DxOMark review - more DR and color at higher ISOs

* I don't think the camera can have more DR in video mode than it has in RAW stills (the only way would be for it to apply very clever NR algorithms, which I doubt such a small body will do)
* I also find dxomark numbers somewhat close to my subjective experience with Canon, Sony and Nikon cameras, in broad terms. But my measures of usable DR are even more conservative than theirs (I rated the NEX-5N at 11 stops for RAW stills, dxo said 12.7 stops).

Looking at the dxomark numbers for a7s, 5N, 5D2, and looking at my tests with pocket, 5D2, 5N, I would expect the a7s RAW stills to have at least one stop more usable DR than the blackmagics in RAW mode (which means 3 stops better than the blackmagics in prores mode). Because dxo says it's clearly better than the 5N RAW, which was clearly better than the pocket in my test. If slog2 saves all that DR in the video stream (instead of clipping it, as it happens in every other DSLR), it's a winner. I don't care if it's 15.3 stops (as Sony says), 13.2 stops (as dxomark says), or 11 stops (as I found in my test for the 5N): "better than BlackMagic" is mighty impressive.

But yes, this dxo number surely is a kick in the balls for Sony's PR.

(also: I may get a review sample from Sony at the beginning of july, so I don't have to wait for my camera to arrive before I put it to the test)
 
Samuel H said:
* I don't think the camera can have more DR in video mode than it has in RAW stills (the only way would be for it to apply very clever NR algorithms, which I doubt such a small body will do)
* I also find dxomark numbers somewhat close to my subjective experience with Canon, Sony and Nikon cameras, in broad terms. But my measures of usable DR are even more conservative than theirs (I rated the NEX-5N at 11 stops for RAW stills, dxo said 12.7 stops).

Looking at the dxomark numbers for a7s, 5N, 5D2, and looking at my tests with pocket, 5D2, 5N, I would expect the a7s RAW stills to have at least one stop more usable DR than the blackmagics in RAW mode (which means 3 stops better than the blackmagics in prores mode). Because dxo says it's clearly better than the 5N RAW, which was clearly better than the pocket in my test. If slog2 saves all that DR in the video stream (instead of clipping it, as it happens in every other DSLR), it's a winner. I don't care if it's 15.3 stops (as Sony says), 13.2 stops (as dxomark says), or 11 stops (as I found in my test for the 5N): "better than BlackMagic" is mighty impressive.

But yes, this dxo number surely is a kick in the balls for Sony's PR.

(also: I may get a review sample from Sony at the beginning of july, so I don't have to wait for my camera to arrive before I put it to the test)
Click to expand...

What if Slog2 adds DR? It appears to extend highlights.
 
Last edited:
philiplipetz said:
What if Slog2 adds DR? It appears to extend highlights.
Click to expand...

It doesn't get any better than linear raw DR before encoding. S-log is just a DR compression scheme.

4:2:0 8-bit AVC is good for about 8-9 stops without running into issues with compression artifacts.
Haven't seen enough XAVC to know anything about it though.
 
When they advertise it as 15.3 stops of DR are we sure Sony was not talking about just stills? Personally I never believed that claim. i figured it was stills they were referring to which is why 13 stops of video does not surprise me. Just sayin...
 
Ian-T said:
When they advertise it as 15.3 stops of DR are we sure Sony was not talking about just stills? Personally I never believed that claim. i figured it was stills they were referring to which is why 13 stops of video does not surprise me. Just sayin...
Click to expand...

13 real stops DR in video would blow away the competing cameras.
 
Red Epic X = 13 usable stops
Arri Alexa = 14 usable stops
Sony a7S = 13.2 usable stops (not too shabby if you ask me)
 
Ian-T said:
Agreed. I just wish we had more samples to see how it performs in a decent production setup.
Click to expand...

Hopefully getting mine at the end of the month or the first week of July. I'll shoot some real world test if there aren't enough footage by then.
 
Remember that the dxomark number is for RAW stills. I doubt we'll get 13 stops of usable DR in video mode. If it is 11 usable stops, it's super high... and already stretching the 8-bit codec pretty thinly.

I expect 14 stops of absolute DR, 9 of them usable with the internal codec, 10-11 with external recording. My current NEX-5N tops out at 7 usable stops, and the 5D2 at 6 usable stops, so I would be more than happy with that, specially if it is mantained at a broad range of ISO values (say 100-6400).
 
Last edited:
Samuel H said:
f it is 11 usable stops, it's super high... and already stretching the 8-bit codec pretty thinly.
Click to expand...

Doesn't the C300 have 12stops in Clog with 8bit codec @ 50mb/s ?
That would be my ideal DR point for a camera of this type and if Canon can do it why not Sony in Slog2 or is it just not as efficient as Clog?
 
Yes. The Canon Cinema cameras have 12 stops of DR at 850 ISO.

They also maintain the 12 stops at lower and higher ISO, shifting how the DR is allocated. That includes the $5,000 C100. Canon has a great sensor there. Hopefully they will open up more bandwidth in the next models, now that the memory and processing power are cheaper.

This is from a C100 review:

"PROS:

- 12 stops usable dynamic range, arguably 12.5 stops using an external recorder and ProresHQ (see chart photos below.) This is a ton of room to nail an exposure BTW even without a lighting package.

- sick low light performance, noticeably cleaner than Alexa or Epic at 1600ASA or higher."

http://www.coreyrobson.com/post/51764005014/canon-c100-thoughts-and-rants

They are rumored to be due for a price reduction. Right now they can do 4K with a $1,000 Atmos.

This is also interesting, from the same article:

"- spatial resolution resolves to the Nyquist limit for 1080 (can’t tell the difference between an Alexa or Canons at 1080 for sharpness.)"
 
Last edited:
^ The C100/300 cannot do 4k with an external recorder. They output 1080 downsized from three chromo channel 4K sensor readout.
 
Last edited:
Samuel H said:
* I don't think the camera can have more DR in video mode than it has in RAW stills (the only way would be for it to apply very clever NR algorithms, which I doubt such a small body will do)
* I also find dxomark numbers somewhat close to my subjective experience with Canon, Sony and Nikon cameras, in broad terms. But my measures of usable DR are even more conservative than theirs (I rated the NEX-5N at 11 stops for RAW stills, dxo said 12.7 stops).

Looking at the dxomark numbers for a7s, 5N, 5D2, and looking at my tests with pocket, 5D2, 5N, I would expect the a7s RAW stills to have at least one stop more usable DR than the blackmagics in RAW mode (which means 3 stops better than the blackmagics in prores mode). Because dxo says it's clearly better than the 5N RAW, which was clearly better than the pocket in my test. If slog2 saves all that DR in the video stream (instead of clipping it, as it happens in every other DSLR), it's a winner. I don't care if it's 15.3 stops (as Sony says), 13.2 stops (as dxomark says), or 11 stops (as I found in my test for the 5N): "better than BlackMagic" is mighty impressive.

But yes, this dxo number surely is a kick in the balls for Sony's PR.

(also: I may get a review sample from Sony at the beginning of july, so I don't have to wait for my camera to arrive before I put it to the test)
Click to expand...
that would be awesome. It would rely now more on the codec
 
I have the 36mp A7r and I only have used it to shoot stills stills so far for large 24x36 prints from daytime images under 3200 iso since I use my GH4 and VG900 for videos. The lack of internal cinema 4K like the GH4 was a showstopper for me but the A7s would make a great street still shooter since it has a silent shutter and the best low light performance.
 
Re: C100/C300 and 12 stops of DR

The thing with "usable DR" is that it is relatively subjective, that's why initially I measured "absolute DR" (if the waveform shows a difference, it counts, no matter if I can see actual detail or not). Now, when I ran my usable DR test, I found that weak codecs throw away a lot of that absolute DR, so even if it is subjective it's probably the right thing to measure.

And I probably went overboard, setting the bar a bit too high: in my test, a stop only counted if I could see the small details in a low-contrast almost-black surface (here). So, with the bar so high, I believe the C100 would have scored around 10 stops: one more than the 5D+RAW. The a7s shoudl end up relatively close to that.
 
Samuel H said:
Re: C100/C300 and 12 stops of DR

The thing with "usable DR" is that it is relatively subjective, that's why initially I measured "absolute DR" (if the waveform shows a difference, it counts, no matter if I can see actual detail or not). Now, when I ran my usable DR test, I found that weak codecs throw away a lot of that absolute DR, so even if it is subjective it's probably the right thing to measure.

And I probably went overboard, setting the bar a bit too high: in my test, a stop only counted if I could see the small details in a low-contrast almost-black surface (here). So, with the bar so high, I believe the C100 would have scored around 10 stops: one more than the 5D+RAW. The a7s shoudl end up relatively close to that.
Click to expand...

the key will be if the A7s gives the all important one more stop DR, as so many have speculated
 
I'm confused as to what the codec would have to do with dynamic range of the sensor. Whether it's 8 bit or 14 bit, isn't the DR just the number of stops of light compressed into the picture profile between 0 and 100% (or 110%)? Sure, you will have a lot less captured detail between each stop of DR at 8 bit vs 14 bit, but it's still compressing the same number of stops of light. Also, if you feel that you are losing detail in the blacks or whites due to the codec, can't you just design a profile that bumps up the floor on the blacks and has a log type profile for the whites. I know that you can increase the 0% IRE floor on the Canon Cinema cameras so that the blacks never reach below a certain percentage. In fact, both WideDR and CLOG don't reach to 0% under normal shooting. Sure, it might look terrible, but the detail is still there.
 
Actually, it matters a lot: DR is the difference in real-world brightness between the brightest souce that is not clipped (codec here is mostly irrelevant) and the darkest shadow that still shows detail in the recorded image (if the codec is bad, artifacts can kill a lot of detail). Just look at a C100 internal vs external recording comparison, and you'll see the difference. And the C100 uses a pretty good internal codec to begin with!!
 
I was gonna start my reply with "It's quite simply because", but I don't think a reply starting with that would be sufficient.
So here's a bit longer explanation: The 15.3 stops Sony are claiming are the raw sensor output. DxO measures DR in "photo mode" in order to benefit from the dynamic range of the A7s you would have to record S-Log2. Furthermore how camera manufactures measure DR differs quite a lot and as far as I know there are no standardised way of carrying out DR measurements.

One last thing; I don't think DxO measured the DR at ISO 3200 and as the base ISO for the A7s sensor is ISO 3200 - and the base ISO generally is where DR is highest - one would assume that this could have something to say.

All of this are of course speculations; I'm not a technician, tester or anything fancy. I'm just a humble film maker who enjoys reading about new cameras and testing them out.
 
You must log in or register to reply here.
Back
Top