I have found that repeatability of a single device is very good, but differing devices, although claimed to be equivalent, are seldom repeatable.
That seems to make sense for a perfect point source. However, with a real loudspeaker, such a distance measurement might depend on the angular position of the mic.
I wrote part of what I was thinking. I agree with you Earl. Many times I measure drivers that are made on a line that has very high quality control. Every driver is tested within tight tolerances. Usually plus or minus 1db over the voice range. When they measure with a Clio or an LMS system you get measurements. When I do it here on the same drivers with a Speaker Tester Pro and some high quality condensers or REW and a USB mic I get reasonably the same. Never exactly the same. Bass will be reasonably the same, midrange will usually be the same, anything above 10k toss a coin. Mic placement, boom reflections, all come into play. Even here with the same equipment with different setups on different days getting the same exact measurement is really not going to happen without some processing that averages out many measurements.
I'm currently running my first fully automated angular speaker measurement from Octave.
Thanks to @fluid for introducing me to tic, @jcga for his rotation table advice, @mbrennwa for MATAA and NTK for his sound field separation software and others for advice or showing interest.
Measurements will of course be of low quality given the way too small reflection free time and loopback is not yet working. I'll see if I can share them later.
Thanks to @fluid for introducing me to tic, @jcga for his rotation table advice, @mbrennwa for MATAA and NTK for his sound field separation software and others for advice or showing interest.
Measurements will of course be of low quality given the way too small reflection free time and loopback is not yet working. I'll see if I can share them later.
Thank you for bringing the Weinreich and Arnold (1980) paper to my attention. The last sentence therein, "Although our main interest was in violin acoustics, it appears that the domain of applicability of the method is much wider", seems to be quite an understatement.
In this case it wouldn't be to difficult to have a way of aligning the speaker in such a way that it's perfectly in line with the zero reference of the microphone?
Is the speaker rotating at the front of the baffle or middle of the speaker cabinet?
What I find is that even using the same mic, the math within most programs, how they deal with the mic signal is not identical. We are not using the analog signal, but usually an FFT to allow for processing.
This is great! It would be advisable to have a larger platform on top of your turntable. You want the baffle to be at the center of rotation so that it's distance to the microphone doesn't change with rotation.
I am very impressed.
Have you checked for repeatability? And how well is the floor bounce being removed in the software, you are quite close to the floor and this will be around 1k if my napkin math is correct?
Maybe we are talking about something else, but how is that effecting the reference point?
If the reference point is exactly the same each time, the only variable that is left, is the phase reference.
(or latency rather)
Yeah I was talking about differences but failed to type the entire thought. I was sidetracked. You are correct, physically we can line up out DUT and our Microphones.
I'm wondering how much bearing on the phase measurements slight differences in timing that the DSP will introduce. Conversion directly at the mic via USB or at the computer via internal card, or via USB interface.
Depends on where you put the speaker. Currently it is roughly the front of the speaker.
Angular position repeatability/accuracy is very good and in practise limited by gear play/backlash which can also be circumvented by just going one way.
Speaker and mic are way too close to the floor for a real measurement. This is a proof of principle and to iron out weird stuff. Currently the frequencies I get are not always the same, probably because the IR start is estimated. I would like to work with the same frequencies so I can put the results in an array/matrix. That are the things I wanna learn and fix by this setup.
This is rarely an issue for 3D printers or CNC's with an extremely similar moving system, so I don't see why this would be a problem here?
They all have mechanical anti backlash (hysteresis is a better word) build in.
Even these cheap devices have an accuracy well below 0.1mm, which can be even enhanced with an encoder on the motor.
Which also cost bugger all these days.
I do not have the engineering experience to give a useful answer. I basically tossed this out for anyone that has high speed signal experience and in how the potential difference in clocking can effect the computations. Few mentioned a bit back a few pages that a USB mic has timing problems. But I have never read anything else regarding this potential problem. And nor have I ever considered it to look!
Conceptually a physically measured distance and a chirp should be automatically correcting each other.
Again conceptually a timing difference in microseconds should have little detriment to phase measurements. But I have no real experience in these matters.
Thanks for the answer. I am truly excited to see this working.
For the Klippel clone we need the mic rotating, but that is still the same principle as you have working here. And how your measurements work out will be very instructive.