(The Program Manager thinks all is done if the prototype isn’t in flames)
Hello, DIYers!
This is intended to be a series of posts about several trial batches of DIY measurement microphones that I built. I was playing with the design as a possible product for Dayton Audio. No decision was made yet on whether Dayton will actually pursue this, but they are ok with me writing about how these could be made (by some more skilled DIY constructors).
It will be partly a how-to, partly a technical design discussion, and partly a documentation about obsessive engineering behavior.
The microphone is an ultra-slim, small acoustic profile, wideband device with very flat response even without an individual correction file. A short cord connects the mic wand to its electronics board that provides power (9V battery or 48V Phantom), as well as response flattening circuitry and drive for balanced XLR and RCA outputs. The design capitalizes on the repeatability of MEMS chips’ sensitivity and frequency response.
-------
ERRATA:
- some later posts mention using 38AWG enameled wires for wiring the chip. That should be 36AWG. AWG38 would of course work electrically, but won't have much of any stiffness and so is difficult to handle even with tweezers to position onto the pads and tack-solder. A little stiffness helps a lot. Even 34AWG would probably be ok, but 36 was easier to find. For example, TheElectricGodmine.com item #G27281
https://theelectronicgoldmine.com/p...jlFgBwuZ5ccgkjYya1pgI-1UYeRFbfSuqFWYDJH3ljWOi
[index into relevant later design posts:]
- Frequency Response curves for mics without adjustment
- Arguments Pro and Con
- 3D Printed Mounting adaptor for MEMS chip
- Equalization Method
- Performance Specs
- Assembly Fixture
- Bypassing and Ground wire
- Assembly of the Microphone Wand
- Skill Level needed
- How Calibration was Done on test run mics
- Equalizer/Interface Box and related 3D printed parts
- Files download package
STL, GERBER, Lists, schematic, notes are now available 9/13/2024
Last edited:
Here are high frequency response curves of 9 mics from the most recent batch as built, not adjusted or tuned. Graphs are 400Hz to 80kHz, 2dB/division. The graph labeled “LM1” is for 4 mics built from MEMS chips (Lot Code 26731591) sourced from Mouser. “LD1” is for 5 mics built from MEMS (Lot Code 26731590) from Digikey. I suspect that much of the small variation in the (quite repeatable) ultrasonic portion of the curves within same lot codes is because of small differences in mounting the MEMS chip.
*There were 10 mics in this batch, but one was accidentally assembled with the chip mounted backwards so its aperture was off center. That response curve looked a lot different, though it was still within a decibel beyond 20kHz.
Background:
When I was looking for higher frequency capsules for the OmniMic40k, Demian Martin suggested MEMS type microphones. MEMS are known for consistent unit-to-unit performance based on micro-lithographic solid state manufacturing. I ended up ruling them out then, though, because the ones I found had too-low SPL handling ability and a high, sharp, and scary frequency response peak within the upper audio band. MEMS mics had strong response to ultrasonic frequencies, but weren’t reasonably flat within the audio band or above.
More recently, I ran across a newer chip (SPH1878LR5H-C) from Knowles Sisonic with some enticing specs:
When I was looking for higher frequency capsules for the OmniMic40k, Demian Martin suggested MEMS type microphones. MEMS are known for consistent unit-to-unit performance based on micro-lithographic solid state manufacturing. I ended up ruling them out then, though, because the ones I found had too-low SPL handling ability and a high, sharp, and scary frequency response peak within the upper audio band. MEMS mics had strong response to ultrasonic frequencies, but weren’t reasonably flat within the audio band or above.
More recently, I ran across a newer chip (SPH1878LR5H-C) from Knowles Sisonic with some enticing specs:
- SPL handling level of about 130dB at 3% THD (good, not great).
- Analog output
- The resonant frequency peak moved to up around 36kHz.
- Strong ultrasonic frequency response
- Low frequency -3dB corner respectably down near 7Hz.
- Decently low noise at 27dBA SPL. Ok for home or venue live recording, not great for studio though.
- A sensitivity spec extremely tight for any audio device – within a half dB! Suggesting that it might make measurement microphones that wouldn’t require individual calibration.
Why you might not want to DIY this! :
And why you MIGHT consider DIYing this anyway:
- The very high frequency response shape is extremely sensitive to the mounting arrangement of the MEMS mic chip and its surrounding surfaces. Geometry of the tip end of the build matters. More so if you care about data above 20kHz.
- Do you really need flat ultrasonic response? Electret mics do up to 20kHz pretty well if used with calibration files and don't cost a lot.
- You'll still need an audio input interface or decent soundcard.
- To get the right precise geometry, a 3D-printed chip mounting holder is needed. This requires a resin-type 3D printer for the needed precision and tiny dimensions.
- The MEMS chip and wiring are insanely small and not at all easy to work with! This is tweezer and (about 4x) microscope work. I had to use both and it still took quite a while to make solder connections.
- The assembly process is touchy, I’ve destroyed a number of mics chips and assemblies* in the process. A very fine-point soldering iron tip (like 900M-T-I) is needed. Soldering a chip capacitor to the metal back of the MEMS chip is part (probably the most difficult) of the project.
* though 8 of those were because I screwed up and miswired the chips before installing them with the adhesive - The EQ/interface circuit uses surface mounted parts, requiring a printed circuit board.
- The steel pipe’s mount and board enclosure design also require 3D printing (though that can be filament-type).
And why you MIGHT consider DIYing this anyway:
- The resin-3D-printed part is so small that dozens can be printed together even in the smallest resin printer. Maybe someone could print and distribute them?
- The other 3D printed parts could be done other ways, not critical.
- I’m 70 years old, with significant arthritis in my hands and lousy eyesight. You're likely better suited for this kind of watchmaker scale work.
- The MEMS chips and other parts are cheap, so you can work with a number of them and afford to lose a few.
- I made a very small PC board pattern that makes mounting that chip cap to the back of the MEMS less troublesome.
- The circuit board Gerber files are made available, and can be ordered inexpensively from PCBway.
- I’ve worked out some ways to make assembly practical.
- I’ve made most of the mistakes for you already, so you’ll be able to avoid them.
Working with the MEMS frequency response peak
A point worth mentioning: when the MEMS mic chip makers give you a curve of the free-field frequency response, they do really mean “free field”. As in: with the bare chip dangling from very fine wires, far from any surfaces. If the parts are surface-mounted on a PC board, as they are designed for though, the frequency response will change a lot. The resonant peak will move down appreciably in frequency and go up many dB in height. A board with the chip back-mounted would also make the mic more directional. So for this project, the chip is not mounted on a PC board, it is wired with tiny magnet wires to its also-tiny pads, and mounted with adhesive into a 3D printed holder.
By the way, how small is this thing really? Here’s the chip, with bypass cap and wiring next to a U.S. nickel --
Though not outrageously bad within the <20kHz audio band, the natural response even in the holder should be flattened to avoid a rising high end. The response peak to be removed is essentially a 2nd order resonance, so I designed a (relatively) simple notch EQ circuit.
It’s a shelved single pole low pass filter followed by a series LC style notch circuit placed in shunt with the signal line to cancel the resonant peak. The circuit topology and components preserve the noise floor of the MEMS mic chip, costing only about half a dB.
Physical inductors are low-Q, hard to work with, and expensive. But opamp simulated inductors are cheap, precise, and very high-Q. With the simulated inductor, a full adjustment of the EQ becomes just setting the values of three resistors controlling
For a DIY project, probably not a lot of people have >40kHz reference microphones with good calibration curves and capable speakers for adjusting the CF. So, I made another 10 mics (the batch mentioned previously) to document what could be done without any individual adjustment at all.
A point worth mentioning: when the MEMS mic chip makers give you a curve of the free-field frequency response, they do really mean “free field”. As in: with the bare chip dangling from very fine wires, far from any surfaces. If the parts are surface-mounted on a PC board, as they are designed for though, the frequency response will change a lot. The resonant peak will move down appreciably in frequency and go up many dB in height. A board with the chip back-mounted would also make the mic more directional. So for this project, the chip is not mounted on a PC board, it is wired with tiny magnet wires to its also-tiny pads, and mounted with adhesive into a 3D printed holder.
By the way, how small is this thing really? Here’s the chip, with bypass cap and wiring next to a U.S. nickel --
Though not outrageously bad within the <20kHz audio band, the natural response even in the holder should be flattened to avoid a rising high end. The response peak to be removed is essentially a 2nd order resonance, so I designed a (relatively) simple notch EQ circuit.
It’s a shelved single pole low pass filter followed by a series LC style notch circuit placed in shunt with the signal line to cancel the resonant peak. The circuit topology and components preserve the noise floor of the MEMS mic chip, costing only about half a dB.
- the center frequency (CF) of the notch
- the Q (bandwidth) of the notch
- the depth of the notch.
For a DIY project, probably not a lot of people have >40kHz reference microphones with good calibration curves and capable speakers for adjusting the CF. So, I made another 10 mics (the batch mentioned previously) to document what could be done without any individual adjustment at all.
Last edited:
The EQ circuit and connectors can’t be assembled inside the 5mm diameter of the metal tube, so it sits in a small plastic box with battery and sockets at the end of about a meter of 2 conductor shielded cable. An advantage of having the EQ and connector box external to the mic wand is that the narrow wand causes least possible disturbance to sound fields, and with the want in a 3D printed head-on mounting piece on a boom arm there is little opportunity for reflections off of mic stand hardware.
Performance
Sensitivity 1kHz, +/-0.5dB (per the chip spec)- single ended: 17.83 mV/Pa (-35dBV at 94dBSPL)
- differential: 35.67mV/Pa (-29dBV at 94dBSPL)
Frequency Response, high frequencies (unadjusted, based on 9 units, see previous graphs in the second post)
Measured by comparison method against an ACO 7016 ¼” microphone, recently calibrated using a B&K electrostatic actuation fixture. These MEMS mics proved to be flatter below 30kHz than the ACO 7016 or a B&K 4135 they were calibrated against, though less extended in response.
- 1dB, 17Hz to 32kHz
- 2dB, 11Hz to 37kHz
- 3dB, 8.5Hz to 55kHz
--
If you do have a way to measure the actual response, then adjusting the three trimmer can get:
(after EQ tuning, based on same 9 units
- 1dB, 17Hz to 45kHz
- 2dB, 11Hz to 47KHz (?)
- 3dB, 8.5Hz to >80kHz
The SPH1878 data sheet says the low frequency corner is nominally 7Hz.
I measured closer to 8.5Hz (the difference may have been from boundary effect). All units measured the same for this, within repeatability limit.
Off-Axis Frequency Response (“Omni-ness”):
The assembly presents a very small acoustic cross-section area of about 33mm^2. Positioning the tube 30 degrees off axis shows insignificant drop in response even to very high frequencies (which is surprising, but I rechecked several times). Even 90 degrees off-axis it stays within 3dB up to 17kHz.
MEMS microphones are said to be very stable in measuring SPL level over time (though I don’t have any data or way to prove that). This microphone chip is specified to be better in high SPL handling than almost any MEMS (except the laser based Sensibel digital mics), but is still more than 20dB below Omnimic40k or type 1 microphone performance for that.
MEMS 1% THD at 127.7 dBSPL (measured using modified OmniMic circuit)
OmniMic40k 1% THD 148.8 dBSPL
Near the resonant peak (about 36kHz) the SPL handling will be about 25dB worse due to internal electronics overload.
But 103 dBSPL sound at 36kHz isn’t something you're likely to encounter anyway...
Noise:
I haven’t measured noise, I don’t have a quiet enough place to easily do it. The MEMS chip is specified at 27dBA and the gain and EQ circuit degrades by just 0.5dB (according to SPICE simulation), so it’s reasonably good. It’s not “recording studio quiet”, but still a few dB quieter than high voltage quarter inch (6mm) Type 1 microphones like 4135 or 7016.
Ruggedness:
The assembly is exceptionally rugged and light weight. I dropped one onto its nose from 8ft to a concrete floor several times (on purpose) with no noticeable change to frequency response. Compare that to a type 1 high voltage reference type mic capsule which will typically die if the mic stand falls over and the tip hits much of anything, even with a protection grid on. With the protection grid off, merely touching a Type 1 diaphragm with anything more than a very soft brush will destroy it. The measurements of these MEMS mic doesn't require taking off any protection grids, either, like the type 1 mics at super high frequencies.
MEMS microphones are said to be more susceptible to damage from sudden extremely high air pressures. That won’t likely happen from any sound encountered BUT it might happen if static pressure is increased dramatically from pushing a MEMS mic into an SPL calibrator! When using a calibrator, insert or remove the mic VERY SLOWLY. It would probably be safest to instead have a large air leak (such as a threaded screw hole) in the calibrator that is open when the microphone is being inserted or pulled out, but which can be closed when the mic is in place during the SPL calibration.
Given the SPL spec (+/- 0.5dB) of the MEMS chips, though, it doesn’t seem that SPL calibration would be necessary anyway.
MEMS microphones can be degraded by dust particles getting into the aperture, so some care should be taken to keep the mic end reasonably protected when not in use. A sock or baggie over it would do. Don’t get water into one (or into any mic for that matter). This microphone design uses a small plastic narrow-weave silkscreen cloth as a grille to reduce chances of dust getting in and to improve appearance. There is very little effect on frequency response from having this cloth over the chip, less than 0.4dB at the resonant peak and basically immeasurable much outside the peak.
Attachments
Last edited:
How this mounting and wiring scheme came about
It might seem this should be simple: a chip, a bypass capacitor and three wires, what’s to design? But the way the chip is mounted has a tremendous effect on frequency response at higher frequencies. And these parts are really tiny, not simple at all to handle or mount. (Touch a soldering iron to it sitting free, and it will stick to the iron’s tip and possibly overheat while you try to scrape it off. Trying to keep it in place while you solder a fine gauge wire to it isn't much fun either).
Back-mounting the chip on a circuit board via zone soldering (as it was designed for) dropped the resonant frequency significantly even with the hole in the board for the aperture made almost as wide as the whole chip. If we want to EQ out the peak for all units with the same circuit, it’s important to keep the frequency of that peak high and consistent unit-to-unit. And the only way to get that is to have the chip sit right at the surface of the microphone tip and for that surface to be the very smallest area possible.
The chip area can fit onto a circle of about 5mm diameter, which is a handy size for putting it on the end of a 6mm tube. “Reusable” 6mm stainless steel drinking straws are easy to find and inexpensive, and about the right length for the barrel of a low cross-section microphone wand, with wideband true “omni” performance. I’ve used both the 8.5” and the 10.5” varieties of these available on eBay or Amazon, use what you prefer or find. The 10.5” length looks a bit cooler to me.
I tried various ways to easily mount the chip on the end to get repeatable results. Sticking the chip onto the surface of a glob of epoxy putty and sticking that at the end of the tube worked, sort of, but was difficult and messy to do and frequency response wasn’t well controlled. I also tried surface mounting to a small round board to be later attached on a printed plastic plug, which was extremely difficult to do and moved the resonant frequency quite a bit.
Here are some photos of failed attempts
The only effective way I found to do it and get consistent response curves was design and use a 3D printed plastic holder for the chip that would hold it precisely right at the top surface and then slide the holder onto/into the tube. The holder has ledges to position the chip so that the aperture is centered in the tube with gaps to either side for running fine enameled wires from the pads on the chip. If you try to come up with a different scheme, the EQ circuit values I'll give here will have to be changed for decent curve flatness, so you'll need a way to calibrate.
The holder is printed with a resin type 3D printer (I used Elegoo Mars4 9k at 20 micron layer height). A filament printer doesn’t have enough resolution. You can get many dozens of the holders with each resin print session though. So, there’s an excuse to get a resin printer if you wanted one already.
The solder connections to the MEMS chip surface should be kept as low profile as possible and the wires should be thin to minimize adding surface irregularity to the assembly. For the bias and signal wires I used AWG 38 gauge heat strip-able bifilar magnet wire. For the ground wire, I used stiffer AWG 30 magnet wire because it’s easier to stuff, with the others twisted together with it, down a long tube than to try to push three fine loose wires.
Parts and wires are almost absurdly small for free component assembly. A low-powered microscope, fine tweezers, smallest possible soldering iron tip and a tip cleaner, some special “fixturing” (see below), and patience are essential. I used lead-tin solder to minimize soldering heat, but don’t know if that’s really necessary. It'll probably be easier though.
It might seem this should be simple: a chip, a bypass capacitor and three wires, what’s to design? But the way the chip is mounted has a tremendous effect on frequency response at higher frequencies. And these parts are really tiny, not simple at all to handle or mount. (Touch a soldering iron to it sitting free, and it will stick to the iron’s tip and possibly overheat while you try to scrape it off. Trying to keep it in place while you solder a fine gauge wire to it isn't much fun either).
Back-mounting the chip on a circuit board via zone soldering (as it was designed for) dropped the resonant frequency significantly even with the hole in the board for the aperture made almost as wide as the whole chip. If we want to EQ out the peak for all units with the same circuit, it’s important to keep the frequency of that peak high and consistent unit-to-unit. And the only way to get that is to have the chip sit right at the surface of the microphone tip and for that surface to be the very smallest area possible.
The chip area can fit onto a circle of about 5mm diameter, which is a handy size for putting it on the end of a 6mm tube. “Reusable” 6mm stainless steel drinking straws are easy to find and inexpensive, and about the right length for the barrel of a low cross-section microphone wand, with wideband true “omni” performance. I’ve used both the 8.5” and the 10.5” varieties of these available on eBay or Amazon, use what you prefer or find. The 10.5” length looks a bit cooler to me.
Here are some photos of failed attempts
The only effective way I found to do it and get consistent response curves was design and use a 3D printed plastic holder for the chip that would hold it precisely right at the top surface and then slide the holder onto/into the tube. The holder has ledges to position the chip so that the aperture is centered in the tube with gaps to either side for running fine enameled wires from the pads on the chip. If you try to come up with a different scheme, the EQ circuit values I'll give here will have to be changed for decent curve flatness, so you'll need a way to calibrate.
The solder connections to the MEMS chip surface should be kept as low profile as possible and the wires should be thin to minimize adding surface irregularity to the assembly. For the bias and signal wires I used AWG 38 gauge heat strip-able bifilar magnet wire. For the ground wire, I used stiffer AWG 30 magnet wire because it’s easier to stuff, with the others twisted together with it, down a long tube than to try to push three fine loose wires.
Parts and wires are almost absurdly small for free component assembly. A low-powered microscope, fine tweezers, smallest possible soldering iron tip and a tip cleaner, some special “fixturing” (see below), and patience are essential. I used lead-tin solder to minimize soldering heat, but don’t know if that’s really necessary. It'll probably be easier though.
Awesome set of posts!
I'm curious if you considered a flex PCB to mount the microphone and cap. Either surface mount as intended with the PCB folded around the sides of the housing or just as a way to give some structure to the wiring and a place to mount the cap. FPC can be extremely thin (0.025mm!) and might have less effect on the resonant frequency and directionality compared to an FR4 board.
I'm curious if you considered a flex PCB to mount the microphone and cap. Either surface mount as intended with the PCB folded around the sides of the housing or just as a way to give some structure to the wiring and a place to mount the cap. FPC can be extremely thin (0.025mm!) and might have less effect on the resonant frequency and directionality compared to an FR4 board.
N Brock --
It had crossed my mind to reflow solder the chips to strips of flex with the chip midway, with a big hole over the chip's aperture. The narrower strips on either side could tuck into a plastic piece like my 3D printed one with an indent to position the chip. It would sure make wiring easier. The little round circuit boards I had tried were just 0.6mm thick, but when I reflow soldered them (kind of crudely, on a hotplate), the solder layer between chip and board didn't end up a lot thinner than the PC board was. I don't know that super thin flex would drop the overall thickness all that much, the aperture would still be sitting in a pocket, (though a more shallow one).
Having the reflow soldering of the flex done by professionals could possibly get the solder thickness more reasonable, but it's not something I'm capable of in a DiY environment.
If mass production were to be planned, it might be worth looking into that again -- having flex strips made, with die-cut outlines within a larger sheet, assembled maybe by PCBway or someone else with the right equipment and skills and custom tooling for the pattern. The resonance frequency and height would certainly be worse than with the bare chip (that has only gold immersion coating on the pads) but maybe could be tolerated. Repeatability would be essential, so not DIY. A new EQ design, or at least new resistor values, would be needed, so it would need another study of multiple units. On my own, it's not something I'd want to take on.
It had crossed my mind to reflow solder the chips to strips of flex with the chip midway, with a big hole over the chip's aperture. The narrower strips on either side could tuck into a plastic piece like my 3D printed one with an indent to position the chip. It would sure make wiring easier. The little round circuit boards I had tried were just 0.6mm thick, but when I reflow soldered them (kind of crudely, on a hotplate), the solder layer between chip and board didn't end up a lot thinner than the PC board was. I don't know that super thin flex would drop the overall thickness all that much, the aperture would still be sitting in a pocket, (though a more shallow one).
Having the reflow soldering of the flex done by professionals could possibly get the solder thickness more reasonable, but it's not something I'm capable of in a DiY environment.
If mass production were to be planned, it might be worth looking into that again -- having flex strips made, with die-cut outlines within a larger sheet, assembled maybe by PCBway or someone else with the right equipment and skills and custom tooling for the pattern. The resonance frequency and height would certainly be worse than with the bare chip (that has only gold immersion coating on the pads) but maybe could be tolerated. Repeatability would be essential, so not DIY. A new EQ design, or at least new resistor values, would be needed, so it would need another study of multiple units. On my own, it's not something I'd want to take on.
Fixturing
The main “fixture” I mentioned is for holding the MEMS chip while wiring. The chip is way too small and fragile to mount in a vise or hold with hemostats. So use a 1mm thick sheet of silicone rubber (sold as a baking mat or for industrial purposes, check eBay or Amazon) to make a simple soft holding fixture. Take an approximately 150mm x 150mm piece of the sheet and using a fine razor knife cut a rectangle out if it near the center leaving a hole only slightly smaller than the outline of the MEMS chip housing. This might take a few tries. A slit at the bottom can help with holding the wires during soldering.
Or even better, if you have a 1/8" punch set, perforate the sheet with holes every inch or so to give yourself lots of places to hold MEMS chips. When it is right, you should be able to push the chip down into the hole so that the tension in the sheet holds it in place. This photo shows a several chips pressed in, during the build of the most recent mics. With circular holes, the wires can be fed in the opening at the sides.
If you use only the silicone sheet, there's nowhere for the heat to go, so the whole MEMS chip will heat up fast during soldering. I even had the metal back of the chip come off on one! An effective mitigation is to put a sheet of aluminum foil underneath the silicone sheet (it will contact the MEMS chip body) while soldering anything to it. You want the pad to heat up for soldering, not so much the whole chip.
The very tips of the wires (no longer than the pad dimensions) as well as the pads at top of the MEMS chip should be pre-tinned. For the chip, I had the soldering iron at about 330C at least for SN60 (maybe not as critical, if aluminum foil is used).
The wires tinned and blistered-off the enamel insulation better with 350C. When doing that, touch the wetted iron tip to the exposed metal at the end of the cut magnet wire to heat it enough for the solder to climb up it -- heat doesn't conduct enough through the enamel itself.
Then tack-solder the wires onto the pads -- be sure the connection is good and shiny, but don't leave too big of a solder bump. The top left terminal, as shown, is the 3.6V supply wire. Top right is the signal output. The gold plated back can (the housing, which solders very easily) is the common ground. Flip the sheet over to get at that, with the chip facing into the foil.
The main “fixture” I mentioned is for holding the MEMS chip while wiring. The chip is way too small and fragile to mount in a vise or hold with hemostats. So use a 1mm thick sheet of silicone rubber (sold as a baking mat or for industrial purposes, check eBay or Amazon) to make a simple soft holding fixture. Take an approximately 150mm x 150mm piece of the sheet and using a fine razor knife cut a rectangle out if it near the center leaving a hole only slightly smaller than the outline of the MEMS chip housing. This might take a few tries. A slit at the bottom can help with holding the wires during soldering.
Or even better, if you have a 1/8" punch set, perforate the sheet with holes every inch or so to give yourself lots of places to hold MEMS chips. When it is right, you should be able to push the chip down into the hole so that the tension in the sheet holds it in place. This photo shows a several chips pressed in, during the build of the most recent mics. With circular holes, the wires can be fed in the opening at the sides.
If you use only the silicone sheet, there's nowhere for the heat to go, so the whole MEMS chip will heat up fast during soldering. I even had the metal back of the chip come off on one! An effective mitigation is to put a sheet of aluminum foil underneath the silicone sheet (it will contact the MEMS chip body) while soldering anything to it. You want the pad to heat up for soldering, not so much the whole chip.
The wires tinned and blistered-off the enamel insulation better with 350C. When doing that, touch the wetted iron tip to the exposed metal at the end of the cut magnet wire to heat it enough for the solder to climb up it -- heat doesn't conduct enough through the enamel itself.
Then tack-solder the wires onto the pads -- be sure the connection is good and shiny, but don't leave too big of a solder bump. The top left terminal, as shown, is the 3.6V supply wire. Top right is the signal output. The gold plated back can (the housing, which solders very easily) is the common ground. Flip the sheet over to get at that, with the chip facing into the foil.
Last edited:
Bypass cap and Ground connection
The data sheet calls for a 0.1uF capacitor connected between the supply circuit and ground near the chip. On the first batch of mics I did that by flipping over the silicone “fixture” and end-soldering a 0805 or 0603 chip cap on-end to the back of the MEMS mic housing, quickly, using SN60 solder to avoid overheating the chip. The stiffer ground wire gets soldered to the housing also, which turned out to be a problem. The on-end capacitor wanted to move or stick to the iron when the MEMS housing was reheated, very frustrating.
So, I’ve made a tiny circuit board design (just a horizontal-mounted capacitor over a ground plane) which helps a lot. The Gerber files for this will be downloadable too, soon, and also posted at PCBway. The pattern is smaller than the minimum board dimension that they can cut, so the board is designed a few mm wider. The board thickness is 0.6mm, with plated through holes. The solid back pad will get bonded to the metal back of the MEMS chip, by conducting heat applied at the plated-through holes on the other side. Then a capacitor will get mounted to the board.
You’ll get hundreds of them from a minimum board order (which is fortunately not expensive). Use some (of your not-favorite) wire clippers to cut off excess board from the width, then on the sections you'll be using, tin the pads on both sides of the board, being sure solder gets into the plated-through holes. The final clipped-out pattern needs to fit within the outline of the metal back of the MEMS mic chip, smaller is better. It is easier to position and cut out an individual capacitor mount pattern after tinning the pads because the wire clipper blade positions between the pads on the board.
With the silicone sheet (with MEMs chip seated within) flipped over so that the terminals of the chip face the aluminum foil, pre-tin a small spot of the gold back of the chip with the soldering iron.
Position the capacitor board over the metal can’s back. It’s best but not essential if the plated-through hole end of the pattern is at the aperture end of the MEMs chip. Hold it in place with the point of some sharp tweezers and heat the plated through area with the iron until it bonds to the solder on the MEMS chip can.
Tack-solder one end of the 0603 chip capacitor horizontally to the pad that doesn’t have the plated-throughs. Solder the other end of the capacitor to the ground (plated-through) pad next, quickly, so that the board stays in position. You get no extra points for appearance, this will get encased in silicone caulk, and strength isn’t very important.
The wire from the +3.6V pad needs to be soldered to the (ungrounded end of the) capacitor. Determine where it will contact when the wire is pulled across the pad and then carefully scrape some of the enamel from the wire there and heat it with a wetted iron tip till it’s tinned. Then solder it to the capacitor or its pad.
Tin the end of the stiffer AWG 30 magnet wire and (again, quickly) solder it to where the capacitor is bonded to the ground pad. Then carefully twist all three wires together so they’ll later be able to be fed down into the stainless steel tube. (I probably should have mentioned it earlier, but all the magnet wires pieces need to be long enough to reach to at least a few inches beyond the end of the stainless steel tube).
The data sheet calls for a 0.1uF capacitor connected between the supply circuit and ground near the chip. On the first batch of mics I did that by flipping over the silicone “fixture” and end-soldering a 0805 or 0603 chip cap on-end to the back of the MEMS mic housing, quickly, using SN60 solder to avoid overheating the chip. The stiffer ground wire gets soldered to the housing also, which turned out to be a problem. The on-end capacitor wanted to move or stick to the iron when the MEMS housing was reheated, very frustrating.
So, I’ve made a tiny circuit board design (just a horizontal-mounted capacitor over a ground plane) which helps a lot. The Gerber files for this will be downloadable too, soon, and also posted at PCBway. The pattern is smaller than the minimum board dimension that they can cut, so the board is designed a few mm wider. The board thickness is 0.6mm, with plated through holes. The solid back pad will get bonded to the metal back of the MEMS chip, by conducting heat applied at the plated-through holes on the other side. Then a capacitor will get mounted to the board.
Position the capacitor board over the metal can’s back. It’s best but not essential if the plated-through hole end of the pattern is at the aperture end of the MEMs chip. Hold it in place with the point of some sharp tweezers and heat the plated through area with the iron until it bonds to the solder on the MEMS chip can.
Tack-solder one end of the 0603 chip capacitor horizontally to the pad that doesn’t have the plated-throughs. Solder the other end of the capacitor to the ground (plated-through) pad next, quickly, so that the board stays in position. You get no extra points for appearance, this will get encased in silicone caulk, and strength isn’t very important.
Tin the end of the stiffer AWG 30 magnet wire and (again, quickly) solder it to where the capacitor is bonded to the ground pad. Then carefully twist all three wires together so they’ll later be able to be fed down into the stainless steel tube. (I probably should have mentioned it earlier, but all the magnet wires pieces need to be long enough to reach to at least a few inches beyond the end of the stainless steel tube).
Wow, that's so tiny! I'm impressed you're able to solder to it. You must be using very powerful magnifying glasses & super fine solder tips. I'm sure it would be impractical for me to even try. Even in my youth, it would have been tough to hold things steady enough and see clearly enough.
I recall another innovative measurement mic setup the late Chris Strahm of LinearX devised -- something like a parallel network of cheap Panasonic electret condenser mic capsules that together would end up exhibiting wide, linear FR & low noise/distortion. Can't recall exactly how it was done, and a quick google search didn't turn up anything. At least not on the first couple pages.
Assembly of the Microphone Wand
The MEMS chip, thin enameled wires, and bypass capacitor are attached with a glob of silicone glue within the holder, the holder is glued onto the stainless tube, and the cable (after splicing and thickening with heat-shrink) is glued inside the other end of the tube. The wires need to be fed through the tube, and spliced to the heavier conductors of the shielded cable. Heat shrink tubing is used to maintain insulation and help the cable fit snug into the tube when pushed in before gluing, so all has to be assembled in order.
The resistors are any high-value 1/8 watt sized types (100k or more), used only for structure to splice the very fine magnet wire to the much thicker cable wires and keep it all narrow enough to slide into the tube without shorting. There’s no room for splices to sit side-by-side inside the tube, so the resistors keep all the joints in a line. Cut the cable’s wires so they are only long enough to reach their splices or you’ll have difficulty pushing them into the tube.
There’s a lot to go wrong, but I think I’ve made all the major mistakes already, so you can avoid them! So here are some additional tips to avoid failures or lost work:
- Buy spare MEMS and capacitor chips. They are cheap and too easy to ruin or send flying across the room when tweezers slip.
- Electrically test every step of the way after wires get attached to the MEMs chip before gluing anything.
The test is easy, you just need a 3V power supply or batteries and a voltmeter, with some clip leads or pads to solder to. There should be about 0.7V between Signal wire and Ground wire when 3V is applied as shown. Check it while there's still time to fix connections.
- When gluing the MEMS chip end of the wand, keep the wand standing vertically with the MEMS on top. That’s to keep sealant or adhesive from oozing into the aperture on the MEMS chip (which would be ‘game over’).
- The magnet wires are springy and make the MEMS chip subassembly want to pop up before the adhesive bonds, so consider making some kind of weight to hold it down. You might add some silicone sheet at top to help protect the aperture and so that the weighted piece doesn’t get glued to the weight if adhesive pops up. I've hacked a 3D printed piece to do this, as well as a tube holder since I was making a bunch of mics for the test. You can use those if you want (I'll put the stl files somewhere for download),. Or just be creative.
Not too powerful, just 3.5X microscope -- you need field-of-view to move tweezers and iron around there. Very fine solder tip and thin gauge solder, though, yes. But those are easily found on eBay or Amazon or elsewhere for Hakko or compatible copies.
Don't be so sure. As I mentioned, I'm going on 71 yrs old, I have very bad arthritis in my hands (handwriting totally illegible even if I'm trying hard), and my eyesight isn't good at all (which a microscope you can focus helps a lot). And my girlfriend and kids will attest to that fact that I'm kind of a klutz. Patience is the most important skill.
These aren't too incredibly difficult once you figure out the tricks and get going
(if you look carefully, you can see the one with the MEMs chip mounted backwards/off-center at left).
(if you look carefully, you can see the one with the MEMs chip mounted backwards/off-center at left).
(Not important for a construction article, but --- )
A MEMs microphone can only be calibrated by comparison methods like side-by-side or substitution, comparing response against a known accurate reference microphone. I use a ¼” type 1 high voltage microphone with capsule type ACO 7016 (a clone of the Bruel and Kjaer 4135) for the reference microphone.
[I also have a 4135 which has since gone too noisy to use. But it had been calibrated a few years ago by an outside lab (using same method) and lived long enough for me to verify my Electrostatic Actuator and equivalence with the ACO 7016 when compensated for their calibration curves].
The reference microphone needs to have a valid calibration curve.
Ideally, I’d calibrate the reference 7016 microphone by the Reciprocity Method, but that method out of my reach (though Demian Martin has been working toward getting that capability). The next best is the Electrostatic Actuator method, directly driving the diaphragm of the reference microphone capsule with a high voltage signal to get the capsules “pressure response”. Curves provided by the manufacturer are then used to derive the “free-field response” corresponding to that actuator induced response and the capsules’ mechanical dimensions. This is the standard process used in manufacture and calibration of high grade measurement microphones. I used a Bruel and Kjaer Electrostatic Actuator device with diy high-voltage electronics and a modified version of Praxis measurement software to calibrate reference microphones and preamps. I was able to verify results matching a calibration of a 4135 capsule (Special thanks to Demian for helping me get this arrangement working).
I made a setup allow measuring a speaker with wideband output (Sony SS-CS5, with slight crossover modification) with the reference microphone and it’s determined calibration using a 5 second “log chirp” (or “exponential sweep”) and impulse response windowing. Then the MEMS mic being calibrated is placed in the same position within about 1mm and it is measured the same way. The point in space for microphone placement is marked by a felt-tip marker on the string of a plumb line hanging from the ceiling. Multiple measurements are made for each and averaged to reduce noise. The modified PRAXIS software figures out the difference and determines the response of the MEMS mic from about 70Hz and up.
Below 70Hz I calibrate against a DC-responding accelerometer chip that is riding on the cone of a sealed woofer. The nearfield pressure of a microphone placed in front of the woofer should have the same shape at very low frequencies as the acceleration of the cone.
It’s a little tricky to do, because the accelerometer is a sampling type and energy it sees above 1kHz has to be avoided or aliasing messes up the measured response.
The curves are combined and judicious smoothing is used to remove obvious noise and ripples and irregularities remaining from suppressed reflections.
Measurement and Calibration Methods:
A MEMs microphone can only be calibrated by comparison methods like side-by-side or substitution, comparing response against a known accurate reference microphone. I use a ¼” type 1 high voltage microphone with capsule type ACO 7016 (a clone of the Bruel and Kjaer 4135) for the reference microphone.
[I also have a 4135 which has since gone too noisy to use. But it had been calibrated a few years ago by an outside lab (using same method) and lived long enough for me to verify my Electrostatic Actuator and equivalence with the ACO 7016 when compensated for their calibration curves].
The reference microphone needs to have a valid calibration curve.
Ideally, I’d calibrate the reference 7016 microphone by the Reciprocity Method, but that method out of my reach (though Demian Martin has been working toward getting that capability). The next best is the Electrostatic Actuator method, directly driving the diaphragm of the reference microphone capsule with a high voltage signal to get the capsules “pressure response”. Curves provided by the manufacturer are then used to derive the “free-field response” corresponding to that actuator induced response and the capsules’ mechanical dimensions. This is the standard process used in manufacture and calibration of high grade measurement microphones. I used a Bruel and Kjaer Electrostatic Actuator device with diy high-voltage electronics and a modified version of Praxis measurement software to calibrate reference microphones and preamps. I was able to verify results matching a calibration of a 4135 capsule (Special thanks to Demian for helping me get this arrangement working).
I made a setup allow measuring a speaker with wideband output (Sony SS-CS5, with slight crossover modification) with the reference microphone and it’s determined calibration using a 5 second “log chirp” (or “exponential sweep”) and impulse response windowing. Then the MEMS mic being calibrated is placed in the same position within about 1mm and it is measured the same way. The point in space for microphone placement is marked by a felt-tip marker on the string of a plumb line hanging from the ceiling. Multiple measurements are made for each and averaged to reduce noise. The modified PRAXIS software figures out the difference and determines the response of the MEMS mic from about 70Hz and up.
Below 70Hz I calibrate against a DC-responding accelerometer chip that is riding on the cone of a sealed woofer. The nearfield pressure of a microphone placed in front of the woofer should have the same shape at very low frequencies as the acceleration of the cone.
It’s a little tricky to do, because the accelerometer is a sampling type and energy it sees above 1kHz has to be avoided or aliasing messes up the measured response.
The curves are combined and judicious smoothing is used to remove obvious noise and ripples and irregularities remaining from suppressed reflections.
Equalizer and Interface Box
I did a PC board layout for the circuit, including XLR connector, RCA connector, battery switch and LED. All equalizer parts are inexpensive, probably under $10 per in quantity. Mostly surface mount parts other than the connectors and some electrolytic capacitors. There are no special layout requirements for the circuit. The layout has been slightly changed from the one I built to help with difficulty I had in hand soldering opamp U1 (an SC70-5) without shorting pins.
There's also a 3D printable box for the board and a 9V battery (the box can be printed with either a resin or filament type printer).
The holes for the screws should be tapped for 6-32 or M4 screws (or self tapping screws). The board itself mounts via two screws into the XLR connector. The connectors holes also need to be tapped, either 4-40 or M3.
The 3D printed parts for holding the steel tube (because mic clips that hold 1/4" bodies are hard to come by) uses two pieces held together at one end by a 3/4" (19mm) screw and at the other by 1/2" heat shrink tubing around the clamshell pieces. The picture shows each part twice to view both sides of them.
Use a small piece of rubber sheet (could be from a rubber glove) in the area for the tube and the cable clamp to ensure they are gripped firmly.
NEXT UP sometime soon: This has been compiled from my scattered notes (fortunately typed in, since I can't read my own handwriting), drawings, and photos. Next is to collect all the files, and a parts and materials source list and posting them online so they could be downloaded. Should anyone want to take on making this mic. I assume no one is in any great hurry for those, though!
Thanks for reading
Bill
I did a PC board layout for the circuit, including XLR connector, RCA connector, battery switch and LED. All equalizer parts are inexpensive, probably under $10 per in quantity. Mostly surface mount parts other than the connectors and some electrolytic capacitors. There are no special layout requirements for the circuit. The layout has been slightly changed from the one I built to help with difficulty I had in hand soldering opamp U1 (an SC70-5) without shorting pins.
There's also a 3D printable box for the board and a 9V battery (the box can be printed with either a resin or filament type printer).
The holes for the screws should be tapped for 6-32 or M4 screws (or self tapping screws). The board itself mounts via two screws into the XLR connector. The connectors holes also need to be tapped, either 4-40 or M3.
The 3D printed parts for holding the steel tube (because mic clips that hold 1/4" bodies are hard to come by) uses two pieces held together at one end by a 3/4" (19mm) screw and at the other by 1/2" heat shrink tubing around the clamshell pieces. The picture shows each part twice to view both sides of them.
Use a small piece of rubber sheet (could be from a rubber glove) in the area for the tube and the cable clamp to ensure they are gripped firmly.
NEXT UP sometime soon: This has been compiled from my scattered notes (fortunately typed in, since I can't read my own handwriting), drawings, and photos. Next is to collect all the files, and a parts and materials source list and posting them online so they could be downloaded. Should anyone want to take on making this mic. I assume no one is in any great hurry for those, though!
Thanks for reading
Bill
This was a really remarkable achievenment and using MEMs mikes does virtually eliminate the calibration challenge. Getting consistemt measurments in free space above 20 KHz is really difficult even with the best tools. Every little thing alters the acoustic space. NIST's setup for free space reciprocity calibration https://www.nist.gov/system/files/documents/calibrations/aip-ch8.pdf probably cost taxpayers $100K and is really tedious to use. The consistancy of the MEMs mikes can make DIY speaker measurement more consistant and available.