GH1 Stereoscopic Rig Collaboration

[g.l]

Panasonic engineers presumably had a reason for shutting off composite output while recording.

The last theory I heard was that their processor simply isn't powerful enough to do capture work and output the image at the same time (I can't think of another good reason).

Instead I get a message on the LCD VF like "cannot use this battery".

This might actually be 3rd-party battery detection. Panny disallowed them from their v1.2 firmware onwards - the firmware hack can defeat this.

Maybe detecting voltage is part of it - what voltage does the fully charged Panny battery measure as?

It warms up and becomes very stable after about 30 minutes (that is, the oscillator frequency stops rising).

What was the difference between cold & stable?

I will only use battery power for other tests and actual use, if any, as AC adapters introduce noise that introduces randomness in startup times. Alan

Great work Alan.

 

[AlanHunter]

GH1 frequency change over warm-up

GH1 frequency change over warm-up

>>>This might actually be 3rd-party battery detection. Panny disallowed them from their v1.2 firmware onwards - the firmware hack can defeat this.

I haven't applied the hack yet - maybe that will let me use my 7.4V batteries.

>>>Maybe detecting voltage is part of it - what voltage does the fully charged Panny battery measure as?

I wrote before that the battery says 7.4 volts - it actually says 7.2 volts. It looks like a fully charged Panasonic battery puts out about 8.24 volts. The Panasonic AC adapter puts out 9.3 volts.

>>>What was the difference between cold & stable?

About 115 microseconds per minute faster when warm. That is, the change in disparity between the phases of the composite output video sync of the GH1 and a (believed) stable composite video source changes by about 115 microseconds per minute as the GH1 warms up. I don't have a frequency counter, just a cheap blackburst generator I bought on eBay for about $50. I don't know how close it is to 59.94 Hz, but I assume that it is stable, as I leave it on all the time. The drift (change in disparity) I read between the GH1 and the BBG when the GH1 was cold was 286 microseconds per minute, and that rose to about 380 per minute after 13 minutes, 390 after 18 minutes, 395 after 25 minutes, 400 after 41 minutes, 401 after 63 minutes, 402 after 70 minutes. Of course, a well-matched pair of GH1's that warm up at the same time will drift much more slowly with respect to each other.

Alan

 

[David Cole]

... the battery says 7.4 volts - it actually says 7.2 volts. It looks like a fully charged Panasonic battery puts out about 8.24 volts. The Panasonic AC adapter puts out 9.3 volts.

The user manual says,

Power Source: DC 9.3 V
Power Consumption: 2.9 W (When recording with LCD Monitor)
2.7 W (When recording with Viewfinder)
2.0 W (When playing back with LCD Monitor)
1.8 W (When playing back with Viewfinder)

 

[David Cole]

Crunchy's suggestion of holding down the "trash" button for 10 seconds works. ...

Alan:

Have you figured out how to toggle the video out back to the "off" setting? In my case, it has stuck. Also, did you observe that the blanking pulse remains on the video out line even when recording? I'm seeing this with an oscilloscope - not sure if the LM1881 will be able to detect.

Thanks for all !!

 

[AlanHunter]

Alan:

Have you figured out how to toggle the video out back to the "off" setting? In my case, it has stuck. Also, did you observe that the blanking pulse remains on the video out line even when recording? I'm seeing this with an oscilloscope - not sure if the LM1881 will be able to detect.

Thanks for all !!

I have not yet tried to reset things so that composite video is no longer output while recording. Just cycling power sure doesn't do it. Maybe there is a way to restore factory defaults? Maybe hold the "trash" button down for 10 seconds again? I'm afraid to be the first one to try anything with my new baby.

Composite sync is still output while recording. I have not tried plugging the AV cable into a TV set to see if good video is there. That would be a good experiment to do. I would expect the LM1881 to be able to read it, although I'm using the more expensive EL4581 sync separators, mostly for lower power consumption.

Alan

 

[David Cole]

I have not tried plugging the AV cable into a TV set to see if good video is there.

The video line is actually quiescent while recording, except for the pulse... no video whatsoever. Here it is on an oscilloscope:

 

[AlanHunter]

The video line is actually quiescent while recording, except for the pulse... no video whatsoever. Here it is on an oscilloscope:

Yep. It just outputs video black when recording. I just tested it by connecting it to a TV monitor. I also double-checked that the sync phase is not affected when recording starts or ends. I'm actually happy about the video black - it makes me think that the video quality may not be taking a hit of any kind from the "trash" button trick. Alan

 

[g.l]

I also double-checked that the sync phase is not affected when recording starts or ends.

Ah, so it is clocked by a global running timer. So the firmware could help simply by resetting this timer when recording starts.

If that's possible, then a dual-release may be all that's required to start in good sync. After all it works with stills, so the trigger polling frequency is probably high enough.

 

[g.l]

Ah, so it is clocked by a global running timer. So the firmware could help simply by resetting this timer when recording starts.

If that's possible, then a dual-release may be all that's required to start in good sync. After all it works with stills, so the trigger polling frequency is probably high enough.

The more I think about it, this would be the best solution - do you agree?

Of course it would still drift over long takes, so it's not genlock - but every new shot would start in sync, no external hardware or hardware hacks would be required, and you would never have to reboot.

And it only needs a single firmware mod, assuming the timer can be reset somehow.

 

[David Cole]

The more I think about it, this would be the best solution - do you agree?

A clock reset would be ideal for syncing - as the cameras could be powered up and thermally stable. This would probably achieve a more deterministic initialization of the system clock than a cold boot.

The obvious caveat is how to signal the firmware to reset simultaneously. The remote auto-focus switch (even with a fast, uniformly-serviced polling routine) might introduce some dissymmetry. Ideally, we'd find an interrupt to fire off (e.g. like cracking open the case and finding a hardware reset).

By the way - we are going to test the correlation between composite video blanking sync and actual measured stereo sync over the next few days.

 

[g.l]

A clock reset would be ideal for syncing - as the cameras could be powered up and thermally stable. This would probably achieve a more deterministic initialization of the system clock than a cold boot.

Right, another plus.

The obvious caveat is how to signal the firmware to reset simultaneously.

It makes most sense to do this right where the shutter-release cable triggers video recording, so that the timers are simultaneously reset before any capture setup code can introduce random delays.

As stills already produce good sync with a dual-release cable, this should work just as well (or better, because with stills some setup code may be getting in the way).

By the way - we are going to test the correlation between composite video blanking sync and actual measured stereo sync over the next few days.

Great.

EDIT: I slightly misread you (it's tricky as I don't have any hardware to play with). You have two switches on the remote right, one for focus (simulating the half-press of the shutter button), and one for recording? Can you directly start recording without the half-pressed state? If no, then I would do it at the half-press. If yes, you could sync at both states to catch all possibilities.

 

[David Cole]

... stills already produce good sync with a dual-release cable

Understood... but we should probably devise a solid test to make sure. The reason I'm suspicious is that there are so many conditions (e.g. histo calculation) that could effect the remote polling service priority. Also, I've seen instances where the cameras are in sync, but the frame count is off by one or two frames (for video... I've never tested stills).

You have two switches on the remote right, one for focus (simulating the half-press of the shutter button), and one for recording?

Yes... and it might be difficult to trigger reset and record from the same button press event.

 

[g.l]

Understood... but we should probably devise a solid test to make sure. The reason I'm suspicious is that there are so many conditions (e.g. histo calculation) that could effect the remote polling service priority. Also, I've seen instances where the cameras are in sync, but the frame count is off by one or two frames (for video... I've never tested stills).

Not sure how this could be tested conclusively without hacking the firmware - any ideas?

The whole frame delays you're seeing are interesting, but I think it's much more likely (as Vitaliy suggested) that this is caused by capture setup code (accessing the cards & creating the initial files + anything else). So if the timer is reset before any of that happens, the frames should be in sync (even if they're out by a few whole frames due to variable setup times).

Finding a way to reset the timer in firmware is key - once we have it, it's easy to try different places to use it.

Yes... and it might be difficult to trigger reset and record from the same button press event.

Why? The idea is to insert the reset code at the point where the release triggers capturing. I think it should work.

 

[David Cole]

Not sure how this could be tested conclusively without hacking the firmware - any ideas?

We could start by taking 50 or so stills of Peter Wimmer's sync tester pattern and verifying that we have great sync for all.

 

[g.l]

We could start by taking 50 or so stills of Peter Wimmer's sync tester pattern and verifying that we have great sync for all.

Oh you meant stills. Yeah, good plan.

 

[g.l]

We could start by taking 50 or so stills of Peter Wimmer's sync tester pattern and verifying that we have great sync for all.

I've just spoken with our Japanese friend inoue_k3D, and he confirms that the still sync stays good over every shot:

"I have no quantitative measurement environment (no CRT anymore near me) for sync, however I believe it is within several sub-milliseconds for every still shot with my special dual cable modified based on UN double release"

So my firmware solution is looking promising.

 

[David Cole]

In my first tests with Sync Shepard circuit, I have seen that the GH1s initialize nearly every time with a .01 millisecond offset and drift to .41 milliseconds of offset over 5 minutes. By way of comparison, my Sony HDR-CX550vs + LANC Shepard drift to .55 milliseconds of offset in 5 min, when they power up at .01 milliseconds.

I'll test over the weekend to see the correlation between the Sync Shepard reported sync and actual camera sync as measured with Peter Wimmer's software.

 

[HHL]

Hi Guys! I built a SUPER CHEAP beam splitter for $50. I'll be refining it (as well because it needs it), but I suspect that some people will be heading your way that are interested in 3d video. It's early, but it looks like our beam splitter will be used for HDR VIDEO first...maybe by then your efforts will have advanced even further. Shooting 3d for less than $150 would be pretty interesting.

Anyway...I figured that for you all to have MORE PEOPLE with beam splitters would be good for your efforts too. (for testing purposes) It's not the AMAZING piece of engineering that your monster is....but I will work. See you guys around.

http://www.dvxuser.com/V6/showthread.php?t=222799

 

[David Cole]

OK - first tests are done. I don't see any clear correlation between composite video sync and actual measured sync using Peter Wimmer's software. Need to do much more testing...

All times are in milliseconds.

 

[g.l]

OK - first tests are done. I don't see any clear correlation between composite video sync and actual measured sync using Peter Wimmer's software. Need to do much more testing...

All times are in milliseconds.

Hmm, if it bears out then the composite frame sync is independent of the capture sync. Seems strange - I could understand if it was slightly delayed, but still consistent. That happens with many (all?) consumer HD camcorders, as the additional processing / scaling for the composite output delays the frames slightly. But to have no correlation at all seems odd.

Did you test this by live-viewing, or actually recording?

BTW, I've spoken to Vitaliy offline and he sees zero hope in finding these timers in the firmware, at least without a lot of chipset documenation. I'm not sure he's right, if enough people with different skillsets & backgrounds stare at it long enough & correlate with the known board signals & pins, someone may figure it out (a point I made to him). But we'd need the firmware opened up to review for that.