GH2 hack HDMI output testing

Vitaliy_Kiselev

@brianluce
Why do you think that this is deliberate bug?
You must remember that 60Hz is used as camera also must output to screen or EVF.

As for HDMI, we'll move to this after encoder.

sam_stickland

@Vitality
I presume he's referring to the way the sound is 0.7 seconds out of sync, and the 4:2:2 is mucked out with to make it look like 4:2:0, the doubled frames etc.

You probably already know this, but apparently there's avisync script that can reconstruct the original 4:2:2:

http://marvelsfilm.wordpress.com/2011/04/19/panny-gh2-hdmi-recorder-bingo/

alcomposer

Hi all, not that I know that much- I have been in contact with the developer of JES Deinterlacer, after looking at some footage captured from HDMI - he was able to conclude that 24p has issues with irregular cadences + 50/60i has a jumpy bottom field (top & bottom written at same time once a second etc)...
This is without the hack- if anyone wants the numbers I can post.

@VK great news regarding HDMI schedule - whoooo hooooo!

kinoglass

It's seems so as V. pointed out, strange scripting around hdmi output. Anyway nothing is sure yet

brianluce

"Why do you think that this is deliberate bug?"

@Vitaliy, it's the conventional wisdom from forum gossip afaik, that panny deliberately limited functionality. I don't have a basis beyond that.

@sam
Yes I have read some of Ralph's avisync script tests but I didn't know it was fully working and it seemed like a lot of hoops to jump through. Also, thought Ralph had said it's a 420 signal via hdmi?

sam_stickland

@brianluce

Indeed. I've not purchased my GH2 yet but apparently the script produces a clean 4:2:2 1080p24 signal from the HDMI port :D

This seems to be the most concise getting started guide: http://colorbyjorg.wordpress.com/2011/04/20/gh2-hdmi-avisynth-no-hassle-quick-start-archive/

I would suggest we start hassling Atomos to produce a version of the Ninja that can remove the weird 7sec cadence and chroma issues and then the AVIsynth scripts will be unnecessary ;)

A sample video is here: http://marvelsfilm.wordpress.com/2011/07/15/gh2-hdmi-output-recorded-with-atomos-ninja/

alcomposer

Hi @sam_stickland, there is no reason why Atomos should fix this as the issues are nothing to do with them. The Atomos Ninja is an amazing device - with lots of potential. Atomos I have been told have no plans on fixing this from this end- they do want to release their software called 'stripper' which will de-interlace 24p from a 50i stream. Other than that the repairing of the 24p cadence issue is a very large one, because it is not following any known pattern.
This issue is best attacked from the camera- and then if it is discovered that there is nothing that can be done- (issue too large even for VK-possibly hardware related) then we should look for other solutions. Other than that the AVIsynth script does work- (as far as I am aware).

Best current outcome is that this problem can be fixed by the man himself. Other than that- the current videos are looking awesome and are really step ups from even the highest stock standard encoders! And remember this is only the beginning!

Vitaliy_Kiselev

>it's the conventional wisdom from forum gossip afaik, that panny deliberately limited functionality. I don't have a basis beyond that.

May be parts of it. May be. But most probably they can't output video at different rates to output devices.

sam_stickland

@alcomposer

Of course, I didn't mean to suggest that this is a problem with the Ninja, just that it would be a very useful product addition if the Ninja could remove the cadence while writing the file :) But perhaps it requires more processing power than the Ninja has.

alcomposer

@sam_stickland

There are some portable HD-SDI/HDMI recorders that do this already, but they work on the established pull-down (or is it up?) patterns... I think that the Atomos has lots of legs and will be a great product once they get the firmware up to scratch, (maybe they should send a large cheque to VK and a SDK... he seems to make more progress on firmware than a team of engineers)

:-D

dkitsov

The AVysynth script does not reconstruct anything from anything. It merely gives a bit of a blur to chroma channels to smooth them out. To reconstruct something it has to be there at first place.

brianluce

@ Sam
At Marvelfilm they're saying the HDMI out is legit 422 but the AVsyn converts it to 420. Whyyyyyyyy!!

EOSHD

It's just a suspicion but here are more details about the 'purposeful' cadence bug http://www.eoshd.com/content/637/how-a-purposeful-bug-prevents-24p-external-recording-on-gh2

sam_stickland

@brianluce
Hmm.. I just noticed that too in the comments. To be honest, a lot of the blog posts I'd read where saying that the HDMI output was a messed up 4:2:2, that came out looking like 4:2:0 but could be restored to 4:2:2. I dunno, I haven't thought about this too much before, but now I have my doubts. The chroma fix in the avisynth scripts sound like it's just smoothing out the red channel.

Only way to be sure would be to record the same image a bunch of different ways and compare the chroma resolution.

EOSHD

When you hit record the gamma change seems more noticeable via HDMI than in live view (AVCHD). I've also noticed in MJPEG mode the gamma change when recording starts is the least stark of all. Maybe that will give us a clue what is going on when the HDMI sensor mode is activated relative to other modes, and why there is a shift in gamma.

Vitaliy_Kiselev

@EOSHD
Again. Sensor mode changes.
And mode used for MJPEG is closer to liveview mode.

misterhu

I am also very interested in the HDMI output of the Gh2! I have the Blackmagic HDMI Capture card and can comfirm, that at the moment there is no difference between a hacked and a unhacked Gh2 in terms of HDMI output format.

The Blackmagic card only recognizess the Gh2 when the card is set to 1080/50i. Its no difference if you select the 24p mode or the 1080p patch! You can clearly see the interlaced effect in either mode!

I can also confirm, that there is a clear quality difference between the live preview mode and if you press record on the GH2. The live preview mode is nearly unusable!

I am on a Mac, so I cant use Streamparser....

I also hope very much, that someday we can use the GH2 as a "studio cam" and capture a good quality progressive picture from the HDMI output without the need of pressing record!

Thanks Vitaliy for all your great work!!!!

alcomposer

@misterhu have you been able to check the ExTele Mode in PTool.
Also- there is 10's of settings in PTool - so can you maybe list what you have been able to test. Already people have noted that in MJPEG mode outputs similar to Live view?! Can you confirm this? (not that its needed-but be good to confirm)

Cheers, and great work!

EOSHD

The gamma shift in MJPEG mode is less severe than in AVCHD that's for sure. AVCHD seems to change the live-view mode (some might say downgrade it). It is not necessarily representative of what's being recorded in AVCHD to the card actually.

lenuisible

@EOSHD (from Zacuto post regarding ETC while monitoring) :
>it was only one setting that got it to work otherwise it just gave you the prompt but not the feed. Oh and it only comes on when you hit record.

Could you specify what setting you're speaking of ?

Txs !

misterhu

@ alcomposer: ExTele Mode works in HDMI mode and I tested some different patches and bitrate up to 65mb/s but the difference between the live mode and the record mode is always the same.

I attached 3 pictures so no extra compression is added when uploading to youtube or vimeo:

1. HDMI output in live mode (you could get rid of the little status icons but I left them in to undestand immedeately which is what)
2. HDMI output in record mode
3. 24p recording from SDHC card (65mb/s patch)

As you can see when you look at it in full size, there is quite some dramatic change between the live mode and the record mode. (the HDMI output in record mode and what is captured on the card is pretty close, but not the same)

I dont really know what happens between live and record mode.... but its very hard to rely on the live mode to setup lighting, when the record mode is so different!

Also when I compare very closely the HDMI output in record mode to the file which is recorded on the card (65mb/s patch), i think the HDMI capture is a tad better...Is that possible? Do we already know what the GH2 really outputs on the HDMI in terms of compression codec?

EOSHD

Thanks @misterhu that test is very useful.

I've applied the same boost to the mids on both of these shots, the live view grab and the record mode grab, look what happens below. Anyone who thinks the live view mode is superior to the record mode is about to have reality check :)

Look at the truncation of the shadows in live view. On the first frame with the green, it looks terrible. On the record mode grab, everything is present and correct.

There's also something weird going on at the top end as well, look at the histogram in Preview below.

danyyyel

From what I understand is that we are getting full uncompressed video out of the hdmi. What is the problem and that have been solved in a certain way by this workaround is the cadence

I just hope that the hack could solve this problem because as I understand things like 422 won't be able even with the hack.

misterhu

Guys, I am not very techsavy and still pretty new to video in general, but please let me know what tests i could make so we can trace down the HDMI problem! I am very interested in getting a good quality progressive material out of the HDMI one day!

Vitaliy, do you need anything special from me? Can I be of any help?

As I said, I have the GH2 connected to a Blackmagic Decklink HD Extreme 3D card with HDMI input on a Mac.
With this setup I can only shoot with "studio lighting" and the black background...

alcomposer

Hi @misterhu

Great work!! (I hope that this helps @VK- even a small degree).

Cheers,
Al

Howdy, Stranger!

Categories

Tags in Topic

Top Posters