Epic-Alexa-F3 from Hannover tests

[Michael Olsen]

Funny how none of the Redheads complain about you not using ArriRAW for the Alexa or SLog for F-3 (which is supposed to add almost two stops of DR), LOL!.

I dont know if i qualify as a redhead or not, but...I could see a one major reason for not including ARRIRAW for the Alexa or SLog for the F3 - neither is the defacto standard of the camera, but are instead, optional extras which come with a considerable price tag. Whereas Epic is just Epic. It comes with REDCODE and it comes with HDRx.

That being said, I don't want to just turn a blind eye to the fact that other cameras exist. Qm certainly not for the sake of "winning" in some way. I would have liked to see ARRIRAW and SLog in the tests. Heck. It would be interesting see a 5D mark3 and the f65. But as will always be the case, every test has it's limitations...perhaps it is for someone else to make the HDRx-ARRIRAW-SLog test.

 

[Brad Allen]

Funny how none of the Redheads complain about you not using ArriRAW for the Alexa or SLog for F-3 (which is supposed to add almost two stops of DR), LOL!.

I was wondering about this too Peter. It would seem to me that a test would aim to test the maximum or best case performance for each of the cameras.

Also regarding the dynamic range of ProRes on Alexa, I must confess I am a little confused. How is it that standard EPIC-X (non-HDR) shots have significantly less dynamic range than Alexa when Alexa is shooting to ProRes? I thought the difference between EPIC and Alex was roughly between half and one stop?

Does the Alexa not lose as much dynamic range when shooting ProRes as we thought?

 

[M Most]

Also regarding the dynamic range of ProRes on Alexa, I must confess I am a little confused. How is it that standard EPIC-X (non-HDR) shots have significantly less dynamic range than Alexa when Alexa is shooting to ProRes? I thought the difference between EPIC and Alex was roughly between half and one stop?

Does the Alexa not lose as much dynamic range when shooting ProRes as we thought?

Dynamic range has nothing to do with the recording format. It is a function of the camera's imaging capabilities.

 

[Peter Moretti]

Michael,

I don't believe that's exactly the case. First off, RAW recording allows you to recover highlight information from non-clipped channels w/ greater chance of success. Now how much you can actually recover and how much that should be added to a camera's DR rating is certainly open to debate. But when it comes to the F3, there is another reason why DR is related to recording format, and the rationale is hardly subjective.

Simply put, XDCAM-EX does not record the full DR of the sensor. When recording XDCAM-EX, Sony clips the image more than a stop below the sensor's actual clipping point. Sony claims it has to do this b/c trying to fit the sensor's entire DR range on an 8-bit scale would cause banding (and maybe b/c they want to sell more expensive cameras and options). S-Log, on the other hand uses a 10-bit Log scale, so it can handle the full DR of the sensor. Sony claims that S-Log recording results in about a two stop gain of actual highlight information.

I'm not regurgitating some Sony marketing copy. I've heard that this is indeed the case from people who know Sony cameras quite well, and I've seen test that--FWICT--bear this out. So I tend to believe it.

 

[Uli Plank]

Yes, that's correct. While Arri can squeeze their excellent DR into ProRes 4444 (12 bit) or even HQ (10 bit) pretty well, you can't squeeze the full DR of any good, up-to-date sensor into 8 bits. Remember, 10 bit is not 25 percent more than 8 bit, but 400%.

So, that ain't a dirty trick by Sony, but they should rather sell their F3 with S-Log as a standard and leave the 8 bit to the FS100.

The F3 is a pretty nice camera, Sony will probably sell a lot of these until the budget cameras from RED are coming.

 

[Patrick Tresch]

...until the budget cameras from RED are coming.

Eyes wide open here...

Pat

 

[Brad Allen]

Dynamic range has nothing to do with the recording format. It is a function of the camera's imaging capabilities.

Thanks for the clarification Mike.

In the back of mind I always understood this, but was regurgitating what I though I had read on forums (I misread Peters post). I thought that perhaps there was some limitation of the ProRes codec (or on Arri's implementation of it) that prevented the Alexa from translating its full DR.

So just to clarify then the only limitation is a resolution hit? (plus missing out on all the flexibility that comes from a RAW workflow )

Thanks again Mike - appreciated

 

[Casey Green]

Wow, there is a new king of the hill... I am so impressed by the RED EPIC's dynamic range, color fidelity, and overall "film look". Wow, just... wow.

 

[Robert Berger]

Geoff, you can tell me now, i can keep a secret......
your 3D LA student dec 2010 at RED studio's best regards Robert

 

[geoffboyle]

I think a year from now it will be far easier to get great results from HDRx using tools from RED and 3rd parties.

No question, we're just barely scraping the surface here at the moment.

It'll be interesting to see what happens with Arri's built in HDR and RED's effectively external HDR.

Whilst I expect many upgrades and improvements from Arri I suspect that the RED approach will be much easier to expand in the long run.

 

[geoffboyle]

For various reasons ArriRaw wasn't possible and anyway we were testing dynamic range not resolution here and I doubt that this would change significantly.

As for Slog, well Sony told us it was SLog...

We tried and I'm going to remove the F3 material as "suspect"

 

[Pawel Achtel]

...OK, now you can start telling me how I've faked the results and all the usual stuff.

OK, so how did you fake the results, Geoff?

Good stuff, thanks for sharing.

 

[David Battistella]

WHen I tested teh Alexa you could only shoot 2K with ARRIRAW. the SXS cards were limited to 1920x1080. Is it still this way?

We ended up recording 422 to a Kipro for our tests because the SXS cards weren't working yet.

David

 

[Patrick Tresch]

...

Whilst I expect many upgrades and improvements from Arri I suspect that the RED approach will be much easier to expand in the long run.

I've heard Arri is also working on a HDR development similar to RED's with second track grab.

Pat

 

[Martin Stevens]

I've heard Arri is also working on a HDR development similar to RED's with second track grab.

Pat

I assume Red would have a patent on this. ???
Seeing that this is a sensor design issue.

 

[M Most]

WHen I tested teh Alexa you could only shoot 2K with ARRIRAW. the SXS cards were limited to 1920x1080. Is it still this way?

We ended up recording 422 to a Kipro for our tests because the SXS cards weren't working yet.

David

The Alexa sensor's RAW format is 2880x1620. That is what most industry types would call closer to 3K than 2K. You can debayer that to its full pixel resolution if you want, just like you can make 4K images from the Red One's 4K sensor, but Arri recommends debayering to 2K for the best results.

 

[M Most]

I assume Red would have a patent on this. ???
Seeing that this is a sensor design issue.

Arri has been using a double flash technique on its ArriScan film scanner for a number of years. The Alexa sensor technology is very similar to the imaging technology in the scanner.

Arri has lots of patents, too. :sneaky2:

 

[M Most]

So just to clarify then the only limitation is a resolution hit? (plus missing out on all the flexibility that comes from a RAW workflow )

That, and as Peter pointed out, possible quantizing issues with low bit depths. In actual practice, the 10 bit Prores 4444 format has been shown to be robust enough, and with enough available levels, to handle Alexa's dynamic range without any real banding issues. Which shouldn't surprise anyone who's dealt with 10 bit film scans over the last 20 years or so.

 

[Emmanuel Decarpentrie]

In actual practice, the 10 bit Prores 4444 format has been shown to be robust enough, and with enough available levels, to handle Alexa's dynamic range without any real banding issues.

I thought the Prores 4444 codec was 12 bits (or at the very least "up to 12 bits")... Why would Arri try to squeeze their dynamic range into 10 bits when the codec they chose supports 12 bits?

 

[Patrick Tresch]

Never heard of Prores 444 being 12 bit... but a new codec development could make it very easy... just implement Prores 444 12 bit...

Pat