Take the 2-minute tour ×
Video Production Stack Exchange is a question and answer site for engineers, producers, editors, and enthusiasts spanning the fields of video, and media creation. It's 100% free, no registration required.

Some time ago, I did a project where the solo guitarist was sitting in a different location. I uploaded rough down mixes to guitarist, who would load the files into his DAW software, record his guitar parts to separate tracks and send the guitar tracks back to me. It worked really well save for one thing: His tracks always seemed a teeny tiny bit off key. Not only that, but he also seemed to drift very slightly tempo wise. The drift became more noticeable on longer takes and that's where it occurred to me that perhaps his and my sampling frequencies were not in sync. After a little trial and error I was able to identify the drift between our sampling rates and subsequently, I would always resample his submissions to get the tracks in tune and sync with my tracks.

Edit: To be clear, the files were sampled at 44.1KHz at both ends. An example: One file is two minutes long. At the beginning of the file, the transient of the guitarists strum is right on top of my drums and bass. But over the course of 2 minutes, the strum consistently lags behind a little further each time. I know is isn't because my guitarist can't keep time--it's too consistent for that. Looking at the last strum, about two minutes into the song, I see the transient of the drum and the guitar is now 200 milliseconds apart. There are supposed to be 44,100 samples every second so that's a lag of 8820 samples. A two minute sample at 44,100Hz is 120*44,100 = 5,292,000 samples so relatively speaking, the lag is less than .17 percent. All things considered that might sound like a pretty good tolerance, but a .17 percent drift after two minutes is 200 milliseconds and very clearly audible.

I know that you can use external clock sources to sync up digital streams, but we didn't have that kind of equipment at our disposal. Besides, although two external clock sources would probably be spot on 44.1/84/96KHz, we'd still have to make sure.

In movie and video production, it is common to use a so-called clapperboard to annotate each take, and to synchronize image and sound. I wonder if maybe marking each take with a reference signal could perhaps help identify equipment that's not sampling the signal at exactly 44.1/48/96KHz. I know I could just add, say, a 1000 HZ square signal at the beginning of each take, but then I would manually have to look at the waveform to see if it is spot on. Are there better ways to deal with this kind of problem?

Is it a common problem or just a case of really bad hardware on my and/or the guitarists side?

share|improve this question

2 Answers 2

up vote 1 down vote accepted

Yes, sample clocks vary, and this can cause recordings on different devices to drift over time.

If you're trying to figure out the frequency at which something was recorded, you could do a bandpass filter around 50 Hz or 60 Hz (depending on which country you're in), and then analyze the hum which is picked up in the recording. If you're in the US, the hum will be at exactly 60 Hz. If you analyze it and get 60.1 Hz, then you know the sampling frequency of that device was 0.16% slow.

http://en.wikipedia.org/wiki/Electrical_network_frequency_analysis

http://www.theregister.co.uk/2010/06/01/enf_met_police/

share|improve this answer
    
I like your suggestion. For once, the dreaded hum might actually turn out to be helpful, if only a little. On the other hand, I know that the frequency of alternating current from your local power plant is also prone to drifting. AFAIK, it is pretty common for power plants to only guarantee the number of oscillations over, say, 24 hours. So if they're falling behind or overshooting at the end of a 24 period they will simply increase or decrease the frequency to meet the guaranteed number. –  Kim Burgaard Dec 16 '10 at 8:15
    
Oh, and that's exactly what the links you provided talk about. So unless I happen to know what the drift was at any given time at the remote recording location, I'm really no better off, yeah? –  Kim Burgaard Dec 16 '10 at 8:18
    
The frequency is maintained within a few parts per million, and the correction is 0.02 Hz (0.033%): en.wikipedia.org/wiki/… –  endolith Dec 16 '10 at 15:14

What file format are you passing back and forth? The sample rate is stored in the file. No need for a reference tone.

One of you is likely recording at a different sample rate than the other. Probably one is recording at 44.1kHz, and the other 48kHz, or 88.2/96, etc.

Now, audio interfaces do vary in sample rate frequency. 44,100Hz can be 44,099Hz on one, and 44,0102Hz on another. This won't throw your tuning off by any perceptible amount. Given enough time, you can start hearing a difference. However, if you are recording against a track already recorded, this isn't an issue. The playback sample rate on his DAW while recording will be the same rate as the record rate.

So, if you put together a few parts and send him the files, and then he records a couple tracks and sends you the files, there should be no issue here.

share|improve this answer
    
I just added an example calculation above. I guess one of the things that make audio fidelity hard, is that our ears are very finely tuned to detect differences. Even though a lot of us don't have absolute hearing in terms of detecting notes, most people can tell when to sounds are just a few milliseconds apart. So even with your example of one device sampling at 44,099Hz and another at 44,102Hz, the relative difference is about .0068%. For a five minute recording that means the length of the files will be about 20 ms apart and, depending on the sound material, beginning to become audible. –  Kim Burgaard Dec 14 '10 at 5:07
    
@Kim, thanks for clarifying. Again, if he is playing back one of your tracks while recording, it won't be an issue, as the playback/record clock come from the exact same internal clock source. If he records against nothing, and you record against nothing, and then you try to merge the files, you will get trouble. –  Brad Dec 14 '10 at 14:18
    
your response and comment makes sense to me, yet this is what I saw. Stretching the guitarists samples upon import fixed minuscule key and tempo drift issues. –  Kim Burgaard Dec 15 '10 at 3:05
    
@Kim, so he is playing something back while recording? Very odd. what kind of hardware set up is he using? I'm assuming he is recording on the same box as playing back? If he has multiple audio interfaces hacked together with something like ASIO4ALL, then sync issues can occur with no external sync. Other than that, I'm fairly stumped. Sorry I can't be of more help! –  Brad Dec 15 '10 at 16:32
    
I actually don't know what kind of audio interface or software the guitarist used, but I believe it was fairly basic. –  Kim Burgaard Dec 16 '10 at 8:10

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.