How to measure a Room Response

What methods can we actually use

• MLS (Maximum Length Sequence) technique: a program playbacks a periodic pseudo-random signal, which is almost the same as white noise, but the program is able to correlate it’s acquired copy with original signal in order to recover room impulse response.
• IRS (inverted repeated sequence): the sequence consists of MLS where every even sample is inverted.
• Time-Stretched Pulses: “this method is based on a time expansion and compression technique of an impulsive signal. The aim of using an expansion process for the excitation signal is to increase the amount of sound power emitted for a fixed magnitude of this signal and therefore to increase the signal-to-noise ratio without increasing the nonlinearities introduced by the measurement system” [2]
• Logarithmic SineSweep technique: probe signal is built with a sine-sweep with logarithmic changing frequency, so as to diminish harmonics influence on the result.
• MLS 60.5 dB
• IRS 63.2 dB
• Time-Stretched Pulses 77.0 dB
• SineSweep 80.1 dB

Farina’s method

Probe impulse

• it has to contain all frequencies you want to measure,
• it has to be self-orthogonal, so as to compress itself into Dirac’s impulse during auto-correlation.

How to build a correct reconstructing filter

The generation of the inverse filter is simply matter of time-reversing the excitation signal, and then applying to it (the reversed impulse) an amplitude envelope to reduce the level by 6 dB/octave, starting from 0 dB and ending to

The Code

How to use it on your own

• Measured Room Response is the reason I did this piece of code,
• Deconvoluted Room Response is the result of Wiener Deconvolution of room response with itself just for check,
• Recorded log-sine sweep is the sound the microphone has just heard,
• Spectrum of the record is the spectrum of the previous signal,
• Deconvolved sine sweep is the result of Wiener Deconvolution of record with the measured room response just for check,
• Spectrum of deconvolved sine sweep is the spectrum of the previous signal.

ARC

by the author.

--

--

--

More from Mikhail Baranov

Embedded Software Engineer and Digital Signal Processing specialist

Love podcasts or audiobooks? Learn on the go with our new app.

Mikhail Baranov

Embedded Software Engineer and Digital Signal Processing specialist