gene said:
All sound in small rooms suffer from standing waves
Using chirp tones simply masks them (it doesn't eliminate them) since they lack the resolution of conventional test tones.
One should never use an SPL meter to do frequency response testing since the meter itself is very nonlinear. This is why we use calibrated, high precision mics with our LMS and/or RTA based systems.
Hi Gene,
The purpose of the test signals used is not to mask standing waves. In fact, the test signal doesn't do anything to standing waves. The signals used for measurement must be such that the impulse response of the (assumed) Linear Time Invariant (LTI) system called the room is measured as accurately as possible.
There are many factors that affect the accuracy of this measurement including background noise, whether it is correlated or not to the test signal, microphone calibration, speaker directivity, etc.
Below are some notes that I use in my graduate course at USC on audio signal processing to introduce the topic.
As you said, measuring with an SPL meter and test tones is woefully insufficient. I am well aware that Audioholics employs professional methods. Unfortunately many consumers don't have access to good equipment and try to draw conclusions from SPL meter "measurements". They should instead rely on technical reviews such as the in-depth ones found on this site for their information.
Best regards,
Chris
CTO, Audyssey Laboratories
Methods for measuring room responses
1. Maximum Length Sequence (MLS)
This method is based on cross-correlation of the signal that is measured in the room with the a pseudo-random sequence that is generated in advance. It is not a statistical, but rather a deterministic method. MLS signals perform better than pure white noise (gaussian) signals because of their binary nature that allows for very fast processing. The MLS sequence is repeated with an intervening silence signal and the average measured signal is used to deconvolve with the original sequence to obtain the room response (also applies to loudspeaker measurements). The length of the sequence must be kept sufficiently high so that time aliasing problems in the deconvolved response (arising from late reflections or long reverberation times) are minimized.
2. Sweep Signals
This method uses circular deconvolution because of the periodic nature of the test signal. A sine sweep is played through the system and then measured by the microphone. The Fourier transform of the measured signal is divided by the Fourier transform of the input signal and then the inverse Fourier transform of this ratio is taken. The result is the impulse response of the loudspeaker-room system at the point of measurement. This will, of course, vary as the microphone is moved to other locations.
An improvement to the linear sinusoidal sweep is a logarithmic sweep that provides higher signal-to-noise ratio at lower frequencies. Furthermore, a logarithmic sweep performs better than MLS because it is possible to separate the loudspeaker distortion terms from the the room response components.
Further reading:
1. Bharitkar, S. and Kyriakakis, C., "Immersive Audio Signal Processing", Springer Science, New York, NY, (2006).
2. Stan G-B, Embrechts J-J, and Archambeau D, Journal of the Audio Engineering Society, 50(2) 249-262, (2002).
