conclusion. Eg if you need lots of dynamic range - use HDR or blending. Of course if something is moving you cannot do this, so more dynamic range is good barring all other factors but the a7rii dynamic range is still not enough for me.
2) The a7r has an apparent boosting at some ISO's. It appears that sony is doing some processing of the results to achieve their dynamic range gains above base ISO. This may be why they don't provide raw raw's because they cook them first? The question I have is - you could do the same with the raws of canon for example to match the higher dynamic range.
3) The related question is does such processing decrease something other than dynamic range (usually there is some trade off) like color, iq, ....
Scott
p.s. the purpose of this post is not to re-re-re start the debate on which camera is better. There is more to a camera than dynamic range capability. Like the ability to us my ts17 that does not exist natively in sony or nikon. I just wanted to understand what sony is up to - to see if it can be achieved by canon in post processing.
p.1 #3 · Interesting Reference Data on Dynamic Range - 5dsr
1. Agreed, all cameras benefit from bracketing. Bill's PDR threshold for the A7rII is not 20:1 - it's 2.47:1, which is very close to DxO's 1:1. The 20:1 SNR threshold for PDR is before normalizing for pixel count (see this post for details).
2. The "boosting" of the A7rII's DR is not raw cooking but instead an analog gain optimization, the same type of optimization seen on most cameras including all Canon DSLRs. The only difference is Sony implements a two-phase optimization that better balances low ISO DR with high ISO DR. Do a web search for "Aptina DR-Pix" for details on a sample implementation.
p.1 #4 · Interesting Reference Data on Dynamic Range - 5dsr
mttran wrote:
Just simply wait for sony BSI EXMOR RS version if you want more DR. Who cares what they are doing as long it fits our needs, right?
But my question is whether it could be done in post processing, to achieve the same results as the sony in canon (e.g.. boosting dynamic range at higher iso's. And whether it trades something else off.
So I care. I am committed to the canon system with 10's of thousands of dollars of lens.
p.1 #5 · Interesting Reference Data on Dynamic Range - 5dsr
snapsy wrote:
1. Agreed, all cameras benefit from bracketing. Bill's PDR threshold for the A7rII is not 20:1 - it's 2.47:1, which is very close to DxO's 1:1. The 20:1 SNR threshold for PDR is before normalizing for pixel count (see this post for details).
2. The "boosting" of the A7rII's DR is not raw cooking but instead an analog gain optimization, the same type of optimization seen on most cameras including all Canon DSLRs. The only difference is Sony implements a two-phase optimization that better balances low ISO DR with high ISO DR. Do a web search for "Aptina DR-Pix" for details on a sample implementation.
p.1 #7 · Interesting Reference Data on Dynamic Range - 5dsr
Scott Stoness wrote:
Some interesting questions that arise to me:
1) My view is that the a7rii (best Iso landscape camera in low ISO) is still too low and needs HDR. The redefining of dynamic range to 20% snr is a better test and supports that conclusion. Eg if you need lots of dynamic range - use HDR or blending. Of course if something is moving you cannot do this, so more dynamic range is good barring all other factors but the a7rii dynamic range is still not enough for me.
I've stipulated all along that the Sony sensors seem to have somewhat more dynamic range than Canon sensors. I've also said much that same thing that you say here — no current sensor has enough dynamic range to handle all possible wide dynamic range scenes, so the greater DR capability produces an incremental improvement that means the area where one might have to use a blending or HDR technique moves a bit but the need for these things won't be eliminated. (For those who think it is, check back with us when Canons continuing improvement in DR reaches Sony's current level and let us know that DR is no longer an issue... ;-)
[quote)2) The a7r has an apparent boosting at some ISO's. It appears that sony is doing some processing of the results to achieve their dynamic range gains above base ISO. This may be why they don't provide raw raw's because they cook them first? The question I have is - you could do the same with the raws of canon for example to match the higher dynamic range. ...Show more →
Several things struck me along these lines as I looked at the chart's curve representing dynamic range of a variety of cameras. (I plugged in quite a few — Sony Ar7 and Ar7II, Canon 5DsR and 1Dx (and others), all of the Fujifilm X-trans cameras, Nikon D810, both Phase One Backs, Pentax 645z, and probably others.)
Many cameras' curves have "waves" in them — that might have a period of approximately one "whole" ISO value. (E.g. - 100, 200, 400, 800, etc.) The 5DsR does not do this. Its response curve is very smooth. I'm not sure that this is especially significant, though I wonder what it means. (It does eliminate the sense that some have that certain 1/3-stop ISOs will be less noisy.)
The A7rII curve is very unusual. There is a significant measured "bump" in the dynamic range right at ISO 100 —a much larger jump away from average values right at that point. Off-hand, I can't come up with an explanation —though the answers to your third question will be interesting in this regard. It does allow Sony to claim a more significant DR difference at ISO 100. Then, as is always the case, the DR curve drops as ISO rises. However, unlike the 5DsR, there are anomalies in the curve at two points. Part way along on the increasing ISO scale, there is a singular bump in DR. Then about the same distance further along the curve there is a second similar bump. This also leads to your third question.
3) The related question is does such processing decrease something other than dynamic range (usually there is some trade off) like color, iq, ....
I think it would be interesting for someone to compare images at that very high DR ISO 100 point with images at something like ISO 200 or ISO 400 where that spike isn't happening. I also think it will be interesting for someone to compare images on either side of the other two anomalies that I mentioned above.
Of course, the bottom line is that when it comes to image quality the Sony sensor seems to do an overall very fine job. (Things are not perfect in Sony Land, as the 12-bit image issue reminds us.
Regarding the "debate" about which camera is better — both are very good cameras that will appeal to somewhat different users. Both have significant pluses and both have some minuses.
p.1 #8 · Interesting Reference Data on Dynamic Range - 5dsr
RobDickinson wrote:
Sony are not above cooking raws sadly. Have done in the past Will do in the future.
The a7s has noise reduction applied to higher ISO raws making comparison hard.
Sony's raw-based High ISO NR doesn't start until 102,400 on the A7s and 32,000 on the A7rII, which is beyond the point of usability for many situations. So for High ISO I would only compare from around 25,600 and below.
p.1 #9 · Interesting Reference Data on Dynamic Range - 5dsr
RobDickinson wrote:
Sony are not above cooking raws sadly. Have done in the past Will do in the future.
The a7s has noise reduction applied to higher ISO raws making comparison hard.
snapsy wrote:
Sony's raw-based High ISO NR doesn't start until 102,400 on the A7s and 32,000 on the A7rII, which is beyond the point of usability for many situations. So for High ISO I would only compare from around 25,600 and below.
Untrue. I have seen NR on sony a7s raws at 6400. Its obvious if you look at them.
p.1 #10 · Interesting Reference Data on Dynamic Range - 5dsr
RobDickinson wrote:
Untrue. I have seen NR on sony a7s raws at 6400. Its obvious if you look at them.
Any NR that's obvious to the naked eye would be very obvious on a noise FFT or frequency plot, whereas the inverse is not always true. This means the analytical methods will detect NR at levels not noticeable to us. And no analytical evaluation of present Sony sensors shows evidence of NR at ISOs below the ones I've indicated, at least for the normal 14-bit single-shot mode.
p.1 #12 · Interesting Reference Data on Dynamic Range - 5dsr
By the way, the notion that testing files "without any NR applied" will produce more accurate comparisons sort of falls apart with at least one test case has applied NR in-camera. In that case, a "no additional NR" test has other cameras competing with one arm tied behind their backs, so to speak.
My idea of a more useful real world test is to take the raw files and process them (for NR, sharpening, etc) in whatever ideal way you would if you were using the cameras, and then compare those results. That, after all, is how you would actually use the cameras, no matter which one your chose.
(My adjusted crop above — the much lighter one — was treated this way. I do show the version without NR or sharpening at the link, but I think the one shown here is the more useful one to consider.)
I saw some very high ISO examples from the A7rII posted in the Sony forum, and they did look quite good. As we would expect, there is noise and the DR is compressed — but considering the ISO setting the test images looked quite good and were certainly usable.
p.1 #14 · Interesting Reference Data on Dynamic Range - 5dsr
Scott Stoness wrote:
Some interesting questions that arise to me:
1) My view is that the a7rii (best Iso landscape camera in low ISO) is still too low and needs HDR. The redefining of dynamic range to 20% snr is a better test and supports that conclusion. Eg if you need lots of dynamic range - use HDR or blending. Of course if something is moving you cannot do this, so more dynamic range is good barring all other factors but the a7rii dynamic range is still not enough for me.
2) The a7r has an apparent boosting at some ISO's. It appears that sony is doing some processing of the results to achieve their dynamic range gains above base ISO. This may be why they don't provide raw raw's because they cook them first? The question I have is - you could do the same with the raws of canon for example to match the higher dynamic range.
3) The related question is does such processing decrease something other than dynamic range (usually there is some trade off) like color, iq, ....
Scott
I just wanted to understand what sony is up to - to see if it can be achieved by canon in post processing....Show more →
p.1 #16 · Interesting Reference Data on Dynamic Range - 5dsr
gdanmitchell wrote:
By the way, the notion that testing files "without any NR applied" will produce more accurate comparisons sort of falls apart with at least one test case has applied NR in-camera. In that case, a "no additional NR" test has other cameras competing with one arm tied behind their backs, so to speak.
My idea of a more useful real world test is to take the raw files and process them (for NR, sharpening, etc) in whatever ideal way you would if you were using the cameras, and then compare those results. That, after all, is how you would actually use the cameras, no matter which one your chose.
(My adjusted crop above — the much lighter one — was treated this way. I do show the version without NR or sharpening at the link, but I think the one shown here is the more useful one to consider.)
I saw some very high ISO examples from the A7rII posted in the Sony forum, and they did look quite good. As we would expect, there is noise and the DR is compressed — but considering the ISO setting the test images looked quite good and were certainly usable.
Yes, the final image is what counts...who cares what the images look like straight from the cameras. That test of the new Sony at high ISO was really quite amazing. The noise was very even without any kind of banding and with careful noise reduction applied, those ultra high ISO images would be very usable. This is one area the A7R lacked which seems to have been greatly improved on in the A7R2.
p.1 #18 · Interesting Reference Data on Dynamic Range - 5dsr
Ok, I'm well out of my level out of expertise here especially since optics was never in my university curriculum (got too fascinated with statistical analysis). So would someone (maybe in small words ) help me out here?
What I think I know is as follows:
(1) No camera (or for that matter any visual sensing device) can come anywhere near to the human eyes' ability to see both dark and light simultaneously - ignoring of course the extremes of "dark as a coal mine at midnight" and staring at the noonday sun.
(2) Variations in sensor technology do exist, and I'd guess the magnitude of variations from say an iPhone to the Hubble approach infinite.
(3) It is a standard canard that the Canon (pick a current model to beat on) is not as good as camera X. (Was the Nikon D800 for awhile, now seems to be the Sony 7rxxx.) This is a constant refrain around here. I can't really comment on the validity of this line of thought.
(4) Sony's sensors seem to be superior(?) at least to some folks and by some tests, but there's a strong suspicion that they're also throwing in some sort of "secret sauce" that produces a not "true" raw. This might not be a bad thing, and I wouldn't be surprised if all makers don't do something in their A/D transformations.
(5) Finally, it does seem clear that Canon doesn't do as well in resolving detail in the shadow at low ISO. I know Fred demonstrated this more than a year ago with a D800. Doubt that has changed much.
So to me, the question is: what is the bottom line? I nearly always shoot at least a 3 AEB with any scene that has extremes of light to dark. Judicious merging (I use Photomatix) produces richer photos without the garish overdone images that can be produced with HDR. Of course multiple exposures of moving objects is worth snot. Those of us who shoot movement in various shades of dark are forced to wider apertures, as slow a shutter as necessary to stop action, and higher ISO.
So here's the $64,000 question: does this make a difference in how the human eye perceives the photo particularly printed at a reasonable size? I know you can pixel peep to 200x and up, but come on, that really doesn't have much to do with what the photo looks like to other humans.
How serious/significant is this reported superiority of the Sony 7rxxx? Inquiring minds (well at least mine) want to know
p.1 #19 · Interesting Reference Data on Dynamic Range - 5dsr
RustyBug wrote:
Link
Sorry. I'm battling a little server issue* and the site was temporarily offline. I have fixed it and the link seems to work now — let me know if you still have problems accessing it.
Thanks,
Dan
* For the past month the site has locked up early on Saturday mornings! Time to dig into some server logs, reinstall some plugins, and more. Oh, fun! Not. ;-)
p.1 #20 · Interesting Reference Data on Dynamic Range - 5dsr
OntheRez wrote:
How serious/significant is this reported superiority of the Sony 7rxxx? Inquiring minds (well at least mine) want to know
From all the samples I have seen from the D810, 5Ds/R, and A7rII, I don't think anyone could reliably pick out what camera did them in a blind test. They are simply all very fine camera's with outstanding IQ.
) help me out here?