A7s first impressions

[michael]

Bart,

Sony uses Engineering Dynamic Range. DxO uses 1:1 S/N. So, not directly comparable.

The important point, to my view, is not the absolute measured number under either schema, by rather the comparative numbers.

DxO now has S/N analysis for some 270 cameras of all brands. These are self consistent and done by one lab presumably using the same methodology.

On that basis the A7s simply is not what many people hoped it would be. That's really what the whole debate is about.

Michael

[ErikKaffehr]

Hi Michael,

The definition of egineering DR is SNR=1.

I would agree that the DXO data does indicate that the A7s is a bit weak in DR at base ISO. On the other hand it seems to excel at high ISO. It is more like Canon than Nikon/Sony/Pentax in that sense, somewhat fishy in my view.

It may be that the A7s is more oriented towards high ISO and motion than low ISO and stills.

Let's put it this way, nothing in the DxO data makes me long for an A7s.

Best regards
Erik

Bart,

Sony uses Engineering Dynamic Range. DxO uses 1:1 S/N. So, not directly comparable.

The important point, to my view, is not the absolute measured number under either schema, by rather the comparative numbers.

DxO now has S/N analysis for some 270 cameras of all brands. These are self consistent and done by one lab presumably using the same methodology.

On that basis the A7s simply is not what many people hoped it would be. That's really what the whole debate is about.

Michael

[dreed]

Hi Michael,

The definition of egineering DR is SNR=1.

I would agree that the DXO data does indicate that the A7s is a bit weak in DR at base ISO. On the other hand it seems to excel at high ISO. It is more like Canon than Nikon/Sony/Pentax in that sense, somewhat fishy in my view.

It may be that the A7s is more oriented towards high ISO and motion than low ISO and stills.

And thus the sensor/camera is built differently - rather than have an on-sensor ADC (like it is with the 36MP sensor), the ADC is offboard like it is with Canon's DSLRs?

[ErikKaffehr]

Hi,

I don't have the slightest idea. The DR at base ISO seems a bit bad for a fat pixel sensor. I guess Sony still has on chip ADCs but they play with on chip noise reduction.

Possibly, the chip is optimised for fast readout (for video), giving up a bit on maximum DR?

Best regards
Erik

And thus the sensor/camera is built differently - rather than have an on-sensor ADC (like it is with the 36MP sensor), the ADC is offboard like it is with Canon's DSLRs?

[Vladimirovich]

DR is hard to measure, mostly because of lens flare affecting the results.

as far as I remember DxO photos illustrating their process they do not use lens during that test - they do illuminate sensor in camera directly and they do not measure sensor DR - they measure "camera's" DR, so whatever reflections are in chamber housing a sensor are not supposed to be totally absent.

[Vladimirovich]

And thus the sensor/camera is built differently - rather than have an on-sensor ADC (like it is with the 36MP sensor), the ADC is offboard like it is with Canon's DSLRs?

ADCs can still be on board but might be designed in a different way vs regular Sony tech

[Vladimirovich]

Possibly, the chip is optimised for fast readout (for video), giving up a bit on maximum DR?

fast readout for video (and electronic shutter for stills) is achieved by switching ADCs in 10bit mode

[Ben Rubinstein]

What would be interesting is to see what DR Sony claimed for the A7r and then compare to their claim for the A7s. With whatever methodology they use the ratio should be similar should it not?

[Jim Kasson]

ADCs can still be on board but might be designed in a different way vs regular Sony tech

Not only the ADC could be different. Take a look at the white paper pointed to by this link:

http://www.aptina.com/products/technology/DR-Pix_WhitePaper.pdf

Unfortunately -- at least for my perspective -- it seems that it was written by a product manager, not an engineer, but it does offer two interesting -- again, at least for me -- insights.

The first is the separation of dynamic range into Intra-scene DR and Inter-scene DR. I've been working with ISO-less sensors almost exclusivly for the last couple of years, so the distinction is a little jarring to me, but those of you who work with Canons will be able to get your head around it more easily.

The second is the idea that the operation of the sensor ship could be configured with the ISO knob in ways other than changing the pre-ADC gain and the ADC resolution. This latter notion may explain some of the a7S performance, because it appears to me that the inter-scene DR can't be explained easily by read noise and the intra-scene DR.

I have an a7S on order and will test and report when it arrives.

Jim

[dbateman]

Micheal, You have a great article.  And I respect you.

The think I want to comment on is DXO is not showing us really what the pixel peepers want to see in the summary table which really doesn't reflect their data.  If you do a comparison of A7s, A7r and Olympus EM1 and look at the actual graphs you will see on the default print setting the A7r is above in DR.  But this is taking a 36Mp sensor in shrinking it "magically" to 8Mpixels for a print.  Cropers like me and the pixel peepers (your peanut gallery) like to click the Screen option for all the graphs.  Now not surprisingly the A7r is the same as the Olympus EM1 and the A7s actually is the best.  And by the best by a large margin.  So if you don't print a 8Mpixel or you display your stuff on high resolution screens or you crop you photo to get the correct frame.  Then you would see the A7s is very good. 
But you are fully correct!  The A7s does NOT have 15 stops DR is more like 13 and equal to the A7r at base ISO.  But at 800 ISo the DR is 11.7 for A7s and 10.7 for A7r, and the A7r just matches the EM1!  Amazing the DR at 25600 is 8.75 for the A7s, where as its 6.4 for both the EM1 and A7r!
So click the screen option to see the real results!
All the best,
David.

[barryfitzgerald]

Evidently DxO have not been able to measure the DR of Fuji sensors in the past (from what I gather)
What counts more isn't the numbers, but real world shooting.

Charts bore me..real stuff is far more revealing

[ErikKaffehr]

Hi,

Screen is pixel peeping and print corresponds to print, simple as that. But if Sony says 15.3 EV without qualification, it is a marketing trueth AKA known as a lie.

Best regards
Erik

Micheal, You have a great article.  And I respect you.

The think I want to comment on is DXO is not showing us really what the pixel peepers want to see in the summary table which really doesn't reflect their data.  If you do a comparison of A7s, A7r and Olympus EM1 and look at the actual graphs you will see on the default print setting the A7r is above in DR.  But this is taking a 36Mp sensor in shrinking it "magically" to 8Mpixels for a print.  Cropers like me and the pixel peepers (your peanut gallery) like to click the Screen option for all the graphs.  Now not surprisingly the A7r is the same as the Olympus EM1 and the A7s actually is the best.  And by the best by a large margin.  So if you don't print a 8Mpixel or you display your stuff on high resolution screens or you crop you photo to get the correct frame.  Then you would see the A7s is very good. 
But you are fully correct!  The A7s does NOT have 15 stops DR is more like 13 and equal to the A7r at base ISO.  But at 800 ISo the DR is 11.7 for A7s and 10.7 for A7r, and the A7r just matches the EM1!  Amazing the DR at 25600 is 8.75 for the A7s, where as its 6.4 for both the EM1 and A7r!
So click the screen option to see the real results!
All the best,
David.

[fdisilvestro]

But if Sony says 15.3 EV without qualification, it is a marketing trueth AKA known as a lie.

Nothing new, Epson has always claimed their V700 scanners capable of 6400 dpi resolution and I have never seen any test that reached even half of it.

I think that the high ISO capability of the A7s is impressive, at least 1 stop above the closest competitors.

My take:
Marketing: how much better is this camera?
Engineering: 1 stop better at high ISO
Marketing thought: if previous cameras had 14 stops DR then this must be at least 15 
 
Jokes aside, if you extrapolate the curve from high ISO (as if it were a ISO-less sensor) this camera could actually reach easily above 15 stops DR

[Vladimirovich]

Evidently DxO have not been able to measure the DR of Fuji sensors in the past (from what I gather)

I 'd not use the wording "have not been able to measure", they simply do not want to invest time/effort yet (may be they hope that Fuji will come to senses and drop x-trans marketing trick when moves to 24mp)

[Vladimirovich]

But if Sony says 15.3 EV without qualification, it is a marketing trueth AKA known as a lie.

and ".3" instead of ".99" shall make that more technically sound 

[dreed]

Jokes aside, if you extrapolate the curve from high ISO (as if it were a ISO-less sensor) this camera could actually reach easily above 15 stops DR

You'd need a raw file that was capturing 16bits per colour channel to do DR of 15.3.

[fdisilvestro]

You'd need a raw file that was capturing 16bits per colour channel to do DR of 15.3.

Only if you use linear encoding. Sony has been using non-linear encoding in other cameras already

[bjanes]

Correct. It's the maximum possible DR that the sensor can record, it's called engineering Dynamic Range, and is universally understood throughout the industry.

The only confusion possible is about how much of that maximum range is usable to a photographer, and that is because of the shadow noise, which for different folks is still acceptable at different levels. The acceptability of a certain level of noise also depends on the amount of noise reduction one is willing or able to apply without hurting the finest detail too much.

Bart,

There is one point about the DXO dynamic range calculation as compared to engineering DR that has always confused me. A common method of determining the read noise of a sensor is to put the lens cap on and take a dark frame. Read noise and shot noise combine in quadrature and with the lens cap on there is no shot noise and the read noise is the output of the sensor (provided that the read noise is not clipped as with Nikon cameras). A SNR of one indicates that some signal is present. Does the signal refer to the input or output of the sensor?

As an example, assume that the read noise of a sensor is 10 electrons. If the photons incident on the sensel releases 10 electrons, the shot noise would be sqrt(10) or 3.16 electrons. Total noise would be sqrt(10^2 + 10) or 10.49 electrons. If one uses 10 electrons as the signal, the SNR would be 1.049, hardly different from 1.0.

Bill

[Fine_Art]

Only if you use linear encoding. Sony has been using non-linear encoding in other cameras already

Please explain what you mean by that. I thought RAW was the numerical representation of the pixel charge. Jpg was a log scaled translation into 8 bits. What do you mean by encoding in raw?

[dreed]

Please explain what you mean by that. I thought RAW was the numerical representation of the pixel charge. Jpg was a log scaled translation into 8 bits. What do you mean by encoding in raw?

It is a linear representation of the pixel charge.