[SpHEAR-devel] latest musings (PCB, new Octathingy, calibration and more)

[Fernando Lopez-Lezcano]

Hi all,

I have been meaning to send an update about the project for a loooong 
time... Many many changes (not all of them in current git yet). I'll try 
to summarize...

I have a new design for the Zapnspark phantom power printed circuit 
board. I decided to make a double sided board as I was going to 
manufacture a few (double sided would be much harder with our small 
mill), and then went overboard and tried to make it as small as possible 
without using surface mounted components. I found some smaller form 
factor capacitors, which helped shrink it.

Using those, the four capsule microphone is much smaller, and the 
Octathingy benefits as well (smaller but not quite as much). There is a 
control variable that lets you switch between the two PCB sizes, and the 
assemblies mostly scale (but work is not finished yet - for example I 
need new windshield and shock mount 3d models).

Speaking of the Octathingy. I have been working on the calibration side 
a lot. Looking at many graphs made it more apparent that the initial 
design, which is a straight scale up from the four capsule design, was 
"clever" but not the best possible due to the resonance from the cavity 
formed by the capsules (but then Rode, SPS, Core Sound and others of 
course already knew about that - hmmm, but not Senheisser? :-).

So I went ahead with a newer "flower power" or "trumpets of doom" design 
with individual conical capsule holders and the smallest practical 
spherical core they could plug into. A measurement of a test 3D print 
(all dummy capsules except for one) showed better performance at high 
frequencies. Just yesterday I finished building the first full 
prototype, now I have to measure it (see pictures, all three prototypes 
and the comparison of the array between the last two).

Measuring... oh so not easy. I have to work on compensating the low end 
of the measurement frequency range as that is currently (I think) the 
weakest link in the calibration data. Juan Sierra here at CCRMA helped 
me understand Eric's paper on the matter, and it looks like I was doing 
the inverse filter and first reflection trimming in the wrong order (to 
begin with!). I have to see which type of window is best (right now 
using a full blackman centered on the impulse)...

The current Octathingy calibration includes equalizing the 8 individual 
capsules, deriving an 8x8 matrix of A to B filters for low and mid 
frequencies, and finally creating 8 individual filters for the B format 
components at high frequencies, all collapsable into a single 8x8 matrix 
of filters.

More about measuring...

I (well, CCRMA) bought a small robotic arm. And customized it to be a 
better measurement rig. The inverse kinematics are now working correctly 
and the arm is now pointing the microphone in the right directions. The 
arm is a bit short (longer would have been much more expensive), but 
with the current dimensions, and depending on how it orients the 
microphone, I can get vertical coverage up to around +/-40 degrees of 
elevation (and complete horizontal coverage of course), or +60-20 or so. 
We'll see what's best, or if both can be combined.

The arm is controlled from a SuperCollider program which does the 
inverse kinematics and will also will play back the sine sweeps and 
record the output of the capsules. Hopefully easier and more repeatable 
at the end of the road (which is still long). I really need better data...

Comments welcome...
-- Fernando
-------------- next part --------------
A non-text attachment was scrubbed...
Name: octa_family_1.jpg
Type: image/jpeg
Size: 277359 bytes
Desc: not available
URL: <https://cm-mail.stanford.edu/pipermail/sphear-devel/attachments/20180623/e99e66cb/attachment-0004.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: octa_1.jpg
Type: image/jpeg
Size: 169049 bytes
Desc: not available
URL: <https://cm-mail.stanford.edu/pipermail/sphear-devel/attachments/20180623/e99e66cb/attachment-0005.jpg>