[Sticky] USB-C Microphone (official topic)

Hello everyone,

I’ve been exploring the forum extensively, but I must admit I’m feeling a bit overwhelmed by the amount of information scattered all over. I’m planning to build an XLR version of the microphone, and I see that the OPA Alice design is intended for that. However, I’ve come across several optimizations (especially by @marcdraco), and there seem to be countless suggestions regarding capacitors. Is there a summarized and updated information or a thread that specifically addresses the XLR version?

Additionally, I have another question that I haven’t seen explicitly addressed — if it has been, that would be nice! It concerns an amplifier with XLR input and output. I understand that the amplifier built by Matt was optimized for a the usb-setup, but does anyone have insights or suggestions regarding an “extracted” microphone amplifier?

Apologies if these questions have already been covered!

Simon 🙂

Varee (detailed above) outperforms Scott's Alice and OPA Alice designs in several key areas, most notably the locked current loop which controls the JFET, and a 20dB boost. Alice is essentially an op-amp version of the 1960s design by J Wuttke of Schoepes, but does away with the discrete JFET and drives balanced lines using the op amps.

Noise wise, I don't have an Alice design to test and but I've used the OPA2134 in an alternative USB design which I've talked about; however it needs an update to the new Varee design. Minus the capsule's self-noise (varies depending on which one you use) and with the stock, basic JFET (2SK208) is almost inaudible on a 2nd Generation Focusrite Solo with everything maxed out.

Swapping out the stock JFET for a better model such as an LSK170 may improve matters but I doubt it will make much difference either.

The original USB mic was never intended for P48 (phantom power) as such but used a variation of the normal biasing scheme to develop a differential output over a short balanced cable. Distortion in this configuration rises sharply with increasing input voltage. This affects the Wuttke's original too but in the 1960s it wasn't something they worried unduly about compared to a valve and transformer-based solution.

Varee's stock version is set for P12 power, which is what the DIY Perks design uses, although it only uses +12V.

The difference between the two is a single 0805 resistor which can be swapped out to 6k8 with a little care. It will work with P48 without modification but the extra current draw may bring down some non-compliant devices. I hope we'll be able to find an eBay seller to keep both in stock as they have to ordered a minium of five which is obviously not cost effective for a single project.

Discrete transistors lack the advantages of negative feedback (per an op amp) but the much larger die sizes reduces noise levels. The feedback loop controlling the current serves a similar function, but I've used medium-power transistors where possible to reduce noise.

What really separates Varee from the various Alice designs is that it is designed specifically for "floating" designs like Matt's. As noted up the thread, I've added a couple of screw holes to mount it to the typical Neewer BM800 bodies. It's a quick way to convert a so-so Chinese mic into something that will give a Rode NT1a run for its money.

This is a working Varee connected to an XLR (there's a cheap 25mm capsule soldered to the other side. It doesn't need a huge amount of screening as the PCB is already shielded. It does pick a little hum so it will need a screen (like a head basket) but it doesn't need anything massive.

And of course, it has a power LED. 😉

I developed this version alongside the full V2 for my own use since I have some P48 equipment but lacked decent mics to use with it. 😉

Michelle/Mary/Quinn/Alexandra/Helios/Hathor/Chimera

Although Alice is named for Alice Through The Looking Glass, due to its small size. I use various friend's names so it's easier to know which board we're talking about.

In all there is are up to four more boards for the main pre-amp now and a couple of effects projects. Most users will only need two. The main board (still in development but close) uses an improved version of the THAT1512 pre-amp and adds regulation, optional fully isolated power and an optional balanced output for other pre-amps (using... drum-roll... an OPA2134 although it could be almost any dual-channel op-amp with low-noise and low-current: NOT an NE5532, they're far too power hungry.).

Alexandra provides extensive filtering, including an adjustable low-pass with a high-frequency kick.

Quinn and Mary are two optional boards which I'll detail when the rest are done.

Chimera is the JFETed head with an OPA2134 with a low-pass filter and 20dB amp intended to drive something like a Soundgrabber directly in the BM800 body. It uses a source follower but the locked loop in Varee is more performant so I might revisit that at some point.

Helios is a circular Vu meter using Neopixels - I should have the prototype back in a week. Nothing esp. exciting, it's more about "ooo look, flashy lights although it does help set the levels. Hathor hasn't even made it to the test bench yet which is a shame since IF it works it provides all sorts of fancy digital filtering.

What doesn't it do? The main difference between Michelle and a "proper" professional pre-amp is that it doesn't output "professional" level audio. I've diode  limited the output to a maximum of around 1V P-to-P vs. the usual professional nominal level of +4dBU - 1.23 volt (up to +30 dBU about 25 volts in the peaks). Consumer gear sits around 100mV to 300mV but "Line" is supposed to be able to handle 1 volt - or 0dBU.

The THAT151xs are more than capable of blowing way past consumer audio and easily handle professional output, resulting in yours truly blowing up two Audiograbbers before I'd realised. (Tsk, tsk, slaps self with large, cold haddock.)

Michelle can be converted for professional audio by removing the clamping diodes and bypassing the filter board as that is designed for consumer audio and is powered from the USB supply. The differential output is driven from the +/-15V supply so will deliver professional, balanced audio with a standard 200 ohms impedance. There are several configurations for the Michelle pre-amp and it's better described in a video plus instructions than me trying to waffle through each.

Capacitors - to be or not to be.

Capacitors are a very large topic indeed with specific types excelling in particular areas. Modern "multi-layer ceramic capacitors" MLCC offer very large values >10uF in very small packages. I've tried to keep the cost down here by using the X5R and X7R types although they are not generally considered to excel in audio. I know a lot of people will get sniffy over this but recall that this is design to mount on the back of a small microphone capsule - up to the largest 34mm in fact, so space is at a premium: including that it has to expose the back of the mic to air pressure for the differential cardioid pattern!

It is possible to get high-quality (i.e. NP0) MLCCs but they are only available in larger bodies/smaller values. The "true" audio capacitors like polystyrene and Teflon are much larger still and tend to be available in through-hole only. Electrolytics are the largest of all, available in very large "capacitances" are the weakest of all. Some of this also has to do with the minute DC leakage through MLCCs which isn't found in the plastic polymers. Such a capacitor would be essential to separate the 48V bias voltage used in unbiased capsules. We're using electrets here so the capacitor isn't necessary.

But the real shocker (as noted by the great Rod Elliott of Sound-Au.com) is that every piece of music you've ever listened to has been through a LOT of capacitors. And I mean a lot; and if your amp is Class D (many are now) it's been through inductors too.

The main gripe against MLCCs is the fact that the can cause audio-frequency noise to occur if they are exposed to vibration. In practice, I've found that in order to "hear" this effect, you have to remove the microphone capsule, drive the amplifier, in my case the Focusrite Solo, all the way up and bang the mic body on the desk.

Such "vibration" is so severe the the microphone proper would be unusable so the actual effect is negligible in practise.

And there you have it - another essay to answer a simple question. I haven't done a dedicated post for V2 yet because Matt hasn't signed off on the final designs and I'm still fixing the odd gremlins that creep in when I look the other way. One time my cat walked over the keyboard and managed to switch a bunch of parts to "do not place" which meant I had to solder them when the prototype came back. True story!

@marcdraco 

Hello! I’m really interested in learning more about the Varee preamp. As a beginner hobbyist, I find myself a bit overwhelmed by the various ideas and versions of this amazing device. I have a few questions that I hope you could help me with:

1. Will it work with dynamic microphones?
2. Are there any potential issues if I use a USB connection as the output instead of an XLR?
3. Can I use a 5V to 12V step-up converter to power the preamp while using a USB connection?

I apologize if these questions seem obvious or if they have been answered elsewhere; I’ve had a hard time finding answers to these specific queries. I’m truly impressed with the tweaks and upgrades you’ve made to the DIY Perks version! Thank you so much for your help!

I apologize for the confusion earlier. I realized that Michelle is the appropriate preamp for a USB connection rather than Varee. With revision to my question:

1. Will Michelle work with dynamic microphone?

2. Can I connect my dynamic microphone to Varee first, then connect Varee to Michelle, and finally to my PC?

3. Which boards (and their specific configurations/arrangements like Varee -> Michelle -> etc.) from all your iterations would you recommend for maximizing a home recording setup?

Thank you!

Dynamics?

Now that's an interesting one. Dynamic microphones are different from condenser in several key ways.

<SIDE NOTE>:

A quick segue into fields because they are something of a dark art unless you have a Ph.D in applied mathematics and can run Maxwell's equations in your head. You can skip this if you want but as a beginner this stuff is useful to know when you're starting out.

I forget (someone will remind me) who discovered this, but in the earliest days of electronics when humanity was just developing these ideas, someone realises that if you MOVE (movement is key) a permanent magnet near to a piece of copper, energy is transferred from the magnetic field into the copper wire causing a current to flow. The word induct (induction) comes from the Latin, inductus meaning to initiate. Hence the old boys determined it was initiating current.

They also figured that when the magnet was stationery but near to the wire then the disturbed electrons would be held in place because it the magnet was removed rapidly current flowed but in the opposite direction. This property gives us electric generators like dynamos (DC) and alternators (AC) and powers our entire world.

Nuclear? Coal? Gas... all do the same thing by using the energy from the material to boil water and drive huge turbines that create the AC we take for granted.

The energy comes not from the magnet, but the magnetic field - and in essence the movement of the field pushes the electrons along the conductor.

The curious thing is that when you do the reverse and put electric current into a wire it creates a magnetic field. Concentrate that field a little using, oh, a coil of wire or some iron-like material and you can store a small amount of energy. When you remove the power that's creating the field, it doesn't just vanish in a puff of quanta, it collapses back into the conductor and forces the electrons back the way they came.

A tiny amount of energy is wrapped around a bit of wire but you have to put in energy to keep it there? This doesn't sound very useful does it? Unlike a battery which has loads of current ready when you need it.

And that is the gotcha. These tiny (and I mean really, really tiny) amounts of energy can really spoil your day although many circuits: switch-mode power supplies, boost and buck converters, even modern "Class D" amplifiers can only exist thanks to these feeble little fields. We just change their state really, really fast.

When I'm laying out a PCB, I always remember the words of the greats "the energy is in the fields".

Fields are invisible, we can't see them but they are everywhere. (OK, light is also a type of energy transferred in a field but that's a special case and we don't see the field, rather the quantum effect created on our retina.)

Every piece of wire, every trace, every conductor and IC is surrounded by complex electro-magnetic fields and to control noise, etc. we have to steer the fields the way we want them to go by designing a layout so they don't get much chance to interact with other conductors that lay nearby.

I rather suspect (and this is physics outside of my area) that ripples of energy disturb the quantum realm and go on forever but for our purposes even a relatively strong magnetic field (and you can feel that if you put the two identical poles of a magnet together) has a very limited range. The magnetic field around even a coil of wire with a modest amount of current flowing is quite feeble. Even the mighty electromagnets we see in scrap metal merchants can only grasp a car from a few inches away.

If you remember the inverse square law from high-school physics you'll have a better idea. If not, think of shining a torch into the gloom at night. The brightness fades away at the same rate.

</SIDE NOTE>

So after that physics review, you will note the difference between condenser and dynamic microphones is the way the fields are moved.

From my scribblings up the thread, you'll see that capacitors/condensers hold charge by putting electrons and an ions at on the opposite side of two conductive plates where a quantum effect holds them like little magnets that are close but not quite able to reconnect. Like being at the zoo and touching the glass with a man-eater on the other side trying to eat you. If you thought magnetic fields were weak, that's nothing compared to the electric fields that hold those electrons in place but at close range they are incredible powerful. Capacitors can store hefty amounts of charge because of this.

Recall that when we speak into a condenser we're moving one of the plates, thus adjusting the value of the capacitor a tiny amount. This affects how much charge it's able to hold and therefore the physical movement is changed "transduced" into electric current.

Dynamic mics work using a moving coil - just like you'd see in those early physics experiments where people would wow as Faraday moved a magnet through a coil and made needle move. We can still do this experiment today but few people bother outside of high-school, we just take if for granted and then forget the rule of thumb - more on that here: https://www.pasco.com/resources/articles/right-hand-rule

In the dynamic microphone a very light (low-mass) coil is fixed to a piece of material and allowed to float in a magnetic field. As sound his the diaphragm it moves the coil of wire through the field and that induces a current in that wire.

On the surface it looks like we have the same effect, right? We have something that converts varying air pressure (sound) into varying electric current.

If it's not obvious, dynamic microphones are difficult to make. That's not to say condensers are "easy" (although you could make one at home with the right parts, that's for another day) but they face a huge problem of mass. In order to make the device sensitive to small changes in air pressure, the mass of the coil has to be small (fewer windings, lighter diaphragm) in orders of fractions of a gramme but in order to make sufficient current to be useful, it needs more windings. It's like the light speed problem in many regards - so we have to make trade-offs.

Now compare that to the condenser microphone. The diaphragm can be small (modern MEMS ones are measured in thousandths of an inch!) which, even in an LDC (Large Diameter Condenser) with a 1" diaphragm has a tiny fraction the mass of the coil assembly. This means that condenser microphones, even nasty cheap ones, are incredibly sensitive to changes in pressure and produce detailed sound that the poor dynamic mic can only dream of. This is like pushing a bike versus pushing a car. The car, with it's greater mass takes much more energy to move than your little push bike.

So why do we even use them?

Dynamic mics are useful for vocal use - mostly for singers (rock stars in particular) and conference use but the experts in this field like Sure have produced ones which (they claim) are sensitive enough for instrumental use. Not only that, but they have much lower impedance than a lower-cost unit because the lower mass coil (the number and type of windings determines the impedance).

Even then they are about 1/10th the sensitivity of a condenser so they need around another 20dB of gain (x10) to achieve similar performance to a condenser, especially those like Varee and some mid-range condensers that have some internal gain (most don't).

It's worth noting that although dynamic mics can be used for instruments, particularly percussive ones, they pale when compared to condensers - even "cheap" ones. Although condensers can be a noisier than dynamics for the same output (due to the impedance matching circuit) they cannot reproduce the detail.

If you're curious as to what I'm blithering about, see if you can "audition" a pair of really nice headphones. I treated myself to some Beyerdyamics 990DTs which legendary, open back phones. I won't try to explain why this is important (except to note it's the "ideal" infinite baffle as imagined by speaker designers) but you have to hear these things to understand why.

Even everyday movies and television produce excellent sound from orchestral instruments like strings) to the absolute position in space produced by an HRTF codec for your copy of Doom, GTA 5, Battlefield, etc. these fine details matter. Audiophiles describe the usual sound (and this includes many headphones, including gaming headsets) as muddy or lacking in detail. This is because, like the coil mass in your dynamic microphone, they have to expend a lot of energy to create the changes in air pressure. If the mass is too high (or there's a pneumatic effect from a closed back) then little details are lost. And your ears (actually your brain) will notice.

I'm skipping all the maths here to keep the explanation sane. Also, I'm rubbish at calculus so I'd probably make a fool of myself so there's that

Varee is purely for electret condenser microphone capsules. (I'm working on something that removes that limitation but it's so "out there" that it might never see the light of day. The original proposal came from an early 1960s paper by the great Peter Baxandall and has been reproduced since but I'm trying to realise a more modern version. It uses the variable capacitance of the capsule to control either phase or frequency in a radio frequency oscillator; 1MHz in Peter's patent.) Peter used variable phase to create an AM signal - and, as with everything he did, his solution is as elegant as it is beautiful. See: https://www.jp137.com/lts/Baxandall.RF.mic.pdf

If AM sounds like - "wait a minute you mean those crappy old radios?" yes, yes it is but if you can work through Peter's paper he explains why this is preferable over the FM bands we're used to using in portable radios for high-quality reception.

The transistor symbols are different to the ones we use today and he talks about "mega cycles - Mcs" which we call Hertz (Hz) now but the techniques remain valid.

TL;DR

Short answer you don't need Varee for a dynamic capsule - it wouldn't be much use.

Michelle should work with a quality (low or medium impedance at a pinch) dynamic mic BUT you must not connect up the "phantom" power which is designed for (and required by) condenser mics as it will energise the voice coil and destroy it in a matter of seconds. I don't have a decent dynamic mic to hand to test this. A Sure SM58 would be ideal and remains the most popular mic in the UK and probably across the world due to its reliability and robust build.

As designed, the Michelle board doesn't supply any DC to the XLR connection - there are jumpers to select the various combinations either bipolar (original design) or unipolar (Varee only) 12V supply or none at all. In that case you would just connect up the XLR to the ground and the two inputs hot/cold to the inputs because Michelle's input stage was designed by the THAT Corp. for low impedance mics.

Here's a shot from the 3D render. + and - pins are hot (in phase) and cold (inverted phase). Those SMD pads are shorted 1-2 or 2-3 according to the voltage required at the output, so for a dynamic mic leave those untouched.

Ground needs more discussion but either mode should work since here's a little dirty secret, ground isn't needed on unpowered XLR because the signal carries its own ground reference! Don't worry about that right now, I promise I'll write all of this up in details when we get to the assembly portion.

If you don't want to go the trouble of building your own, Sure has the remarkable MVXU2 which uses LSI to cram a 5->48 V boost converter and an ADC into an XLR line module. I haven't heard them but at the price around £130/$170 ands://www.youtube.com/watch?v=1YfgIj-mYiU it looks hard to beat although the performance may not be as good as Michelle overall.

OK, so I just realised I missed one of your questions.

Which combination?

Varee really developed from Matt's idea to put the JFET at the capsule and run wires from that. You'll have noticed that it originally started out as just a PCB with some experimental JFET designs, some worked, some failed terribly for various reasons, but my own error/ignorance usually because I have a nasty habit of bypassing things that take a lot of setup (simulations for example) and breadboards.

So Varee is the result of all of those experiments although some designs that came out of it are still quite valid for other applications such as putting a digitiser directly into the body of a BM800. I haven't developed that idea as far as I could due to time and budgetary constraints.

Varee comes into "her"* own in two applications:

1. With the current limiting resistor (marked on the board) set around 1K it operates on 48V phantom power supplied by pretty much every piece of professional audio gear out there. You'd still need to mount it on something because although it's inherently screened, small amounts of noise can creep in. Beyond that you can use a Focusrite Solo or something you found on Facebook marketplace provided it has that 48V phantom power. Put another way, if it doesn't offer that, it's not good enough to bother with anyway!

2. With the resistor left at the default (10R - but subject to change in future) it works with Michelle in the fully isolated mode. At least, that's the idea. I'm still running experimental designs about one per month to perfect this. Isolation completely removes any crud coming from your laptop by separating the power supplies completely and removing the ground from the main circuit ground.

Michelle was, where practical, meant to be an upgraded version of Matt's original so I've fought tooth and nail (and you should see the state of my poor toenails now!) to keep as many of the original parts as practical. So it uses a THAT1512 - but can uses others including the Quinn discrete INA or a THAT1510. The values of the single gain resistor (or the ones on the rotary switch) do change depending on which one you use, but I'll put the recommended values in the build log and not repeat all that here.

Advanced builders can use SMD parts should the DIP versions become unavailable in future too. So while the board is still experimental at this stage, when I have one I'm sufficiently happy with, it should be future proof(ish).

As an Open Source project (commercial licensing is reserved for DIY Perks) other builders are free and even encouraged to make improvements. One change I looked at but ultimate went a different road was the isolation. It's possible to isolate the THAT, mic power, etc. but when Varee is added into the mix - (not to mention other little bits that I've added here and there) the NMA0515 just doesn't have the output to drive everything.

It's possible to get a 2W and 3W industry standard fitting converter and drop it in place but (at least this is my logic) many people will already have them AND an original DIY Perks head assembly so they won't want to go out and buy a whole new set of parts. That's wasteful (and bad for our planet).

In view of this I split the power into two sections - one mimics Matt's original design but uses regulators in place of large capacitors. Ironically this is cheaper, uses less space and reduces feed-through of switching noise. SMD gives us the benefit of compact placement of lots of other components and it's more cost effective to order them pre-made from China than it is to buy the separate through-hole components and assemble it here.

Dancing with the Differential

The other point of note here is that in order to *fully* isolate a section of a circuit like this you have to remove the ground. (This isn't the same as ground lifting. Ground lifters remove the earth circuit that's there to stop people getting electrocuted and it's dangerous)

We isolate the audio section of the board from its digital section by regenerating the ground reference - this is done by a transformer inside the NMA0515s; and this is where a differential line comes in.

What seems like black magic is actually very simple. We know that a battery has two terminals so it's clear that we only need two wires to hook it up. Two wires is sufficient to form a circuit - keep that in mind and always remember "ground/0V" is always relative to something else.

The single-ended way of sending a signal works like this: we have some AC, perhaps riding on top of some DC. That takes one wire. The other wire (often a screen) carries the return current. Phono cables are line this. Terminated with the example of how NOT to do it, the RCA connector ***.

Not very exciting and "so what?" this is basic stuff. Use a sufficiently large capacitor to block the DC (the impedance of a capacitor is theoretically infinite at DC) and the AC just marches on through to the next stage.

What's missing piece here is explained by the fields I mentioned earlier. That's why it's better to think of electric fields rather than current. Use current and voltage on the schematic but fields when you have to apply that to a working entity.

You see those fields aren't static. Any energy in those wires causes a field and that field can impress itself onto other wiring in your circuit. (This is why we use screened cables for very small signals.) And the ground is more than a reference now - it's a lump of wire that's all over your PCB, flying around in the cable screens... it's everywhere.

And worse than that - it's carrying the returning fields from other parts of your circuit and that end up reflecting of stuff. Not to mention all the mains hum it picks up. Ground will only remove "hum" if it's actually "earthed" - earth and ground are used often interchangeably (guilty, your honour) but they are not the same thing.

This all matters because if a current is induced into ground, the effective reference isn't 0V any more but slightly more. This one is tough to wrap the old noggin around but it's like that little bit of peanut butter you can't spoon from the jar. It's not much, but it's there and it's annoying.

Now if you imagine that your entire circuit actually relies on the fact that (over time) the ground reference remains at 0V; but in reality, it actually wobbles around by a few millivolts. Now since your amplifier is pushing 100x or 1000x gain it suddenly becomes a huge problem. Because we're sharing a ground reference with (ultimately) the power supply in your PC, any power supply noise on the USB goes right through.

You can think of it as riding a bicycle over a ploughed field.

PSSR?

"Ah but...", you say waving a scolding finger, "what about PSRR!?"

OK as a beginner you might not. This means "Power Supply Rejection Ratio" and it's the ability of an amplifier or electronics to buffer small changes in the ground reference. (HINT: It's all done by really smart people with current mirrors, we just use it!)

In other words a well-designed IC with a good PSRR will absorb the worst of the wobble but not all of it. And not everything that shares that ground reference will be as well behaved as (say) a THAT1512.

Even a THAT1512 can't remove everything. What we need is a new piece of road. And that's what the isolation does; it does other things too but they are outside of this discussion. Isolation provides a clean ground reference for our well-designed circuit to use and because I've routed everything so carefully** and added loads of screening to steer those currents it all works flawlessly. See note below.

My "naked" Varee test module [I'm not in mein host's league when it comes to metalwork] is usable without any screening at all (and the right capsule) provided that it's fully isolated from power supply noise. Even the amazing Focusrite Solo still passes a very audible amount of mains hum through the grounds.

Right so we've isolated our ground reference (0V) - it's important to remember this is a reference that other voltages are measure with respect to. So how do we send our nice clean signal to the circuit with a "dirty" ground without infecting our nice clean ground? In a single-ended system, we're sunk because we have to connect share our clean 0V reference with.

But in our balanced system we have two wires - so (and this isn't exactly what happens) one of them can take the place of a ground. Remember that ground is only a concept - it only exists so that we can predict voltages in the rest of the circuit. What matters and all that matters is that there is a difference in voltage between the two wires. That's it. Our signal wires are their "own" ground. No matter what one signal line does, the other does the exact opposite.

In that way, any noise that has infected the circuit we're feeding can't infect our ground.

This technique is found all over the place - even the humble USB data connection and wired Ethernet use differential signalling to provide clean, high-speed data to connections with equipment that doesn't necessarily have the same idea of ground. It's quite possible to have a circuit with a ground (referenced to actual earth, recall the difference from earlier?) of thousands of volts connected to some other piece of equipment with a ground (reference to earth) of a few millivolts. But the signal, being it's own reference, is passed between the two at voltage levels as predicted by the designers.

But what about ...

If you picked anything from that discussion you'll have gleaned that differential signalling (differential pair, balanced line...) is a very important technique across all major electronic devices.

Using it internally is less common but it can be done and this is the other option I had with the Michelle layout.

Had I gone that route, all the ICs on the main board would be driven by an isolated +/- 15 volt split supply. The final signal would then driven through a unity gain phase splitter (inverter) generating a mirror image of the signal.

On the USB (5V) side that signal would either be recombined or more likely used in single ended mode. Either way it would then be referenced to the USB's ground and all is well. You can probably imagine that's could mean extra cost - although it's a toss-up which is better. I used the dual NMA0515 design to allow people to re-use parts and add stuff at their leisure rather than all at once.

For this reason I've included the option for a 200 ohm differential output to be added using a simple, 8-pin dual op amp. Naturally, a decent quality amp is needed to get the best from this and I'd suggest something like the OPA2134 from Texas. The NE5532, while a cracking IC, is just too much of a power hog to trust. The OPAx134s (the X is the number of amps in the package) chips have been around few years now so they're no longer premium price and the performance is superb.

* Varee is a Thai Buddhist name meaning "rain" or "shower".

A more flowery description from the web - I think it fits rather well, "serene and peaceful".

"Varee is a name of Thai origin, meaning "rain" or "shower." It is primarily used as a feminine name, though it can be unisex in some contexts. The name evokes imagery of refreshing rain and the life-giving power of nature. It is a relatively uncommon name, offering a unique and beautiful choice for parents seeking a name with a strong connection to nature.

Varee is perceived positively, evoking feelings of tranquillity, purity, and renewal. It is easy to write and pronounce, making it a practical and memorable choice.

While not widely used in popular culture, the name Varee holds a special significance in Thai culture, where rain is often associated with prosperity and abundance. It is a name that resonates with nature lovers and those seeking a name with a serene and peaceful quality."

** Because I never make mistakS, right? [Splutters, coughs, spits tea in British while trying to contain laughter].

*** The RCA plug and socket represent perhaps the pinnacle of bad design in the interconnect world. Why? Because the "live" terminal - the centre conductor - makes contact first (and several millimetres) before what passes for a ground connection mates. If there was ever an example of where the teacher should say to the class, "this right here is how not to do it" this is probably it. The pain this design has caused over the years is boundless. People can wire up "live" equipment thinking these are just audio cables but then miss the target by a hair and short the centre terminal (carrying the signal) to the ground. And since the ground of both devices will be connected via the house earth wire ...Cue some Brittany Spears music, and a 1- 2- 3-  "Ooops, I shorted again. I touched the earth and blew my amp in". You get the idea. It probably sounded better in my head. Perhaps it's time for my medication. 🤣 🤣 🤣 🤣 

@marcdraco Now I understand why my desk speakers were damaged two years ago when I connected them directly to my PC using an RCA to 3.5mm jack haha!

Also, thank you for taking the time to discuss how these devices work starting with just a bit of energy and how this outcome to the differences between dynamic and condenser microphones.

The reason I am considering using a dynamic microphone is that I already own two dynamic mics (the Samson Q2U and Fduce SL40X), and I want to use them for voice recording on my computer. However, the output sound is too noisy, even at lower volumes.

I understand that there are affordable inline preamps available, but for the sake of DIY and the satisfaction I get from soldering, I’m hoping to use the concept of Matt's preamp, combined with your upgraded version, to add gain to my dynamic mic with less noise.

I’ll try to attach my recording once I’m home, as I am currently at my internship.

If it’s possible for me to use Michelle, I am considering the through-hole version.

1. Just to clarify (sorry, I’m really new to this!), should I connect the GND/HOT/COLD of my dynamic mic to the 0/+/- input of Michelle?
2. Since I won’t be connecting the preamp to an interface (as a precaution for the P48 and i just only want additional gain with less noise), will there be any issues using the USB-C of capture card as the output audio of the preamp?

Thank you for taking the time to answer my questions! I’m really looking forward to you and Matt releasing your USB Mic V2 video soon!

Hey, no problem that's what we're here for.

The Samson is a seriously odd fish isn't it? I don't think I've ever seen a USB interface built onto a dynamic mic like that - the technique is common enough but electret capsules are so much cheaper, it doesn't seem to make much sense. But as I alluded to earlier, the greater mass of a dynamic moving coil dramatically reduces sensitivity for more subtle background sounds (which can be intrusive) and even some of the unwanted sibilance (hiss, lip smacks etc.).

Neither unit seems to specify the impedance at the output but it's reasonable to assume that if they follow the usual standards they should be fine.

The stock Michelle (it's supplied with almost all the minor parts, inductors, resistors, regulators what have you) should work. You would connect HOT (in phase) to "-" and COLD to "+" - this seems a bit back to front but it's because the layout worked out easier. It doesn't matter if you're using a single mic (or two mics on two boards) but if you are working as part of a larger set up, phase does matter.

The Audiograbber (or another USB digitiser) connects directly to the "Aux" or "Audio Out" of Michelle and that then drives the USB directly. There are a number of small digitisers and I think I must have tried all of them (most are the same parts sold by other people). Of those the Audiograbber has been consistently excellent.

The FDUCE looks like it has potential to be a donor body for people who want the "look" of a Rode or Sure podcasting mic but can't stretch that far.

I'll be happy when I can put this design in my rear-view. 😉

@marcdraco

I agree that the Samson Q2U does look a bit odd, but I must say, its USB interface delivers decent sound quality! It’s been a cost-effective solution for me, especially since I don’t have a dedicated audio interface at the moment. Thank you for explaining how to connect the hot and cold lines of the mic to the input pins of Michelle. I’m now just waiting for the resistor values to be used in the rotary switch, as it will also be mentioned in your build log.

I have a few questions that came up after reading your previous responses:

1. When you said, "Those SMD pads are shorted 1-2 or 2-3 according to the voltage required at the output, so for a dynamic mic leave those untouched," does this mean that I can't easily switch between dynamic and condenser mics without adjusting these SMD pads depending on the microphone I'm using?

2. Regarding "two mics on two boards," can I use two boards but output them through a single USB audio capture card? Or would I need separate capture cards due to power delivery concerns?

3. I know this was addressed in previous responses, but I’m still not entirely familiar with the schematics. If I want to add an XLR output in addition to the USB, can I directly connect the hot and cold pins of the XLR output to the audio out of Michelle?

Thank you again for your patience and for taking the time to answer my questions! I’m looking forward to starting this project soon and sharing the results with everyone 😊 

It's OK. If anything isn't clear, it's best to ASK rather than invite the "magic smoke genie" around for tea.

1. This design was specifically intended for "phantom" supplies and had to retain backward compatibility with Matt's to be able to save people money. Most of these parts are inexpensive, but it adds up. The +12 side for Varee should work with other P12 mics but I don't have any to try. The other side (assumed default because Varee needs extra parts) is compatible with Matt's original split supply design. There's still a niggle with my regulator implementation so I'm going to try breadboarding it again to get to the bottom of it. Naturally this means I need to do another update so I'll look at this issue for you.
2. Yes you can. Michelle is a "mono" (single channel) design so two boards will give you two "line" level outputs.  That's a maximum of 1V peak to peak. Some (not all) digitisers have two channels. The "audiograbber" Matt used has four but I've never tried all of them at the same time. It's quite possible there's a mixer in there but YMMV.
3. The schematics change across iterations - I should have perhaps made that a little more clear earlier. I just never imagined I'd have to go through so many prototypes learning my way around SMD manufacture and the oddities of cloned (same part number) chips. The uA- and LM- 741s are operational amplifiers from the very early days of discrete ICs. They follow exactly the same layout as an OPA1234 and can do many of the same things. Not as well, sure, but they are all the same part. The point is that the uA741 and LM741s (and many clones) have the same pin configuration. Not so with the TL431 and CJ431 they are both adjustable voltage regulators with a 2.5V reference and 2.5 - 36v regulation range. But the two of the three pins are reversed. A mistake that really has cost me a lot of time, money and confusion. Stupid error on my part but that's how we learn.

XLR out depends on what you mean by "XLR out". XLR is just a three-pin connector used as a professional interconnect. The dedicated XLR out on Michelle (an option that requires the extra NMA0515 and an OPA2134 or similar) has a 200R output impedance with a level determined by the THAT's gain.

This should work but I don't have a way to test it at present.

For instance, if you plugged the XLR out into another mixer (not one with 48V phantom, the blocking capacitors aren't rated for that!) and set the THAT for 0dB gain, the other pre-amp will "see" a 200 ohm signal from the mic - (condenser or dynamic) and at "mic" level for that type of mic. So this should (emphasis should) work on an Low-Z pre-amp input.

I've added a simple "phantom" switch to the design. It's a simple matter of hard wiring the two switch positions for anyone who is only using a powered (electret) head. One of those resistors is the wrong value in this screenshot though.

In case I forget, here are the resistor (1%, E96) calculations for the Quinn DIY instrumentation amp.

For this table I've shown the relationship between gain in decibels and the actual voltage gain delivered. Quinn should be considered experimental at this stage and is for advanced users since it requires soldering two 8-pin SOIC ICs which is challenging at first.

The reason we use decibels is because our hearing has a logarithmic response but working in decades (as we are here) might not be ideal. Most condenser mics require a gain of about x100 (40dB) and most low-Z dynamic mics require (around) 60dB for a good result.

For this reason the second table offers some alternatives bunched up around 30-60dB.

Alternative Gains

In reality though, these gains are only accurate to the accuracy of the resistor network (1%) and the gain resistor. The values supplied here are from the 1% set because the accuracy of the resistors used in the Quinn. Factory built, with standard parts, it's "only" 1% which is nowhere near as good as one might expect from an proprietary INA like the THAT1512 but the cost of a Quinn board plus parts is expected to be very attractive, esp. with if low-cost, low-noise op-amps are used.

It's not a great leap to include the gain resistor on the board when the best value is known for a particular microphone, so this is something I'll add to the new designs.

@marcdraco 

Sorry for the late reply. I’ve been caught up with my internship over the past few days. I really appreciate your response to my questions and infos about the gain resistors.

I’ve gone ahead and ordered a custom PCB for Michelle Through-Hole. Now, I’m just trying to figure out where to get the components. It’s been a bit tricky since most of the legitimate sites either don’t ship to the Philippines or have crazy high shipping fees, like five times the cost of the components, which just isn’t practical for me.

I’ll keep posted on how things go once I start working on it. Thanks again for all your help and patience! 😊 

@renzevan Don't worry about that, the through-hole one isn't feature complete by a long way. Where did you pick that up from? There was one kicking around but that was a very early design. I haven't published the one that suits dynamics yet as I'm waiting for the latest batch of boards.

I understand things are difficult in the Philippines and some other parts of the world so I will do a full, through hole design. Perhaps using just 5V as a low-budget variation. It's possible, if a little fiddly. Let me know which parts you're finding difficult to source - THATs are very tricky in some areas for instance. The boost converters are industry standard but I understand they can be difficult to get too.

We want as many people to benefit from this as possible. Varee (technically) should just about operate from 5V but I've never tried that as it means going well below the design voltage.

@marcdraco

Hello! I recently came across a PCB layout in some old comments here and decided to use it.

I realize now that it was my mistake to order it without confirming if it would be suitable for my project. But on the bright side, I have these for future DIY Condenser (if my budget will allow me haha) to keep me busy, and I got the boards at a fraction of the cost thanks to some vouchers from JLCPCB, haha.

I'm having trouble finding a specific part—the THAT preamp. As you might expect, it's quite hard to find. While there are a few listings for these ICs on our local online stores, I'm pretty sure they're all knockoffs. That's why I was planning to buy from Mouser or Element14, which both offer local delivery in our country. However, the shipping fees are just too high.

If I'm not mistaken, the "Michelle" PCB that seemed almost suitable for my dynamic mic is the one with Quinn and Alexandra on the PCB file, right?

Just nipping through (I got called out to an urgent electric cooker installation - and jiggered a pair of really expensive side cutters doing it.) DOOP!

Yeah, that's an old Michelle (May, 24) which was when I moved away from those large caps (which, strictly speaking, shouldn't be used). It's Matt's circuit original design so it should be fine.

The THAT1512 is a huge issue in many parts of the world.

Quinn will do the job but it's very difficult to construct and should really be left to people with SMD ovens as it's double-sided! Only lunatics like yours truly are crazy enough to deal with 0603 and (worse) 0402 parts. There are smaller ones but the 0402 needs a very steady hand. A little too much solder paste and they "tombstone" (this also happens in production, but worse for us). And I have to work under my old "dissection" microscope which I don't use for dissections any more. (Cackles evilly.) I did have a Crayfish to do but he's grow into a really neat pet. 😉

ANYWAY... can you give me some idea of which op amps you can get in your part of the world.

The OPA2134 is excellent but any of the Texas Instruments Burr-Brown amps are good. They are widely available in 8-pin DIL (duals).

We *can* pull this off with something more generic - the TL072 is very old and a bit noisy at higher gain which will likely spoil your day. My usual go-to in this case is the NE5532, also an old but a very good quality amplifier for audio.

I've knocked together a 5V - yes JUST 5V, single supply - version of the Quinn which looks like this (the power supplies aren't shown).

You should also note that biasing the amp in this way does result in a 50% rail bias on the output terminal.

This is because it uses the "spare" amp to push the ground reference pin to 2.5v (half-supply). This function is identical in the THAT1512 - shown as "REF". We *can* do the same trick with the THAT but not at 5V sadly as it needs a minium of 10V to work as the designers intended. However, after a nice chicken chow-mein I realised that we can do both (selectable), so I did. The jumper selects either external reference - in which case the amps work as a normal split rail OR internal bias (at which point pin 5 becomes a virtual ground for the rest of the board.

Anyway since 8-pin DILs are most likely what you'll be easily able to source AND solder, I've converted Quinn into through-hole. Let's call that one another Donna (one of my lovely friends). I'll tell her later.

Assuming you're able to get hold of an NMA0515 (or similar); several manufacturers produce them on a standard SIP (single in-line package) so you're not just stuck with Murata. They are also available in 5v and 12v versions - all with spilt rails and isolation. Even a 12V one would work. 5V *might* but I wouldn't bet more than a few cents on it working well and Varee would just turn her nose up in disgust.

Kinda funny story, I recently realised that I can use the same trick these devices employ (flyback transformer boost converter to generate 48V with enough current to drive Varee. Which is a right PITA. Now I won't be able to stop myself spending my grandkid's college fund upgrading Michelle to a fully isolated P48 system; and still runs from a USB port (sorry Matt, that will be another 18 months). I'm not developing the current Michelle now - just "bug" fixing. Unlike software though, a bug during development can be rather expensive. Ouch.

I'm gonna slip five of these in my current order if I can get them done in time. It should work, but I'm reticent to let people experiment with experimental hardware - if you do, it's at your own risk. But why dangle temptation?

@marcdraco 

I hope that either the joy and experience of working with that electric cooker or a good service payment will make up for your side cutters, HAHA!

The Burr-Brown op amps that I can assure are legitimate here in our country are the OPA2604 and OPA2134. Other op amps available here, which are a bit expensive but have been praised in local audio groups, are the Burson V5i and Muses 02. I must ask, if single or dual channel op amp will matter?

I'm finally starting to understand some of the boards you've worked on in Michelle version 1.4. Sorry it took me so long, haha. If I'm correct, does Quinn serves as an initial socket for an op amp before it's inserted into Michelle's op amp socket? I'm might learn Alexandra along the way, thanks to the detailed explanations you've provided, for which I'm really grateful!

I did manage to source a Murata NMA0515, and I've already ordered it (just in case people here start competing with me for the stock, hahaha).

Even if we have to wait for months, or even a year, for your and Matt's final creations, it won't be in vain. We all know the attention to detail you put into making these boards. The updates you provide are incredibly detailed, too. I'll be eagerly waiting for more updates, and I apologize for the additional work regarding dynamic mic compatibility, haha!

It happens, and thank you for your lovely reply. It's certainly a journey!

This is a long post. A very long post, so sit down, pour a large glass of your favourite beverage, roll your sleeves up and let's get crackling or not. We don't want any noise..

So yeah, Quinn (the SMD version) is intended as a replacement for the THAT151x or similar instrumentation amplifier.

You'll need two OPA2134s for the Quinn/Donna board. I did consider the OPA4134 - the quad version since it uses three amps of the four (Donna uses all four) but I've settled on the duals as they are generally easier to source.

You don't HAVE to use Burr-Brown audio, it's just it's a poorly kept secret of how amazing they actually are - and even then, they've been superseded by more recent designs.

The main point of Quinn/Donna is that you can drop in something "so-so" and everything will work. The NMA0515 might just about be able to cope with two NE5532s, although I've found that this one is a bit "suck it and see" since it's literally right at the extreme of what the NMA0515s can cope with and more than I've designed the power regulator for.

A newer, better version from TI is available the TLV9362 which is a drop-in replacement with much lower power requirements. The RC4850, also TI looks similarly useful.

You can use pretty much any low-power (this bit is crucial) dual op-amps in this circuit. Bipolar amps like the NE5532 tend to have lower noise but for low power during testing the TL072s are fine. They might be a little noisy for a dynamic mic however.

Assuming I got my math right this time and the wired the CJ431s as CJ431s and not TL431s (ugh) you have about 12mA to play with (per side, so 24mA in all) before the supply buckles. This is well inside the performance envelope for the boost converters but 5532s can eat that up in a single package without too much difficulty. Even if I tried it here and it works there's a better than 50% chance that it won't work for you, so better stick with devices that don't gobble power like a ravenous dog faced with a plate of finest beef. 😉

Strictly, this solution isn't as ideal as an IC designed for the job but they do come at a cost premium. Also (strictly) I could have done this with a single or configuration but not with a classic gain resistor and with much poorer performance.  High CMMR, "common mode rejection ratio"  is absolutely crucial in this application which is a fancy way of expressing the amount of unwanted signal rejection.

CMMR appears on the spec sheets in decibels where the larger the number is, the better the amp performs. CMMR is  dramatically affected by the input impedance which is why it's crucial to use the best resistors you can find in this part of the circuit. The three-amp solution benefits from very high input impedance leaving most of that work to the input filters. For the 5532 (in isolation) that varies from a poor 70 dB to a decent 100 dB which is far and away too much variation for reliable operation.

So I'll take a quick break from my day job to try and explain this weird sounding figure. I mean, what the hell is a common mode anyway? What's even a "mode" right?

The dictionary defines a mode as

• (Physics) Any of the distinct kinds or patterns of vibration of an oscillating system.

It's one of those words with multiple meanings but it originally derives from a Latin word meaning "to measure".

In our case our oscillating system is the sound coming from the pressure transducer (the microphone) and it enters the input stage in two distinct modes: phase and anti-phase.

If you're familiar with, say Star Trek, you'll have heard Mr Scott decry mixing matter and antimatter because they will annihilate one another and release a huge amount of energy. (Astrophysicists are still trying to work out why we're here using this very logic. When our universe was created out of nothing, if you can wrap your head around that) before even the first particles condensed out, there must have been an equal amounts of positive and negative (two modes) energy but someone "upstairs" got their math off a little and darn me, there's a tiny amount left over. That tiny (note: physicists have a very odd idea of tiny) amount of excess energy condensed into everything we see and a lot that we can't. You, me, our entire universe and everything in it came from the bit that was left over. Kinda makes your brain want to squeeze out of your ears and take a long holiday in the sun doesn't it?

Segues aside, it's fairly clear this is a very powerful system. Indeed if we just add together the two signals they will, in theory, cancel one another out. I should write this in terms of dy/dt (voltage over time) but that's just a fancy way of saying something like this:

2 + -2 = 0

Or in modes:

Mode 3 = Mode 1 + Mode 2.

Modes 1 and 2 are our two complex signals from the balanced line and Mode 3 is the final result.

But looking at this example, we end up with Mode 3 being zero. That's not good. I fact it's the complete opposite of what we want! We want to do this:

Mode 1 + (-)Mode 2 = Mode 3.

In other words we flip Mode 2 (doesn't matter which) on its head and mathematically that comes to:

2 + (-)-2 = 4

Now we're talking!

This assumes that both signals are exactly equal and here is where the clever bit comes in. Let's assume a real signal has a component we'll call noise - which may be quite significant. We could be looking at something like this.

Mode 1: +2 + Noise

Mode 2: -2 + Noise

And this is where the magic happens:

Mode 3 = (Mode 1 + Noise) + ((-) Mode 2 + (-) Noise).

Oh wait, that still looks like Greek, but it's all explained using waves.

A "fundamental" idea of sine waves is that a single, pure sine wave cannot be broken down any further. That's why a pure wave is called the fundamental. But so what?

So imagine if you will (in my warped mind) we have bucket and someone has filled it to the top with a mixture of fine plastic and steel dust. Although it might seem like one amorphous blob, we can separate out the two using nothing more than a magnet as a sort of selective filter.

In terms of a wave functions, we have two waves at different fundamental frequencies. The steel wave and the plastic wave. Or, if you prefer, the sound wave and the noise wave.

Now imagine that our bucket is filled with a huge amount of different types of plastic dust (noise) with a small amount of steel (signal) powder mixed in. You can imagine that a magnet works just as well here: separating out the steel from all that other rubbish. (Environmentalists and their recycling 🤣 .)

This is remarkable. It's almost as if we put a needle (signal) in a haystack and used the world's largest electromagnet to pull it out.

The steel, recall represents a pure sine wave and the mixed plastic dust represent a mess of other pure sine waves each one oscillating at a different frequency. (Quantum physicists believe that our entire universe is, at some level of magnification, nothing more than oscillating patterns of energy at different fundamental frequencies, which is likely what happens when you consume DMT with 18th C. French mathematicians. 😜 )

Forgive me that's a terrible joke, but If you wind further back up the thread you'll see charts of Fourier analysis I've done occasionally. Fourier (FFTs or Fast Fourier Transforms some very clever math) isolate the amount of energy of each fundamental frequency: which is the height of the wave (voltage in our case) and represent that as the height on the chart. Sine waves in fact. A fundamental concept in physics and across all of mathematics - and they're closely associated with circles, which is why good old irrational constant PI keeps popping up in our calculations. A pure sine wave:

Looks like this when analysed with an FFT.

This is the first A over middle C - 440Hz - called the Stuttgart Pitch ( https://en.wikipedia.org/wiki/A440_(pitch_standard)) generated in Audacity. I'll leave an Audacity file with the functions already done for you to try if there's room here.

You can fiddle with the "window" functions to isolate this peak more succinctly but it doesn't look very interesting.

Let's now call this our Mode 1 signal. Our Mode 2 signal is also at 440 Hz but in the opposing phase - Mode 1 goes up, Mode 2 goes down and vice versa. One is the mirror image - mathematically the negative of the other.

Now this is where we can start adding in some of those annoying magic plastic beans into the mix. Let's do that and see what our incoming signals look like now.

Emmm... right?

I promise, our 440 Hz fundamental tone is in there somewhere. What a mess!

But (and this is the crucial part) all noise is common (it's the same) on both signal lines and since the noise can be called a "Mode" (it's really an infinite mess of them) we call it the (drum-roll maestro) the common mode signal.

We control Mode 1 and Mode 2 because those have been created for us by the microphone (actually the FET) and imposed onto the two signal lines.

The noise gets picked up on the way and the job of the INA (THAT151x, Quinn, Donna) etc. is to separate out the common mode noise from everything else. The CMMR - common mode rejection ratio represents how well the amplifier separates out the Mode 1/2 signal from the Mode 3 noise signal.

It's literally just that.

If we know the amount of noise at the input and the amount of signal at the input (Modes 1 and 2) and then measure how much of the noise (Mode 3) remains we can calculate a ratio of the two. So if our pure sine waves go in at 1 volt and our noise (remember, that's a collection of infinitely small differences in frequency at all sorts of different energy levels) into something that calculates the DIFFERENCE between the two we get something like this.

Output Signal = 1 V

Output Noise = 1 uV

So:

1/0.0000001 = 1,000,000

The ratio of signal to noise here is therefore, a million - or, in dB), 20*log(1,000,000) = 120dB

Neat right? So if an INA advertises a CMMR of 120dB, it can separate out our noise (which includes an DC incidentally) from the signal and make the signal a million times larger. That's pretty cool and it is (by the magic of reading the data-sheet is, the predicted CMMR of a THAT1512 at 60dB (1000x) gain is between 105 and 120 dB.

By comparison the TL072 comes in at 95 dB (sure it does, TI, sure it does...) and the CMMR for the NE5532 is a paltry 70dB (worst case) which is around the 3,000 mark. Expressed in terms of signal that would deliver a signal of 1 V  for a noise contribution of around 300 uV (300 times WORSE).

Doing the math:

Div by 20: 70/20 = 3.5

Antilog: 10^3.5 = 3,200 (approximately)

Convert to volts: 1V / 3,200 = 0.000333 V

And trust me on this, your ears are far more sensitive than you might imagine. A few thousandths of a volt impressed upon 1 volt of signal doesn't sound much (no pun intended) but it will drive you insane - that noise might even include some energy at the fundamental and that's really going to ruin your day.

Most of the noise we're interested in preventing comes from the mains - that loud hum of 50-60 Hz that's created by the huge generators at the power station. About the only thing we can say about that is that it's stable. In fact we used to run mains-operated clocks from it using the ultra-stable, if rather noisy, fundamental to determine the timing. These days of course, it's all batteries and crystal oscillators with timing accuracy measured in terms of a few seconds a month for a cheap device.

But you can't beat a practical experiment. So fire up your copy of Audacity and let's hear this in action.

Uncompress this and you'll be able to load a file I've prepared earlier (British folk might remember this phrase from the kid's TV show, Blue Peter - a title in today's world is nowhere near as rude as too much Internet might suggest).

Make sure you can hear sounds from your PC but don't put the volume up too high before you play this the first time because you won't believe your ears.

For those following,  @renzevan is going to be listening to this:

 Hideous mess... which isn't quite so hideous when you magnify it:

Aha! There's our sine wave right there with a noise signal riding along on top. Note the two modes are exact mirror images - which you can see from the actual quantized data produced by the ADC.

The disturbance on our fundamental isn't massive but it's there - and if you listen to the sample it's easy to hear just how "dirty" this signal is.

I encourage everyone who ever wanted to know about differential line receivers to download and try this example, it's gonna blow your mind!

OK @renzevan, I want you to select one of the two "mode" tracks and then select INVERT

Play the clip again and you won't believe your ears!

Now that's a simple example because, hey, I'm only separating  a fundamental from some noise. I've hidden a sewing needle in some trash and pulled it out with a huge magnet. What about if there were loads of needles of different sizes and shapes (music)? Wouldn't that be more difficult? How about something a bit more meaty - say some rapid orchestral strings (which are rich in harmonics) - that'll sort the men from the boys (and it's one of the few bits of music I can get "copyright free". Credit for this clip goes to Gregor Quendel and it's an extract from Vivaldi's the Four Seasons*.

For reasons of space, example and my own sanity I've down-sampled the two modes as MP3 files. This is to illustrate where things can go a bit pear-shaped for reasons I'll explain in a moment.

Right click on each of these to download them. It sounds like a very, very badly tuned radio. You might just be able to make out the sound of the string section over the hideous roar of random "white" noise.

As before, load both samples into an Audacity project - be careful, you'll need to add each one as a MONO track and then import the sample into that track.

Again and as before, use the invert function to flip one of the modes and play the sample.

Not what you were expecting?

What's gone wrong here - why doesn't this produce music like I hinted at?

I know the music is present but it sounds all squeaky and kinda weird. (Distorted.)

It's better than the utter mess we started out with but it's far from pure.

You can even see the effect of combining the modes in Audacity using the "mix and render" function.

Spoiler

A Real (but faked) Example

In reality, we've done pretty well to separate the complex harmonics of strings from the hideous noise but it just doesn't sound "right" does it?

The reason is that I've compressed the two samples down into a FM radio quality MP3 file. MP3 compresses sound by discarding "high frequency" information which we can't hear. (We could hear it but our brains don't notice a little bit of difference.)

Now we know that MP3 introduces distortion because the Mode 1 and Mode 2 signals are no longer the same, despite being mixed with identical levels of crud.

But hang on Marc, we're not using MP3 here, this is a pure signal from a microphone isn't it?

Now this is all technically true but the problem is that we're not working with perfect components and this is particularly irritating when it comes to capacitors.

At DC, all we have to worry about are resistors but add capacitors into the mix and we're suddenly into reactance - which is resistance that varies with frequency. Rather than looking a the math which involves complex numbers and is hard unless you understand the symbols (which I  can't even type). So let's take a shallow dive into where this deviates from perfection and why component tolerances are going to ruin your beautiful design.

Here's a simplified version of the input stage with the protection diodes removed:

The first thing our signal hits are some small capacitors which have a very high impedance (at least 16K) at audio frequencies dropping to a few hundred ohms when we hit the low end of the RF spectrum. In other words, very high frequency noise is rejected before it even reaches our differential amplifier.

You get the reactance from the frequency and the capacitor using a simple formula but it's quicker to use an online calculator like Omnicalc: https://www.omnicalculator.com/physics/capacitive-reactance.

Moving along we have a series capacitor of 22 uF and here's when the wheels start to come off our calculations because we're only considering the impedance of a perfect component - and capacitors, they're far from perfect. Even good ones. Electrolytics are the absolute worst here and while one of the plastic dielectrics would be preferable, I've used ceramic (MLCCs) for their compact size. The great thing with Open Source hardware is you can change this stuff up at your leisure!

Ok, so a typical 22uF capacitor has can measure anything from 20 to 24uF and that throws our calculations off. Due to how this all fits together, even if you bought 0.01% resistors, if you use these capacitors... the network will still have an impedance that varies by as much 10%! Generally speaking this won't happen but it can and in ultra-sensitive equipment it really is an issue we have to consider.

You might be wondering what would happen if we removed the capacitors entirely? They're only blocking the DC right? And if the DC is common to both Modes (it is) then it gets zeroed out by the THAT1512 and we're laughing.

The answer to this is a shade more hairy. Looking at the data-sheet we can see the THAT can actually handle +/-13V so with a P12 setup we might (just) be able to do away with them for (potentially a) significant improvement in CMRR at low frequencies where the reactance is at its height. These capacitors are an absolute MUST if you find a way to power (say) a Varee capsule from a 48V DC supply and then try to feed that to this circuit. If the capacitors (rated at 16 V) don't go pop, you can be sure that your $30+ THAT will. Irreparably so.

OK so voltages aside, let's look at the actual problem using 22 uF caps with a tolerance of 10% (and tell those pricey 0.01% resistors to stop giggling at the back of the class).

Imagine a worst case of 24 uF on one leg and 20 uF on the other.

20 uF at 20 Hz = 398 ohms.

20 uF at 20 Hz = 332 ohms.

If everything else is perfect we have a disparity of input impedance of 66 ohms. Which doesn't sound all that much but any error here is going to be amplified up to 1000x! And it gets worse (groan) because everything has inductance and resistance and ... this is why I've tried to keep everything the same length when routing it.

Looking back up towards the input signal (not shown here) you might guess that's why it's vital to have these wires as close to each other as possible and the same length. We use low impedance because noise signals are generally quite feeble and they are swamped by the signal. The impedance of our microphone forms a voltage divider - and in effect a frequency dependant voltage divider at that.

This is compensated by the T-bias circuit as suggested by THAT Corp. which raises the input impedance considerably, thus swamping the negative effect of capacitive reactance. (*It may prove necessary to alter this circuit slightly due to the low impedance of the load resistors in P12 phantom power, but I'll look at that when I have at Michelle with a working power section.)

Due to the common-mode weirdness the impedance of this is 45 K with the values shown (2 x 22K + 1K) which appears in parallel with the "phantom" load resistors giving the 5k9 impedance per leg mentioned in the THAT data-sheet. With the 2k2 loads for the original that drops to 2 k per leg and a fairly measly 670 ohms for the smallest, P12 setup. Better than a kick in the head but not marvellous. Fortunately for us, that's mostly to balance out the difference in reactance at very low frequencies and things relax a bit as we get into audio frequencies of interest where the capacitors appear more like a small resistor.

In case you missed my bit of silliness about closely matched resistors, consider that the specified (factory-fitted) components are "only" 1% tolerance themselves and yet, this project sounds pretty good. There's always a trade-off between cost of components and circuit performance. It's why the specifications for modern ICs are a bit silly. THD of just 0.000001% the data-sheet trumpets, but if that's been driven by a microphone the distortion from the transducer alone (perhaps as high as 1%) swamps that.

And the THAT is no exception here - which is why you might (and I stress might) be able to produce a quality sound with a more competitively priced alternative like a Donna board with say, some LM358Bs which are way cheaper, assuming you can get them. The gain bandwidth isn't as good at 1.2 MHz and you have to account the greater the gain, the worse the frequency response.

I use a rule of thumb by dividing the GBW by the desired gain to find the frequency response.

So an LM358 with a gain of 100 (40 dB) is already starting to lose lose performance at 12 KHz, pushed to 60 dB... you might as well give up! Even the better specced OPA2134 (which is fine up to 40dB, rolling off at 80 KHz gives it gain to spare) but will suffer a serious bout of asthma at 60 dB with a 3dB point at just 8 KHz which. While outside of the speech frequencies is definitely going to ruin your day if you hook up a dynamic mic.

How's your brain doing? Mine is fairly cooked I'll admit - all of this math makes me want to crawl into a hole and hibernate until it's all gone away.

You might find yourself wondering why you'd even use a 3-amp differential input stage when chips like the THAT can give you oodles of lovely clean gain with a CMMR that makes the otherwise great NE5532 and later derivatives blush?

You know the answer of course. Cost and availability.

If I was designing specifically for a solution like this, I'd set the main three-amp stage to a gain of 1/10th of the intended gain and follow that with a x10 gain stage on the fourth amplifier. But in situation like this, you know ahead of time what you're setting out to do. Neither Quinn nor Donna offer this option but if you're up for a bit of DIY on Veroboard (and I'll warn you this is fiddly) you can wire up something like this:

Note this is a standard 3-amp layout (for split supplies) but the difference is there's a fourth (inverting) amp at the end with a 20db (x10) gain. This, effectively shoves the GBW of the assembly by a factor of 10 since all of the amplifiers are operating at much lower gain. The capacitor is additional here - it has a reactance of about 11 K at 20 KHz which is in parallel with the 10K feedback resistor effectively choking high frequencies and operating a single pole filter with a 3dB of 20 KHz.

We can't hear these frequencies of course, but they do suck energy from our power supplies so it's better to get shot of them at any opportunity.

I've included an extra resistor to compensate for offset bias in the bipolar transistors caused by base leakage current. This is calculated as the impedance off the gain setting resistors (10k and 1K in this example) in parallel. FET and BiFET amps like the OPA2134 don't need this resistor because they (almost) don't leak current and don't require a path to ground.

If this resistor is omitted the bias error (a few mV) gets amplified and ruins your day. At the very least you're likely going to need a DC blocking capacitor in this case and (another spoiler) op amps don't really like driving capacitive loads as I found out when I screwed up doing exactly that. (I might have skipped that class, but it's decades ago...)

Voltage Dividers as Filters

You mentioned the Alexandra filter board so let's take a quick look at filters.

Starting with a DC "filter" we can derive a lower voltage from a higher one just by using ohm's law V = R * I.

OK so let's put some values on those to see what's going on. To keep the math simple I'll just unrealistic values. Say RD1 and RD2 are both 1 ohm, the total resistance (impedance) in the circuit is therefore:

12V / (1 + 1) = 6 amps.

That means that 6 amps is flowing through each of the resistors so we can calculate the voltage across that resistor using the same law:

1 amp * 1 ohm = 6 volts!

Perhaps surprisingly, no matter what impedance you make the resistors, if they are the same, the voltage at the output will always be 6V.

Now if that sounds simple enough, watch out for flying spanners, like this one.

Keeping it simple let's see what happens when we connect our simple voltage divider to another circuit (called a load). R_load here is the "effective" impedance of the circuit you've connected to. If R_load is also 1 ohm, lets see what happens.

RD1 is unaffected.

RD2 however is now in parallel with R_Load. Identical resistors in parallel always work out as divide by two - so that's 0.5 ohms in the lower part of the divider.

The divider now draws 8 amps from the power source and eight amps through 0.5 ohms is just four volts. The load has (in effect)) dragged our voltage divider down by some 33%.

In the real world (at least in preamps) we're dealing with much smaller voltages and currents but the principals still apply (most designers will strive to keep the impedance of the load at 10x that of the driver. Let's see how that works out.

RD2 is now in parallel with 10 ohms so that works out at about 0.91 ohms for the lower leg, the circuit still loads the power supply by around 6.3 amps but the load voltage is determined by the current flow through that parallel combination and that works out:

12 / 1+ 0.91  = 6.3 amperes.

6.3 * 0.91 = 5.6 volts

Much better.

As an aside this is why you can't really use resistive voltage dividers to split the power supplies for operational amplifiers. A notable exception to this seems to be the excellent OPA-Alice which uses just this system to generate a ground reference at half-supply.

Right, so back to filters.

The simplest type of filter (and the one I'll look at here) is a voltage divider formed with a resistor and a capacitor. Simply take one of the resistors from the circuit show, replace it with a capacitor and you magically have a frequency dependent voltage divider.

And just like that we have a high-pass filter. It's easy to spot a high pass filter (although they are more usually drawn like this):

Why? Because the first element is a capacitor and capacitors block DC. Flip the resistor/capacitor around and (since resistors pass DC) you have a low pass filter. Higher frequencies are passed through the capacitor to ground.

Putting some numbers in to this for fun, let's design a high-pass filter with a -3 dB point of 15 Hz. At these low frequencies we need a fairly large input capacitor, so let's try 10 uF. I suggest using Omicalculator

The resistor works out at 1K for a predicted -3 dB point of 15.9 Hz which is quite sufficient.

Unfortunately it's an imperfect solution as the rate at which the unwanted frequencies are removed is only -6 dB per octave. It's not really practical to make a brick wall filter (millions of words have been written about this) but we can improve it by adding more stages.

But there's a gotcha here that will ruin your day. Recall how a load affects the circuit it's connected to? Same thing here. The (frequency dependant network) load formed by the second stage affects the impedance of the lower leg so it doesn't work as you might expect.

Here's our filter graphed (in LTSpice). We can see how the 3 dB point meets at our design guide of 15Hz and the filter is effectively a wire above about 200Hz. It rolls of at 6 db per octave OR at 20 dB per decade. Recall that an octave is eight steps with a doubling (or halving) in frequency. The frequency of the A one octave higher than the Stuttgart Pitch is 880 Hz. You'll have to ask a skilled musician why they thought using octal was a good idea. Maybe they counted their fingers and didn't include the thumbs? And, on a piano at least that's further divided into sharps and flats giving us 12 tones. But that's only Western music and I think my brain is melting. So here's the chart.

Let's load it now without allowing for the load impedance of the next stage. The -3 dB point is now raised to 35 Hz and the graph is staring to move upwards in frequency. Here's the same passive filter with three stages modelled (6 poles).

Sure we have a much steeper drop off in unwanted frequencies but we've also pushed the calculated cut-off from 15 Hz all the way to >80 Hz.

Not very useful - which is why you don't usually see multi-pole RC filters because it's easier and less stress to use an Op Amp - which is, of course, what Alexandra does - with predictable results.

More of that another day.

@marcdraco 

Looks like I'm going to need some snacks to get through all this info! I might start by asking some basic questions and come back to the later discussions afterward, as I tend to get a bit overwhelmed with all these details. But I’m really enjoying the learning process and challenging myself!

I think I’ll stick with the Burr-Brown op-amps for now since I find it hard to tell if other op-amps in our local stores are genuine or fake. A lot of the five-star reviews seem to focus on the low prices rather than the quality, whereas the original Burr-Browns are easier to spot based on their target market, reviews, and price points.

Just a heads up, the only op-amps I can find here are in DIP-8 packages. I’ve been searching for days and can’t seem to find any SOP-8 OPA op-amps, which is different from what Quinn is using. I’d love to try modifying the PCB, but I know it will take some time for me to learn the basics of KiCad. (Not to brag, but as a civil engineering grad, this is way outside what I studied—haha!) So, I’ll be patiently waiting and hoping a through-hole design gets released.

The NE5532s are quite popular here, but like you mentioned, they do consume a lot of power, which I’m not a big fan of either.

I’ll also take some time to learn more about CMMR since I only have a bit of free time tonight because of my internship hours. It’s all pretty new to me but really interesting! From what I understand so far, it’s about a device’s ability to reject noise or interference, especially in audio signals. This is definitely going to be a deep dive for someone like me who just started out wanting to make a device and now I’m learning so much more about audio! Thanks again for your help and all the information!

You're welcome, your questions and intelligent and insightful. Many more people will benefit from them and I don't mind doing a deep dive into this. The new version with Matt's new designs is going to blow everyone away (no pressure then @diyperks ;)) but I hope will bring more people to this hobby.

Practical Stuff

The "Donna" board has completed production in China but I have to wait until the others complete and ship. Maybe by the start of next week? I'm tied up with other stuff but I can give "her" a quick test without too much trouble. Funny thing was I did consider the OPA4134 (quad, 14 pin) which is better in terms of space and cheaper, but the availability is spotty. So OPA2134 or similar, low-power FET, BiFet or perhaps bipolar op amps will suffice. You can always test it with some TL072s which are peanuts to make sure it works before you drop the real ones in there.

It's not ideal and yeah, the SMD versions are hard to get in many parts of the world so I went with the 8-pin DIL which is the worst in terms of space and cost, but the easiest to get hold for more people and this is what we want. To make this thing more accessible.

Ranty, ranty...

What I really detest are these clever folks who go to Wikipedia pages on electronics (or math) and then launch into advanced calculus. It's supposed to be accessible and yet it reads more like a patent document. From what I've seen they appear more intent on looking smart than actually helping people work through these problems. Not everyone is blessed with a natural ability to do calculus, hell some of us have difficult with a pocket calculator. But should that be a bar to enjoying making your own stuff?

It's impossible to dodge a little bit of math when showing how to design even simple filters, but I don't think everyone needs to know about poles and zeros for example. I can show you a pole or a zero on a chart and I can explain WHAT they will do to your audio and the pictures really do convey thousands of words.

I'm more in the mould of Horowitz and Hill (The Art of Electronics) because while electronics has a lot of math heavy theory, a lot of it is fairly intuitive when you can see the worked examples and hear/see for yourself.

In this entire thread I think we've only really discussed Ohm's Law (which is a magic triangle) and reactance of capacitors. I've discussed moving charge in relation to condenser mics too - but without needing to dip into Coulombs and how many charge carriers are on a given area of plate. For me, it's better to understand why capacitors exist and how they work (by opposing, therefore attractive fields) than the usual explanations which never made much sense to me.

Likewise grounding is often though of as black magic (some just rely on the "ground Mecca" and crossed fingers) but when you understand this concept of fields - and that they travel through the insulators - it's much easier to conceptualise them moving around your project. It's not immediately obvious but once you see it in action it's quite beautiful.

I recall first being taught of "reflections" - where electrons would "bounce" or reflect from the open end of a cable and travel backward - even reflecting at a steep bend in the cable. Worse, in some cases. I should say this affects higher frequency information well above the audio range but it's worthwhile knowing.

How does an electron bounce of something? It doesn't.. at least not in the way that we imagine it hits the air at the cut end and bounces off like a mirror! As I understand it, the it's the FIELD that's moving it does that as it "looks for" a easy path; so the two are intimately tied together. The real answer is (as I think I mentioned) electrons aren't little particles moving through the metal, they are oscillating spheres of energy - wave functions in their own right. The fields change the state of the electron (it's wave function). Quantum physics is way too weird for me. Hell who am I kidding, I'm still struggling with pocket calculators.

Sorry, I segued into a rant. I'm happy to do the "hard" work (it's not hard, just long and expensive if I screw up) and create an intermediate "almost ready to fly" board/s so more people can enjoy Matt's work. Veroboard is very twitchy stuff - PCBs are far stronger and more reliable. SMD components are often an improvement to electrical characteristics despite being more compact and lower in cost. The downside is they are hard to work with hand soldering. I've done 402 parts with a microscope and a TS80 but it's not fun. Easier to put a ship in a bottle. 0603 parts are doable (just) and although larger ones do exist, they are harder to source and many parts aren't available in the larger packages.

I think I have the postal addresses for the beta testers so if you're one of them, you qualify for a Varee adaptor at no cost - but you'll have to drop me a PM so I know you're still on the project.

Hey, I'm planning on ordering the Varee 1.51 PCB through JLC PCB, but I'm not sure which 2SK208 to select. They have various ones that seem to match, but I don't want to resolder these SMD components later! Does anyone know which 2SK208 to select or the specifications it should match in the datasheet? I could also order the part from else where, but I still wouldn't know which one to get.

Here's the list of options:

 Thanks in advance guys!

JLC part number: C112988

The board is designed for the SC59 pinout and I completely agree - better to have all that tricky stuff done at the factory.

Any similar "condenser mic" JFET will drop into this board but the key to making Phantom power work with the serious restrictions it imposes on us, is to get the lowest current one we can. The circuit is a voltage amplifier/buffer so large currents aren't necessary and are actually a nuisance.

The difference between a JFET used for a mic and one for, say, RF oscillators is in the parameter. The voltage change at the microphone input isn't very great (perhaps a tenth of a volt at most) so ideally we want small range of Vgs (off). But JFETs are laughably bad in this regard because of manufacturing spread and seem to be stuck that way. Perhaps because the majority of development cash has been ploughed into other tech like power MOS, IGBTs and other transistors that are suitable for a wider range of tasks.

The difference between the F and G versions seems to be the VDS (30 V vs. 50 V) but we're only operating at 10 V so that's not really an issue.

Might I ask which version you're getting?

The artwork for P48 suggests an Ri of 3k3 to 6K8 (from simulation) but in practice a smaller current (680R to 1K works better). 10R is specified for the P12 Michelle boards at the moment but I'm waiting for the revised boards to get back before I can verify that.

Ri (R14 on the schematic) is an 0805 although it's not terribly difficult to re-route for a larger size if you think you might want to tinker and don't want to be fooling around with a microscope and a hot-air gun. A 1210 size will fit and but larger parts will require a little more work. That's the wonderful thing about Open Source - you can make it how YOU want it.

@marcdraco 

I just got home, and the first thing I was excited to check out was the discussion in this thread, even before resting after a two-hour commute haha!

I wanted to ask if Donna is readily available here in the thread? If so, I’ll just scroll up to check and download it. Also, just to confirm, does Donna use the DIP8 package?

I totally get what you mean about things being overcomplicated. It’s not that I’m ungrateful for not understanding it; it just makes me a bit sad thinking about those who might struggle with unfamiliar concepts, even if they try their best.

I do have a bit of background in resistance, capacitance, and the like from my senior year physics, but that was about six years ago, so I’m definitely a bit rusty—haha! But these refreshers are really helping me recall what I learned.

If the early batch PCBs are only available with SMD components, I won’t mind at all since I can order SMT services from JLCPCB. I’m just trying to see if I can solder things myself for fun—haha!

I’d really love to try Varee! I’ll message you to see if it’s something that could work out for me. Thanks again for all your help!

Afternoon my friend, "Donna" is with DPD right now - on the way to me from JLC. While I expect it will work, I would really rather test it than make a mistake that costs everyone. I really don't fancy getting chased around the Internet because I published a non-working design! It's really a case of the removing the last few gremlins (or cat-related cockups).

I designed it with as many people in mind as possible - so it is dual DIP8s - the least efficient use of PCB space but by far the easiest one to get parts for and build so it's all through-hole components.

I've found that sometimes things work on the schematic and in the simulator but real parts will throw you - hence I've been very conservative in designing the power on Michelle. However, the OPAs will work down to 5V total so even if the supply dips a bit, it should still work and will work at lower voltages than the THAT151x and other INAs require.

The production Varee (apart from having a minor error on the silkscreen) has two paralleled PNP transistors - one of which should really be removed (more of that another time) and it has a 0805 voltage dumping resistor which determines how much load the system can excerpt on the phantom power (P12 from Michelle or P24/P48 from a professional deck).

Neither of these are difficult to remove - re-soldering the 0805 resistor to swap out for a better value is more challenging. So that's some experience for you. I strongly suggest you source a small amount of solder paste.

I did some early versions of Michelle as a through hole board, but SMD gives us a lot more options and they perform better due to the much lower lead inductance/capacitance. It's also a lot larger. However, once I get this thing working to my specification I'll route up a new board with 1/4W 1% resistors and better capacitors. Skilled builders can mix and match - and to a large degree, KiCAD's Design Rules Checker has an option (which should be turned on at all times!) that will ensure that the routing matches the schematic. While that's not 100% guarantee it will work (as I discovered to my great shame) it most certainly better than crossing my fingers and hoping the darn thing works.

The alternative is to stay with SMD but use much larger components - some of which are relatively easy to handle. (No one in their right mind should try soldering a 0402 (imperial measurements) part with a soldering iron! It's 1mm long an 0.5mm wide. I know, I've tried and it looks like a train wreck in there. Even 0603 (1.5 mm x 1 mm) can stick to the end of the soldering iron - held in place by the surface tension of molten solder. Here's a pic for scale. Not my finger but it's close.

Solder Paste

It's worth noting here what solder paste is an how it works.

This stuff is like that scene in T2 where the T1000 (liquid metal) puts itself back together in the climatic scenes. But HOW?

Solder paste is actually a colloid with microscopic balls of tin/lead (Sn/Pb) mixed up with a liquid flux.

When it's heated to the correct temperature the stuff floats on the solder resist and flows together in little rivers which are attracted by surface tension under the component leads. This is why you'll see some appalling work on YouTube where some repair guys lash it on with gay abandon. You don't need very much at all and while you don't need to only hit the pads, it's far better to do so.

A fine pointed soldering iron can be rested near the pad where the component is placed and the heat will melt the solder allowing it to flow. You'll note I said Sn/Pb which isn't great for the environment (due to the lead) but it melts at a much lower temperature and it flows better too.

I'll PM you.

@marcdraco Thanks! It's been awhile since I took my electronics course, so I appreciate the breakdown! I somehow always forget what to look for in datasheet specifications. I make sure to order the 2SK208-Y(TE85L,F).

As for the Varee version, I'm using the Varee 1.511 from your Neweer BM-500 upgrade post (the one with "For Garrie" on it), but R14 is listed as 10 ohms. Given what you mentioned, I might change it to somewhere in the range of 3k3 to 6k8 for P48, but I'll double-check for an affordable and compatible audio interface later. I want to try to both BM500 chassis and the DIY Perks Art Deco style to see which I prefer. When the Michelle board is verified by the experts, I'll be tinkering with that configuration!

Quick question, could I also swap out Ri/R14 to make it USB-C? And can I still use the noise suppression circuit from USB-C mic video from DIY Perks? I'm thinking of gifting a USB-C version mic to my friend.

I'm finding (since I put that on the back of the silkscreen) that it works better with a 1K resistor. I think the LED (not required for the BM800 upgrade) is hogging more current than I expected. I'm backed up the wall right now so I can't do a deep dive into what's gone a bit off the wall. The FET current is limited by the clever circuit based on one appearing in the Art of Electronics 3rd edition - so the extra drag is something I need to investigate properly. We're really getting in to the weeds here but I'm a perfectionist where possible.

It does work with P48 with a 10R in that place but it's really putting too much of a unnecessary strain on the supply.

BUT - and this is important - it still needs a pre-amp like a Focusrite Solo. This is the poor-man's way of getting a cracking microphone setup in the shortest amount of a time as the 2nd Edition Solos are quite reasonably priced now.

The P12 version (with the 10R) is intended for the Michelle board. It won't work with Matt's original because it's designed for a single-voltage supply. I expect you could chop the schematic a bit. I'd be tempted to drop the 2K2 resistors to 680R which is standard for P12 and ensuring each one is connected to the POSITIVE 15V rail.

Sounds a bit nuts - to get a AC signal out with a single-rail supply but it's actually surprisingly easy using the genius front end dreamt up by Jorgen Wuttke. (*I hope I spelled that right).

@marcdraco

Michelle Through-Hole has finally arrived at my house! I’m so excited to try it out (even though it will only work strictly with Matt’s capsule design—haha) because I can finally use my soldering tools again.

The only issue I’m running into is with the 22µF capacitors. As you mentioned on the back of the board, the ideal capacitor is a 22µF 100V MLCC, but all I can find in our local stores are either 22µF 100V electrolytic capacitors or ceramic capacitors with a maximum voltage rating of 50V. Based on your insights, which one would you recommend I use?

I’m also really excited to see Donna in its full form and can’t wait to try it out!

Swapping SMD resistors will definitely be a challenge for me. I do have a hot air station, but I’ve never worked with SMDs before. I’ll try to source some good solder pastes first since I’m only familiar with good soldering lead brands.

For Matt's through-hole design, you can use any non-polarised caps. Actually, polarised ones are fine provided you put them the right way around. 😉

You can get leaded MLCCs but good quality electrolytic caps will do the job just nicely. The two input capacitors have their respective plates connected to the microphone input side. Remember that Matt's design uses a split rail supply to the FET. The positive plate of the capacitor on the positive input goes to the mic. The negative plate of the other capacitor goes to the negative rail to the mic. If that makes sense - it's probably why Matt went with bipolar capacitors.

It's questionable if we need a capacitor at the output at all (this board is really just a PCB version of the original). Reason being the THAT's output is referenced to (and swings around) 0 volts. The  DC block is really "belt and braces" as the actual DC offset is in the order of millivolts. I've left them in for completeness but they can be omitted. Some might say should...

Matt's board only runs at 15V per rail so 16V is a bit tight - so I'd pick a 25V one. 

The newer Michelle boards deviate significantly from that earlier design (there's no negative supply out to the mic for example). I clung onto as much as I could but in the end I had to swallow hard and make some serious changes.

As you know (others don't) I'm also researching a completely different  method that's rarely seen in the DIY space - mostly because it's quite challenging to design. And I need something to distract me from the day to day grind, right? 🙂

So I soldered everything as shown in the tutorial and when I plug the mic into my pc it recognizes it but there is no input. The only thing that is different is that I used a different rotary switch than in the tutorial and that I didnt use a usb-c break out board and instead I supply the circuit directly from the USB audio interface and use its USB.

So this is my circuit - and that is the rotary switch that im using (is it soldered right?)

and this is how i supply my circuit - is this right?

So could anyone please help me and point out whats wrong or what to do? Btw I also have a multimeter but I dont know how and what to measure to find out what isnt working.

First of all, welcome to our part of the web.

Good choice of meter - if that's your first anyway. Should deliver years of service if looked after.

Let's start with the (potentially) bad news - and don't feel too bad, I did this myself while testing the Michelle (V2) of this project which is nearing end of development.

Uncontrolled, the output from the THAT1512 is capable of destroying the digitiser in a less than a second. Wired properly  and used conservatively it won't but a sight slip and pow, just the hum from the uncovered capsule is enough to blow the input stage which is likely only safe to 5V at best (the USB supply rail).

I did post an addendum a while back which explains this in more detail. The answer is to wire a pair of diodes in "antiparallel" across the output to ground. Antiparallel just means back to back so that signals in both directions are effected. Any voltage which exceeds the turn on voltage from the device is immediately shorted to ground (harmless to the 1512 although it's better to have a resistor there too) like this:

Finding the Fault

So the first thing to do is remove the Audiograbber (digitiser) from your circuit and test it in isolation. No point going any further until we establish that is still in good order.

Carefully de-solder everything and use Audacity (or whatever you're using to record with) and see if the device is still working. You can touch the input with your pinkie and you should be able to see the noise pickup quite easily.

Don't be put off if that has bitten the dust, I know they're not cheap but this is the hard side of learning.

Let us know how you get on with that and we'll take it from there.

@marcdraco so sorry for answering this late i didnt know you answer so fast 😀
So the digitiser is the component that component that I plug into my pc right?
And how exactly am I supposed to test if my component is bricked? Should i use my multimeter or do I litteraly just plug in the digitiser and touch one of the ends?