Wednesday, November 17, 2010

Frequency Analysis of On-Air Audio Recordings

I used the FFT frequency analysis feature in my audio editing software to produce some views of audio from several members' transmissions as received on my Heathkit SB-104A. The results are presented in this photo album. Click on the first thumbnail image to display it full size with its caption. You can use the two arrowheads above the image to go backward and forward to compare images. Or you can select the slide show mode and navigate using controls below the photo. All graphs were made with the same settings in the audio software. Two overlay images show a pair of curves on a single slide.

Here are some additional notes:
  • All audio recordings are influenced by the receive passband and audio filtering of my SB-104A 
  • A local peak appears around 3 kHz in all graphs. This is a birdie in my receiver and should be ignored.
  • Each horizontal line represents a 3 dB amplitude step -- a factor of 2 power ratio
  • Each graph is the result of FFT analysis of the speech sample that was recorded, and the frequency content varies according to what slice of time you choose to analyze. Speech is a complex, continuously changing waveform. Pure, steady tones would give more consistent results.
  • My software can present a real-time FFT of the audio as it plays. In that case the frequency display varies continuously as expected for a complex, continuously changing waveform.
  • Increasing the FFT size from the 1024 used in these graphs to higher numbers (2048, 4096, etc.) reveals increasingly fine details, but does not affect the overall shape of the response shown in the lowest (1024) setting.