Sound file protocols
(notes to self)

  • Data preparation (in sLab):
    • For 25 bit data, squeeze into 24 bits:

      % squeeze gram_in gram_out 0 24

    • save grams to audio files at desired sample-rate conversion (e.g., ×245):

      % audio gram_in -s 24 -t aif -C 245

  • Data preparation (continued):
    • Microseismic noise. The natural microseismic background noise peaks at around 0.15-0.2Hz (5-7 second period). Depending on the time scaling factor, you'll eventually want to either insert a notch or a high-pass filter in this vicinity. Make a note of the scaled microseismic noise band.

      In this example, we have a time-scaling factor of 245. The bounds of the microseismic noise band scales as follows:

      0.15 Hz × 245= 37 Hz
      0.2 Hz × 245 = 49 Hz

      You can therefore expect to see a broad noise peak in the range 37-49Hz.

  • Data assembly (in ProTools):
    1. Import audio files to tracks
    2. Rename each track to a simpler name (just the station id)
    3. Group the tracks
    4. Duplicate the playlist for the group
    5. Suspend groups
    6. in Slip mode, snip out obvious glitches (clips, offsets), applying fade-in/outs at clip boundaries.
    7. Unsuspend the groups
       
    8. Duplicate the playlists
    9. Apply to each track: EQ ("Q6 mono"; "heavy lo-cut" settings (4 corners at 100Hz) — this takes care of the microseismic noise band, and then some.)
       
    10. Duplicate the playlists
    11. Normalize each track to itself
       
    12. Duplicate the playlists
    13. Review all tracks, snipping out any remaining clips, etc., and applying fade-in/outs at clip boundaries. Re-normalize any tracks that needed editing here.
       
    14. Duplicate the playlists
    15. Apply to each track: compressor "C1 Comp Mono" ("IRIS Basic" settings)
    After this, you'll have the following tracks:
    Track
    suffix    		Contains
    -raw audio
    01edited to remove clips and other garbage
    02EQ: Q6 Mono ("heavy lo-cut")
    03Normalized
    04Compressed: C1 Comp Mono ("Iris basic")
    05Normalized
  • Evaluate the data set (in ProTools):
    1. Listen to each station and evaluate quality of recording. Rank each on a scale of 0-3:
      1. Useless. Incomplete data, too many dropouts, impossible microseismic noise level, etc.
      2. Possible. Has problems, but there's something of interest there that might be worth digging out later.
      3. Good. Nice signal, interesting sounds, but needs some additional tweaking of compressions, etc.
      4. Excellent. Wow.
      Record the ranking (and any notes) on a printed copy of the network checklist.
    2. On the azimuthal planning map, highlight the stations that received a ranking of 2 or better.
    3. Select eight stations with rankings ≥2 that also have satisfactory azimuthal coverage. It will probably be difficult to find eight stations that satisfy these criteria. Maybe a narrower time window will have better data. Maybe you can trade station X, with a low ranking but good azimuthal position, for station Y, with a high ranking but poorer position. It's an iterative process, a bit like shuffling cards. Depending on your sonic goals, trade-offs will invariably be required.

    The result of this process is a set of eight workable recordings that have good spatial distribution.

  • Fine-tune the recordings (in ProTools). Work with each track individually, occasionally pairing them up in stereo for comparison, then going back to the individual track. Do not edit individual tracks while listening to the full 8-track playback — there's way too much going on! Things to listen for:
    1. Test tones. → Apply an EQ notch.
    2. Anthropogenic noise (trains, etc.). → Is it fleeting or distracting? If the latter, consider snipping it out and replacing it with a quiet section. If there are too many such edit points, consider replacing this track with one from a quieter station. (If you evaluated the recordings properly in the previous step, you shouldn't have to resort to this!)
    3. Microseismic noise. Some coastal stations (e.g., SHEL, KDAK, etc.) may require additional microseismic noise filtering. Let your ears be your guide here. Earth noise that was above the reach of the "heavy lo-cut" filter that we applied earlier will appear as a tiresome rushhh in the neighborhood of 125Hz; apply a broad notch (Q=6.3) at around 95Hz. Do not apply another high-pass filter here, as we still want to hear the stunning lows from high-amplitude surface waves.