Audio drift with a long DV capturing session at DVinfo.net

Ryan P. Green · February 3rd, 2009, 12:28 PM

Before you ask, yes, I checked to make sure everything audio-related is set to 48khz.

What I'm doing is capturing live off a SDI feed to a Sony DV deck in the field (I didn't make the setup, don't ask me...). It records audio in 48 khz, and then I connect it to a Sony DSR-25 deck. FCP is set to capture at 48 khz (I can screenshot it if people want), however over the course of an hour-long capture would the audio drift? It plays back normally straight off of the tape, so I don't really know what's going on here.

Shaun Roemich · February 3rd, 2009, 02:33 PM

Dropouts on tape can cause audio to go out of sync in FCP, regardless of input device.

John Miller · February 3rd, 2009, 03:16 PM

Your software is responsible. Despite popular myth, DV cannot drift. The specification is very explicit about audio sampling vs. video sampling rates. It is up to the software to correctly interpret the DV stream.

Is the drift during playback from the timeline within FCP or from rendered files? How do the captured files play within other applications such as Quicktime? Or if you send them back to the device?

Mike Barber · February 3rd, 2009, 05:22 PM

Quote:

Originally Posted by Ryan P. Green

however over the course of an hour-long capture would the audio drift? It plays back normally straight off of the tape, so I don't really know what's going on here.

During the capture, did you leave the computer alone or were you checking email, surfing the net, etc?

The very first time I captured a whole tape I didn't know better and went surfing while waiting for all 50 minutes to capture... when it was done and I started scrubbing through single clip of footage in FCP I discovered the audio was getting way off. A recapture while leaving the computer alone to do this single task resulted in a new clip in perfect sync all the way through.

Steve Oakley · February 3rd, 2009, 07:13 PM

DV audio can drift. there are 2 DV specs, one for Pro gear, and one for consumer. consumer gear is NOT clock locked and is allowed to drift a bit in samples / frame as long as it averages out after so many seconds. there was a huge problem in the days of FCP 1-3. at one point, apple finally added code to adjust it out automatically after capture, and the problem went away. apple also had 2 patch files for as I recall, the VX100 and the other I don't remember.

so back to the original question. how much is the sync going out ? same direction ? or does it vary over time, sometimes even catching up ?

given that you are using SDI, it should be less then a frame out, and not accumulate.

could be a deck problem / setting

could be a camera setting problem

could be a combination of specific camera & deck

it could also be a weird FCP problem. does your video card have a control panel ( AJA ) to indicate what it detects as the incoming video ? sometimes the card doesn't lock right and it sees the video at one rate, the audio at another, and then the audio drifts out getting worse the further down you get, even with SDI.

could maybe even be some weird SDI video signal problem with the deck

and have you tried FW ?

John Miller · February 3rd, 2009, 08:36 PM

Quote:

Originally Posted by Steve Oakley

DV audio can drift. there are 2 DV specs, one for Pro gear, and one for consumer. consumer gear is NOT clock locked and is allowed to drift a bit in samples / frame as long as it averages out after so many seconds. there was a huge problem in the days of FCP 1-3. at one point, apple finally added code to adjust it out automatically after capture, and the problem went away. apple also had 2 patch files for as I recall, the VX100 and the other I don't remember.

The specification is very clear. Irrespective of locked vs. unlocked, the average number of bytes of audio per video frame must be exact (e.g., 48000 x 2 x 2). There is no wiggle room for this. The difference between the two is that locked specifies a very rigid sequence of numbers of bytes per frame (for NTSC, this repeats on a 5-frame cycle) and for PAL every frame has the same number of bytes. Additionally, the first frame of the cycle (for NTSC) must coincide with the first frame of the color frame sequence. Locked audio is essential for mixing audio tracks between DV sources easily, particularly with real-time equipment since the number of bytes per frame is predictable and the same between sources. Unlocked audio can have a more lax sequence but must still come out to the same rate when averaged. Captured files that play just fine with one application may drift with another. If the software does its job properly, no drift will occur. On Windows, a common cause is the use of the Type-2 DV AVI file. If drift were a true phenomenon of unlocked audio, it would manifest itself with DV hardware, not just poorly implemented software. Such drift would also render the hardware out of specification.

Reference: Tables 17 and 19 of IEC 61834-2 (commonly refered to as the DV specification).

Steve Oakley · February 3rd, 2009, 08:59 PM

specs are great, but that doesn't mean they are followed. feel free to quote any specs you like, but back in the real world companies building the gear don't always follow them all the way, if at all.

Mr. Adam Wilt put it nicely -

Unfortunately, such precisely-locked audio clocks are expensive. Since DV was designed as a consumer format, unlocked audio was allowed as a cost-saving measure. In unlocked audio, the audio clock is allowed some imprecision, such that there can be a variation from the locked spec of up to +/- 25 audio samples written to tape for every frame, instead of a precise and exact number.
This economy measure is simply one of allowing the audio clock to "hunt" a bit around the desired frequency; the phase-locked loop (or other slaving method) used to keep the audio sampling in sync with the video sampling can have a bit more slop in its lock-up, with the audio sampling sometimes running a bit slower, sometimes a bit faster, but always staying in sync over the long run. The total amount of sync slippage allowed in unlocked audio is +/- 1/3 frame -- not enough to really worry about.

-------------
the full read is here dated 2005

The DV, DVCAM, & DVCPRO Formats -- tech details, FAQ, and links.

so hardware can keep up. so regardless of the bytes / frame reserved for audio samples on tape & in the data stream, that doesn't mean they get filled all the time. yes its a software issue, but in the case of SDI, the DV audio has already passed thru the audio hardware in the deck and its in sync +- 1/3 of a frame. so the original poster needs to provide some more details

John Miller · February 3rd, 2009, 09:41 PM

Adam is saying the same thing as me. Regarding unlocked audio, he states "but always staying in sync over the long run" - i.e., no drift. Over very short durations, there will be deviations but beyond a second or two the audio frequency sample rate *must* be exact - otherwise the equipment is not DV-compliant.

The +/- 25 samples per frame is the range permitted for deviation within a given frame for each channel. It does not permit every frame to be, say, 1 sample short of the target. Adam is incorrect regarding the +/- 1/3 frame slippage. The maximum amount allowed equates to approx. 1.3 percent of a frame (makes me wonder if it was just a typo).

DV hardware has to prove adherence to the specifications in order to get the DV logo. Most DV editing software doesn't carry the logo and so is not necessarily validated against the specifications. The unlocked audio leading to drift myth comes about because of misleading (or ill-informed) marketing and/or misunderstanding by software developers of the finer points of the specifications. Having written DV encoding/decoding software from the ground up, I bear the scars of misinterpreting the specifications. I have the luxury of open-ended project completion dates. Companies such as Apple, Adobe, Sony etc do not. Actually, one of the reasons I like Sony's software (including Vegas) is that they (nearly always) get it right as far as DV is concerned (which they should). For years and years, Adobe failed to fix their most fundamental mistake regarding audio that invariably led to drift even with non-DV sources. The fix was trivial (changing the NTSC frame rate from 29.97 to 29.97002997 in the AVI header...)

I can't comment on SDI and, I agree, we need more info regarding this thread's particular issue.

Mike Barber · February 3rd, 2009, 09:55 PM

Quote:

Originally Posted by Steve Oakley

yes its a software issue, but in the case of SDI, the DV audio has already passed thru the audio hardware in the deck and its in sync +- 1/3 of a frame. so the original poster needs to provide some more details

I took the OP to mean that the footage was captured to tape via SDI, and that the picture and sound on the tape is in sync. Then upon capturing that tape (my presumption is he did this via Firewire, but that needs to be confirmed by the OP) the result was a clip with picture and sound out of sync.

So Ryan, can you please clarify for us what you are doing and on what machine (technical specs)?

Ryan P. Green · February 6th, 2009, 10:48 AM

yes, capture was via firewire. Doing it all on a Mac Pro with latest version of FCS. I'm going to be doing some more testing today with other decks/machines, to narrow down where exactly the problems are.

Pete Cofrancesco · February 6th, 2009, 07:21 PM

Quote:

Originally Posted by Shaun Roemich

Dropouts on tape can cause audio to go out of sync in FCP, regardless of input device.

this has been my experience too

February 3rd, 2009, 12:28 PM	#1
Ryan P. Green Regular Crew Join Date: Jun 2004 Location: USA Posts: 35	Audio drift with a long DV capturing session Before you ask, yes, I checked to make sure everything audio-related is set to 48khz. What I'm doing is capturing live off a SDI feed to a Sony DV deck in the field (I didn't make the setup, don't ask me...). It records audio in 48 khz, and then I connect it to a Sony DSR-25 deck. FCP is set to capture at 48 khz (I can screenshot it if people want), however over the course of an hour-long capture would the audio drift? It plays back normally straight off of the tape, so I don't really know what's going on here.

February 3rd, 2009, 02:33 PM	#2
Shaun Roemich Inner Circle Join Date: Sep 2002 Location: Vancouver, British Columbia (formerly Winnipeg, Manitoba) Canada Posts: 4,088	Dropouts on tape can cause audio to go out of sync in FCP, regardless of input device. __________________ Shaun C. Roemich Road Dog Media - Vancouver, BC - Videographer - Webcaster www.roaddogmedia.ca Blog: http://roaddogmedia.wordpress.com/

February 3rd, 2009, 03:16 PM	#3
John Miller Major Player Join Date: Oct 2005 Location: Hillsborough, NC, USA Posts: 968	Your software is responsible. Despite popular myth, DV cannot drift. The specification is very explicit about audio sampling vs. video sampling rates. It is up to the software to correctly interpret the DV stream. Is the drift during playback from the timeline within FCP or from rendered files? How do the captured files play within other applications such as Quicktime? Or if you send them back to the device?

February 3rd, 2009, 07:13 PM	#5
Steve Oakley Trustee Join Date: Apr 2006 Location: United States Posts: 1,158	DV audio can drift. there are 2 DV specs, one for Pro gear, and one for consumer. consumer gear is NOT clock locked and is allowed to drift a bit in samples / frame as long as it averages out after so many seconds. there was a huge problem in the days of FCP 1-3. at one point, apple finally added code to adjust it out automatically after capture, and the problem went away. apple also had 2 patch files for as I recall, the VX100 and the other I don't remember. so back to the original question. how much is the sync going out ? same direction ? or does it vary over time, sometimes even catching up ? given that you are using SDI, it should be less then a frame out, and not accumulate. could be a deck problem / setting could be a camera setting problem could be a combination of specific camera & deck it could also be a weird FCP problem. does your video card have a control panel ( AJA ) to indicate what it detects as the incoming video ? sometimes the card doesn't lock right and it sees the video at one rate, the audio at another, and then the audio drifts out getting worse the further down you get, even with SDI. could maybe even be some weird SDI video signal problem with the deck and have you tried FW ?

February 3rd, 2009, 08:59 PM	#7
Steve Oakley Trustee Join Date: Apr 2006 Location: United States Posts: 1,158	specs are great, but that doesn't mean they are followed. feel free to quote any specs you like, but back in the real world companies building the gear don't always follow them all the way, if at all. Mr. Adam Wilt put it nicely - Unfortunately, such precisely-locked audio clocks are expensive. Since DV was designed as a consumer format, unlocked audio was allowed as a cost-saving measure. In unlocked audio, the audio clock is allowed some imprecision, such that there can be a variation from the locked spec of up to +/- 25 audio samples written to tape for every frame, instead of a precise and exact number. This economy measure is simply one of allowing the audio clock to "hunt" a bit around the desired frequency; the phase-locked loop (or other slaving method) used to keep the audio sampling in sync with the video sampling can have a bit more slop in its lock-up, with the audio sampling sometimes running a bit slower, sometimes a bit faster, but always staying in sync over the long run. The total amount of sync slippage allowed in unlocked audio is +/- 1/3 frame -- not enough to really worry about. ------------- the full read is here dated 2005 The DV, DVCAM, & DVCPRO Formats -- tech details, FAQ, and links. so hardware can keep up. so regardless of the bytes / frame reserved for audio samples on tape & in the data stream, that doesn't mean they get filled all the time. yes its a software issue, but in the case of SDI, the DV audio has already passed thru the audio hardware in the deck and its in sync +- 1/3 of a frame. so the original poster needs to provide some more details

February 3rd, 2009, 09:41 PM	#8
John Miller Major Player Join Date: Oct 2005 Location: Hillsborough, NC, USA Posts: 968	Adam is saying the same thing as me. Regarding unlocked audio, he states "but always staying in sync over the long run" - i.e., no drift. Over very short durations, there will be deviations but beyond a second or two the audio frequency sample rate must be exact - otherwise the equipment is not DV-compliant. The +/- 25 samples per frame is the range permitted for deviation within a given frame for each channel. It does not permit every frame to be, say, 1 sample short of the target. Adam is incorrect regarding the +/- 1/3 frame slippage. The maximum amount allowed equates to approx. 1.3 percent of a frame (makes me wonder if it was just a typo). DV hardware has to prove adherence to the specifications in order to get the DV logo. Most DV editing software doesn't carry the logo and so is not necessarily validated against the specifications. The unlocked audio leading to drift myth comes about because of misleading (or ill-informed) marketing and/or misunderstanding by software developers of the finer points of the specifications. Having written DV encoding/decoding software from the ground up, I bear the scars of misinterpreting the specifications. I have the luxury of open-ended project completion dates. Companies such as Apple, Adobe, Sony etc do not. Actually, one of the reasons I like Sony's software (including Vegas) is that they (nearly always) get it right as far as DV is concerned (which they should). For years and years, Adobe failed to fix their most fundamental mistake regarding audio that invariably led to drift even with non-DV sources. The fix was trivial (changing the NTSC frame rate from 29.97 to 29.97002997 in the AVI header...) I can't comment on SDI and, I agree, we need more info regarding this thread's particular issue.

February 6th, 2009, 10:48 AM	#10
Ryan P. Green Regular Crew Join Date: Jun 2004 Location: USA Posts: 35	yes, capture was via firewire. Doing it all on a Mac Pro with latest version of FCS. I'm going to be doing some more testing today with other decks/machines, to narrow down where exactly the problems are.