[music-dsp] Recording very long sound files

Laszlo Toth tothl at inf.u-szeged.hu
Wed Sep 14 07:41:28 EDT 2011

On Wed, 14 Sep 2011, Richard Dobson wrote:

> How is the recording controlled - some external timer telling the process
> to start and stop?

Actually, we are recording a fixed number of samples (which is calculated
from the sampling rate). So we ask the system time - record N samples -
ask the system time. The difference between the two time instances should
be 8 hours, but in most cases it is somewhat longer. The difference varies

> Are you sure it is not simply a slight difference in start/stop times?

In most cases it's slight, but in eight days of testing it was 1 minutes
in one case on 5 minutes in another case - so unfortunately the whole
system seems unreliable.

> Do you actually find discontinuities in the recording?

We want to avoid listening to dozens of hours of recordings. Also, the
recordings are made for a sleeping center, so they contain the sound
people give during sleeping - in most cases just some very low-frequency

> You are using long long to index this aren't you?

I hope my programmers are better than committing such basic mistakes.
Also, in that case the difference would be fixed, while in our case it
varies from day to day.

> Personally I'd set up a machine with ntp time sync and for each packet
> you read from the sound card make a log of the number of samples you've
> seen so far and the PC time (sycned to network time).

I'm afraid we cannot synchronize to network time, as the whole equipment
is set up in a hospital room. But otherwise the idea is good, we'll try to
do some more detailed logging of packages and times.

Thanks to everybody who replied.

               Laszlo Toth
        Hungarian Academy of Sciences         *
  Research Group on Artificial Intelligence   *   "Failure only begins
     e-mail: tothl at inf.u-szeged.hu            *    when you stop trying"
     http://www.inf.u-szeged.hu/~tothl        *

More information about the music-dsp mailing list