GH2 HDMI Discoveries

R

Ralph B

Well-known member
I've been testing an NTSC GH2 with an Intensity Pro board, and I've made some discoveries.

First, it is possible to get a clean, live output without any icons on the screen. By 'live' I mean you don't have to be recording to the SD card. The trick is to close the LCD screen and use the viewfinder. This turns off the Quick Menu icon. The only other graphic you need to get rid of is the auto focus brackets. You've got three options: 1) manual focus, 2) face detection, or 3) 23 area auto focus. Any one of these will give you a clean screen.

I successfully recorded this live signal with the Intesity card. However, I don't recommend that you actually do this. There is a serious problem. In this live mode, the GH2 clips 0-16 in the blacks. When you hit the Record button, the clipping stops and a full range signal is sent over the HDMI. Apparently, in pre-record the camera is using the digital video standard of 16-235. You definitely don't want to use this signal for recording.

Even if you're using the HDMI out for live monitoring (and not recording), you need to be aware of this problem, because what you see is NOT what you're going to get. This picture is darker and contrastier than what is going to be recorded. I know from working on sets that it's almost irresistible to make lighting decisions based on what you see on the big monitor. But if you do that with the GH2, you're going to be screwed.

There are two work arounds that I can think of. 1) Adjust your monitor VERY, VERY carefully to compensate. Although this could sort of work, you're still not seeing the information in the deep shadows. 2) When you're lighting the scene, simply press Record on the GH2. That way you'll get an accurate image out the HDMI.

Hope this helps.
 
Last edited:
Very useful information. Thanks for passing along what you learned. I noticed that the blacks did seem crushed on a monitor I was using with the camera so thanks for pointing this out. Did you have any luck recording the image when the camera itself was in record?

Norman
 
nbon said:
Very useful information. Thanks for passing along what you learned. I noticed that the blacks did seem crushed on a monitor I was using with the camera so thanks for pointing this out. Did you have any luck recording the image when the camera itself was in record?

Norman
Click to expand...

So far no problems, but I've only done short tests. Beautiful picture. No audio. 60i signal only. 24P has 3:2 pulldown added in.
 
Last edited:
Yeah, I've been playing with it, and here's my opinion on trying to record the GH2's HDMI:

Don't.

Seriously, there's pretty much no point. You can't get any sort of usable progressive out of it. You can't get 720p, because 720p becomes disabled when you plug in the HDMI cord. You can't get ETC, because ETC is disabled when you plug in the HDMI cable. You can't get a real 1080/24p because it's embedded in 60i but with the previously-documented pulldown issues, you're not going to be able to extract the footage. So that leaves 1080/60i mode. Yeah, if you want to record 1080/60i externally, you can. But you'd better have a monitor attached too, because you just may be recording displays and the focus box and everything else. And who wants to record 60i anyway?

And, final insult to injury, of course no sound is recorded either because the GH2's HDMI doesn't output sound. And you can't sync it with timecode because, of course, the GH2 doesn't have timecode...

Done enough experimenting to come to the conclusion that using the GH2's HDMI out for recording, as a practical tool, is a fail. I'm sure someone can come up with some hacked-together way to sort of kind of get something that's potentially maybe possibly useful under some certain circumstances, but ... nah. Not really. The GH2's HDMI output is not made for recording, and even when you cobble it together, it doesn't do it usefully. Use it for what it's intended for -- monitoring. It's not made for recording and it won't output what you want to record anyway.
 
That's really a shame. Lots of people were excited about it. However, looking on the bright side, at least we have HDMI monitoring which we didn't have before. And that's a welcomed change.

One more reason to spring for the AF100 if you absolutely need pristine external recording. But from what I've seen so far, think I'm going to be satisfied with the AF100's AVCHD. No matter how you cut it, between the GH2 and the AF100 things have gotten a whole, whole lot better in a very short period of time. Lovely.
 
There's another issue to the HDMI recording that kind of died in the other thread.
Aside from crippled 24P, which can be recovered, by the way, if you don't mind some manual labor or trusting avisynth to figure it out for you. Even syncing audio is no biggie if you're feeding it directly into the KiPro Mini's audio inputs while capturing HDMI.
The real killer for me is the actual picture seems to suffer from red and blue chroma smearing that looks like interlacing, but actually spreads out in both directions. And it's a patterned interlace, so probably a result of trying to upsample.
Shoot a red chip on a 45 degree angle, and the red will interlace in blocks of four, but not in a phase sense, where the red could be slid back into position. It's actually an expanded line of pixels, which I'm thinking is more of a smear as the encoder tries to deal with the HDMI signal. The HDMI footage shows chroma aliasing where the AVCHD footage has none, so something is definitely inferior in this signal, making the whole prospect of recording uncompressed 24P a moot point.

Posted some examples here: http://www.dvxuser.com/V6/showthread.php?231315-Any-GH2-news-on-the-HDMI-output-front/page27

It's a real shame.
 
You got icons on the screen during recording? I only ever get clean and I actually want some icons.
 
I'm just an amateur, but has anyone more experienced than I tried hdmi out with the new Vetura Capture (free add-on) program on the Matrox Mini? The 1080i out can be captured as 30p Prores at 140 kbs, looks very good to my eyes. No crushed blacks, no icons, and it will take a lot of stepping on. Don't rush out and buy one yet, someone with more experience is bound to have tried it.
 
Just changes display settings on the camera LCD and viewfinder not the HDMI display. This is while recording.
 
Okay, one last thing to add to my glowing review of HDMI recording (glowing as in, nuclear waste dump)...

The captured nanoflash files don't say "29.97" like you'd expect. They say "30.000". Perhaps this is the key and the secret to the "weird" pulldown pattern that folks are experiencing? Maybe the HDMI output is an actual, legitimate, 60.000 fields per second instead of the video standard of 59.94 fields per second? If so, the 3:2 pulldown wouldn't be exact, because the GH2's sensor is running at 23.976, and therefore it'd have to occasionally duplicate fields to match the faster 60.000 rate.

Seems bizarre to me, so I ran a control test, recording a clip off the AF100 -- and yes, Premiere reports that the resulting MXF file is 29.97 "frames" per second. But all the GH2 clips, Premiere reports as 30.00 "frames" per second.

The NanoFlash can record 1080i/59.94, or 1080i/60.00. So I plugged the AF100 in directly, turned all progressive off on the Nano, and the Nano reported the input source as "1080i59". I then removed the AF100 and plugged the GH2 in, and the Nano changed to "1080i60".

So, my conclusion here is that yes, the GH2 is outputting a 60.000 field per second signal. And that could explain the pulldown pattern. Six duplicated fields per second, at (presumably) even intervals, to round out the 59.94 to 60.00. And actually, if that's the case, then that means there's half a prayer that someone could figure out an automated way to extract the 59.94 out of the 60, and once they've done that, the conversion back to 23.976 should be trivially easy.
 
Last edited:
Barry, if this turns out to be the case, then you will have vaulted passed Hunter Richards in the "Cool Bearded Dudes with GH2s" category.
 
Barry_Green said:
Okay, one last thing to add to my glowing review of HDMI recording (glowing as in, nuclear waste dump)...

The captured nanoflash files don't say "29.97" like you'd expect. They say "30.000". Perhaps this is the key and the secret to the "weird" pulldown pattern that folks are experiencing? Maybe the HDMI output is an actual, legitimate, 60.000 fields per second instead of the video standard of 59.94 fields per second? If so, the 3:2 pulldown wouldn't be exact, because the GH2's sensor is running at 23.976, and therefore it'd have to occasionally duplicate fields to match the faster 60.000 rate.

Seems bizarre to me, so I ran a control test, recording a clip off the AF100 -- and yes, Premiere reports that the resulting MXF file is 29.97 "frames" per second. But all the GH2 clips, Premiere reports as 30.00 "frames" per second.

The NanoFlash can record 1080i/59.94, or 1080i/60.00. So I plugged the AF100 in directly, turned all progressive off on the Nano, and the Nano reported the input source as "1080i59". I then removed the AF100 and plugged the GH2 in, and the Nano changed to "1080i60".

So, my conclusion here is that yes, the GH2 is outputting a 60.000 field per second signal. And that could explain the pulldown pattern. Six duplicated fields per second, at (presumably) even intervals, to round out the 59.94 to 60.00. And actually, if that's the case, then that means there's half a prayer that someone could figure out an automated way to extract the 59.94 out of the 60, and once they've done that, the conversion back to 23.976 should be trivially easy.
Click to expand...

Yes, Barry, that's exactly what's going on. The signal coming out the GH2's HDMI port is 30.00P, while what's being recorded in the camera is 29.97.

I did a different sort of test.

Roll camera and roll computer, slate with a clapboard, let run for five minutes, post slate with clapboard. Bring camera and computer clips into editing program. Throw them on timeline, one above the other. Align first slate. Go to the end and see if post slate is aligned. It is not. Computer clip is 9 frames longer than camera clip. A 9 frame difference is exactly what you'd expect after running 5 minutes if one clip is running at 30 FPS and the other 29.97 FPS. It works out to one frame every 33.3 seconds.

I did this test both for 60i and 24P embedded in 60i, and the result was the same - 9 frames longer at the 5 minute mark. So, there is some consistency here!

Now it's a question of figuring out how to deal with it.
 
If it's field duplication, which I would expect, then surely some tool can be written to strip out duplicated fields? Maybe some wizard with avisynth could work something out? Strip out any duplicate fields you find, and then provide a 3:2 pulldown removal, and the result should be an original 24p file. The big question at that point, in my mind, is if the chroma will be royally screwed up. If it outputs true pure 4:2:2, as processed on the embedded file, then it should probably be okay. But if it's 4:2:0, then you're pooched six ways from Sunday and I don't think anything's gonna bail that mess out.
 
Sandbox said:
There's another issue to the HDMI recording that kind of died in the other thread.
Aside from crippled 24P, which can be recovered, by the way, if you don't mind some manual labor or trusting avisynth to figure it out for you. Even syncing audio is no biggie if you're feeding it directly into the KiPro Mini's audio inputs while capturing HDMI.
The real killer for me is the actual picture seems to suffer from red and blue chroma smearing that looks like interlacing, but actually spreads out in both directions. And it's a patterned interlace, so probably a result of trying to upsample.
Shoot a red chip on a 45 degree angle, and the red will interlace in blocks of four, but not in a phase sense, where the red could be slid back into position. It's actually an expanded line of pixels, which I'm thinking is more of a smear as the encoder tries to deal with the HDMI signal. The HDMI footage shows chroma aliasing where the AVCHD footage has none, so something is definitely inferior in this signal, making the whole prospect of recording uncompressed 24P a moot point.

Posted some examples here: http://www.dvxuser.com/V6/showthread.php?231315-Any-GH2-news-on-the-HDMI-output-front/page27

It's a real shame.
Click to expand...

Yes, I saw it, but this is pixel peeping to the extreme. It's fine to run tests like these, and I'm all for it. But, don't lose sight of the forest for the trees. The uncompressed picture is still superior because it has more real information. Look at real world footage and draw your conclusions from that.
 
A commenter, Makoto, on P Bloom's site when Phil announced GH2 had "clean" HDMI, said (about half-way down the comments page) he had spoken to Panny engineers, who'd reported that recording HDMI output wasn't possible because, "It happens because of the difference between 60 fps, which is the frame rate at which the sensor reads the light, and 59.94 fps, which is the frame rate of the HDMI output."
The fact that Makoto was a lone voice speculating that HDMI capture would be a problem off the GH2 makes him quite credible now, in my mind.

But even if that's fixed, the chroma issues are actually bigger. Or have you found a solution for that? HDMI should look better, not worse, than AVCHD.

And Ralph, if the digitized footage is longer, then it is recording at a slower frame-rate than the camera, which is interesting . . .

As to pixel peeping, there is a purpose to it for some. One doesn't record to HDMI, except for the reasons supported by pixel peeping. Cleaner edges. Real details, not fake ones added due to aberrations in the signal path. This isn't about comparing codecs, but justifying the extra expense in recording an image that is actually inferior to what's being captured on board. So, it's actually very annoying to see Panny did this. I understand wanting to protect AF100. But still annoying to mismatch sync in camera and say nothing about it. It's taken a month to figure it out, thanks ONLY to the people here at DVX User, btw. Cheers to all of you for not waiting for someone else to do it.
 
Last edited:
Sandbox said:
And Ralph, if the digitized footage is longer, then it is recording at a slower frame-rate than the camera, which is interesting . . .
Click to expand...

Incorrect - it means the frame rate is faster. Think of this - record for one second at 24P and one second at 60P. Which one has more frames ( or is "longer")? Answer: 60P. Which one has the faster frame rate? Answer: 60P.

So, the HDMI clip is 9 frames longer than the camera clip, and they're both recorded in the same amount of time. This means the HDMI signal is running at a faster frame rate than the camera. In this case, 30.00 vs 29.97.
 
Last edited:
You must log in or register to reply here.
Back
Top