10bit with card

[Michael Hodges]

Advertisement (gone after registration)

I'm doing some testing with my SL for LUTs etc.and noticed that I am consistently getting 10bit 4:2:2 signal from the camera (On Odyssey 7Q) regardless whether there is a memory card in camera or not.  I was under the impression that 10bit was not possible with card inserted.  Has something changed in firmware or is this simply an internal 8bit signal converted to 10bit for HDMI?

[digitalfx]

10bit is not possible to internal card. It’s always been possible to record 10bit from HDMI to an external recorder.

 

Whether a card is inserted or not doesn’t matter

 

 

Sent from my iPhone using Tapatalk Pro

[frame-it]

could be a conversion from 8 to 10 [for HDMI] or vice versa [for the card] internally

[Michael Hodges]

10bit is not possible to internal card. It’s always been possible to record 10bit from HDMI to an external recorder.

 

Whether a card is inserted or not doesn’t matter

 

 

Sent from my iPhone using Tapatalk Pro

 

I hope this is true.  Can you confirm?  The manual seemed to state ( and I believe there was some past discussions on this) that while the card is inserted all video, including HDMI, is processed in 8bit.  I would love for this not to be true, or not true any longer, as that was going to be a pain to work around.

[digitalfx]

Well I could be wrong in my last sentence, I no longer have the SL to test. Why do you need to leave the card in?

 

 

Sent from my iPhone using Tapatalk Pro

[Michael Hodges]

My guess would be that the processor might only be setup to do one conversion at a time.

Either raw data to 8bit stream for the card or raw data to 10bit stream for HDMI.

 

I'd love to be wrong and find it will do both simultaneously.

 

That was just my understanding.

 

The stream I'm getting out of HDMI does seem to be 10bit regardless of the card being present but I'm hesitant to trust that (unless someone can confirm) without more testing to see if it truly is 10bit without degradation.

 

Of course it would still be doing an additional conversion regardless ( if my testing is correct) to re-stream 8bit back to 10bit so this would be a very archaic setup if it were indeed the case.

 

I just need to clearly understand it's behaviors before sending into a production environment.

Edited December 30, 2017 by Michael Hodges

[scott kirkpatrick]

Advertisement (gone after registration)

Others who shoot10 bit over HDMI are careful to take the cards out to ensure 10 bit.  The manual says to do this. I don't have an external recorder.  I would suppose that the recorder is adding the 2 bits, but is it really 4:2:2?

[Michael Hodges]

Others who shoot10 bit over HDMI are careful to take the cards out to ensure 10 bit.  The manual says to do this. I don't have an external recorder.  I would suppose that the recorder is adding the 2 bits, but is it really 4:2:2?

 

It does report a 4:2:2 signal which prompted my wondering.  The Odyssey 7Q is generally pretty good with it's diagnostics so I tend to believe what it reports.

 

I'll see if I can do some color charts reference tests to see if I can discern an image quality difference.

[Irakly Shanidze]

The way to answer this question once and for all is to put the file through DaVinci Resolve and look at metadata.

 

I do not use internal SD cards while shooting into a recorder for a different reason. When the card is in, the "record" button works with a delay

[Michael Hodges]

I did some tests to try and solve this and decided to post the following images which (far from being 100% scientific) should be good reference.

 

These along with numerous other tests I did have resolved the issue as far as I'm concerned.  

 

 

While I can't trust Davinci Resolve to know whether any 10bit file actually originated as a 10bit stream (or was simply a re-recorded 8bit signal to a 10bit file) I can can use color correction tools to see if there is much color information present.

 

My conclusion is that, currently, 10bit 4:2:2 is retained via HDMI regardless of the card being in the slot.  

 

However, I'd be happy to hear any other opinions or actual technical specs to the contrary.

 

There is, however, some HDMI gamma shifts while internal recording is occurring which may be reason enough to be cautious if attempting simultanious internal/external recording.

 

The following are frame grabs of a color chart extremely corrected (contrast and saturation) to reveal gaps in the color information to give clues as to how much color information may have been retained from the original sensor data.  The blurred images are present to accentuate banding apart from file compression.  You can usually see this most in the red channels.

 

*NOTE: I realize after uploading that the jpg compression makes accurate study a little more difficult but hopefully it's still apparent to the observer.

 

Edited December 31, 2017 by Michael Hodges

[Michael Hodges]

Hopefully this highlights the issue much more clearly then my last post.  This is a detail of a color chart through various types of recording formats from the camera.  

 

1. Internal card recording.

2. HDMI signal with simultaneous internal card recording.

3. HDMI signal with internal card installed but not recording.

4. HDMI signal with no internal card in camera.

 

All external recording was done on a Odyssey 7Q+ recorder.

 

The color sampling is visually demonstrated by keying the red channel in Resolve and adjusting only those values to an extreme.  

 

I think it shows pretty clearly that the card does not need to be removed to obtain full HDMI video quality.  

However, active recording to internal media (simultaneous recording) does reduces the color sampling of the HDMI output.  

 

This degraded stream is oddly encapsulated within a 4:2:2 10bit signal.  Most likely to avoid resetting the HDMI "handshake" with an external monitor recorder each time the camera records.

 

My guess is that this "preprocessing" step (for the card recorder) prior to the HDMI data stream is the reason for the note to removing cards (to ensure no one ends up with unintendedly degrading footage by recording internally also).

 

Interesting how much additional information is lost through H.264 compression/artifacts with the internal recorder.

 

The animation should play when you enlarge the image.

 

Enjoy and I hope someone else finds this as helpful as I did.

 

[Irakly Shanidze]

What is really super annoying with HDMI recording, regardless of bit depth is that zooming a variable aperture lens results in visible jumps in brightness, whether floating ISO is on, or off. Getting rid of it in post is super hard, as the jumps are abrupt with a smooth tail. At least I haven't figured a way to key frame this thing. So, for all practical purposes, zooming can be used only with internal recording. This is clearly a firmware issue.