Equalization to Remove Speaker and Room Effects

[analog_guy]

In my day job I have been working with various types of equalization to mitigate the reflection and attenuation of PCB traces for data rates in the 8-10 Gbit/sec range. These techniques include FIR and DFE and CTLE. It would be interesing to see if these techniques could be applied to audio.

The basic idea would be to use a microphone to measure a pulse response of an audio chain and then select the appropriate equalization coefficients such that the signal measured at the microphone closely replicates the original voltage pulse response. Coefficient selection could be done by various optimization techniques: LMS, zero forcing, etc. The equalization length would need to be fairly long, essentially the room and speaker's decay time times the sample rate. Sample rate would need to be at at least 2x Nyquist. So we are talking about filter lengths of several thousand taps, although each tap would not need to be unique.

To the extent that the audio chain and room acoustics are linear with signal amplitude, the coefficients will remain independent of signal amplitude.

Work of this nature was done years ago by telephone companies to implement adaptive echo cancellation. For audio, however, the equalization would not need to be continuously adaptive.

[Key]

It's a little different when you are dealing with 3-D reflections and there is already a processor (your brain) doing some unknown corrections to the received signals. I think most likely you will cause more harm than good.

[Speedskater]

The more you adjust the equalization curve at one listening location, the worse it gets at all other listening positions.
The more you adjust the transient response curve, the worse the sustain response curve gets.