3 soundcards, 3 times different samplingrate! Interesting...

[Marijn]

Hi all,

last week we were recording some sound from a DAT-tape to the PC using Cooledit. After a few minutes we noticed that the recordtime of Cooledit and the playtime of the DAT-recorder started to differ. When analysing the difference, we noticed that the DAT-recorded was perfectly synchronous with the time, but that the PC was slow! It got like a second behind in about five minutes!

Since we got curious, we tried two other PC's. One of them (laptop) was having like 1 sec of difference in 10 minutes, while the other was like 1 sec in 8 minutes.

I don't (yet) know if this is related to Cooledit, but it seems strange though.

I don't know about the soundcard in the laptop, but the other two were SB PCI128. Sound was captured at 8 kHz.

This might be something related to synchronisation- or dropped frames problems.

Marijn


[This message has been edited by Marijn (edited 29 January 2001).]

[Marijn]

I totally agree with you, but if 8kHz ain't 8kHz, how reliable is the 44kHz sampling frequency we are using?

If this time base is used as a reference for NLE cards, that are skipping frames in order to keep everything synchronised...

Ah well, it maybe nothing.

[Dr Mordrid]

Variances in the audio sample rate from a standard setting isn't limited to sound cards.

Many DV cams vary from the typical 48k sample rate by enough to cause an AV synch loss. Ex: some Canons sample at 48.009 khz, which can cause the audio to get out of synch with the video by about 1 second per hour.

Dr. Mordrid

[Michel Carleer]

Marijn,
The standard MS defined frequencies of 11025, 22050 and 44100 Hz are probably close to the true value, but I doubt very much the precision of the other frequencies. All these frequencies are derived by division from one single high frequency oscillator. If the frequency of the oscillator is chosen to be an exact multiple of 44.1 kHz, it probably is not an overtone of 8 kHz.
Remember, this is only one possible cause. The only way to know for sure is to measure the freqs.
As you say that your dat is very precise, you could use it to make the measurement. The ratio of the time indicated by Cool Edit by the true time would be a measure of the true sample frequency. You can do it at the various sample rates and see if it is better for some of them.
Anyway, a 1 second error on 3 minutes is only about 0.5%.
Michka

[Michel Carleer]

Or maybe the fact that when you ask for 8 KHz sampling frequency... well it ain't exactly 8 KHz. The total time of a sound file is calculated by multiplying the sampling frequency by the sample number. If the real sampling frequency is not exactly what you asked, but only the best approximation from the division of the frequency of an oscillator on the sound card, then the calculated time will be incorrect. It would be interesting to actually measure precisely the real sampling frequency.
Michka