[Sticky] USB-C Microphone (official topic)

@oz I forgot to mention. My DM (design mark) is named for and designed as a homage to my old cat Garrie. You'll see her on most of my designs. She was quite the character a true "Naughty Tortie".

@marcdraco wow that was quite an explanation !

I figured it was worth a bit of detail since I used to increase the output level of one of the capsule head adaptors coming in the next couple of weeks and also as a cautionary tale that bad information (even if its highly mathematical) tends to get copied around the web. A classic pre-amp doesn't need any gain at the capsule but there might be cases where we're feeding something like an Arduino A2D where we don't have the benefit of a dual supply, etc.

@marcdraco Hi there, I've only just become interested in this project and want to try and make one. Though I am a massive noob... do you have an updated version of the microphone? Also any idea on costs and time to make? 

The final "beta" units of the V2 (although they are numbered slightly differently due to the amount of development each one has had) are in production now, but it'll be a couple of months before Matt can run up a video due to the vast number of options we have now. Almost too much choice in fact... We wanted to give everyone an upgrade path from the existing design and I've been rather annoyed with the problems caused by the amount of fake THAT1512s on the open market too. It's such a performant chip that (predictably) the clones are everywhere, esp. Ali Express and eBay.

In the event, I cribbed a (fairly recently) expired patent from designer Douglas Self (Padless Microphone Amplifier, "Small Signal Audio Design, pp. 336" and adapted it for our use. In the event the cost for the whole pre-amp is lower than the original IC, about $6 per board plus shipping and import duties etc., which vary depending on where you live. The US is likely to catch a fairly hefty tariff if Donald Trump imposes duties on Chinese imports. That's outside of our control but there's nothing to stop someone making the boards inside the continental USA.

This design (which I'll make available in a couple of weeks, because I needed to iron out a couple of last-minute "whoopsies" that have to be fully re-tested) has almost 100% backward compatibility with Matt's V1 unit. This for folks who already have a working "head" but have had problems with the Vero/stripboard original. The circuit is deceptively simple (while heavier on component count) because a lot of the weirdness of a traditional discrete design is done by a very clever loop control Doug patented in the early 90s. Here's his original paper:

https://patentimages.storage.googleapis.com/52/76/46/8792fb86ea4215/GB2242089A.pdf

although it's quite heavy reading if electronics isn't your thing. That and the fact that patents use flowery language for reasons... (When my first was written up by a lawyer some time back, I didn't even recognise my own work!)

Self's design appeared in a number of high-end preamps, some of which are probably still in production. A low-cost operational amplifier fills in for some of the trickier bits of loop-control but leaves the grunt work a pair of discrete monolithic PNP transistors with lower noise due to being physically larger (more silicon). Self's version used a marginally higher supply voltage for 70dB overall gain, whereas ours is limited to +/-15V for simplicity and backward comparability, delivering 60dB (x1000), but that's more than any condenser capsule needs. Even ones with lower-than-average output levels only need a gain of around 40dB (x100) to push the digitiser to its beyond its clipping point.

Things not applicable to the patent like "P12 phantom power" and the power supply are all integrated too, but those very large capacitors have gone, replaced by more complex circuitry to do the same job at lower cost to us. It's not a practical design to do with Veroboard due the size and number of through-hole parts, even so the SMD version is smaller than the original and has allowed me to incorporate some trickery to isolate the noise from (in particular) isolated switched mode laptop supplies which aren't referenced to mains earth. These fellas inject a lot of nasty crud into the 5V input and that invariably creates annoying noise feed-through.

This is especially the case on Veroboard and even perfboard because there's so much copper floating around leaking unwanted magnetic fields all over the place. Matt's KISS design employs more brute-force method (nothing inherently wrong with that) but this hasn't translated well to a modern PCB design - and trust me, I've tried as have several others! In the end, I through the whole lot away and started from the ground up. (I'll probably update a THAT based design eventually with the new PSU filters, but early tests seem to suggest that it's not going to make a whole world of difference.)

The main advantage is there are fewer ways to get this one wrong and blow something accidentally - things like the NMA0515 inverter only go in one way around for instance and there's only one (optional) part that needs to go in the right way around. There's much less soldering involved too. The actual production units might be upgraded with shielded inductors and SMD headers to make life a little easier. It's a trivial change to do this but it does add a little to the cost due to those components needing a person to manually load the machines with the parts. JLC refer to these as "extended" components so I've kept those to an absolute minimum.

The head adaptors (the parts that mount direct to the rear of the TSC/JLI2555 mic capsule) are available in several flavours, you can chose the one that suits you best.

"Happy" which is already on GitHub is the simplest by dint of just being a more robust and slightly more versatile method to hold the JFET in place. Best performance noise is likely gained from the original JFET (2N4416) or the LSK170 but there several other totally useable devices which cost considerably less, some of which I've used on the more comprehensive "head" modules, but even the most complex one costs The trade-off is much lower distortion, at higher signal levels around the same as a single 2N4416 but slightly more electronic noise (mostly only measurable with test gear).

The truly adventurous have a couple of extra options with a hand-solderable, hybrid THT/SMD design and one with >20dB of gain that can be used without needing a pre-amp. But to stress, these are mostly for builders with a bit more experience or people upgrading P12 phantom standard pre-amps. They all use the same form factor so they can be fitted inside a donor body like the Neewer BM800 and similar. Not all of these will be in the video (or we'll be here another couple of years) but they'll be fully documented on GitHub. This is the Ultimate USB C mic on steroids!

We're open to a supplier to take on distribution long-term because JLC's boards are produced in batches of five boards (or two assembled and three bare boards). I can't really say for sure what the retail is likely to be due to factors beyond our control, but as an Open Source Hardware project, anyone can order some and sell the "spares" on eBay, etc.

How long any of these take to make is hard to say. A fully-SMD head and pre-amp (i.e. with factory soldered headers) is only restricted by the amount of time taken to construct the case and solder a single capacitor and the NMA0515. That device is available in SMD but I'm unsure on costs right now. There are lower-priced "clones" because it's an industry standard layout so for now I've left it with the through-hole SIL version per Matt's original. It's relatively trivial to swap that out for an SMD version, however, it's not a super-low-cost part (no getting around that) although it's cheaper and easier than putting a discrete dual-voltage SMPS on the board. Engineering is all about trade-offs so I've put a lot of time in trying to keep this as low-cost (and therefore accessible) as possible.

While not officially supported, the Jaime Quicksilver capsule adaptor can run from a single 9V cell and outputs a signal to a those lower-cost digitisers. The most recent version has had a complete overhaul with a better JFET and balanced output.

If that's made your head spin, imagine poor Matt having to wade through it all and make us a video! (Rather you than me mate! 🤣) I could literally write a book on this and if I can find the time, I might just do that a supplementary documentation for the fatally curious. I hate maths as much as the next guy (more than many) so I've done a lot of work to eschew that and developed techniques that don't need a college degree in higher maths to understand. It's surprising just how much you can do without having to dig into the "S-plane" with it's 3D spacial representations of poles and zeros.

Most engineers have college degrees so they forget the rest of us just want to make something that works and curiosity will only get you so far before the equations start to look like a spider smoked a bowl, dropped some acid, fell into an inkpot and then crawled all over the page.

If you can deal with some occasional silliness like that, you'll probably enjoy it. I'll include my own mistakes to - and I've made plenty, some of which are so embarrassingly silly, I've done a "Victor Meldrew":

For the uninitiated, "One Foot in the Grave" was British black situation comedy about a grumpy old man ... now almost embodied by another grumpy old man (me). I don't have a flat cap, but that's about the only difference. A man who never made a mistake, never made anything, and it's annoying to me to see people publish "perfected" designs but aren't transparent enough to show their working. Working out the errors, or having them explained, is a much more valuable learning experience than being presented with the product of several years of work as if someone just produced it from thin air. You don't need to know any of this to make the project though.

Several of of the leading lights in hobby electronics will tell you that there's a lot of poor information on the web and even in some textbooks. The CE amplifier which I explained in some detail up the thread, is a classic example because many designers simply don't know how it works at a fundamental level so they just copy what they've seen, putting their own spin on it. How do we know? Most of us don't... until it's far too late.

Hi! I finally received all the components from China (unfortunately, I can't order from eBay to Ukraine). I assembled the circuit, but there's no sound at all—not even noise. I checked the assembly correctness multiple times before and after testing. The ±15V are present. I noticed that the THAT1512 circuit is heating up, so I connected a USB tester and found that the assembly draws 0.3–0.4A. If I disconnect the THAT1512, the assembly consumes 0.02A. I tested with the microphone connected and disconnected.

The transistor is also from China, but they sent a used one, so there's hope that it's good and functional. Unfortunately, it's not from Motorola. 

Question: Is my conclusion correct that THAT1512 is the cause of the issue?

If my assumption is correct, can this circuit be replaced with something similar for this project? I'm already scared to order from China. Unfortunately, there are components in this project that are not easy to find. In my country, they can be bought in a regular store, but they need to be ordered, which takes more than 3 weeks, and the prices are just astronomical—$20 for the THAT1512, plus $5 for shipping.

Thank you in advance for your help!

You can replace the THAT1512 with other audio grade INAs - the most obvious choice is the 1510, the earlier model which is usually a good deal cheaper. The 1512 is a comparatively expensive chip no matter the supplier. I have an experimental prototype now which does away with it to keep costs down, based on a design by audio master Douglas Self.

Instrumentation amps (INAs) are fairly expensive because they are intended for precision applications often requiring low noise inputs/outputs, better power supply rejection and so on. Not all are created equal with some operating with very limited bandwidths because they are intended for DC or very slow moving signals.

I did a daughterboard that can take a couple of 8-pin DIL op-amps that operates like a THAT but it's never going to be as good as the real thing since it has to sit in the same socket. It's on my GitHub but you can make your own on strip-board with almost any common op-amp like the TL072 (cheap, noisy) or audio amps like the OPA2134 (quiet, more costly). Here's a white paper on how to do this.
TI discrete INA

256 mA is normal max draw for this circuit due to the the 1W supply with the THAT operating at high gain. Any more suggests something is wonky in the state of Denmark. May I ask how you measured it? That might be a hint.

What sort of test gear do you have? (Sorry to ask again, I forget who has what.) An oscilloscope is almost essential for testing this sort of circuit if it plays up. It's very difficult to see signals on a meter.

The JFET isn't super critical for Matt's original design, the 2N4416 is low-noise, high gain (when compared to a typical JFET). If the microphone capsule isn't in the screen it will pick up every bit of mains hum and be utterly unusable, but you'll certainly hear it. It will also show up as a 50/60Hz signal on many modern DVMs when checking the output with the meter set to AC.

If you can check that and get back to me I might be able to infer what's going on.

@marcdraco

Thank you for your prompt response! The thing is, I hardly understand this, I can only follow instructions. The only tools I have are a multimeter and a USB tester.

I measured the voltage without the chip installed:

• On pin 2, I got 14.7V
• On pin 8, I got -14.7V

When I connect the chip, the voltage drops to:

• 10.5V on pin 2
• 8.5V on pin 8

At the same time, the chip heats up within a minute to the point where it becomes impossible to touch. 

 "Audio Capture/Grabber Card Device - Analog Cassette to Digital MP3 Converter." It is working; I tested it on other equipment.

That's why I'm confused by the chip's strong heating and the fact that I have no sound at all, not even noise. I tried changing the connection via a 3.5mm jack, switching between the left and right channels. In desperation, I even tried swapping + and -, but there were no changes.

I would really like to get it working as it is since I have already invested a lot in this project (considering my financial situation).

A DVM is very helpful. Your readings give me a better clue as to what's going on.

The THAT1512 heating is only normal if it's operating at very high gain. I suggest you remove the gain setting resistor and re-test when the THAT has cooled down.

The 1512 has a very high frequency response, well outside of the audio range and I suspect what's happening is that it's picking up HF (>20Khz) or VLF (<100Hz) noise which is pushing the output transistors hard against the rails. This is difficult to detect on a DVM, its only usually visible on a scope.

We can infer this is the issue as the supply rails appear to be collapsing due to the extreme amplification. You might be able to see this on the AC range on your DVM but it's likely that it the RMS filter will not correctly measure it. (It will show something like 10+ VAC if it is.

The power pins are 4 and 7 on the 1512 - best to measure that directly. Pin 1 and 8 are the gain setting (so the voltages there depend on other factors).

Your inputs are Pin 2 and 3. The should have no voltage (they will show a few mV DC from the bases) on them due to the DC blocking caps.

Output is on Pin 6. Pin 5 is the GND reference - so you should measure anything with your black lead on that and the red one to take the readings.

What you've done is quite revealing. The circuit is prone to these problems. I've had to reimagine the whole thing from the ground up due to spurious issues with the NMA0515 leaking switch-mode noise either onto the supplies (this is a known issue) and/or the USB 5V is leaking HF noise onto the supplies. I've done some analysis of my own PCBs and that was such a serious issue I had to separate the grounds physically and only connect them via a small inductor.

I think Matt's original didn't suffer from this despite the generally awful strip board because the noise didn't get through as easily. It's ironic really. My testing also indicates that some PCs generate a lot of noise on the USB which is going to get through to your board.

I usually only work with SMD these days (it's lower cost, smaller footprint and generally quieter due to the reduction in parasitic coupling). That said, I've got a couple of unproduced boards based on the revised layouts one of which is completely through-hole so it will be easily made either at JLC, PCBWay etc. it's not really suitable for home production as it's got extra ground planes, it's quite small though so it shouldn't be terribly expensive.

The isolated ground is used right the way through analogue side and only connects to the USB ground via another inductor. I'll publish these to GitHub in a couple of days. Initial testing showed a dramatic reduction in HF noise even at full gain (a small amount will leak through but it should be outside the audio range and is small enough that the signal from the capsule will swamp it.

@marcdraco

I love smd, this is a 22nF silicon cap, and is considered large (0201) for the stuff I work on at work (400Gbaud data rates). We have custom stuff I can't post that's smaller than 01005. 🙂

Well colour me jealous and call me Shirley! 😶 

I'm limited to 0402 minimum on the standard PCB service of JLC. I did consider going to the precision boards for the capsule adaptors but in the end the extra cost far outweighed the benefits. Partly because it's so hard to work with 0603 never mind 0402 with a hand soldering iron (even one with a fine, temperature controlled tip.

Those are so useful when you've got a lot to pack in a very small space. There's an array of small caps on the back of my GeForce card - presumably some sort of localised storage but I don't have a schematic to refer to.

It is getting a bit ridiculous for home builders though, so if there new reference design works (first prototype did but it needed a bit of love and I busted a track, clumsy oaf I am). Second ones are due back next week and if all is well,  I'll publish the through-hole version of it to Github.

I made a couple of changes to the input side so it should work with P12 mics off the peg. Probably won't play nice with a 48V unbiased capsule as there isn't the voltage to pull the front "plate" nice and tight. I don't have one to check but the new capsule adaptors include a couple of P12 compatible designs.

Just had another member on - turns out they also had a dodgy 1512 too - seems the ****ing things are everywhere. I didn't self-censor, I said "star-star-star-star-ing" - well that's my excuse and I'm sticking to it. (I know it's an asterisk, but that's too much to type and I'm bone idle. 🙂

But totally with you on SMD. I love it. The parasitics are cut to almost nothing and it's possible to get the leads really short so noise issues are much easier to deal with. It's taken me a couple of years to get back up to speed - I'm so out of date, SMD was barely a pipe dream when I left the business. Most PCBs were single sided too, it was that long ago. The only people who did multi-layer boards with the huge names like Tektronix, etc. Plebs like me were stuck with a mix of silicon and (occasionally) still germanium. The first radio I made used OC71s (PNP, Germanium max operating temp 80C!). The reminded me of little liquorice sweets. Cute little fellas. I think Bob Widlar was still alive back then too. What a guy. Died far too young but I guess that was normal back in the day.

A lot of what we did was point to point - and some wire wrapping which was a really smart way to work, esp. for prototypes. Sometimes I miss those days and then I remind myself how much better things are now. It's pretty weird looking at old magazines from the era. Douglas Self has a neat collection of old Wireless Words dating back to the 1930s! It's incredible the stuff the early pioneers came up with considering what they had to work with like oil and paper capacitors. Not many people (myself included) feel comfortable working on those old sets which is a shame, but the arrow of progress always points forward. http://www.douglas-self.com/ampins/library/ampartew.htm

It's a horrible site (doesn't even have HTTPS) but he's an audio engineer not a web designer. We can't all be good at everything. I'm can't claim to be great at anything, my knowledge is broad but shallow which is very useful for what I do which is usually take a complex topic, learn from an expert and then break it down into manageable chunks the rest of us can understand (me included).

Hi, I'm looking into building this project as a gift for a friend. After reading a good bit of this thread (I was skimming through at certain points, and some pages I haven't read, so I apologize if I missed anything important here) I think my best course of action would be to order the Chimera board by @marcdraco from JLCPCB so everything can be consolidated into a single body. This would be a pretty much drop in replacement for what they're using currently (a USB mic). However, when I went to get a quote for assembly on JLCPCB I couldn't find a bom file in any of the zips provided in the original post for the pcb. Did I miss anything? I also wanted to double check if there was anything newer that might be better suited for my application, since that post is almost a year old at this point, and developments seem to be happening regularly still. Thanks in advance!

Hold off for now. The brand new V2 boards just dropped this morning with massive improvements across every area. Chimera has long been deprecated.

This isn't an official announcement, I'm always worried about jinxing it but if you want/need to get a jump on the game.

"Happy" (you'll find that on my GitHub) is a simple capsule mount for multiple FET (from the 2N4116 to SMD types) and works with the original Veroboard design although it's uncommitted so it can work with a crummy PC input too using 1.5-3V power. (Shudder.)

The three I haven't published as yet, are:

• Katie: mixed SMD 0805 and through-hole for hand-assembly is a P12 compatible.
• Jamie "Quicksilver" is P12 compatible with a 20dB gain on both outputs to drive most microphone inputs. The signal can be tapped of via a small capacitor for this purpose.
• Jax "Heisenberg": In theory the top of the range with some extra trickery to reduce distortion at higher input levels.

The original Michelle pre-amp has been split into several different designs, two based on the "Padless Microphone Amplifier" by Douglas Self. A remarkable piece of design for P48 systems that I've adapted for our use. The others are based on Matt's original with a greatly improved power supply that's cheaper and (supposedly) quieter. I use names and sometimes version numbers so we all know which major/minor version we're on. More fun that way.

The reason for the split was that some people are innocently buying fake THAT1512s (at the price, there's always someone who will re-badge up a cheap amp and sell it to the unsuspecting). This way we have one that works without needing any soldering at all, save for the NMA0515 which I can't get at JLC. Everything else, including headers can be fixed at JLC. I'll do versions for each

P12 is a 12V version of Phantom Power and requires two 12V positive lines which the signal is returned down. Matt's original was +15 and -15 volts which is fine with Happy doesn't match the more recent specification but simpler to implement.

The two mainline pre-amps will be available in full or partial SMD (or plain board at your choice) and I'm just finishing off the routing on through-the-hole (THT) in normal and "compact" versions. The compact ones are the same design but the resistors are mounted vertically to save real-estate. This is generally a "no-no" (the Chinese do it a lot) but it's far too easy to snap a component right off the board. The huddled up ones will probably be a bit noisier too.

All that said (and I know there's a lot) Matt will explain in more detail when the video drops, it's a lot less complicated than I've made it appear. Some boards are meant for commercial or expert (SMD) and the through hole ones are for everyone else. Cost-wise the Padless Amp comes in cheaper than the THAT on its own on a per-board basis with performance similar to the THAT. The through-hole versions are better for people learning electronics as you can chop and change some of the parts more easily than SMD.

Finally, the new pre-amps can be configured for the split-rail +15/-15 supply for backward compatibility (for someone who has the original capsule already built) but not at the same time.

Drop me a line here or leave a support request at GitHub/marcdraco and I'll fix you up.

Hello everyone!
Like many others here, I wasn’t able to find the necessary components for this project locally in Ukraine, so I had to order them from AliExpress. Unfortunately, this raises serious concerns about the authenticity of the parts.

I initially ordered a THAT1512 chip, but it just heated up with no audio output at all. Assuming it was defective, I ordered a second one — but sadly, it behaves exactly the same.

As a test, I also ordered an INA217PA. With it, the microphone does produce sound. It’s very quiet, but at least it works, which makes me think the rest of the circuit is fine.

I’m trying to figure out: is it likely that both THAT1512 chips are fake or defective? Or could I be making a mistake in the circuit or setup?

Any suggestions or advice would be greatly appreciated!

I looked at the INA217PA and it has a fixed gain of just 6dB (x2) according to the datasheet so my guess is you've been (yet another) victim of fake chips. It's ridiculous, but the suppliers are often unaware that they are selling fakes. They are inserted quite early in the supply chain. If that seems bad, consider that a large number of commercial airliners have parts sourced from official channels that (when investigated) do not actually comport with the required specifications. This has contributed to at least one disaster with loss of life, so it's a very real problem.

I have a couple of beta boards for the new design which doesn't require a THAT chip and performs as well in most applications but I have no idea what it would cost to ship one over to you with the current state of world affairs, drop me a PM if you're interested.

@marcdraco This is really frustrating, as I’ve already invested $45 into this project with no results.
Thank you for the offer, but at this point it’s easier to abandon the project — I’m not ready to invest more into it.
With shipping and components, your suggestion would cost another ~$20, which just doesn’t make sense for me right now.
Now I need to think about how to prove to the seller that the chip is fake — maybe I’ll be able to get a refund.

Hello everyone,

I have just completed my own version of the microphone as per the video and I have not been able to get any sound out of the microphone.

I checked the breadboard and capsule leads

for continuity and the connections that need to be bridged seem to be. I also probed some voltages on the back of the NMA chip as well as the THAT 1512 chip while powered and I have gotten only a few volts out of the THAT and 5V or close to it from most pins of the NMA chip.

It is worth noting that while probing the THAT pins I could hear static noise from the setup as I ticked the "listen to device" box to try and diagnose the issue.

I am not very knowledgeable about electronics but I have seen other users mention different voltage values for the back of these chips so I'm thinking either I wired up the 5V USB power incorrectly or some component died or is defective.

If anyone has any ideas I'd appreciate it.

Thank you

From the colour code you've wired the 5V supply backwards. By convention we use black for the 0V reference (ground) and red for the positive supply. It gets a hair more difficult when you have a negative supply (referenced to 0V) but I'll assume you've wired the board correctly.

If you pop the THAT1512 out of its socket (I hope you used a socket, that beastie is not the sort of chip I'd solder into place on Verboard)!

Anyway, so you should connect the BLACK lead of your voltmeter to Pin 5 on the socket and that's going to be our reference. Note the way the pins are numbered on any DIL chip like this:

1 -- 8
2 -- 7
3 -- 6
4 -- 5

OK, so from Pin 5 you should get the following:

1 = 0V
2 = 0V
3 = 0V
4 = -15V (it might be as high as 20V but don't worry if it does.
5 --- that's the reference.
6 = 0V
7 = +15V
8 = 0V

If you would be good enough to check these and reply with what you get we can work from there.

Marc

@marcdraco 
My mistake, I don't know the conventions for cable colours but I did make sure the 5V and GND leads went to the right pins on the circuit board.

Yes I used a 8pin socket for the THAT and I assume the volatges are to be measured with an otherwise complete circuit meaning potentiometer, mic cable and audio output plugged? In that case these are the voltages I am getting following your instructions:

1- ~40mV
2- 0V
3- 0V
4- 0V
5- ref
6- ~100mV
7- 0V
8- ~40mV

I really hope this isn't a bad sign or is something that would indicate a dead NMA chip as this part took the longest to ship to my remote location...

Here is a picture of my circuitboard front/back for reference. Maybe you will see something I missed.

Thank you for your time, contribution and help to all these users, Marc.

No need to apologise for something you don't know Dunkan. It's one of those things that's really useful to know like the resistor colour code:

0 = Black
1 = Brown
2 = Red
3 = Orange
4 = Yellow
5 = Green
5 = Blue
6 = Violet
7 = Grey
8 = White

Standard colour codes allow you to follow circuit hookups much more easily. Most electronics follow these codes and when you know the conventions it's very easy to spot (in this case) the supply lines. Even Chinese products tend to use red/black for power and usually green (or green/yellow stripe) for the system ground. Beyond that most bets are off, esp. in Chinese made goods. (Mains wiring has a different coding, but the UK used red/black for live and neutral for many years and we occasionally still see it in very old houses.)

The NMA has a fairly robust output side although I did break one recently when I accidentally shorted one from input to output. The problem with the original (the one you're building) is Verboard is a nightmare for beginners. You'd think it would be easy (it's a prototyping board) but the reality of it is it's a minefield, which is why I've been developing a reliable PC that's (almost) beginner proof. Poor Matt has had to sit on his hands for months now while I was getting it knocked into decent shape. I made a fair number of mistakes myself and I'm supposed to know what I'm doing... hence my profile pic. 🙂

OK, so those voltages are definitely wrong, so let's see if we can figure out what the issue is.

Nothing leaps out at me immediately although it's difficult to be sure from a photograph, so the next test is to check for continuity/shorts. From the photo it appears you might have a short on there between two tracks but it might be a trick of the light. So for this stage can you check each pair of tracks to make sure that they are not connected - use the "buzzer" test on your meter. If it does start screaming, check the actual resistance to make sure it's under a few ohms. Some meters count anything less than 200R as a short, this is by design so it can pick up shorts that only really go short at higher voltages causing too much current to flow in some part of a circuit.

While you're at it, it's worth checking the cuts (under the THAT) are complete by testing across the chip 1-8, 2-7 and so on. They look clear though.

Low voltages like this do imply a short(s) - particularly with voltages in places they have no business being, 1, 6 and 8 since they are ONLY connected to the IC. I can only apologise for being so slow in producing a finalised design, I know a lot of people have been waiting (the main one is ready now, with a couple of alternatives due in a few days).

Things aren't as bleak as they might appear, it might be a short is pulling everything down and the very fact there are voltages in places that should be "dead" in effect strongly suggests that. Hopefully the NMA0515 (inverter) has survived, so you need to check and isolate everything before you power it up again.

The problem of testing the inverter is that you really would need to remove it from the board and that's an utter nightmare. There's a guy on YouTube called Mr Solder Fix who has some cracking videos on how to extract difficult ICs from circuits without risking damaging the boards. He's definitely worth checking out. He has one (you'll have to dig a bit) where he demonstrates using a loop of wire around the offending pins and filling it with a pool of molten solder. It does mean wasting a lot of solder but the cost of solder vs. the cost of an expensive part - it's a no-brainer. He's definitely one of the best out there for repairs, including repairing busted tracks on PCs. Very clever lad.

If you're up for that, removing the inverter would be a smart move so you can test it out of the board.

If not, we'll just have to try and see if there are any shorts and work from there.

@marcdraco 

Thank you for the information about colour codes. Also yes I do hope the NMA has not died in the process... I did find a short across two couples of tracks  but the second one did not make the multimeter buzz the second time I tested it (I had done nothing to the board in the meantime).

I think in a stroke of bad luck it was across two of the tracks on which the inverter is connected... They were the two pins on one side on the NMA, I scraped the space between those two tracks clean with a cutter so should I try powering it on again and seeing if it works? I am referring to tracks 15 and 16 as labelled on the board itself (topside).
They correspond to the two legs of the same 2200uf cap for information.

If powering it up several times would have worsened the situation then it is already too late, I must have done that 4-5 times before writing in this thread. It would be a shame to have to spend more time and money a new inverter but the money aspect is personally not a big concern.
I undertook this project as it seemed fun and I could use a better microphone, and if all else fails I would have walked away having learned new things and dipped my toes into this world previously mostly unknown to me.

Don't worry about it Dunkan. If you know anyone else near you who is interested, the open source PCB version (Michelle 4.0) is now in good shape but it's not that cost effective to get just two made up at JLC. It is cheaper than than a single THAT. Matt will be doing a video in the next few months with more detail on how to make it (and the other parts of the new version) work together. There are several options depending on what sort of mic you're going to make - desktop, wired for studio use or even stereo.

If money is less of a concern, that would be the better solution as we know it works and it's dead easy compared to a Vero solution since almost all of the parts are pre-soldered and the THAT will only go in one way around.

It can't hurt to power it up now (without the THAT) to see if the inverter survived. MOS devices are pretty robust and they will take surprising amount of abuse before they collapse and die. With a bit of luck the short-circuit protection will have "folded" the output drive so the output MOSFETs didn't curl their toes up.

If you can do that for me and run the tests we'll have an idea if it survive - fingers crossed it will have. I've learned a huge amount by getting myself up to speed on SMD. When I left the hobby, double-sided boards were only just starting to find their way into hobbyist spaces. Nowadays even a relative beginner can access and use a board with 16+ layers - something we couldn't have had in our wildest fever dreams.

Any more than I could have dreamt of a CPU that ran faster than 1GHz. In my day a few hundred MHz was utterly bonkers, and yet today, little MCUs routinely run faster than that. It's progress, which is a good thing.

You've already learned new stuff by asking questions and running some tests, so it's not all lost. That's what I tell myself and it's true - but it hurts dropping a bundle of cash only to find you wired a transistor in the wrong way around (as I did recently, because I failed to check the SMD footprint was the same). A sloppy error and I paid for it in cold hard beer money.

But these little shorts are what makes Veroboard and similar tech such a nightmare to work with. It's aimed at beginners but really it's something only experts should use since there are so many ways it can catch you out.

I've done a couple of "bare" experimenter's boards which you'll find at GitHub that are cheaper than commercial ones and, because they have solder mask in place, it's a lot hard to accidentally short a couple of tracks. I don't make anything on these you just take the Gerbers (which you'll find on my GitHub page) and order them from a board house of your choice. It's up to you where you go although I tend to recommend JLC for pure cost effectiveness - and quality too. They are exceptionally hard to beat and lovely people to deal with.

All of my work is Open Source Hardware so anyone with the cash and the business nose can use them for profit. The larger one (Claire) has space for an 8-pin SMD chip on the back too and it's got a LOT of uncommitted tracks that you can use for your own designs. Solder (decent) quality pins into the places you're using the most and you'd make a breadboard that would last a lifetime!

This applies to the current boards too - the only thing I ask if anyone is running a commercial set (it's part of the OSHW licence we're using) that the credits remain on the silkscreen. Matt did initiate this after all so he deserves the credit for the work he put in.

@marcdraco 
I'm not sure what the price to ship the new boards would be but if it's not much more than a THAT it might be worth looking into, for now I would still like to get this one up and running though.

I triple checked shorts across the tracks and haven't been able to find any... Traces under the DIP socket itself are indeed severed.

I ran the tests again and nothing on the pins supposed to receive +/- 15V... I have seen 1.5V dropping quickly off to low values on pin 6 and the two pins with ~40mV still had those values on them. Does this mean the NMA chip died after all? Could it also have been a fake all along?

But yes I still have learned things undertaking this project and I do want to see it through. I might just make the new version when it comes out as well just for fun.

Um.... that does look like the inverter is an issue. (I'll curse under my breath.)

I never intended to do away with the 1512 but it became increasingly clear that there were so many fakes flying around (cough) AliExpress (cough) that it became clear an alternative was necessary. The fact that the entire board, assembled without the 0515 inverter costs less (per unit) than the 1512 was purely an accident of cost.  This only applies if you get it from JLC of course since they're sourcing parts in China themselves. I'd have a cow if I had to get these parts locally!

Michelle 4's circuit descends from a design by Douglas Self, one of the world's pre-eminent audio designers and subject of a mid-1990s patent - so it's got a very professional heritage. Any mistakes beyond that are mine alone. It's got a completely redesigned power section which is quiet without the huge capacitors which can push the devices a little hard.

It's impossible to tell exactly what's going on but as this stage your best bet is to look at one of those videos by Mr SolderFix and remove the 0515 carefully - it's not feeble but those pins will snap if you use too much force. A good solder sucker is essential as is some flux and quality solder wick - you need to get the pins free from the copper tracks or some might just get ripped off. The cheaper versions are far worse in this regard. I've ruined a couple myself.

If you remove that device it can be tested on the bench (or in a socket if you have something suitable). You only need to hook up the power pins and then measure to check it's giving out the right voltages which are >20V per pin when it's not "loaded".

There's not a lot that can go wrong with Matt's design if it's assembled properly except the issues created by prototype board. As a matter of interest, before you do that, you might want to check you're getting 5V USB *IN* - the input capacitor can be a bit hefty for some USB PD chips to drive and the 0515 needs a minimum of 4.5V to operate. 

You can (not that I recommend it) make up a "fake" THAT on Vero. using a couple of everyday operational amplifiers. I've done this with everything from TL072s through to proper audio op amps but the fact that it needs to be a daughterboard means it's only good enough for simple testing. If you're interested the circuit is described by TI here:

@marcdraco 
Could you explain to me how to probe for the USB power to the board?
Also Is it possible to test the amplifier while it's still soldered to the board? I don't know how I would supply it when it's free floating and I'm not too sure what I would need to look for.

I currently do not have any other equpiment, I bought what's strictly necessary to build this microphone. The only stuff I have extra of are the brass bars for the arm, the pins and corresponding plugs to crimp cables and solder...

I don't feel comfortable to make a fake THAT anyway but thank you for the suggestion

You can test the THAT1512  by measuring some resistances between various pins (I'll get some measurements from a real one tomorrow) but the easiest way to check the 5V is at the input to the NMA0515. If your 5V USB is shutting down, the inverter won't work.

Resistance measurements sound bonkers but it's surprising how much you can figure out because quite a few pins have specific functions which will throw resistances that should be quite easy to confirm against a working device. It's not perfect by any means but it's something that is done with a device called a "Scope Octopus" (although that works visually, it's basically the same test.

The input pins (2 and 3) are pretty much just the bases of a couple of bipolar transistors for example. It's more a case of checking a known working device and seeing which pins respond to being probed with a meter. Not all will of course.

@marcdraco 
I just tested across the Vin and Vout of the NMA amplifier and it is being fed 4.96V and outputting the same exact value. Does this mean the USB power supply is good and that I have a dead aplifier? Because the pins on the DIP socket are still reading values close to that reported above.

Also for resistance testing of the THAT1512, if I understood correctly I need to plug it in and probe the back of the socket to get readings?

While I'm at it, Matt mentions in the video that the 4th leg of the transistor that is soldered to the back of the mic condenser is not needed and can be chopped off.

I think I saw a user mention it can be soldered to ground to improve audio quality? This is what I did, should I just chop it off and remove it or is it fine to just leave it as is?

Sounds like the NMA0515 has bought the farm - or it was a fake but it would be the first one I've come across in this thread. Not impossible but unlikely due to how these things are made and the profit margins, etc.

Resistance testing a THAT (or any IC we can get at the pins of) is usually done outside of a circuit. Even an octopus needs the chip to be tested either in isolation (not plugged in) or against a circuit with a known good chip as the surrounding parts will affect the readings to some degree. Resistors are the biggest culprits but any other component can factor in.

You can do live tests but that assumes that your board is powering and that's not the case here.

I would guess the THAT is actually genuine and something else (perhaps a momentary short) has killed the inverter. As I said earlier, I've done that myself and that was in a working board - I was just sloppy with a test lead and it shorted something before I could stop it. No magic smoke, but no worky any more either...

The NMA0515 is an inverter - more specifically a dual-voltage, isolated boost converter. It takes 5V USB and converts that using an oscillator and some very small toroidal (do-nut shaped) transformers to generate the required voltages. It's a switched mode supply too so it's quite efficient in real terms.

The THAT is the amplifier - that takes a signal and makes it larger - so "amplifying". We could argue that the NMA0515 is a voltage amplifier but due to the way that it works (using inductors) it's more correct to call it an inverter or a boost converter. It's worth trying to remember the various terms in case you happen to be talking to someone who is less patient than the folks here (and there are a LOT of very pedantic people in electronics).

I won't say who, but I saw a post in a well-known forum taking a swipe at Matt for saying that microphone design is both an art and a science. Matt is dead right, but they grumbled that electronics is a science (which it is). What they missed - and this is crucial, is that microphones - particularly ones for vocal and instrument use as here - don't have a "flat" response. Ask any microphone designer and they'll tell you the same thing. We design mics to do a job and, particularly where vocals are concerned, most mics have a little bit of boost in the high-end (>6KHz) to even out how things sound to the human ear.

Doug Ford (formerly Chief Designer of Rode Australia, a little company you might have heard of) and all around good egg, described as a "whoohoo". Rode got feedback from an early design (which was as flat as my singing) along the lines of "it sounds boring". THIS is the art - that and the pure art of Matt's unique desktop design for the original. The new boards (for anyone wondering) will drop into the old one with very little modification but have more features because we could. 😉

It's just something to watch out for. I got absolutely pummelled by one guy for asking about altering the design of a famous discrete operational amplifier. The word used was "cheapen"! The problem is that the design called for a transistor that's been out of production for decades now so making one these days requires changing some parts. That's the sort of flame war that's all too easy to get into.

I try to remember what it was like when I was starting out. No clue, a couple of screwdrivers and some old TV sets which repeatedly made my hair stand on end and smoke. (Don't laugh, I'm not kidding.) How I didn't kill myself and/or set my folks home on fire is something that will go down in history as lucky as  winning the lottery jackpot three-times in a row! I really was very, very lucky indeed. 50-some years later, I'm a lot more careful but I won't ride someone for asking a reasonable question or making the sort of mistakes even the best of us do. This is also why I talk about my own screw-ups which (I know) a lot of designers won't. We should never be afraid to try stuff or we won't learn. I only recently found out I'm neurodivergent which was news to me personally but not one who knows me. Imagine that! I don't have to prove myself to anyone, but if I can show that even someone well-versed (I trained with the a branch of the UK miliary) can still make mistakes and admit to them, then no one needs to feel shamed OR ashamed for making a boo-boo.

Just better leave someone like me to do it first so it's harder for you chaps at the sharp end to go wrong! 🙂

It also gives me an excuse to develop more exciting projects. There are a few in the pipe - but this one has been an absolute beast because it's so impractical to test the capsule ends on a breadboard and that's where I did most of the development. (At the capsule, not the breadboard). The adaptors are in two form factors - a 25mm and a 34mm format and most have around 10-12 components on a low-cost PCB (so I'm restricted into using "larger" SMD parts and more space between tracks). The add in that most of the roughly 80 square mm of board is grabbed by not just the two solder tabs and centre tap, the back of the capsule also has to breathe - so the things have huge holes wherever I could put them. (Strictly I could have gotten away with smaller ones but no one would believe that - and there are other considerations like the distance the pressure wave has to travel etc.)

Matt and I had the idea, fairly early on, to get the entire device on the adaptor - with a 5V in and a signal out - which I did manage but the digitiser was an issue doing that. In the end, I converted most all the capsule end boards into to do everything from just carry (almost) any N-type FET all the way through to a balanced output P48 (48 volt phantom a professional standard) with several variations in between. Most didn't work as well as hoped or at all in some cases, so I've spent a fair amount of time waiting for stuff to come back from China so I could check the latest changes. Some days the boards would sit on the bench for a week while I plucked up the courage to test them. The current one is a bit nerve-wracking for reasons I won't go into here in case I jinx it. :/

While I was still learning my way around KiCAD I accidentally caught a "via" (a tiny hole that's used to send current to the other side of the board - or the inner layers). Anyway, the via moved a few fractions of a mm, shorted another line to ground and ... weeeeee Magic Smoke. Now it's a case of run the checks several times even after locking everything down. Even then I still forget something from the parts list (BOM) - most recently I put a 0402 part on the board and a 0603 sized part on the BOM. JLC's people caught that (like I said, *amazing* service esp. at the price we pay outside of the USA).

To satisfy my own lurid curiosity, where in the world are you?