Notes on Minidisc to Speech Analyzer Transfer

From SurveyWiki
Jump to navigationJump to search

Ladefoged (2003)<ref>Ladefoged, Peter. 2003. Phonetic data analysis. Malden, MA: Blackwell Publishing Ltd.</ref> states that there is seldom need to record at a rate higher than 22,000 Hz. This will theoretically capture frequencies up to 11,000 Hz. Higher sampling rates (e.g. 44,100 Hz) capture higher frequencies, but these may be too high for the human ear to hear anyway. Sampling at higher frequencies increases file size, as well. SonyMD players record at 44,100 Hz (see specifications in the manual). When transferring the data to the computer, Speech Analyzer (SA) lets you choose the sampling rate and the recording volume.

In SA, you do not want to have any “red” in the record window in SA. I did an experiment comparing 44,100 Hz vs. 22,050 Hz, SA recording volume 50% vs. 100%, and MD player volume 100% vs. lowered. The following summarizes the results:

  • Lowering the volume on the MD player helps a lot with waveform amplitudes that otherwise would go off the screen in SA. For some words maybe it helps too much; the amplitudes are so low that it is hard to see things.
  • Lowering the recording volume in SA helps with amplitude, but not very much. It does help, though. It reduces the file size only slightly. Note that it does not look like it has any effect when you are recording as the green/yellow/red indicator still shows the same colors. But the waveform will be different (lower amplitude) with a lower recording volume.
  • Formants: All options gave about the same formants. Not exactly the same, but seemingly close enough to each other.
  • Using a sampling rate of 22,050 Hz seems to do better for formants than 44,100! The formant tracks seem smoother and the formant plotting seems to work better. Sometimes, for 44,100, there will be no blue dot shown at all, but for 22,050 there is one (that seems correct). Also, the file size is reduced by half. In conclusion, I recommend the following when you transfer sound data to the computer via a patch cord using SA:
  • Experiment to find what the best combination of MD player volume and SA recording volume is for your file BEFORE you start recording for real. You may need to change the settings for different tracks. Check them each time!
    • Note the effect on amplitude and on pitch.
    • For the data in this example, the best settings turned out to be about 40% record volume on SA and a little over 50% play volume on the MD player. The right settings may depend on the volume of your recording.
  • Set SA to record at a sampling rate of 22,050 Hz.

<references />