Some technical notes on our lecture recordings

By SM | Published: March 21, 2011

We use a Tas­cam DR-01 dig­i­tal recorder and a Shure SM-58 micro­phone to record lec­tures at the Vedanta cen­ter. We used to plug the recorder right into the sound sys­tem, but our wire­less mikes were drop­ping out from time to time, and the ampli­fier was adding a low but audi­ble hum. This way we have an extra micro­phone at the lectern, but we get a clean record­ing. When record­ing classes in the liv­ing room, we use the built in stereo micro­phone on the Tas­cam recorder, and hide the recorder up on the man­tel­piece. We are record­ing in mp3 for­mat at 256 kbps, rather than wav for­mat, to reduce file size. Though mp3 is a lossy for­mat, we thought that these lec­tures were not so impor­tant that we need to record in wav or another non-lossy for­mat, espe­cially since we are record­ing at a high rate.

Audio pro­cess­ing: we use the free open source pro­gram Audac­ity to pre­pare the lec­tures for online pod­cast­ing. It’s a great pro­gram, though it takes some time to learn how to use it. We take the fol­low­ing steps.

  • First, amplify (the micro­phone is not boosted, so record­ing lev­els are very low).
  • Remove back­ground hiss via noise reduction.
  • Apply com­pres­sion (to even out the sound volume).
  • Apply a hard lim­iter to catch any high points that the com­pres­sor didn’t quite level out.
  • Amplify again to -1.5 dB.
  • Fade in at begin­ning, fade out at end.
  • Remove any coughs etc that we can catch (but we don’t lis­ten to the whole thing—often one can see such sounds by their very loud volume).
  • Remove the announce­ments and other unim­por­tant stuff.
  • Export to 256 kbps for our archive, and again at 64 kbps for upload to the website.

Advanced func­tions include the abil­ity to remove hums or tones of spe­cific fre­quen­cies. This is great for remov­ing elec­tri­cal hum, and also for knock­ing out feed­back. We have been able to save some poor record­ings spoiled by feed­back, using the Nyquist notch func­tion. First, one ana­lyzes the feed­back (plot spec­trum) to deter­mine the exact fre­quency, then one runs the notch over the whole record­ing, and the feed­back is gone.

The Nyquist notch for­mat for sin­gle, mono tracks: (notch2 s * Q), where * is the fre­quency to be notched, and Q is a num­ber which deter­mines the “depth” of the notch. We have found 50 to work well; the lower the num­ber, the deeper the notch. For exam­ple, an elec­tri­cal hum in the USA will have a pri­mary fre­quency of 60 Hz, so our notch (via the Nyquist prompt) would be (notch2 s 60 50).

Typ­i­cally a hum will have sev­eral fre­quen­cies; elec­tri­cal hums might have fre­quen­cies at 60, 180, and 300 Hz. To run the notches simul­ta­ne­ously, the fol­low­ing code can be used:
(notch2
(notch2
(notch2 s 60 50)
180 50)
300 50)
The fre­quen­cies to be notched can be deter­mined from the “plot spec­trum” func­tion. One selects a very short (.25 sec or so) por­tion of the track in which the tone to be removed is dom­i­nant, ana­lyzes it by “plot spec­trum”, view­ing the “log frequency”.

For stereo tracks, the notch for­mat is:
(vector (notch2 (aref s 0) * Q) (notch2 (aref s 1) * Q) )