GH1 Stereoscopic Rig Collaboration

[crunchy]

Some time ago somebody published very detailed picture of GH1's main electronic board. Was it on this forum??

Found:
http://www.henryolonga.net/MainBoard.jpg

 

[g.l]

What did you mean??

I meant that it is unlikely that the video-out timing isn't directly coupled to capture timing, because that seems the simplest and cheapest design - everything running off one master clock. ('asynchronous' is a programming term meaning 'independently, at the same time').

 

[David Cole]

And, of course, one important difference is that I am using separate switches for FOCUS and for TRIGGERING shots. Before each shot I first focus (always!!), then I trigger the shot. David, can you try it with your cameras as well?

I have focus and shutter triggers rigged as well. in my first tests, I always pressed the focus first, before I began recording. Didn't give me a reliable sync with the M4/3 lenses. Roughly 30% of the test runs had serious sync offset.

I can't retest w/ M4/3 lenses at the moment, as I no longer have them. I'll have the GH2 14-42mm kit lenses that I could use to test - as soon as my GH2s arrive.

 

[crunchy]

I'll have the GH2 14-42mm kit lenses that I could use to test - as soon as my GH2s arrive.

Great! Until then I'll let you know about the progress on "3DSLR Master". Development will be slightly delayed due to my trip to Colorado (next week), but I assume that you'll also have to wait for your lenses...
Unfortunately, due to development stage of the project, I made experiments only indoor and usually during the evening under artificial light. Maybe it has some (smaller) influence on focusing & synch.

 

[inoue_k3D]

Hi All

I am a late comer to this forum and posting for the first time, however I have been using twin GH1 rig long since just after the product release and suffering from random sync so far .
Recently I tried GH13 firmware to make usable of an external battery, Sanyo KBC-9V3J (9V, 2A) that have one switch and two DC outputs and used it to test REC sync of twin GH13. I use a special remote shutter release (based on UN Double Release) for Rec start, which is almost perfect for the photo sync.

At SH mode I found perfect sync (by sound) can be achieved in 720p60 sequence of Premiere CS5. In some case I need one frame shift of particular track video (actually right camera), I do not know why. After the shift the sync is perfect as seen at least in the timeline, unfortunately I have no CRT anymore and no synchroscope at home to measure the sync at present. Autofocus and OIS mode seem not to affect the sync, but shutter speed less than 1/60 sec seems to degrade the sync. The drift of sync is about 0.3 mS/min for my particular pair of GH1 bodies, which is twice larger than my pair of Sony HC3 HDV video cameras, where its initial sync is always less than 10 micro Sec when powered on and measured with LANC Shepherd.

At FHD, 1080P24 mode, perfect sync in the 1080P60 sequence can be achieved as well. However frame shift causes jitter of disparity in some case due to the 1/60 Sec shift of 1/24 Sec frame in 60P sequence. I am not sure this is acceptable or not.

I look forward to any comments and questions.
------------------------
inoue_k3D
inouek3D@gmail.com
http://v-galleries.com/
------------------------

 

[David Cole]

What kind of lenses are you using? I'm achieving near perfect sync at all frame rates / resolutions - including 720 60P - but - ONLY with manual lenses that have no electrical connection to the camera body. Your "power-up at the same time" and "ganged shutter release" techniques are exactly what we are all doing here.

If you don't have a CRT for Peter Wimer's Sync tester, you can use a fast photo strobe flashed at both cameras from 10 feet or more away - or - a small fan with alternating white and black tape on opposite fan blades. These won't give you the calibrated sync measurement that a CRT test will, but, it will allow you to see any sync offset in detail.

 

[inoue_k3D]

What kind of lenses are you using? I'm achieving near perfect sync at all frame rates / resolutions - including 720 60P - but - ONLY with manual lenses that have no electrical connection to the camera body. Your "power-up at the same time" and "ganged shutter release" techniques are exactly what we are all doing here.

Hi David

I tried kit lenses (14-140mm) first and confirmed perfect sync with other lenses (20mm) at AFS mode and manual focus lenses (Canon FD Lenses with adapter) at MF mode as well. The results are same as reported. I think it is reasonable as the AF procedure is done when half shutter push before the REC procedure start by full push and autofocus seems not to affect the recording start consequently.

From my experience with twin HC1 and twin GH13, two factors, "simultaneous shutter release" and" simultaneous power on" to start clock of each camera are enough to achieve the practical sync for 3D video shooting of several minute clips, I think.

Letting me know how to use external battery for GH2 or Sony CX560V or 700V (a brand new 1080P60 AVCHD model) will be much appreciated. Thanks in advance.
------------------------
inoue_k3D
inouek3D@gmail.com
http://v-galleries.com/
------------------------

 

[MarcusWolschon]

Some time ago somebody published very detailed picture of GH1's main electronic board. Was it on this forum??

Found:
http://www.henryolonga.net/MainBoard.jpg

You did not make photos of the OTHER side, did you?
(I guess it's a multi-layer board.)

Did anyone actually look for a genlock-pin as described in the beginning of this huge thread?

Did anyone actually try the connection of GND+REMOTE of 2 cameras?
Did anyone yet try to use a remote AND a microphone? Where there GND-issues with the sound-quality?

Are there any examples of unsynced 3d-shots or shots >30min running out of sync?
I'd like to have a look at how bad this issue actually is with different kinds of scenes.
(I'm planning on filming an event in 3D that requires shots in the order of 45min.
If 3d does not work out, I can still use the 2d-shot of one camera. )

I'm currently working on a 3d-printed diy shoulder-rig for 3d shots.
( http://sourceforge.net/userapps/mediawiki/marcuswolschon/index.php?title=Projects/Camera/ShoulderRig the resulting files will be posted on thingiverse for anyone to use.)
I thought about mechanically connecting focus and zoom of manual Canon FD-lenses
to the same follow-focus, so I can adjust focus+zoom of both cameras in sync.
Any thoughts on this?

 

[David Cole]

..."simultaneous shutter release" and" simultaneous power on" to start clock of each camera are enough to achieve the practical sync for 3D video shooting of several minute clips, I think.

I've exhaustively tested two GH1s with MANUAL lenses using the above technique and found consistently less than .02 ms of drift at the beginning of each test run and seldom more than .1 ms after 5 min. The drift over time is likely to vary, depending on how well the cameras are paired (oscillators) . In my case, I have 2 sets of 2 GH1s and both sets have sequential serial numbers. This a best practice for non-genlocked cameras.

...Letting me know how to use external battery for GH2 or Sony CX560V or 700V

Can you explain what info you need in a bit more detail? Are you asking how we wire the external batts to the battery-eliminator adapters in the GH1? Note that in the case of GH2, you have to purchase the battery-eliminators separately. in the case of the CX700 - you really don't need to use the power-on trick. A good stereo LANC controller (e.g. LANC Shepherd, ste-fra LANC) will power-up the cameras simultaneously and give you an instant indication of any offset. Works like a charm.

 

[David Cole]

Did anyone actually look for a genlock-pin as described in the beginning of this huge thread?

Highly unlikely that Panny padded out a GENLOCK input on the PCB. The hack that might work is to locate the oscillator that controls sensor integration timing (which may or may not be a function of the camera CPU clock) and tie it to the same part in another GH1. This would be highly explorative surgery and take mad skills and courage. (so... won't be me)

Did anyone actually try the connection of GND+REMOTE of 2 cameras?

Everyone here that is seeing results with GH1 sync gangs the shutter release remote lines from the two cameras - requiring that the GND from both be tied together. Not sure about the mic inputs - but - that should be a piece of cake...

Are there any examples of unsynced 3d-shots or shots >30min running out of sync?

I could give you gigabytes from a RED rig As a rule, I always power-on sync before every take, so, I avoid drift problems. I was rolling on the elephant footage (posted earlier in this thread) for about 10 min - so you can see what 10 min of drift looks like with my camera pairs (your milage will vary).

In general, however, the faster the action, faster the shutter, faster the framerate... the less tolerance there is for sync offset. Sync offset is responsible for the WORST form of eye-strain in stereo production and really should be avoided at all costs. Differential stereoscopic encoders (like the one from my company) will puke all over footage with a sync offset because the disparity between the left and right image pairs is TOO great.

There are no golden rules of sync, but this one is bronze plated... A sync offset greater than 0.50 msec will produce just noticeable disparity for an average city street corner with wind, people walking, and slow moving cars. Action video with considerably faster motion and shorter shutter speeds, needs to have a sync proportionally better.

 

[ekteTV]

Hi! I'm new to this forum, and just read through most of this thread looking for how to sync my two GH2s for 3D. Looks like you have found a technique that works, good to know. But today I found an interesting description of how to sync two 5Ds, by first starting to shoot video, and then using Pocket Wizard Plus II to trigger a still while shooting video. The still forces the sensors to reset, and since both cameras are triggered at the same time they reset the sensor at the same time. After this procedure, you can shoot your video is in perfect sync. Will this work on GH2 as well?

http://www.3dfilmfactory.com/index....d-mark-ii-cameras-and-shoot-3d&catid=42:press

 

[MarcusWolschon]

Does the gh1/gh2 use a separate pll chip+quarz or an internal oscillator?
In the former case, did anyone simply connect them to keep them phase-locked?
Then there should be zero drift, I guess.

 

[David Cole]

I'm not sure how to fire off a still shot while you are recording video on a GH1... the remote port's shutter-release function becomes a "start & stop record function" when the camera is in motion picture mode. In the case of both the GH1 and the Canon 5D (which I've worked with extensively on a stereo rig with a ganged shutter release) - these ports appear to be polled by the camera. That means that polling intervals will impact the sync. I'll give 3D Film Factory a call and see if they have any metrics on Canon sync using this method.

 

[ekteTV]

When shooting video in any of the stills modes by using the dedicated start/stop button, the regular shutter will fire off a still frame while recording. So maybe that will reset the sensor and the polling intervals, which will then be in sync for the consecutive shots? Don't know how this works from a remote shutter, though - maybe those of you who has a couple of PocketWizards lying around can give it a try?

 

[David Cole]

Don't know how this works from a remote shutter, though

That's what I mean... with the GH1 the remote doesn't do double duty - it either snaps a still (photo mode) or starts/stops recording (video modes). In ANY case - it's a polled port (as is the 5D) so I would not count on it for perfect sync.

 

[g.l]

That's what I mean... with the GH1 the remote doesn't do double duty - it either snaps a still (photo mode) or starts/stops recording (video modes). In ANY case - it's a polled port (as is the 5D) so I would not count on it for perfect sync.

The fact that synced stills can be taken with GH1's suggests you may be wrong about that David (ie. the polling freq seems to be good enough)... unless the freq was lower in video mode.

Syncing video by taking a still shot is an ingenious idea, effectively a version of my preferred solution (a firmware hack that resets the video capture timers somehow). I don't have my GH2 bodies yet - if you start video recording in _stills_ mode using on-camera controls (as ektaTV said), does the remote shutter still only trigger video recording? Do GH bodies even support taking a still during video?

 

[David Cole]

The fact that synced stills can be taken with GH1's suggests you may be wrong about that David

You are right - we don't have a precise measure of GH1 shutter release latency with a ganged remote. inoeu_k3D speculates that he is achieving "several tenth of mS". That's what I'd expect when the port is polled for resistor values.

The LANC Shepherd manual says.... A sync offset greater than 0.50 msec will produce just noticeable disparity for an average city street corner with wind, people walking, and slow moving cars. Action video with considerably faster motion and shorter shutter speeds, needs to have a sync proportionally better.

Next3D (my company) has found that ANY motion disparity (which may be compounded by motion estimation in the AVC encoder) can result in eye-strain and discomfort - especially when the object in motion is in negative parallax (popping out of the screen).

The ideal solution to a non-genlocked camera rig hack may be to fire up the cameras, have them init the lenses and thermally stabilize - then - trip an interrupt to reset the sensor integration cycle. Maybe the flash synchro connection ?? Not sure though - the GH1 manual says that the flash synchro speed is equal to or smaller than 1/160th of a second.

 

[David Cole]

I'll have my test bench set back up in a week or so. I'll measure the still photo synch on the GH1 and Canon 5D - as well as the above mentioned 5D "shutter release while recording" technique for video.

 

[inoue_k3D]

Hi David and all,

I learned polling method of shutter release of GH1 for the first time. It helped me to recognise the following picture.

If the polling process does not start simultaneously for both cameras and the REC start accuracy is same as still shutter release (I speculated to be several tenth of mS as David referred), It can be considered that the remote shutter release may cause REC start of both cameras at different REC start cycle of 1/60 sec which is already in sync for both cameras by the simultaneous power on procedure.
And this results in exact one frame shift of particular track (right camera in my case) in the Premiere CS5 timeline which I observed at a frequency of several out of of ten.

I think the above is consistent with the fact that observed sync in video mode (most several tens of micro sec judging from from audio waveform) is better than still sync by approximately one order of magnitude.

For GH2, external battery can be used already or not?
------------------------
inoue_k3D
inouek3D@gmail.com
http://v-galleries.com/
------------------------

 

[g.l]

For GH2, external battery can be used already or not?

Yes, check my post here. The simplest solution is shown by the video at the bottom (remote control car batteries).