GH1 Stereoscopic Rig Collaboration

With additional datapoints and some filtering, a mild correlation is apparent between composite video sync and visual sync analysis of the recorded files (Wimmer's test). The filtering I did was to discard the 3 outliers that seem to correspond to test runs where I adjusted focus before beginning record (more testing needed to see if THIS is the big problem).

sync_tests_filtered.gif


sync_graph.gif


Overall this sync is pretty darn good. If we could exercise the outlier demon, we'd be in business with just using composite sync as a guide.

By the way, I spent the evening helping a friend analyze sync on a Red One - based 3D rig. Red has a known sync problem. It genlocks to Trisync perfectly and the cameras exhibit no drift - BUT- they have a constant sync offset that is orders of magnitude bigger than the worst observation on the GH1 rig. And Red 3D rigs are being used right now on Pirates of the Caribbean IV.
 
David

Did you tried solution using timed turning on and record press?

Other solution is to use PTP interface.
One guy at cinema5d forum made ready firmware for arduino with USB host and PTP library (for Canon follow focus).
I don't know PTP parts, as I did not made research at this direction.
But you can always request firmware and try to locate this parts yourself (using PTP protocol documentation and looking for corresponding constants).
I am sure that PTP has shutter half press and shutter full press (as service program has such functions). And full reset is easy to add.
Of course, if we could find GH1 service program it can be even better :-)
 
Vitaliy Kiselev said:
Did you tried solution using timed turning on and record press?
Click to expand...

Not yet. This was the first step (to measure the composite sync offset and correlate with actual visual offset in the recorded files). I'm going back to the test data now to see if one camera is always offset in the same way, relative to the other. If so, I'll wire up the Arduino and the solid state switches. Ideally, the Arduino could measure the sync offset and make the adjustments to the power and/or shutter-release sequences automagically.

Vitaliy Kiselev said:
Other solution is to use PTP interface.
Click to expand...

I gave the PTP commands a try on the 5D as well. The latency to initialize "Live View" and then start recording was too much ... the sync was way off and non-deterninistic. We were not able to do the power-on sync trick with the 5Ds, however. Perhaps with the GH1s powering up in sync, the USB command latency would be good enough (assuming we marshal the commands simultaneously on the computer side). It SURE would be nice to have a RESET command!!
 
Unfortunately, it doesn't seem like one camera is uniformly slower or faster than the other, using the Wimmer visual comparison. Perhaps I can modify the Sync Shepherd circuit to indicate which camera is lagging the other with regard to composite video sync.

First, however, I'm going to use the existing test setup to try to figure out what causes the outliers where sync is way off. It looks like my initial estimates were wrong - sync is nearly perfect 95% of the time, and off only 5%. Maddeningly close to "good enough".
 
An approach that I am trying with Stereodata Maker is to briefly flash the PRINT led at a certain point in the Canon frame-capture code.


An external, inexpensive, tiny microcontroller can then measure the synch.


It may even be possible to use the left/right keys to introduce delays that improve the synch.


I starting playing with this many months ago with 'crunchy3d' but could not stop the camera crashing as soon as the led is lit.
 
Somewhat encouraging results....

I have found that in 100% of the cases where sync is badly blown - I have manually adjusted focus before beginning record. Conversely, in 54% of the test runs where I manually adjust focus, sync is bad. In ALL other test runs, there is at least mild correlation between composite video sync and actual sync in the recorded files.

This means - if I DO NOT adjust focus before recording AND only record when the composite video mis-sync is less than .05 ms, I get acceptable sync every time.

These conclusions are supported with 85 test runs. I'll try to double that this weekend - to be 100% I'll also test the focus rebound issue and manual lenses.

This seems like light at the end of the tunnel for a usable rig.
 
I think you can build very simple controller using ready and cheap Arduino boards available on ebay.
I just hope that ADC performance is enought.
 
Vitaliy Kiselev said:
I think you can build very simple controller using ready and cheap Arduino boards available on ebay.
I just hope that ADC performance is enought.
Click to expand...

A friend of mine has modified the Sync Shepherd circuit to scale the voltage up such that the area of interest fits the Arduino ADC. Should be enough resolution to do the trick. I'll post the schematic as soon as I build it and make sure it works.
 
I don't know why you need whole circuit.
As I understand only sync separator and few passive elements are enought (as you can use two ADC channels and software to do all else).
 
We're just using the sync separators, a hex inverter and the NAND gate (plus the passive stuff) to fill the 5vlt ADC range with a sync range of 0 to 5 degrees. Simple enough.

I actually hadn't considered using 2 ADC channels and doing the logic in software... duh!
 
Vitaliy:

Is there any chance that the cameras are calibrating the autofocus motors in the 4/3 lenses at power-up (even when autofocus is switched off) and that is causing the blown focus? Do you think this behavior can be patched out?

It would be VERY handy to be able to use the Olympus and Panasonic Micro 4/3 wide-angle lenses for 3D. We're having a difficult time sourcing manual wide-angles.
 
Last edited:
David Sykes said:
I starting playing with this many months ago with 'crunchy3d' but could not stop the camera crashing as soon as the led is lit.
Click to expand...

David, has anybody tried powering up two SDM cameras at the same time (e.g. pressing both "power-on" buttons and then applying external power supply to both cameras)? Namely, delayed power signal can be made relatively simple with some additional electronics. If yes, what happened to video synch?
 
David Cole said:
I have found that in 100% of the cases where sync is badly blown - I have manually adjusted focus before beginning record. Conversely, in 54% of the test runs where I manually adjust focus, sync is bad. In ALL other test runs, there is at least mild correlation between composite video sync and actual sync in the recorded files.
Click to expand...

This is very important result. Maybe we'll be able to figure out internal mechanism. However, it's important to know that video can be quite reliably synched. Some time ago I ordered the second camera to make measurements, but I had to send it back, since it wasn't 100% functional. When/if I get another camera, I'll let you know about my results.
 
David Cole said:
Vitally:

Is there any chance that the cameras are calibrating the autofocus motors in the 4/3 lenses at power-up (even when autofocus is switched off) and that is causing the blown focus? Do you think this behavior can be patched out?
Click to expand...

This is very low level code. Plus it requires very extensive research.
Any error in patch will result in bricked body.
As for AF modes and focus motor I can't tell you anything as I don't have any AF lenses for m43.
 
Should have listened to Vitaliy earlier when he recommended sticking with manual lenses....

It seems that there is a relationship between the position of the focus ring on the M4/3 lenses and camera startup delay (even with AF turned OFF). Difference positions in the focus travel result in different delay durations. With manual lenses - actual sync closely covaries with the composite video blanking sync (I have 25 test runs confirming this).

There's more work to do (e.g. build the microcontroller to monitor composite video sync and control camera power and shutter release)... but... I'm going to pull our rig off the bench and go shoot some 3D first.

Hopefully I'll have some footage to post in a day or so.
 
Last edited:
hybridvision said:
Hi, I take it you are aware of Panasonic H-FT012E?
Click to expand...

Yes, we know about it. The problem is that you get less pixels per one picture, lens separation is relatively small and FOV (per one lens) is small as well. Therefore you can get 3D only for closer subjects and 3D can show, due to narrower FOV, cardboard effect.

Having two independent cameras is significant advantage, since you get whole resolution, distance between the cameras can be modified if needed (depends on the closest subject), FOV can be adjusted by choosing suitable lenses, etc.

Downside is complicated handling and not perfect synch after some time.

David Cole, have you get any footage? I'll probably make some measurements when I order electronics (including power sockets).
 
You must log in or register to reply here.
Back
Top