James Miller adds £167 hardware 4K scaler box to Sony A7S with stunning results

[Andrew Reid]

Ah ok. So we need to see a comparison of straight to Ninja vs scaler+Ninja?

That is exactly what is shown in the article.

[Michael Tiedtke]

So does anyone know the dynamic range of the A7s in video mode? DxO Mark rated the sensor at 13.2 stops for stills, but with S-Log2 will you be able to use all those 13 stops in video? 

[Luke Mason]

The perceived sharpness comes from inferior scaling algorithm. My guess is that the scaler box uses "Nearest Neighbour" and the resulting aliasing is severe. It's fast though so it can be done on the fly.

 

Regarding dynamic range, DXO Mark rated 10.5 stops at ISO3200, which is the base ISO for S-LOG2.

[Guest d5f8611fa423d0e628c016f9d5c93b47]

That is exactly what is shown in the article.

Oh ok, that explains my confusion. I thought it was scaler+Ninja vs standard in-camera (no Ninja). 

[Markastellor]

It seems that this could really change one's workflow. Recording in enhanced 1080 would save an enormous amount of hard drive space and in some ways simplify things.

 

The GH4 essentially does this in camera, but of course you lose the low light capabilities of the A7s.  I wish I could see more tests.  

 

I'm also wondering if the Atomos Shogun or any similar recorder could do this in a single unit.  Could it take a 4K input and downscale to 1080 and record.  If this would increase the detail while keeping file sizes low, it would be useful.

[levisdavis]

Andrew:

 

I read your remark about the A7S / GH4's internal downscaling... Could you please help clarify? I've been wrapping my mind around this particular downscaled shooting scenario quite a lot and am wondering if it's actually effective with the GH4.

 

For example:

 

I've shot some test footage with the GH4 and the BlackmagicDesign Hyperdeck Shuttle 2... I was shooting in 4096 x 2160 at 10-bit output through the HDMI at 1080p... In theory, this produces the same results as the additional downconverter box, right?

 

Of course, I'm totally taking your previous comment into consideration... But, I truly wonder if it would be much of an advantage with the GH4 having 10bit output turned on.

 

10bit vs 8 bit is about 4 times the information.... 8bit UHD vs 8bit FullHD is about 4 times the information... In essence, this provides an external recorder with a true 10bit 422 recording. I think.

 

I've also read on your site that UHD 4:2:0 is essentially close to a 10bit 4:4:4... Which also goes to prove why the downconverter chained to a 10bit 4:2:2 full-HD recorder is such an awesome tool for the Sony A7S. But what about the GH4 with the native 10bit 4:2:2?

 

Mathematically, there is a possibility of gaining a 400% image increase in Full HD with the Sony A7S... While there would be a potential of what??? with the GH4? Maybe 110% to 150% max...?

 

At best, we would have a UHD 10bit 4:2:2 to FullHD 12bit 4:2:2 conversion factor... However the external recorder already made it's match at 10bit 4:2:2 in FullHD when shooting at 3840x2160 or 4096x2160.

 

Anyway, what I recorded to the BM Hyperdeck 2 produced footage that appeared highly detailed for 1080p, little to no noise floor, great sharpness, no moire or alliasing, and also featured the 1.9:1 aspect ratio embedded in the 1.78 image area from shooting 4096x2160. The results produced a great key, however, the native 4k 4:2:0 keyed better... Totally did not expect that.

 

All said and done, I suppose this might be a good test factor for the A7S and GH4's ongoing cost factors...

 

Will the GH4 save shooters from the external downconverter option by providing FullHD 10bit 4:2:2 natively?

 

Great discovery by the way! Easily the cleanest 1080p footage to date! WOW!

[hans87]

just a theory:

 

it would be nice to program this scaler box to crop full frame 4k image into 4x3 s35 and then scale it to 2.39 image.

this should be cheapest way to get true cinemascope hd image with anamorphic lenses.

or just wait atomos shogun and do this in computer.

[Guest Ebrahim Saadawi]

Can the 1080p recorders accept a 4k signal, and record it in 1080p? That would be very neat if they did!

Anyone tried connecting a GH4 to a ninja ( whilst in 4K mode without enabling the "Down convert to 1080p" option) ? Does it not accept the signal at all ?

Having the external scaler between the recorder and the camera is kind of a pain and I think if recorders did the downscaling it would be a great option to save time, money instead of downscaling 4k in post, while still maintaining highest 1080p quality possible with 10 bit, 4:2:2 and 1:1 sensor readout. 

[Chris Santucci]

Does anyone else think the color science in this camera is still complete crap? I haven't seen anything to make me think otherwise. 

 

 

No offense to the authors, but these still frames are horrendous from the standpoint of color. "Stunning results?" Really?

 

Here, I fixed the dog image for ya:

 

[hmcindie]

 

Regarding dynamic range, DXO Mark rated 10.5 stops at ISO3200, which is the base ISO for S-LOG2.

 

The base ISO is measured to be ISO 3200 (and the camera reports that) but it's the same as ISO 100 in photography mode. Same thing as the FS700 has a way higher base iso at S-log. It's just a measuring way, why are people so confused?

 

And DXOMark probably measured the camera all wrong with an old pre-release firmware. But DXOMark sucks anyway, why do people keep bringing them up? You guys think a bmpcc would measure well with the DXOMark tests? It would maybe measure 10 stops.

[SleepyWill]

No offense to the authors, but these still frames are horrendous from the standpoint of color. "Stunning results?" Really?

 

Here, I fixed the dog image for ya:

 

Ohhh... proof positive that colour is in the eye of the beholder (and their monitor)

 

Yours, on my monitor and to my eye looks over saturated and like you've put it through a tv's "cinema" mode. I find the original (and I share some peoples concerns about the A7s colour rendering) much more natural, much more pleasing and far less Michael Bay.

 

But please don't take that as criticism, I'm almost certain if I looked at it on your, doubtlessly scrupulously calibrated monitor (as is mine) I would share your opinion, it's just the difference in rendering and taste.

[sunyata]

I'm confused. Why can't you just output 4K straight to the Ninja and record 1080, as with the GH4? Why does the A7s require the scaler and the GH4 doesn't?

 

I don't think the Ninja Blade has an internal scaler. In a previous post about exporting 10bit from the GH4, the test was outputting a 10bit signal via HDMI scaled 4k to HD 1080 before it went to the Blade, which was recording at 1080 10bit. I also noticed, doing a quick luma test, that it looks like the GH4 is undersampling, not exactly horizontal line-skipping but something similar. 

 

So does anyone know the dynamic range of the A7s in video mode? DxO Mark rated the sensor at 13.2 stops for stills, but with S-Log2 will you be able to use all those 13 stops in video? 

 

Assuming you can get 13 stops in video mode, "usability" would be limited by the 8bit 4:2:0 color space.

 

Can the 1080p recorders accept a 4k signal, and record it in 1080p? That would be very neat if they did!

 

I was told that the Ninja Blade didn't have a scaler, so you couldn't give it 4k.

[Andrew Reid]

Ohhh... proof positive that colour is in the eye of the beholder (and their monitor)

 

Indeed! I am wondering how wildly different people's display calibrations are.

[hempo22]

Indeed! I am wondering how wildly different people's display calibrations are.

 

I don't want to know... I go crazy every time I see my parents' TV (all blue and motion plus +1000) or my own grading on someone else's laptop.

[blafarm]

Indeed! I am wondering how wildly different people's display calibrations are.

 

I would wager that some of you are on PC and others are on Mac.

 

I will share that on a calibrated PC monitor, the OP's image appears washed-out (elevated gamma).

 

I have quietly thought this since the inception of this thread -- but it is now worth pointing it out. 

[Andrew Reid]

Yes PC displays have different gamma setting to the Mac default, but they should read the gamma from the file and show it correctly, having said that this is Windows we're talking about...  :wacko:

 

I much prefer James' original grade to the one posted later by someone on the forum, the elevated blacks are really silky and film-like.

[blafarm]

The gamma issue between Mac and PC has been a problem in the film industry for decades -- especially in terms of the QuickTime 7 system framework, which has been deprecated.

 

Unfortunately, media applications on these two disparate computing platforms *do not* read the file-based gamma data and compensate accordingly.

 

And aside from file-based gamma differences, you also have issues surrounding the calibration (or not) of the user's display.

 

Color grading is very definitely a subjective art, but I can tell you that I would never let the image at the top of this thread out my door -- unless of course, it was an intentional stylistic decision.

 

On a somewhat related topic, it is interesting to note the number of comments on the Internet questioning the color science of the A7s.  Interestingly, when that color science has been called into question, I have found myself in general agreement after watching the clips on my calibrated monitor.

 

However, it's hard to know if these issues are related to computing platform gamma differences -- or users posting S-Log2 footage on Vimeo (or an alternate picture profile) -- without having pulled it back down to Rec. 709 or Rec. 2020. 

[Markastellor]

I would love to see more tests or information on this.  Does anyone have the shineybow scaler?  http://www.bhphotovi...stribution.html   It appears to do the same thing for a lower price.  I'm debating between the A7s and the GH4 right now and the idea of downscaling while recording with the A7s will effect my decision.  Could change my work flow.

 

I know nothing about the technology behind downscaling but I'm wondering if there are some custom chips specifically for the process that might be used for most equipment and the cameras that do it.  I also wonder if Blackmagic Designs has something along these lines.  Sounds like something right up their alley.

[Michael1]

The gamma issue between Mac and PC has been a problem in the film industry for decades -- especially in terms of the QuickTime 7 system framework, which has been deprecated.

 

Unfortunately, media applications on these two disparate computing platforms *do not* read the file-based gamma data and compensate accordingly.

 

And aside from file-based gamma differences, you also have issues surrounding the calibration (or not) of the user's display.

 

Color grading is very definitely a subjective art, but I can tell you that I would never let the image at the top of this thread out my door -- unless of course, it was an intentional stylistic decision.

 

On a somewhat related topic, it is interesting to note the number of comments on the Internet questioning the color science of the A7s.  Interestingly, when that color science has been called into question, I have found myself in general agreement after watching the clips on my calibrated monitor.

 

However, it's hard to know if these issues are related to computing platform gamma differences -- or users posting S-Log2 footage on Vimeo (or an alternate picture profile) -- without having pulled it back down to Rec. 709 or Rec. 2020. 

Does any camera support Rec. 2020 yet?  That's a much wider color gamut.

 

Michael

[dahlfors]

Since the release of Snow Leopard, Macs default to gamma 2.2, just as Windows.