An Inexpensive Measurement System

In the following, you will find a description of my acoustical measurement system, which is based on very inexpensive parts and the use of my PC as a signal generator and detector.

Thoughts on Acoustical Measurements

In the design of loudspeakers I have untill recently relied on intuition and my hearing preferences rather than measurements. This is partly because I never thought measurements could represent what I could hear and partly because I never did bother to make a measurement system :0).
The ear can be described as an A-weighting filter combined with critical band filtering, which operates in the joint-time frequency domain. Our hearing is sensitive to both the time domain and the frequency domain coherently and this coherence is based on an unlinear weighting. One thing is what we hear at the listening position, another thing is how it sounds if we move an inch down, backwards etc.
For example: If the tweeter is delayed due to a misalignment in the enclosure baffle it will sound dark and therefore it should be louder than the midrange driver. This will look terrible on the frequency magnitude response but probably sound better.
What really matters in that respect is what you think, -not the measurement.
I would never use measurements for anything but pointers in a design, simply because they are incomplete for true representation. As a reference, I prefer to use my headphones, HD600.

The Acoustical Measurements

My computer is equipped with a sound card that has a microphone input. Via tha windows control panel, it can be decided whether a phantom power should be applied to this or not. Therefore, I purchased a panasonic condensor microphone at the local electronics shop for the price of 12dkr (less than $2 US). I hooked the microphone up to the computer using a microphone (shielded) cable.

minijack connection
The mini jack connection.

Due to poor electromagnetic shielding, there is quite a lot of noise on the microphone signal. A frequency domain representation revealed that most of this noise was 50Hz and therefore, I decided simply to cut the signal at 100 Hz and thereby improve the SNR significantly. (I have no chance of measuring direct low frequency sound anyway due to the physical boundaries of my testing environment)

Use of test signals

I generated a pink noise signal, a logarithmic chirp signal sweeping from 20 Hz to 20 kHz, a white noise signal and a couple of MLS signals. It proved difficult to perform the measurement due to the reverberant surroundings (my living room).
I found that the chrip signal and the MLS signal revealed the best indication of the response, and I therefore used these in conjunction to estimate the free field response.