When you are capturing audio from a speaker, you are rarely capturing the actual direct output of such a system. There are reflections and artifacts caused by anything and everything in the environment that make it to whatever detector you might be using. With the modern computation age, you would think there would be a way to compensate for such artifacts, and this is what [d.fapinov] set out to do.
[d.fapinov] has put together a code base for simulating and reversing environmental audio artifacts made to rival systems, entirely orders of magnitude higher in cost. The system relies on similar principles used in radio wave antenna transmission to calculate the audio output map, called spherical harmonic expansion. Once this map is calculated and separated from outside influence, you can truly measure the output of an audio device.
The only problem is that the project needs to be tested in the real world. [d.fapinov] has gotten this far but is unable to continue with the project. A way to measure audio from precise locations around the output is required, as well as the appropriate control for such a device.
Audio enthusiasts go deep into this tech, and if you want to become one of them, check out this article on audio compression and distortion.
This is obviously incredible, but I’m not much knowledgeable about sound recording. What are the usual reasons to capture sound from a speaker rather than plug in directly into the stereo channel of the amp the speakers are playing from ?
It’s precisely because the OP wants to know what the speakers sound like, free of interference from the room that they’re situated in.
“called spherical harmonic expansion” or rather spherical harmonics expansion, because ist an expansion in spherical harmonics, a function set?
Record the characteristics in more than one environment then find the difference.
interesting. its the exact opposite of what a impulse response is trying to achieve: adding artificial reverb to a dry signal. maybe that route is also possible.
Wow! Just what I need. I’ve been a Hammond/Leslie tech for many years and have been working on a way to accurately document the sound characteristics of Leslie speakers on location – mainly in church auditoriums. Occasionally, I’ll run across one with a “sweet” sound but room acoustics just don’t allow for accurate measurements. I have some Leslies that I’ve considered testing in an open field so I posted a question to AI about what to expect. The response I got was that there are known techniques for obtaining relatively accurate measurements (on a par with an anechoic chamber), but were limited to around 500Hz and above. The Hammond organ goes down to about 30Hz so that’s as far as I got. But, I’m ALL IN for working with others on putting a Leslie (or any other speaker) on a carousel and/or attaching a calibrated mic to a robotic arm if the techniques discussed here will give reliable results.
For an open field the main problem is refection off the ground, which has 2 effects: both the direct speaker output and the reflected output are detected by the microphone, and the ground raises the acoustic impedance seen by the speaker (raising the low frequency response.)
Both problems can be addressed by raising the speaker off the ground. Most of the acoustic impedance changes are minimized by getting the woofer at least 1/6 wavelength high; that’s 6.1 feet
or 1.86 m at 30 Hz. Minimizing the effect of the reflected wave at the microphone requires making the direct path to the microphone much shorter than the reflected path. There are difficulties with microphone placement because even the direct response varies wildly with position with respect to the speaker. Get a ladder and experiment with microphone positions.
Alternately, don’t bother raising the speaker off the ground. If your application has the speaker on the floor, then the ground will affect performance similarly to a floor, especially at low frequencies.
It’s not invalid to just get a half-space response. Unless the speaker is vented in back, just lay the speaker flat on the ground and make measurements with the microphone 6 feet or so above the speaker. This will boost low frequency response by as much as 3 dB compared to a free space measurement.
It’s possible that what you hear as “sweet” sound is due more to room acoustics than speaker response, so keep that in mind.
I needed to measure speaker performance without reflections some time ago. I didn’t have access to the fancy analysis tools (heck they didn’t exist). Nor did I have an anechoic chamber handy.
I did have a nearby open field with clear sky above though. A perfect empty hemisphere. No echo, no reverberation. Worked great.
So this solution relies on a static map of the environment around the speaker for calibration and lots of computation to obtain a signal; what’s the utility in that? Except in a few very narrow cases, or in controlled environments, the world is a very dynamic place; always changing, even in small ways that would invalidate the mapping, and all that computation would no longer output a signal with fidelity.
What advantage does this method of audio input have over a more conventional sensor that does not require this computational overhead and awareness of the environment?
This reminds me of those old plate cameras that required long exposures to obtain a picture. “Stand still while I take your picture!” They might have been able to obtain an image, but couldn’t really capture, with fidelity, the likeness of the dynamic life in front of them. Not many landscapes or pictures of animals back then.