3D Stereo Photography: camcorder frame sync (forum newbie)

jjanes · 2009-02-07 18:07

I found your company and forum because I've been looking for a way to determine the frame sync difference between two Canon HV20 camcorders for optimizing 3D stereo video recording.

It certainly seems that your USB Oscilloscope could do the job.

But I was wondering if a much simpler and cheaper new product could be developed using your expertise.

Here is a link to a product that is no longer available where in the section called, How it Works,·is explained just such a device.

http://www.ledametrix.com/lcdsync/

I believe there is a market for such a device since 3D video photography is becoming popular.· And most consumer camcorders do not have a means to lock frame sync between two camcorders.· Such a device as I am looking for is needed by these enthusiats.

Here is another link to a video where this Sync Shepherd is demonstrated so you can see how it is used in practice.

http://www.youtube.com/watch?v=9FhU64xhC3Q

Perhaps someone would care to simply offer up a circuit to gt the job done.

Thanks

JJ

Phil Pilgrim (PhiPi) · 2009-02-07 19:59

Do you need to know the time between fields (i.e. vsync-to-vsync) or between frames (i.e. odd field-to-odd field)?

Also, just out of curiosity, how does this information help, given that it's not possible to synchronize the two cameras?

-Phil

PJAllen · 2009-02-07 20:21

Genlock.

Phil Pilgrim (PhiPi) · 2009-02-07 20:35

PJ,

Yes, "genlock" is, indeed, a buzzword for this topic. But it's very nearly a non sequitur in this case, since the OP was referring to cameras that, lacking a sync input, can't be genlocked. Producing the circuit he needs is easy enough, once I have the frame vs. field answer. But my second question to him has more to do with how to deal with two signals that are not genlocked, since 3D artifacts can often result.

-Phil

rjo_ · 2009-02-08 01:10

jj... you are on the right track... take a look at Bean's site: hittconsulting.com

http://www.hittconsulting.com/hcosd.htm

Sir Bean offers an overlay board, which is just too sweet... and he shows you how to hook up and use a sync separator chip.

AND over at mydancebot.com you'll find a very modestly priced 1kx1k camera that is prop ready... syncing the acquisition shouldn't be a problem and you will get all the help you need.

http://mydancebot.com/kits/s.php?k=camera

Rich

Post Edited (rjo_) : 2/8/2009 1:17:48 AM GMT

rjo_ · 2009-02-08 01:20

By the way, there are companies that offer lenticular screens for computer displays... which allows you to see in 3d without glasses of any kind. The only problem is that early on... in the 3D field... a patent was granted for vertically integrating multiple frames... which is what you need to do to create a lenticular display[noparse]:([/noparse]

BUT once you have synced two cameras... there is no reason not to sync 7 cameras and then all you need to do is publish a new method in the public domain for interlacing 7 frames... and the world will beat a path to your door.

Rich

Phil Pilgrim (PhiPi) · 2009-02-08 02:31

For jjanes' app, it will take two LM1881s and some logic to create a single phase detector output. I'm waiting for his answer to the frame vs. field question, so I know which outputs from the '1881s to use. From there, a BASIC Stamp could measure and output display the phase difference.

Again, this has nothing to do with synchronizing two cameras, so far as I can tell, since they cannot be synced. It's more to do with just measuring how much they're out of sync. I suspect this info is fed to the PC so the capture program can tell how many bottom lines from one camera's previous field and top lines from its current field to pair with a full field from the other camera. This is not the optimum solution that genlocking would be, were that possible, but it should help to reduce 3D artifacts from objects moving across the field of view. At least that's my speculation.

-Phil

rjo_ · 2009-02-08 02:36

jj... there are a couple of guys here who are always right, I'm not one of them but Phil is[noparse]:)[/noparse])

Phil... it looks like Parallax is shopping. I put a word in for a linear array sensor[noparse]:)[/noparse]

Rich

jjanes · 2009-02-08 04:43

Phil Pilgrim (PhiPi) said...
Do you need to know the time between fields (i.e. vsync-to-vsync) or between frames (i.e. odd field-to-odd field)?

Also, just out of curiosity, how does this information help, given that it's not possible to synchronize the two cameras?

-Phil

The idea is to get the cameras synced as closely as possilbe to make any 3D stereo artifacting unnoticeable.· Most consumer camcorders do not have genlock.· The way this is done in practice with most consumer camcorders , as far as I know, is to measure the difference in the two camcorder syncs.· Then turn one camera off and on and measure the sync again until you get a difference that is very small.· In the video I linked to, he showed about 3 milliseconds and he noted that this was good enough for certain subject matter.· I will be experimenting to find out what measured syncs are good for various subject matter.· And that subject matter characteristic would be speed of movement.

Here is what the description on the other linked web page said and I hope this helps:·

"... the LCD Sync Shepherd uses a pair of video sync separator chips to generate a pair of square waves corresponding to the video fields. A microprocessor is then used to measure the time difference between the field square waves, and displays this information on the LCD in both fractional second and millisecond formats."

Obvious to me is that each chip creates a square wave from each input from each camcorder.

I believe that I read, perhaps somewhere on this fellows website, that certain characteristics of the input waveform are better to sync on than others, such as the falling edge of the vsync(?).· Foregive me but I am not up on the technicalities of video electronics.· There are probably a few different points in the general waveform that are consistent that one could·sync on more predictably than others between the two camcorders.

Using my limited common sense I'd say optimally it would be frame to frame sync since in a video editing program one aligns / syncs frames.· In the quoted description above he does say "fields" but he may have said this to be intentionally vague or obscure or he was just imprecise or he meant "fields".· I have addressed problems that on first glance common sense might say one thing but actually it didn't matter.· I'll leave the best approach for you to decide keeping in mind the purpose or goal of this discussion.

I am really impressed with all of your enthusiastic and thoughtful responses.· I have been a long-time member of the HV20.com forum with 18,000 members.· We are always looking for ways to increase our capabilities with this camcorder as well as other newer models in this Canon HD consumer line.· You can be sure I will be sharing any useful information from all of you·in that forum.

JJ
·

jjanes · 2009-02-08 04:46

rjo_ said...
jj... you are on the right track... take a look at Bean's site: hittconsulting.com

http://www.hittconsulting.com/hcosd.htm

Sir Bean offers an overlay board, which is just too sweet... and he shows you how to hook up and use a sync separator chip.

AND over at mydancebot.com you'll find a very modestly priced 1kx1k camera that is prop ready... syncing the acquisition shouldn't be a problem and you will get all the help you need.

http://mydancebot.com/kits/s.php?k=camera

Rich

I will check these sites out.

JJ

jjanes · 2009-02-08 04:50

rjo_ said...
By the way, there are companies that offer lenticular screens for computer displays... which allows you to see in 3d without glasses of any kind. The only problem is that early on... in the 3D field... a patent was granted for vertically integrating multiple frames... which is what you need to do to create a lenticular display[noparse]:([/noparse]

BUT once you have synced two cameras... there is no reason not to sync 7 cameras and then all you need to do is publish a new method in the public domain for interlacing 7 frames... and the world will beat a path to your door.

Rich

I prefer cross viewing with no equipment.· My next preference is anaglyph with red / cyan glasses.

I can make both video types with Adobe Premeire Pro CS3 and or Adobe After Effects CS3.

Also, if you use progressive scan you won't need to interlace, if I'm thinking correctly.

JJ

jjanes · 2009-02-08 04:59

Phil Pilgrim (PhiPi) said...

Again, this has nothing to do with synchronizing two cameras, so far as I can tell, since they cannot be synced. It's more to do with just measuring how much they're out of sync. I suspect this info is fed to the PC so the capture program can tell how many bottom lines from one camera's previous field and top lines from its current field to pair with a full field from the other camera. This is not the optimum solution that genlocking would be, were that possible, but it should help to reduce 3D artifacts from objects moving across the field of view. At least that's my speculation.

-Phil

Feeding this data to the PC is a possibility.· Or a seperate battery operated device with an LCD could be another possiblity and probably preferred since it would be most portable.

Artifacts from moving objects across the frame, if slow enough and where the cameras are closely synced, it is hoped, will not produce any noticeable 3D stereo artifacts.· This is what I hope to accomplish for these types of consumer camcorders that do not have genlock and nothing more.

There is one other point I want to make but it slips my mind at the moment.

JJ

jjanes · 2009-02-08 05:02

I remember the point now: it is that the difference of the sync between the two cameras should hold relatively closely for a length of time that will be useable for shooting short clips. The drift over short periods of time should be negligible.

JJ

rjo_ · 2009-02-08 05:33

jj ... between the time you asked your question and an hour ago, Hanno apparently decided not to offer the specific product that I mentioned. Hanno is very close to offering another solution, which I suspect will be capable of doing what you want and far more. I would guess, somewhere between tomorrow and two weeks... but development isn't a linear beast... so, whatever it takes... that's how long we wait.

In the meantime, if you haven't looked at ViewPort, this would be the time to take a gander at it.

I also use cross viewing... it is difficult for young people to learn... in young people the linkage between focus and convergence is very strong.

My comment about interlacing was related to lenticular imaging to form the 3D image from a flat screen without glasses of any kind.

Indeed what you need for motion is genlock... there are fairly cheap cameras that have the ability to automatically sync to each other... many times this feature isn't listed in the product description... anything with a serial connector is a good candidate and deserves a closer look. You can then record the images to 2 different recorders... or digitizers and actually sync up your final movie using a flash to indicate a start frame for each set of images.

on the other hand these guys get nice results doing exactly what you are talking about with no sync signals at all!!!

http://www.dvinfo.net/conf/3d-stereoscopic-production-delivery/94478-my-first-3d-hd-video-hv10.html

There is a very strong 3D perceptual clue in the focal characteristics of an image... this may overcome the 3d distortions you would expect from two images not perfectly sync'd. My interests are mostly quantitative 3d... for that everything I say above, based on photogrammetry... is true... but based on perception... well maybe not[noparse]:)[/noparse]

As far as I know the 3D perceptual effect that I am talking about doesn't have a name... and I couldn't find it when I googled it... but if you would like to see it... simply take an image that has lots of blur clues (with distance) and then duplicate it and cross view it... it looks like 3D!!! but it isn't. There was a company that claimed to be extracting 3D from a single view on the fly... I suspect this effect is what they were using...

Post Edited (rjo_) : 2/8/2009 6:10:59 AM GMT

Phil Pilgrim (PhiPi) · 2009-02-08 06:31

Thanks, JJ, that's the information I needed.

The "square wave" comes from the LM1881 sync separator's odd/even pin. Here's a schematic you can use with the BASIC Stamp, which I've built and tested. You should omit the 75-ohm resistors if your signals are looping through to a capture board. Also I've spec'd a 2N3904 transistor, but any general-purpose NPN will do.

attachment.php?attachmentid=58533

The width of the output pulse will give the relative phase difference between the two video signals. Here's what it looks like on the scope (Ch 1 & 2 are the camera inputs, Ch3 is the the phase output to the BASIC Stamp):

attachment.php?attachmentid=58529

The width can vary from 0 to 33.367 ms. However, once the width goes past halfway (16.683 ms), it's more useful to subtract it from 33.367 to get the smaller phase difference value. Here's a BASIC Stamp program that monitors the output pulse and displays the phase difference in milliseconds (dd.ddd) on the DEBUG screen:

' {$STAMP BS2}
' {$PBASIC 2.5}

Phase VAR Word

DO
  PULSIN 15, 0, Phase
  Phase = Phase * 2
  Phase = Phase MAX (33367 - Phase)
  DEBUG HOME, DEC Phase / 1000, ".", DEC3 Phase // 1000, " ms", CLREOL

LOOP

I used two cameras of different makes and noticed that the phase difference drifted quite markedly. So I'm not sure how useful this will be for getting two cameras in sync if they won't stay in sync.

-Phil

Post Edited (Phil Pilgrim (PhiPi)) : 2/8/2009 7:24:50 AM GMT

Phil Pilgrim (PhiPi) · 2009-02-08 06:42

JJ,

Addendum: One inexpensive way to get synchronization between two cameras is to get a pair that use "line lock". These typically have a 24VAC power requirement and lock the vertical sync to the 60Hz power line frequency. This causes a 0.1% mismatch to the NTSC vertical frequency spec, but it will keep both cameras in lock-step with each other (plus or minus one field interval: i.e. the phase difference will be either 0 or 16.667 ms).

Further info: Apparently, line-locked cameras have a phase control that can be used to fine tune their phase relationship with the AC line and, hence, with each other. So a circuit like the one above would still be useful for aiding this adjustment. BTW, there are professional-grade "video timing meters" on the market that display the same measurement and cost over $1000.

-Phil

Post Edited (Phil Pilgrim (PhiPi)) : 2/8/2009 7:17:57 AM GMT

jjanes · 2009-02-08 07:33

Phil Pilgrim (PhiPi) said...
Thanks, JJ, that's the information I needed.

The "square wave" comes from the LM1881 sync separator's odd/even pin. Here's a schematic you can use with the BASIC Stamp, which I've built and tested. You should omit the 75-ohm resistors if your signals are looping through to a capture board. Also I've spec'd a 2N3904 transistor, but any general-purpose NPN will do.

The width of the output pulse will give the relative phase difference between the two video signals. Here's what it looks like on the scope (Ch 1 & 2 are the camera inputs, Ch3 is the the phase output to the BASIC Stamp):

The width can vary from 0 to 33.367 ms. However, once the width goes past halfway (16.683 ms), it's more useful to subtract it from 33.367 to get the smaller phase difference value. Here's a BASIC Stamp program that monitors the output pulse and displays the phase difference in milliseconds (dd.ddd) on the DEBUG screen:

I used two cameras of different makes and noticed that the phase difference drifted quite markedly. So I'm not sure how useful this will be for getting two cameras in sync if they won't stay in sync.

-Phil

Great!· But I only have a first year university education in electrical engineering.· I know resistors, capacitors, diodes, transistors, truth tables and logic gates, calculus, etc.· But I don't know chips.

As far as a capture board is concerned, I think the average user just wants to know the sync difference in milliseconds, such as, 2.031 ms.·· And they want it portable and versatile, like as in handheld with an LCD display.· I like the idea of connecting two outputs, one·from each camcorder's AV jack into the device in question and having the sync difference displayed.· It should be that simple.

I can tell you right now that I would be willing to pay $25 (easily) - $50 for such a device.

I think that two cameras of the same make and model, like in the linked video in my original post, should hold sync fairly well and well enough for short clips.

Of course, you or others may think of other capabilities useful to consumer camcorder users but for now I have just this one capability in mind.

JJ

·

jjanes · 2009-02-08 07:44

Phil Pilgrim (PhiPi) said...
JJ,

Addendum: One inexpensive way to get synchronization between two cameras is to get a pair that use "line lock". These typically have a 24VAC power requirement and lock the vertical sync to the 60Hz power line frequency. This causes a 0.1% mismatch to the NTSC vertical frequency spec, but it will keep both cameras in lock-step with each other (plus or minus one field interval: i.e. the phase difference will be either 0 or 16.667 ms).

Further info: Apparently, line-locked cameras have a phase control that can be used to fine tune their phase relationship with the AC line and, hence, with each other. So a circuit like the one above would still be useful for aiding this adjustment. BTW, there are professional-grade "video timing meters" on the market that display the same measurement and cost over $1000.

-Phil

It's hard to say what the real capabilities of the Canon HV20 HD camcorder are.· All we know is what we are told about the capabilities provided by the firmware.

A Russian going by the name of VitalyB somehow hacked the firmware of the Canon Powershot line of point and shoot digital cameras and the CHDK.com people have developed a monster of a firmware hack.· For instance they have the capability to have shutter speeds of 1/60,000 sec., they can sync twin cameras and with flash, and it goes on and on and on.

The HV20 probably has true capabilities we'll never know, such as genlock, but the firmware doesn't allow users to access it.

JJ

jjanes · 2009-02-08 10:36

Phil Pilgrim (PhiPi) said...
Thanks, JJ, that's the information I needed.

-Phil

Tell me, Phil.· What o-scope are you using?

Thanks.

JJ

Phil Pilgrim (PhiPi) · 2009-02-08 20:36

JJ,

It's a Tektronix TDS3034.

-Phil

jjanes · 2009-02-09 13:46

Phil Pilgrim (PhiPi) said...
JJ,

It's a Tektronix TDS3034.

-Phil

Wow.· So I am assuming·your screen shot is from that Oscope.· That's why I asked.· Okay.· Very nice.

Is the Parallax USB Oscilloscope adequate to measure the frame sync timing difference·betweem two digital camcorders?

I'm thinking that it is:· easily.

If someone comes up with a dedicated device for my purposes that is reasonably priced I will buy it.· If not I'll buy the Parallax USB Oscilloscope if you confirm its usefulness here.

Thanks

JJ

·

Phil Pilgrim (PhiPi) · 2009-02-09 18:20

JJ,

This is not something I could do — even with my 'scope — without the external cricuitry cited above. Thre reason is that the even fields are impossible to distinguish visually from the odd fields without zooming in on the area around the vertical sync interval. But once you've done that, the vertical sync from the other trace may well be out of view.

Do not be afraid of constructing the LM1881 circuit yourself. I did mine in the breadboard area of a BOE. It took all of five minutes.

-Phil

jjanes · 2009-02-09 22:18

Phil Pilgrim (PhiPi) said...
JJ,

This is not something I could do — even with my 'scope — without the external cricuitry cited above. Thre reason is that the even fields are impossible to distinguish visually from the odd fields without zooming in on the area around the vertical sync interval. But once you've done that, the vertical sync from the other trace may well be out of view.

Do not be afraid of constructing the LM1881 circuit yourself. I did mine in the breadboard area of a BOE. It took all of five minutes.

-Phil

If need be I may tough it out, build the circuit, and buy the oscope.

I don't mean to be a pest.

(Detailed discussion removed.)

In short, if my total visible horizontal scale was 4 or 5 milliseconds and my useable sync was 3 ms or less then I could see the offset.· Otherwise it would be off scale and out of view and unuseable anyway.

Anyone else, feel free to chime in.

JJ

Post Edited (jjanes) : 2/9/2009 10:54:30 PM GMT

Phil Pilgrim (PhiPi) · 2009-02-10 00:37

JJ,

If you build the circuit, you won't need the 'scope, just a BS2 to measure and display the output pulse width. If you don't build the circuit, you will need a two-channel 'scope that has a special video trigger mode. With my 'scope set at 400us/division (i.e. 4 ms across the screen), I can barely make out the difference between the odd and even fields visually.

-Phil

jjanes · 2009-02-10 02:49

Phil Pilgrim (PhiPi) said...
JJ,

If you build the circuit, you won't need the 'scope, just a BS2 to measure and display the output pulse width. If you don't build the circuit, you will need a two-channel 'scope that has a special video trigger mode. With my 'scope set at 400us/division (i.e. 4 ms across the screen), I can barely make out the difference between the odd and even fields visually.

-Phil

I hear you.· Check.· Check.

JJ

Thanks.

rjo_ · 2009-02-10 18:00

AND there are lots of apps that require something like this[noparse]:)[/noparse]

jjanes · 2009-02-10 23:03

I was looking for a solution that I could purchase:· a ready made device of some sort.··It appears that I will have to start a new hobby almost from the ground up with the·"What is a Microcontroller" kit and then on to the next kit, and so on and this will take some time.· It would certainly be worth it.· And I am glad I found this site.· But I cannot make this decision right now.· It will take time for me to consider my options.

JJ

·

jjanes · 2009-02-19 00:39

Phil Pilgrim (PhiPi) said...
JJ,

This is not something I could do — even with my 'scope — without the external cricuitry cited above. Thre reason is that the even fields are impossible to distinguish visually from the odd fields without zooming in on the area around the vertical sync interval. But once you've done that, the vertical sync from the other trace may well be out of view.

Do not be afraid of constructing the LM1881 circuit yourself. I did mine in the breadboard area of a BOE. It took all of five minutes.

-Phil

If I decide to go the·BS·route, would you recommend the Basic Stamp Discovery Kit?· If USB is strictly for programming the BS then I guess USB only is the way to go:· not Serial (with USB adapter and cable.)

I've only programmed one PL device once in my life and that was about 12 years ago.· I'll need to brush up on the BS.

Thanks

jj

Post Edited (jjanes) : 2/19/2009 1:03:53 AM GMT

jjanes · 2009-02-20 16:49

I've decided to get the Basic Stamp Discovery Kit.

I'll go to Fry's this afternoon and pick it up.

Then in a few weeks I'll make the circuit you describe and program the Basic Stamp as you also have kindly provided and I'll let you know how it goes.

Thanks.

jj

Carl Hayes · 2009-02-20 17:41

In general it is not possible truly to synchronize the output from two unsyncable vidicams.· But there is a workaround.

Workaround #1:· find, or build, equipment that will receive the video (two channels of it), record each frame, and send it out.· For each frame it will wait until it has complete frames from both cameras, and send them out synchronously.· When one camera gets a whole frame ahead of the other, our box inserts a duplicate frame in the video from the slower camera.· Maximum out-of-sync is then 1/30 second.

Workaround #2:· create a video processing program that can do #1 in software after the fact (in postprocessing), working with two AVI files on a PC.· This is no task for a beginner, or even for most experts.

Me, now, I'd go look for a vidicam that has a GenLock input.· You don't need to buy two such cameras because you can GenLock the newly obtained one by feeding its GenLock input with the video from the other one.

About six months ago I sold two such cameras on eBay.· One was a studio-quality Panasonic that originally sold for about $7000 and brought less than $200 on eBay, in like-new condition.· The buyer was ecstatic.· The other one was much less sophisticated, much less expensive originally, and brought over $300.· Go figure.· But either one would do what you want to do.

I'll bet you can find a camera with GenLock much more cheaply than you can build a workaround.· It'll be easier to use, too.· Right now, with NTSC fading into the sands of time, originally expensive NTSC equipment ought to be going begging on eBay.· Hmmm, I've got three TBCs, two complete Panasonic SVHS editing systems, a Kramer digital motion stabilizer that corrects NTSC video for camera vibration (really neat box)·-- maybe I should dump all that.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
· -- Carl, nn5i@arrl.net

Phil Pilgrim (PhiPi) · 2009-02-20 18:20

Carl,

I agree. Even two cameras that are the same brand are not going to stay in sync long enough to be useful in this app. If the relative drift is only 5ppm, two cameras that begin in sync will be 90 degrees out of sync in less than 15 minutes. Unfortunately gen-lockable cameras tend to be rather expensive. And the professional-level cameras that support gen-locking are typically too large to clamp cheek-to-jowel for 3D apps with a reasonably small lens separation. That's why I suggested line-sync cameras, which are commonly used in security applications and are fairly inexpensive. These are synchronized to their AC power supply and come with a vernier to fine-tune the synchronization. Once that's done, two cameras using the same AC source should stay in sync.

One thing I've been tempted to try is to get a couple cheap board cameras, remove their crystals, and drive the clock inputs with a Propeller. By advancing or retarding the clock going to one camera, the two could rapidly be synchronized and would stay in sync. It's on my list, but it's a long list.

-Phil

3D Stereo Photography: camcorder frame sync (forum newbie)

Comments