One thing that is very important to me at Affordable Home Audio is that when I talk about speakers, I try and do so from as informed a position as possible. Within monetary restrictions, I talk about speakers which I can hold in my hands and test myself. An important part of any discussion of speakers is speaker measurements.
There were many techniques developed years ago for making engineering measurements of speakers. Gated sine, white noise response, all sorts of things made it possible to understand both frequency, temporal, linear and non-linear aspects of audio equipment.
These techniques necessitate the use of either very large spaces, like your backyard, or a very expensive anechoic room, in order to perform. Reflections in the room make it impossible to analyze speakers using these traditional techniques. Most businesses focusing in audio (the company I work for included), and university labs have an anechoic room for acoustic testing. Below is a great example of an anechoic chamber, made by Paradigm Audio.
With the advent of mass computing, the speaker measurement became cheaper and more widely used. This, combined with computer aided design has been revolutionary in the development of audio systems.
The most important element, however, was the development of the MLS measurement algorithm in 1989. This paper described the method for the MLSSA tool, which was launched alongside the paper. It has become the de facto standard in the industry and has been used by many famous companies.
Without getting too into the nitty-gritty, it works by sending pseudo-random sequences through the speaker and interpreting the results via cross-correlation of the transmitted and received signals. The key benefit is it allows for the effect of the room to be removed from the measurement. This means any room can be used for speaker measurements, making it cheaper and easier for anyone to do speaker measurements.
That being said, the better the room, the more information can be extracted from the MLS method for a speaker. With MLS you are limited to the amount of time between the direct path from the speaker arriving at the microphone and the time the first reflection off a wall arrives. In a long paper by John Atkinson, also available from Stereophile in three parts (1, 2, 3), an example of a good measurement system using MLSAA is described. It also forms the basis for the types of measurements I make.
That said, MLS is now 26 years old. There have been developments since then, and though they haven’t displaced MLS, there are reasons to go elsewhere if you are developing your own software.
Since my specialty in graduate school was digital signal processing, writing my own audio analysis software was somewhat of a given. I wanted things to work exactly how I wanted, and specially tailor everything I was doing to my own setup.
I use a technique using swept sine waves from Farina. It has the benefit of separating out distortion from the response, but is more sensitive to transient noises in the room. If you’ve used Audessey room correction, you’ve heard the technique, it works for measuring rooms and speakers at the same time. I remove the effects of the room, but otherwise, what I do is likely very similar to Audessey’s method.
My space is an unfinished room in my basement with several sound absorption panels on the floor which eliminates ground reflections. With this I can measure down to 300 Hz.
I use a Dayton UMM-6 USB Microphone hooked up to a custom built silent PC. The microphone is a good, cheap measurement microphone that comes with a calibration curve to adjust measurements. To send my test signals to the speaker I use a Schiit Modi for an ADC hooked up to a Emotiva Mini-X A-100 Amplifier. This combination is something you would use in a high-end setup, and the reason I use these is because they are so good I know they aren’t the sources of problems I see in the speakers. That’s a very important thing about a testing setup.
That’s a full description of the setup I use for doing my measurements. I’ll add some guidance on the meaning and interpretation of various measurements and graphs in a future article.