Hi,
I've searched around this forum for a while, now, and I haven't found any topics which seem to address this specific question, so here goes...
In the absence of a word clock input on seperate PCI cards you want to keep in sync, is it conceivable to bypass both of their internal clock circuits in favor of one master clock that controls them both?
In particular, I have heard that many sound cards and video capture cards may run on nearly identical clocks, but the smallest deviation (eg 48.0000...9 KHz versus 48.0000...1KHz) can cause the sound to drift from the video when capturing to your computer's hard drive.
The obvious (and expensive) answer is to have bnc word clock inputs on both cards for sync, but for the prosumer user like myself, that is prohibitively expensive.
I see that many folks have either built clocks from scratch, or modified their CD player's existing clocks to nudge their speed in one direction or another.
1. Has anyone worked on computer clocks, first of all?
2. Are there resources out there like this one:
http://www.lcaudio.com/index.php?page=66
3.Can someone point me to a thread or resources where I may bone up on this specific aspect of clocking?
Thanks,
Erik.
I've searched around this forum for a while, now, and I haven't found any topics which seem to address this specific question, so here goes...
In the absence of a word clock input on seperate PCI cards you want to keep in sync, is it conceivable to bypass both of their internal clock circuits in favor of one master clock that controls them both?
In particular, I have heard that many sound cards and video capture cards may run on nearly identical clocks, but the smallest deviation (eg 48.0000...9 KHz versus 48.0000...1KHz) can cause the sound to drift from the video when capturing to your computer's hard drive.
The obvious (and expensive) answer is to have bnc word clock inputs on both cards for sync, but for the prosumer user like myself, that is prohibitively expensive.
I see that many folks have either built clocks from scratch, or modified their CD player's existing clocks to nudge their speed in one direction or another.
1. Has anyone worked on computer clocks, first of all?
2. Are there resources out there like this one:
http://www.lcaudio.com/index.php?page=66
3.Can someone point me to a thread or resources where I may bone up on this specific aspect of clocking?
Thanks,
Erik.
Thanks
Ok, that's kind of what I was thinking.
Are there any software applets out there that allow you to slave your PCI cards clocks to the MB clock? Or one card to another?
I was also thinking that if I'm using S/PDIF input stream as the master clock for the soundcard, perhaps there is a way to dictate to the audio/video recording software to lock to that signal. Would the video capture card lock to that clock as well?
-Erik.
Ok, that's kind of what I was thinking.
Are there any software applets out there that allow you to slave your PCI cards clocks to the MB clock? Or one card to another?
I was also thinking that if I'm using S/PDIF input stream as the master clock for the soundcard, perhaps there is a way to dictate to the audio/video recording software to lock to that signal. Would the video capture card lock to that clock as well?
-Erik.
Explanation
Neutron7 Thanks, I'm on the same track. I should have detailed what I'm trying to do for the benefit of the discussion.
I've got an M-Audio Audiophile 2496 soundcard with S/PDIF input. The card actually does allow you to use the incoming digital signal as the Master Clock source. I intend to capture a particular favorite Sci Fi Trilogy from Laserdisc. I would use the S-video analog input on a video capture card, and the digital AC3 Surround signal via the S/PDIF input on the audio card. Unfortunately, A/V sync is a problem when you don't have pro gear that all locks to a "House Sync" Black Burst clock source from which each piece of gear derives it's required clock speed. Due to the slightly different clock speeds of various equipment, drift is likely to occur, especially during longer captures.
What I'm trying to figure out is a way to jerry rig a Video Capture Card to reference the same clock that the audio card uses so that the Audio doesn't drift from the video. If one can be assured of that, then any offset bewteen the two can be rectified by nudging the audio file just a bit in an editor program. With different clock speeds, no amount of alignment will correct the issue totally.
I've imagined bypassing the clock of the video capture card in favor of that of the audio card or S/PDIF input clock. I don't really know how to do that. The article I linked to in my earlier post opened me up to the possibility of bypassing the clock of an AC97 soundcard in favor of a better one. What I need is to learn more details about how video capture cards and audio cards are clocked in the first place and how to mod them.
For instance, is there a popular resource of information on the standard clock rates, circuits, and applications? I've found several references to these facts here at DIYAudio.com, but nothing complete or definitive yet. I'll keep searching.
I'm unable to afford more expensive capture cards with word clock sync, but perhaps through my misguided efforts, I can create a sort of "poor man's A/V sync lock" between a limited number of devices. Really all that is required is for the clocks of the audio and video capture cards to be locked to avoid drift between the two. If two seperate crystals aren't going to be precise enough, how about a single crystal in a master clock to control them both? Is there any issue with physical distance between a master clock and it's slaves? Can it be an external unit in a box, or must it be built onto one of the cards?
I also wonder if there is a way of introducing a commercial word clock signal to consumer or prosumer gear, just by finding the correct pins on the PCI card and hooking them up (It might require a bit more circuitry to get them to play nice)?
These are the ideas I've been kicking around. If anyone has any ideas finding resources on these technologies I would be very appreciative.
-Erik.
Neutron7 Thanks, I'm on the same track. I should have detailed what I'm trying to do for the benefit of the discussion.
I've got an M-Audio Audiophile 2496 soundcard with S/PDIF input. The card actually does allow you to use the incoming digital signal as the Master Clock source. I intend to capture a particular favorite Sci Fi Trilogy from Laserdisc. I would use the S-video analog input on a video capture card, and the digital AC3 Surround signal via the S/PDIF input on the audio card. Unfortunately, A/V sync is a problem when you don't have pro gear that all locks to a "House Sync" Black Burst clock source from which each piece of gear derives it's required clock speed. Due to the slightly different clock speeds of various equipment, drift is likely to occur, especially during longer captures.
What I'm trying to figure out is a way to jerry rig a Video Capture Card to reference the same clock that the audio card uses so that the Audio doesn't drift from the video. If one can be assured of that, then any offset bewteen the two can be rectified by nudging the audio file just a bit in an editor program. With different clock speeds, no amount of alignment will correct the issue totally.
I've imagined bypassing the clock of the video capture card in favor of that of the audio card or S/PDIF input clock. I don't really know how to do that. The article I linked to in my earlier post opened me up to the possibility of bypassing the clock of an AC97 soundcard in favor of a better one. What I need is to learn more details about how video capture cards and audio cards are clocked in the first place and how to mod them.
For instance, is there a popular resource of information on the standard clock rates, circuits, and applications? I've found several references to these facts here at DIYAudio.com, but nothing complete or definitive yet. I'll keep searching.
I'm unable to afford more expensive capture cards with word clock sync, but perhaps through my misguided efforts, I can create a sort of "poor man's A/V sync lock" between a limited number of devices. Really all that is required is for the clocks of the audio and video capture cards to be locked to avoid drift between the two. If two seperate crystals aren't going to be precise enough, how about a single crystal in a master clock to control them both? Is there any issue with physical distance between a master clock and it's slaves? Can it be an external unit in a box, or must it be built onto one of the cards?
I also wonder if there is a way of introducing a commercial word clock signal to consumer or prosumer gear, just by finding the correct pins on the PCI card and hooking them up (It might require a bit more circuitry to get them to play nice)?
These are the ideas I've been kicking around. If anyone has any ideas finding resources on these technologies I would be very appreciative.
-Erik.
well depending how much time it will take for you to do the jerry rigged version, maybe it would be quicker to just open it and "nudge" like you said (since it is only for 3 movies) otherwise you might spend more time making a time saving device, than the actualk amount of time you save.
apparently this is a very common problem
http://www.inventa.com.au/Audio-Video Out of Sync in PC Video Capture.htm
and
"Another thing that Can cause Sync problems is the Latency between the Time Clocks on your Video Capture card and your Sound Card...This Latency is usually only about a second or two per hour but over a whole movie it can cause a Pretty Major Sync problem....
Good Capture Software Should Compensate for this Time Latency and for Dropped Frames....Maybe Try Captureing directly to Mpeg-2 useing "InterVideo WinDVR 3.0" as it can compensate for Dropped Frames and the Latency between your Sound and Capture cards and it will also Save you a Bunch of Time you would spend encodeing to Mpeg-2 from AVI.....
Cheers
PS: If you want to capture to AVI look for a Setting in your Capture software that is called the "Master Stream" setting and set it to "Video" and this should Make the capture software Properly Compensate for Dropped Frames and any Latency...." [Minion from afterdawn.com]
apparently this is a very common problem
http://www.inventa.com.au/Audio-Video Out of Sync in PC Video Capture.htm
and
"Another thing that Can cause Sync problems is the Latency between the Time Clocks on your Video Capture card and your Sound Card...This Latency is usually only about a second or two per hour but over a whole movie it can cause a Pretty Major Sync problem....
Good Capture Software Should Compensate for this Time Latency and for Dropped Frames....Maybe Try Captureing directly to Mpeg-2 useing "InterVideo WinDVR 3.0" as it can compensate for Dropped Frames and the Latency between your Sound and Capture cards and it will also Save you a Bunch of Time you would spend encodeing to Mpeg-2 from AVI.....
Cheers
PS: If you want to capture to AVI look for a Setting in your Capture software that is called the "Master Stream" setting and set it to "Video" and this should Make the capture software Properly Compensate for Dropped Frames and any Latency...." [Minion from afterdawn.com]
Thanks for the tips
Neutron7... Are you by any chance the same "Neutron" over at Hexus.net?
Thanks for the info. Especially that last one. If there's a software fix to this, it'd certainly save me some time money and effort. I would only question why if it's so simple, that most people don't seem to know about it? Or why more Designers aren't incorporating this feature into their software? Perhaps it only accounts for straight latency and not sample clock differences? I'll have to have a looksee...
Regarding the Nudge technique, I don't think it's possible if the cause of the problem is drift and not straight latency. If the sound lags ahead or behind the video by a consistent period, then yes, a simple realignment would do the trick. In my studies to become an audio engineer, that is one of the easy jobs. However, when the sample clocks are running at slightly different speeds, you end up with an audio file that is either too short or too long for the accompanying video file. That's the troublesome part. Pitch is correct, but aligning the two at any point along the timeline, causes an equal misalignment somewhere else. The only solution I can see is to get the recordings right from the get go. That would require something like locked clocks, or a function of the software that somehow recognizes this error as it occurs and resamples one of the files (without any degradation, hopefully) on the fly.
That being said, I will explore the info you've posted and see if it addresses the problem the way I need. I hope so.
If not, in a perfect world it may turn out being one of these DIY mods that solves one of the great plagues of the enthusiast video capture crowd.
I'll post my discoveries here. Thanks very much for the responses and please keep the ideas coming!
Cheers,
Erik.
Neutron7... Are you by any chance the same "Neutron" over at Hexus.net?
Thanks for the info. Especially that last one. If there's a software fix to this, it'd certainly save me some time money and effort. I would only question why if it's so simple, that most people don't seem to know about it? Or why more Designers aren't incorporating this feature into their software? Perhaps it only accounts for straight latency and not sample clock differences? I'll have to have a looksee...
Regarding the Nudge technique, I don't think it's possible if the cause of the problem is drift and not straight latency. If the sound lags ahead or behind the video by a consistent period, then yes, a simple realignment would do the trick. In my studies to become an audio engineer, that is one of the easy jobs. However, when the sample clocks are running at slightly different speeds, you end up with an audio file that is either too short or too long for the accompanying video file. That's the troublesome part. Pitch is correct, but aligning the two at any point along the timeline, causes an equal misalignment somewhere else. The only solution I can see is to get the recordings right from the get go. That would require something like locked clocks, or a function of the software that somehow recognizes this error as it occurs and resamples one of the files (without any degradation, hopefully) on the fly.
That being said, I will explore the info you've posted and see if it addresses the problem the way I need. I hope so.
If not, in a perfect world it may turn out being one of these DIY mods that solves one of the great plagues of the enthusiast video capture crowd.
I'll post my discoveries here. Thanks very much for the responses and please keep the ideas coming!
Cheers,
Erik.
Admittedly, it's many things...
Tiroth,
It's many different things all lumped into one. Many of the software programs that offer to resample audio or drop frames cause degradations in the final product on their own. So, it can be like robbing Peter to pay Paul. "Fixing it in the mix" (eg afterwards) is often an inferior solution to recording it right from the start. Thus, I'm trying to reinvent the wheel, so to speak...
Regarding your question about firewire boxes. I did indeed look into this option first. They offer locked audio and video with no drift as they have both sets of inputs on the box and can use the same clock for both signals. Encode together and voila! no drift. Nice... All Firewire boxes and DV cameras use the same codec: DV. It's efficient and handy and the hardware can handle it very well. It gives nice results for the most part, but...
...And this is a pretty big "BUT":
Look here at this post:
http://www.videohelp.com/forum/viewtopic.php?t=183502
The whole first few pages are interesting reading, but FulciLives' post on the first page (about 20 posts down) was the bit that convinced me to avoid these boxes and the DV codec they employ.
I'm looking to capture the highest quality I can using a top of the line LD player (Pioneer CLD-99) and from my investigations, uncompressed AVI format will give the best results. Since it's uncompressed, there's no loss of information and I can edit and retouch to my heart's content. Yes, in the end I would have to compress to MPEG2 (DVD standard) but not before it's necessary.
-Erik.
Tiroth,
It's many different things all lumped into one. Many of the software programs that offer to resample audio or drop frames cause degradations in the final product on their own. So, it can be like robbing Peter to pay Paul. "Fixing it in the mix" (eg afterwards) is often an inferior solution to recording it right from the start. Thus, I'm trying to reinvent the wheel, so to speak...
Regarding your question about firewire boxes. I did indeed look into this option first. They offer locked audio and video with no drift as they have both sets of inputs on the box and can use the same clock for both signals. Encode together and voila! no drift. Nice... All Firewire boxes and DV cameras use the same codec: DV. It's efficient and handy and the hardware can handle it very well. It gives nice results for the most part, but...
...And this is a pretty big "BUT":
Look here at this post:
http://www.videohelp.com/forum/viewtopic.php?t=183502
The whole first few pages are interesting reading, but FulciLives' post on the first page (about 20 posts down) was the bit that convinced me to avoid these boxes and the DV codec they employ.
I'm looking to capture the highest quality I can using a top of the line LD player (Pioneer CLD-99) and from my investigations, uncompressed AVI format will give the best results. Since it's uncompressed, there's no loss of information and I can edit and retouch to my heart's content. Yes, in the end I would have to compress to MPEG2 (DVD standard) but not before it's necessary.
-Erik.
Erik,
If you are looking at CCIR 601 4:2:2, then you are definitely beyond the level of a DV input box. By the same token, you are almost certainly looking at a very high-end solution, since most computers are not up to the 20MB/s that uncompressed recording requires. It was probably five years ago now, but I know I had to do some system tuning then to reliably capture 6MB/s MJPEG, and this was on striped 7200rpm SCSI drives.
That said, I have used a (non-DV) 4:1:1 capture card and the quality can be quite excellant. I have to think that the example posted is worst case (non-safe red and/or exagerated). I don't know if DV is as good as MJPEG for editing, although you could always reencode to CCIR with no loss for editing. You need 72GB per hour for uncompressed, though, so it gets expensive fast. It's also really slow to edit and hard to playback without big hardware.
AV sync and framedrops always seem to plague video editors. The fact that you have to get the whole system to play together is what generally kills you -- that's why there are turnkey systems out there.
BTW, I was not referring to "fixing it in the mix"--I know this is generally impossible. What I meant was that I think the issue is at the point where the software interleaves the inputs from the AV feeds. There seem to be a lot of problems that exist even with single-board AV capture systems, which makes me think that the clocks on the boards are the least of your problems.
If you are looking at CCIR 601 4:2:2, then you are definitely beyond the level of a DV input box. By the same token, you are almost certainly looking at a very high-end solution, since most computers are not up to the 20MB/s that uncompressed recording requires. It was probably five years ago now, but I know I had to do some system tuning then to reliably capture 6MB/s MJPEG, and this was on striped 7200rpm SCSI drives.
That said, I have used a (non-DV) 4:1:1 capture card and the quality can be quite excellant. I have to think that the example posted is worst case (non-safe red and/or exagerated). I don't know if DV is as good as MJPEG for editing, although you could always reencode to CCIR with no loss for editing. You need 72GB per hour for uncompressed, though, so it gets expensive fast. It's also really slow to edit and hard to playback without big hardware.
AV sync and framedrops always seem to plague video editors. The fact that you have to get the whole system to play together is what generally kills you -- that's why there are turnkey systems out there.
BTW, I was not referring to "fixing it in the mix"--I know this is generally impossible. What I meant was that I think the issue is at the point where the software interleaves the inputs from the AV feeds. There seem to be a lot of problems that exist even with single-board AV capture systems, which makes me think that the clocks on the boards are the least of your problems.
Tiroth,
Yeah, you're right about needing a lot of real estate to capture uncompressed. I'm looking at a pair of the new Maxtor Diamondmax 10 300GB SATA drives with 16MB buffers in a RAID 0 configuration. That should take care of it in terms of speed and size.
From what I've read, however... Hold on, I need to switch computers to finish this thought...
Yeah, you're right about needing a lot of real estate to capture uncompressed. I'm looking at a pair of the new Maxtor Diamondmax 10 300GB SATA drives with 16MB buffers in a RAID 0 configuration. That should take care of it in terms of speed and size.
From what I've read, however... Hold on, I need to switch computers to finish this thought...
Okay
I'm back. Where was I?
From what I understand, the kind of drift I'm referring to is prevented by doing what pro studios do: using a black burst generator that feeds the clocks of all the equipment in the studio; from analog VCRs, to digital sound and video devices...all the way to the software editors and sequencers (running on SMPTE timecode for instance). With everything locked to the same count, or divisions therein, there is no problem with sync. And these studios use the same software the consumer does: Sony Vegas, Adobe Premiere, Avid, Final Cut Pro, Logic etc... The fact that every pro video studio runs on some form of Master Clock and Timecode must mean it's important. One has to wonder, then, why it's important. The only reason I can think of is Sync.
However, there's still a lot I don't know about this subject so I'm loathe to say I know exactly what I'm talking about.
So, I guess the interleaving errors must be clock related too, if the software in a studio needs to run on word clock or SMPTE.
-Erik.
I'm back. Where was I?
From what I understand, the kind of drift I'm referring to is prevented by doing what pro studios do: using a black burst generator that feeds the clocks of all the equipment in the studio; from analog VCRs, to digital sound and video devices...all the way to the software editors and sequencers (running on SMPTE timecode for instance). With everything locked to the same count, or divisions therein, there is no problem with sync. And these studios use the same software the consumer does: Sony Vegas, Adobe Premiere, Avid, Final Cut Pro, Logic etc... The fact that every pro video studio runs on some form of Master Clock and Timecode must mean it's important. One has to wonder, then, why it's important. The only reason I can think of is Sync.
However, there's still a lot I don't know about this subject so I'm loathe to say I know exactly what I'm talking about.
So, I guess the interleaving errors must be clock related too, if the software in a studio needs to run on word clock or SMPTE.
-Erik.
Further thoughts
I'm starting to think that the better route might be to find a way to force my cards to accept a commercial Word Clock signal -like from a MOTU Midi Timepiece AV unit. It could also feed the video software to prevent any software based sync issues.
Tiroth, I'm glad you mentioned the interleaving issue. I hadn't thought of the possibility of software clocking it wrong until then. But when I mentioned SMPTE running Pro Editors, I had an "I could've had a V8!" moment.
-Erik.
I'm starting to think that the better route might be to find a way to force my cards to accept a commercial Word Clock signal -like from a MOTU Midi Timepiece AV unit. It could also feed the video software to prevent any software based sync issues.
Tiroth, I'm glad you mentioned the interleaving issue. I hadn't thought of the possibility of software clocking it wrong until then. But when I mentioned SMPTE running Pro Editors, I had an "I could've had a V8!" moment.
-Erik.
Re: Further thoughts
I think that instead of feeding WC signal into both video and audio capture boards, you can achieve the same results clocking audio card using PLL synchronized with video pixel clock (or PLL for pixel clock synchronized with audio). As alternative you can try to use different capture software like iuvcr (http://www.iulabs.com/iuvcr/index.shtml) wich implements different AV multiplexere.lectronick said:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.