Difference between output from SDI-A/B vs. SDI

[skimmel]

I have tried to figure this out from the Sony specs but I just can't: When I compared output from the dual link SDI out, recorded at 4:2:2, with output from the "regular" SDI (the SDI that sits right next to the HDMI out) out to 4:2:2 (all else being equal), the difference is very noticeable.

To be clear: both recordings are at 4:2:2 and both use the same codec (ProRes HQ). I would have thought that the 4:4:4 output from the dual link would have little advantage when recording at 4:2:2, but there is, particularly with skin tones. I have now confirmed this based on 2 different, independent sets of tests.

Is there something different in the output from the dual-link SDI than the regular SDI above and beyond just color sampling?

And, if so, I wonder if the SDI-A out with the new firmware will look better than the regular SDI out.

 

[Douglas Villalba]

SDI A & B in my opinion is for external recorders and the SDI Out is for monitoring. If you have the S-log feature, you get 444 out of the dual link (A&B) or the A 3G.
I use a KiPro Mini connected to SDI A and a Panasonic BT-L2150 out of the SDI Out.
The resolution should be the same, but I haven't tested it. What is the use of monitoring if you can't see the same thing that is being recorded?
My 3¢

 

[skimmel]

Is the SDI-A output any different from the SDI-Out output?

 

[nsoltz]

Let me try to summarize:

SDI A-B
For dual link or 3G RGB 444 or RGB 444 S Log
Set to 1.5G and from the A port, you can get clean YPrPb 422 signal
With Firmware 1.3, you can get clean S-Log 422
"Clean" means you cannot display menus nor can you output a LUT

SDI port
You can display Menus
You can display LUT
Only YPrPb 422-- no RGB 444
Can output downrezed HD->SD

Using dual link OR A HD-SDI, for secondary output can choose either SDI or HDMI. Can't use both.

If dual link ports off, then can output through both SDI and HDMI if selected as such in menus.

In S-Log, SDI/HDMI os either LUT on or LUT off. The advantage of the 1.3 firmware allows a clean signal Dual Link 444 or A-port 422 to recorder and LUT output for monitoring.

Does this clarify?

Ned Soltz

 

[scocklin]

No character display from SDI-A output, a good thing when your feeding video to an external recorder.

Sorry nsoltz beat me to it!

 

[skimmel]

Thanks guys, that helps.

Is there a difference between the YPrPb 422 you get from SDI-A and the YPrPb 422- you get from SDI port in terms of quality (assuming no LUTs applied, etc)?

My main question though is that I got better results from Dual-Link recorded at 422 than I did recording from SDI Out at 422 (no LUTs); is there a reason for this? Thanks.

 

[Nate Weaver]

I've had the F3 service manual for some time, and there's circuit block diagrams in it...I've studied them for yuks. I noticed what I'll say below, but I didn't really think of the implications.

The Dual-link ports are derived from a completely different signal path in the camera...that is, they are driven by their own processor that has direct access to the sensor processor block. Whereas the monitor SDI out goes through COMPLETELY different signal path, that has a lot more going on processing wise...and likely has different design goals as well, given there's another set of onboard outputs specifically designed for recording.

It never occurred to me that they would have differing quality levels and all that, but if you saw the block diagram, it's clear how it's not only possible but even likely that the dual-link ports would stay cleaner. By the way, this could explain some differences Alister Chapman saw when recording Gemini vs. Samurai a couple months ago. Gemini would have been driven by Dual-Link A, and Samurai back then would have had to be driven by Monitor SDI out.

Good find, Skimmel! I never thought to check if recording 422 from the A port would actually look better!

I'd post a shot of the block diagram, but I have a feeling Sony should come down hard on that.

 

[Nate Weaver]

I just looked at some Samurai clips from A port. It is indeed nicer. I think you'd have to have a really good monitor (pixel for pixel) to see the difference, but I see it. Thanks for the heads up, I never would have though to really look at it twice.

Very cool.

 

[skimmel]

Wow, thanks Nate. Sounds like this could be another advantage to firmware 1.3.

 

[nsoltz]

Thanks, Nate.

Honestly didn't know that. Great contribution.

Ned

 

[skimmel]

I just looked at some Samurai clips from A port. It is indeed nicer. I think you'd have to have a really good monitor (pixel for pixel) to see the difference, but I see it. Thanks for the heads up, I never would have though to really look at it twice.

Very cool.

Nate, what specifically did you see that differed? I'm wondering if it's similar to what I observed. For example, I noted that S-Log tended to look flatter out of SDI-A and, similarly, without S-Log, seemed like files had a bit better transitions from shadows to highlights.

 

[Nate Weaver]

Actually, I think I'm seeing a tiny bit of finer details, so far. And cleaner.

The block diagram has it looking like the S-Log processing happens concurrently in two processors, as in, one creating the image in the chain that ultimately feeds the entire rest of the camera, and the Dual link ports have their own, separate image processors that get their data from very early on in the "chain". I'm guessing this is necessary because the rest of the camera simply couldn't handle what's necessary to maintain a 444 path.

So far I'd theorize that the dual link ports have their own scaling, DSP, etc etc that is free from the limitations of the rest of the camera.

 

[PaulReam]

For everyone with firmware 1.3 this is a huge discovery... thank you Nate & Skimmel

 

[skimmel]

I'd be curious to see some more testing on this. (I don't currently have an external recorder, was just renting one, so can't test more myself.)

 

[lasikdenver]

I have been using the "A" port to an external recorder (Samurai) using 1.1 prior to upgrading to 1.3 without S-Log and I am not sure that I see any difference with the 1.3 version. Maybe the "A" port now enabled with 422 S-LOG is different than the monitor port with S-LOG but I am not sure that without S-LOG there is any difference from the "A" port. Maybe it is better since I never compared this and never used the monitor port to record.

 

[skimmel]

I have been using the "A" port to an external recorder (Samurai) using 1.1 prior to upgrading to 1.3 without S-Log and I am not sure that I see any difference with the 1.3 version. Maybe the "A" port now enabled with 422 S-LOG is different than the monitor port with S-LOG but I am not sure that without S-LOG there is any difference from the "A" port. Maybe it is better since I never compared this and never used the monitor port to record.

Thanks. Is there a difference between the "A" port and the SDI-OUT port? That's the main issue I'm curious about.

 

[Doug Jensen]

I wouldn't be so fast to call this a "huge discovery". I wouldn't call it a discovery at all.

I've done my own testing today with my F3 running firmware 1.3 and my Samurai set for ProRes HQ
There is no visible difference whatsoever between the output of SDI "A" and the regular SDI port.
Even at 800% magnifcation you cannot see a split-screen moving wipe.

 

[skimmel]

Thanks Doug. Very interesting.

Now, back to my other original question: Why does an image coded as 422 in an external recorder connected to the dual-link (both A and B -- i.e., getting 444 feed) look different (to me) from an image at the same 422 and codec recorded on a recorder connected to only single link (i.e., getting 422 feed)? Problem for me is that I didn't do this with the same recorder so it could be due to the actual recorders.

Put another way: does taking a dual-link 444 feed enhance image, even if transcoded at 422 in an external recorder, over a 422 feed?

 

[Nate Weaver]

It is possible there's a placebo effect going on with my eyes.

But weeks ago, I was a little disappointed with the Samurai output. It was good, but now it does seem better. I guess I'll just have to do what Doug did to see if it's just my brain and desires playing tricks on me. I don't wish to be telling people it's better if it's truly not.

BTW, here's why I thought it was entirely likely the dual-link outputs could/would have a better image, it appears to me that the camera literally generates two separate images because it has two camera processors. One that feeds an FPGA that a lot of images have to be derived from (XDCAM, SD out, SDI out, HDMI out), and another completely different path for the dual-link...one that is single-minded, so to speak.

 

[skimmel]

In case it helps, attached are 2 screen grabs. One on the left is s-Log using 444 dual link out to Cinedeck which is encoding at ProRes Nl 422. Second is s-Log and uses the exact same set up (no change in anything except recorder) except 422 out to Pix 240 encoding at ProRes Nl.

Not sure how well this will project but hopefully you can see the difference.

I found the the 444 out to 422 has more latitude for grading.

This could, of course, be a difference in recorders. Would love to have someone test this with the same recorder (e.g. Cinedeck or Gemini out of dual-link and then out of SDI-OUT or SDI-A if you have firmare update).