[Sticky] USB-C Microphone (official topic)

Flux got some weird properties 

working on the build first then the circuitry 

tried multiple times to stick these damn hooks on

Matt makes solder look like a magic alloy. My brass shell seems to be solder-phobic I've tried wicking and heating both parts first but they just won't be hot enough to melt the solder.

bought some cheap helping hands gonna try again later

A lot of modern solder is lead free and that stuff just doesn't flow like the old (and highly toxic) stuff. That's likely the problem. You'll need to source some SnPb stuff (60:40) as I recall. It's still available but you have to ask for it. I've found I need to raise the temperature of the new stuff which is tin (Sn) plus some other things like copper to 400 C on my iron to get the stuff to flow. It's also extremely brittle.

I should also point out that lead-based solder is highly toxic (due to the lead content). People from my era have this hanging over our heads but there is no reason anyone else should, so you must use a fume extractor when working with the stuff for your own safety and the safety of those around you.

Hello everyone, I have ordered THAT1512 from two different sellers in Aliexpress, both are the same and fakes. With them, I have the same issue and the same measurements as in this post https://forum.diyperks.com/postid/4484/

Spoiler

Orders

Spoiler

Fakes

So I have ordered a genuine one from Farnell, which looks and measures differently.

FYI:

Spoiler

Genuine

Comparison:

Spoiler

First comparison. Diode test mode in multimeter

Black probe on 6 pin, red probe on 3 pin

Genuine:

Fake:

Spoiler

Second comparison. Diode test mode in multimeter

Black probe on 8 pin, red probe on 4 pin

 Genuine:

Fake:

Hate it when that happens, I really need to make an "Octopus" but that will take a little longer. The usual way is to do this with a traditional oscilloscope, the digital ones don't seem to work as well due their improved performance but what we need (those of us who do this) is a reliable and portable one. 

The difference is that an Octopus puts a varying signal through the component under test and gives very specific curves (they are also called "curve testers" if you buy a proper one) but they make this sort of debugging a breeze. 

So, guys, I have a very serious question for you.
How can all of this be converted to a 3.5mm jack connection?
I'll say right away that I don't understand any of it; at most, I know how to desolder a capacitor and solder a new one in its place.
P.S. I have a Creative AE-5 sound card, which is why I want to make it with a jack.

Posted by: @f4lc0nsv

↑

So, guys, I have a very serious question for you.
How can all of this be converted to a 3.5mm jack connection?
I'll say right away that I don't understand any of it; at most, I know how to desolder a capacitor and solder a new one in its place.
P.S. I have a Creative AE-5 sound card, which is why I want to make it with a jack.

Welcome to our little corner.

Don't worry about your skills, everyone learns over time. My own contributions contain errors (which I'll leave here with the corrections) as I fine tune the prototypes from utter failure to pretty good. 🙂 There's a cracking example of one of my foul ups at the end of this post.

OK, so if you want to take the existing circuit and connect it to your sound card you can solder a mono (or stereo) 3.5 mm jack directly to the output from the THAT1512 (that is after the 22uF output capacitor). I've done some PCBs, current ones just on the way back, that have an auxiliary port for this precise purpose.

The key point though is this goes to the Line In port, not the "Mic in" because it's already been pre-amplified.

On Matt's original stripboard layout you just connect it to Line at the point where Matt attached the digitiser. 

Technically, you don't need the stripboard pre-amp as the AE-5 has its own pre-amp and digitiser. The down side is that distortion is a little higher than with these designs.

A typical (TRS = Tip, Ring, Sleeve) mic socket (according to the standard) should provide 5V bias and a mic input. That's it.

Tip: Signal In

Ring: Bias voltage

Sleeve: Ground

With Matt's original head inside the blue line, you wire everything as shown (with the JFET) but then wire it like this:

It doesn't get any simpler and while this is "cheating" it's the way all normal headsets are wired. Typical electret capsules have an internal JFET so it's simply a matter of adding these.

So why the extra complexity?

It's about signal quality. The "differential" design Matt uses is far better at keeping out strain mains hum; and that's not all but beyond that you need some knowledge of JFET amplifiers and I don't want to overload anyone with information.

This circuit works like this:

Current flows from the "bias" pin through a the 2k2 resistor (the value isn't critical and affects the gain but the mic input will be expecting around that sort of value - values around 1K to 5K works but is actually dependent on FETs which are by nature, not that predictable). 

Put (very) simply what you have here is a voltage divider formed from the resistor and the JFET. When the microphone is in operation the tiny movements cause a variable voltage to appear at the gate terminal which, in effect, changes the resistance in the channel.

Thanks to online simulation you can see this in action using a typical microphone JFET.

https://tinyurl.com/26dy3l66

But JFETs are weird, here's something like the classic 2N3819

https://tinyurl.com/285jfs76

See how with the same values, we can go from a fairly chunky signal with the LSK170 to just 1/10th of that with a 2N3819. Horses for courses.

We can improve the distortion quite dramatically by adding some "negative feedback" by limiting the current at the "source" terminal (the lower one in the drawing). 

https://tinyurl.com/28vbft2o

This reduces output (sensitivity) by about 50% but reduces distortion (THD) from 2% in the classic wiring to around 0.2%. Not bad for a single resistor! On the diagram above you just put the second resistor where FET2 is connected to ground.

I should note for completeness this method is used in Matt's version and in the classic Schoeps design. My own designs go further but that's beyond the scope of this discussion and I'll leave them until I've verified that everything works as it should.

I recently found an "game-over" error that I've carried through multiple prototypes that has driven me utterly round the bend, all for the sake of a missing decimal point! A case of not being able to see the wood for the trees has cost me ... a lot. 🙂 But sometimes the best lessons are the hard ones. There's a gotcha 3% THD error on the original too but I'm not going to discuss that until I've produced and tested the full V2 designs, so don't tell anyone. 

I blame it on my old brain (and the simple bone-idleness of not running the whole thing in the TINA-TI simulator. (I use several but I prefer one called LTSpice as it's much faster to draft and test a circuit: but LTSpice lacked a critical component so (to my shame) I bodged it by just putting a voltage source in and assumed... and you know what assume does? Made an ass out of me anyway. 

As a result I'll be eating out of the dog's bowl for the rest of the year. Sigh. The dog is not impressed.

Fire away if anything is unclear.

@marcdraco Thank you for that)
I will continue to try to do something. But first need to wait for all the boxes from amazon)

Here's a tiny section from the new KiCAD 8.0 branch's Spice simulation.

and this one signal, this one little flag could have saved me a fortune in prototypes and a lot of head scratching over the last year.

In fact, this level of simulation has never been available in KiCAD previously but what's the deal? What's half a volt got to do with anything?

If you're wondering it's because that's supposed to be 2v5 (2.48 volts to be more accurate, but hey, it's close enough).

The TL431 is a fairly old chip that replaces Zenner regulators in most application with the addition of just two extra resistors to set a regulation voltage (Zenner equivalent) of anything from 2.5 to 36 volts with far better performance than an actual Zenner diode.

It's also something that I've been unable to simulate in KiCAD until recently, despite KiCAD having Spice integration for several years now. I've always used LTSPice to check my working but like KiCAD it lacked a TL431 model so I (rather foolishly 😧) used a workaround. And assumed that the calculations I'd done by hand were OK.

Well you know what assume did...

I fact, what I'd done is missed a decimal point out and make an upstream resistance an order of magnitude larger than it was supposed to be - meaning that happened rather than this:

The voltage at the 431's cathode is determined by the voltage at the reference pin which should be 2v5 derived from design output voltage passed through the divider. The 431 uses this reference to pull up to 100 mA of current from the supply, causing a voltage drop across a dropper resistor.

So if you have (say) 30 volts in and you need a 10 v supply, you'd arrange for 20 v to be "dropped" across that resistor. The value of that resistor isn't too critical as the device adjusts the current until the correct drop is achieved. If the dropper resistor is too large, then the 431 reaches a limit (it can't generate negative impedance) and the supply voltage drops too. This results in anomalous behaviour as seen in the first image.

I'll include this in the final Varee (capsule adaptor PCB) source so anyone can play with it but note that you'll have to source the TL431 macro yourself as I'm using one from Texas which I can't redistribute. This is an issue across much of KiCAD's simulation but the new version includes a basic Op Amp too so far more circuits can be simulated and tested before they are sent for production.

This is going to save a lot of developers (pros and hacks alike) a lot of money.

Posted by: @yrambler2001

↑

Hello everyone, I have ordered THAT1512 from two different sellers in Aliexpress, both are the same and fakes. With them, I have the same issue and the same measurements as in this post https://forum.diyperks.com/postid/4484/

Spoiler

Orders

-- attachment is not available --

Spoiler

Fakes

-- attachment is not available --
-- attachment is not available --
-- attachment is not available --

So I have ordered a genuine one from Farnell, which looks and measures differently.

FYI:

Spoiler

Genuine

-- attachment is not available --
-- attachment is not available --

Comparison:

Spoiler

First comparison. Diode test mode in multimeter

Black probe on 6 pin, red probe on 3 pin

Genuine:

-- attachment is not available --

Fake:

-- attachment is not available --

Spoiler

Second comparison. Diode test mode in multimeter

Black probe on 8 pin, red probe on 4 pin

 Genuine:

-- attachment is not available --

Fake:

-- attachment is not available --

hi. so is it just a different pinout or is it not working at all? This is a project I probably won't complete because there are a few parts I can't get genuine.
back to the question. If you know function of each pin and you connect it like a genuine chip, will it work?

Posted by: @chinhnguyen

↑

hi. so is it just a different pinout or is it not working at all? This is a project I probably won't complete because there are a few parts I can't get genuine.
back to the question. If you know function of each pin and you connect it like a genuine chip, will it work?

The THAT151x series can be replaced with a STANDARD 8-pin instrumentation amplifier IC - but it's unlikely to be as "low noise" although that doesn't mean it won't work broadly as expected. The problem is that the gain equation (that's the volume control) will need the values adjusted.

The INA126 is one example but there are quite a few since this is a fairly common configuration for, well, instruments... due to low noise, level shifting and so on.

All the 8-pin DILs have the same pinout so they can be swapped for alternatives. At least that is a standard of sorts.

If there was enough call, I can do these for you - other parts like the NMA0515 are also standard pinouts if you can't find that. It's even possible (not advisable) to make a THAT from three separate chips at a pinch.

I’m having a lot of trouble creating enough friction from the hinge. I’ve tried filing down more surface area and adding a silicone layer in between. My failsafe is to add an adjustable support arm but don’t wanna sacrifice aesthetic. Any suggestions?

Hi @Kirby, been a while. Hope you're well. My area is more electronics (I recently completed the final head and it's a doozy but more of that another time).

There are several types of washer you might look at. The simplest is the helical (split) spring which gets tighter as it's flattened.An alternative, disk spring washers, uses flat sprung steel which (again) produced more friction as it's tightened. These are thinner that the alternative and unlikely to be noticeable. Starlock washers (you can't miss them as the look a little star like with the inner teeth) are available in brass-like material but might not work as well. Finally there are "nord lock" washers which use teeth that work against each other but they're likely to make it "click" when adjusted which may not be what you're after.

Hey guys,

I am new to the forum and basicly electronics as well.
My wife stumbled accross the DIY Perks mic in my YT feed and fell in love with it.
So I will try to make one for her.

As her laptop only has a build in sound card I would need to make a 3.5mm version which has a MIC-IN.

I`ve read all posts in this thread (understood not even 10% of it) but it seems like there is currently no version that supports 3.5mm mic-in connections, right?

best regards

@kirby I did not find the best solution, but I can tell you what I tried:

- Nothing: obviously, this doesn't work, not enough friction

- Foam: I tried with some protection foam I found in a package long time ago. This particular one was not dense enough and it had nearly no impact 

 - Foam board: This is like a 3mm thick sheet of paper, but made of dense foam. This work well for frictions, but after a few days, the foam got compacted by the pressure and would not come back to its original shape. So if I change the hinge rotation, then it slowly come back to the previous position where the foam is compacted.

 - Rubber seal: I found a patch of 2mm thick rubber seal. This is really strong so it would keep it's shape for a really long time and friction is pretty good. But it is a bit too strong against deformation, so if your brass is not perfectly flat, then the hinge will have a "preferred" position. For my mic I used it for the capsule rotation, and the rotation hole was not perfectly centered on the brass rod, so some positions are really hard to move due to too much force needed to compress the rubber seal.

Hope this helps

Posted by: @beorak

↑

Hey guys,

I am new to the forum and basicly electronics as well.
My wife stumbled accross the DIY Perks mic in my YT feed and fell in love with it.
So I will try to make one for her.

As her laptop only has a build in sound card I would need to make a 3.5mm version which has a MIC-IN.

I`ve read all posts in this thread (understood not even 10% of it) but it seems like there is currently no version that supports 3.5mm mic-in connections, right?

best regards

Strictly speaking it was never intended to go to a mic input as this is an audio -> USB device. It works with USB 2 and 3 too.

This thread has got a little outa hand due to my long-winded experimental designs. :/ I've come up with a few solutions but, as a stopgap (I'm ordering the new pre-amps today all things being equal) you can do the electronics very simply with just a few parts. It won't have a volume control but that's not very practical for a "mic" input without going to a lot of trouble and the whole point is... 😉

The *simple* (less than ideal) solution is to do this:

In fact, pretty much all low-cost microphones (esp. those little clip on ones) work like this.

Which might have you wondering why all the discussion over JFETs, noise, and a lot of experimental boards? Does the KISS [keep it simple, stupid] principal apply?

The simple fact is that for pure voice applications, we don't (usually) need to be that demanding. The vital component is the first one in the chain - the capsule or the JLI2555 in this case. This is ultimately what will make or break any microphone be that from some old dude in his bedroom to the multi-$1000 behemoths you find in recording studios.

Matt's design sits in a slot somewhere between some generic mic (with an internal FET - the transistor) and the studio quality beast that I've been developing. And Matt's original certainly made that job a lot more challenging in a good way.

I won't got into all the jargon of precisely how this works but here's an overview.

A standard PC mic socket uses a TRS (tip, ring, sleeve) jack which supplies about 5V nominal to power the "impedance converter" which is a fancy name for the JFET transistor.

Electret capsules work by moving a minute amount of fixed electrical charge around. It's so small that we need a JFET with their special characteristics to detect it.

The 5V supply from the PC is fed via a small resistor, classically 2.2k through the channel - pins 1 (Drain) and 3 (Source) in this diagram.

As the diaphragm in the capsule moves, it causes current flow in the channel to change in (roughly) direct proportions.

 That results in an varying DC appearing at the drain (from a few thousandths of a volt) which is "decoupled" from the 5V DC by the capacitor - C1 so only the changing voltage is sent to the computer.

Cheap electrets have the transistor fitted internally but the better ones allow us to select better JFETs and do a whole bunch of neat tricks which I won't go into here.

To do this with the existing design you can make the hardware (capsule, brass parts, suspensions and so on), even the braided wire exactly as described.

But then at the "business" end you can just connect the "source" terminal to the ground/shield and the other two components as shown and you're in business. Volume control for such a device is already built into a modern OS so you need not get concerned over that.

If you need any clarification, just let me know.

@ldoppea Cut up a piece of silicone mat that is similar to what you tried, it started tearing almost instantly, Marc recommended different washers I used the standard ones but this time in between the "joints" while also carefully flattening the walls. which gave it just enough friction to hold up, just for how long is my issue.

Standard washers won't give you a lot of friction. You could line the tube with a solid brass piece (after the holes are drilled) to stop it collapsing I suppose. I wonder if anyone has tried springs (a la those desk lamps) which I think Matt used for inspiration?

homemade square brass washers, with a bit of tinkering they should hold up. Anyway, I've seen some people opening up their desoldering wire, is there a technique to open it up or I gotta get lucky getting the right kind? LSK170B how can I tell which leg is which,

 "Top view" is very vague, diagram makes it worse as it makes it seem the legs are facing you.

Also, can someone explain what Analog means and how vibration signals from the mic get turned into digital info? Also is the audio jack or red/white wire used for the build? 

The middle pin (whichever way you look at the LSK170) is the GATE. The other two terminals (drain and source) are interchangeable as the device, like most JFETs is unipolar - which means the drain and source terminals assume their function depending on what voltage is present on the gate.

A positive voltage applied at either end of the channel (with the gate tied to, say, the 0v line) reverse biases the gate-channel junction which electrically looks like a diode.

In that regard they are the easiest transistors for beginners. Beyond that they are a massive pain in the backside. 😉

If you look at the picture - hold the device so you're looking at it from above with pins away from you and the flat side facing you. The pins are then:

DRAIN - GATE - SOURCE.

But honestly, don't sweat it.

Audio Grabber

This is a little deceptive as it has four inputs for two channels. Any input can be used to digitize sound from the pre-amp. White or red are the simplest as there are fewer wires to fiddle with and separate but you can leave it as is and connect it up via a "phono" male to either red or white.

Microphone

A detailed explanation of how a condenser microphone works is quite fascinating but it requires some knowledge of the physics of electrical charge which isn't fun unless you're really into it.

But let's try to simply it a bit.

A capacitor is "just" two bits of metal (called plates) with a layer of insulation between the two. If we connect a power source to it current can't flow as it does through a wire BUT a very small amount of current does flow caused by charge carriers moving around. This means that one pate is positively charged in relation to the other (electrons move to the other plate).

This only happens because we've applied something to move the electrons, they don't do it on their own.

But then something "strange" happens when we remove the power. The electrons are attracted to the opposite charge on the other plate - and get stuck there.

Now we say that the capacitor is "charged". (Very large capacitors use chemicals to simulate this effect but microphones are tiny capacitors in the order of 50 picofarads. That's 50 x 10e-09 Farads - a minute amount.

OK, so we have a charge on the plates - and if we have a very sensitive voltmeter that has a very high input impedance (like a modern digital meter) we measure see this charge represented as the voltage we put there.

You can try this with a larger capacitor (up to about 10-100 uF) quite easily. Anything bigger might spark a bit when you try it - a 9V battery is usually convenient to do this.

With old multimeters we would be able to see the charge leak away as the device leached charge away while it did the measurement. (The observer's law - we can't measure something without changing it. The Schrödinger's cat experiment makes fun of this.)

Anyway, so we have a voltage on a capacitor which is all fine and nice - and we can measure that voltage (voltage represents electrical charge) by putting a very sensitive ("high impedance") measuring device on there.

In the microphone circuit this measuring device is the JFET which has an input impedance in the order of 10s or 100s of gigaohms (1 x 10e+9) which is huge so it doesn't "drain" the capacitor at all.

But that's a fixed capacitor. What about a microphone?

In a mic there are two metal plates spaced a fraction of a milimeter apart. The actual distance determines the value of the capacitor. I won't go into the physics but you can trust me when I say that the closer the pates are, the larger the value of the capacitor. Also the larger plates (the more surface area) the larger the capacitor is too - which is why many professional microphones use a 34mm (with a 32mm - I think) circular plate. So smaller microphones like the JLI2555 tend to have a smaller raw capacitance unless the plates are even closer together. If those plates were touch, the capacitor would discharge so the engineering that goes into these devices is a work of many decades of development and improvement.

OK so the insulator in a condenser/capacitor microphone capsule - the stuff between the plates is just dry air which has a very high impedance - you need a lot of voltage to get current to flow through air and when it does the results are spectacular - see Tesla coils for a demonstration.

So now we have two fixed metal plates (one of which is slightly flexible) with a tiny air gap between them forming a capacitor at rest of around 40 pF.

Not very exciting on the face of it but consider that sound is just rapid variations in air pressure. Those variations are picked up by the tympanic membrane in our ear and makes that move in and out. The greater the pressure, the more the membrane is pushed in - and if too much pressure is applied (diving underwater more than a few meters down without equalising), or a loud sound like a gunshot or explosion for instance, the membrane ruptures. This doesn't cause bleeding from the ear oddly, but it does damage hearing - sometimes permanently. The smallest bones in our body are the ones in our inner ear. (It's also why movies are so ludicrous as you see actors firing large guns, sometimes automatics, inside buildings. In reality a single gunshot causes temporary deafness at best.)

The front plate on our condenser mic does the same work as our tympanic membrane - moving in as the pressure increases and back out as the pressure decreases.

Note that the speakers in your phone, television, hi-fi, car radio... everywhere do the exact opposite. The create air pressure according to the amount of voltage applied. In fact, it's possible to create a speaker with a condenser and use a speaker as a microphone. Not that you'd do that because they are lousy at it. Condenser speakers are hideously expensive and very, very large as the diaphragms have to move a lot of air with but the input voltage doesn't move them very far. That's really something for a different post though - but if you can work out how the condenser speakers work you'll have an excellent grip on how capacitor microphones work.

SIDE NOTE: It's unlikely that even a 34mm capsule would make any audible sound but it's something that might be worth experimenting with at ultrasonic frequencies.

So now here's how the magic happens.

If you've ever done that experiment where you rub your hair with a balloon and then the balloon makes your hair stand up? I can't due to lack of hair... but the principal applies.

This works because when we rub a couple of insulators together (hair and a plastic balloon) some charge carriers get transferred between the two.

It's much easier to think of "charge carriers" and "fields" rather than current when you get into electronics. Fields - which carry the power, "live" in insulators and the vacuum of space. It's fields that move charge carriers about the place. Electrons (negative charge carriers) don't move at light speed (they're actually rather slow) they are affected by fields. This stuff has to do with Maxwell's equations and more bloody physics so let's give that a miss.

Fields are the things that allow us to have things like everything from mobile phones to WiFi and Bluetooth. Fields are everywhere and they will be the bane of everyone with this build because the field generated by the electrons moving in your house wiring are moved by fields themselves.

TL;DR

I know this is complicated so let's look think in terms of magnets for a moment. The magnetic force that you feel when you put two magnets together is in essence (although caused by a different quantum effect) the same thing - someone will hopefully correct me on which is which, it's been a while since I studied this.

Imagine holding two magnets - near each other with the poles at the opposing ends - you can feel the force (or the opposing one as you try to put two north or two south poles together. That force is a field - a "force-field" but not like in science fiction. I don't know your schooling but most kids are shown the piece of paper with a magnet underneath and iron filings spread out. That pattern is the field - literally the energy bound up within the magnetic poles of the material. As you move further out, the pattern starts to fade as it gets weaker at distance.

It's an electromagnetic field that's going to move our charge carriers around for us and this is where it gets cool.

Recall that I said when we apply power to a capacitor then we move charge carriers around? The field is tearing "free" electrons from the conductor on one pate and depositing them on the opposite one. In truth the electron doesn't move very far at all - it's like a Newton's cradle in effect - the field shifts the entire bunch (trillions of electrons) along a bit.

This takes a finite amount of time but imagine it's instant for now. (I'll skip over some of the physics but HMU if you want more).

At this stage the capacitor is said to be charged.

The electrons are stuck wherever they landed now and held there directly (OK *this* is the cool bit) because they are attracted like little magnets to the little "holes" (positively charged molecules) left behind in the other side.

Recall from the simple magnet experiment that the opposite poles are strongly attracted? Same thing here. The electrons "stick" to the charged plate because there's an annoying bit of insulation there - air in a microphone.

And this is why the gap between the plates has to be so small. The range of this attractive force, while very strong, only has a range of a few fractions of a millimetre in real terms. About the width of an average human hair is typical though much smaller capacitor mics do exist.

So at this stage we have:

Two metal plates separated with an insulator and a field generator (a battery) that has force a bunch of electrons to get yanked off one plate and dumped on the other. A this stage the two plates are weakly attracted to each other but this force is so small we can't feel it. (We can in a very large plate with a very high voltage on it ... and that's your capacitor speaker - https://en.wikipedia.org/wiki/Electrostatic_loudspeaker!)

So what happens when the battery is removed?

The electrons stay stuck where they are - attracted by the holes created by the battery. They cannot move without something to push them.

The fact that they are stuck there, held by this quantum force, means that we've stored a very small amount of energy. (See note **).

Now this next bit might bend your noodle but hang in there.

The size of the capacitor, around 40-50 pF is determined by how far apart those two plates are. If they move slightly closer together the capacitance increases very slightly and if they move further apart, the capacitance decreases.

If I've explained this properly you should be able to see now that when the capacitor's size changes, the amount of energy it can store changes accordingly.

The capacitor now, almost like magic, can move electrons by changing its size and the size is determined by how much pressure is being excerpted on the metal diaphragm. Louder sounds mean more pressure and that means ... the size of the capacitor increases accordingly.

If the capacitor is in a circuit with a power source (say a battery) this change in capacitance, which is going to be in fractions of a picofarad, I don't know the exact amount for a given device, but it's not much, allows the fields created in the circuit to push more electrons onto the plate. The distance is key here remember. The shorter the distance, right up until contact, the greater the force pulling electrons to their positive charge carriers.

As the pressure is removed, the plate moves further away and there's less force to hold the electrons in place so they get yanked back. However the amount of electrons flowing is controlled by the distance between the two plates and the speed the plate moves determines how fast that happens.

This is electrical current - and since it can go either direction, it's alternating current too. (This remarkable effect is used in power converters to "flip" a small voltage from positive to negative.)

The complex frequencies that make up sound cause these minute variations in the material and the microphone has converted ("transduced" is the fancy term) the physical effect into an electrical one.

The pre-amp circuit simply steals tiny bit of that current (recall the Observer effect, this is crucial here) and converts it into a voltage which is then amplified and eventually sent to the ADC - the analogue-to-digital encoder in the little black box.

For anyone following along - you'll note that I've described a non-polarised mic here. It's usual in these (often the most sought after mics) to have a polarising voltage of around 50-80 volts. It's that high to allow sufficient current to flow when the capacitor's value changes to be readable.

Electrets are a slight different animal but the principal remains the same.

If any of this is unclear @Kirby, just let me know and I'll see if I can expand.

*** NOTE: In most capacitors some charge will bleed off over time due to other properties in the material that are imperfect but a large capacitor can hold a lot of voltage for a long time. Large electrolytic caps exhibit a weird electrochemical effect where they self-charge after being rapidly discharged. Freaky stuff when you see it IRL but it's not magic. It's just the electrolyte is actually the storage facility here (more so than the simple plates and insulator). This makes very large capacitors possible with Supercaps going into the 100s of Farads. This is the stuff of dreams, nightmares and very powerful lasers. A large, charged super capacitor can quite easily kill a grown man by stopping his heart in an instant.

The 2200 uF ones here won't hurt you of course but it's worth remember that in some equipment (primarily televisions) some capacitors can hold very high voltages and that's going sting and leave a mark.

This post really needs some pictures but I'm tight on time so if anyone has some relevant ones please drop them in the comments below or I'll add some when I have a moment.

Don't Try This at Home!

Be careful with modern "supermagnets" as they are extremely brittle and the shards could blind you. I'm cautious by nature of being variously: suffered mild exposure sea swimming; having a couple of 240V belts; an explosive blow up in my face; burned (2nd degree); gassed - all before I was 14 years old. I've survived this long largely as a matter of luck. All of this is true and I know how much it hurts. None of those incidents, as serious as the could have been, caused lasting damage except the sulfur dioxide gas - that left me with scarred lungs. And like an idiot I smoked after for near 12 years.

I am an idiot. Just an old one from a time when occupational Health and Safety was still a new-fangled idea that'd never catch on.

Posted by: @diyperks

↑

That was on the old forum installation and it broke unfortunately, but it's still accessible: https://diyperks.com/community/

Hello. thanks for  your great work ! I have the same question about the xlr and the link you give says that the forum has moved, an it leads to the new forum... can't find the answer. anyway, would and xlt coming out from your amp work, if linked to a RME ucx, for instance, or would it be a direct link form the caspule to the RME instead ? In both solutions, which connections have to be soldered ?  

I was surprised , regarding your amp, to see that you added a volume knob, because all the USB mics I have tested are automatically leveled internally in the computer, and this is not a good point if one is a voice talent because you can't choose your entry level by hand like with a mic preamp. How come you can change the level with this knob, will it prevent the automatic leveler to do the job ? thanks. 

Luc 

Hi Luc,

Not sure what you meant in the first half of your question - do you mean an XLR 3-pin with a 48 volt switch? Something like this?

The current design (in the video) won't work with XLR but Version 2 that I've been developing for the last 18 months does. All of my experiments and mistakes are still in this forum, so people understand it's OK to make a mistake. The latest version performs (on paper) similarly to the Rode NT1a. I'll know better when the prototypes get back from JLC. I have tested an earlier version of that and it's at least as quiet as my AKG which, honestly, was a surprise, so I've made a few improvements to the linearity and frequency response. If my simulation is accurate, it should be the final one. I'm still improving the pre-amp section.

So to that - I'm not quite sure what you mean about auto-levelling. USB mics will generally use a very similar circuit regardless of who makes them so the output is generally going to be similar.

Of course, we can have an AGC but causes many unwanted effects including distortion.

Using a simple gain control here we can set the "volume" for different uses, podcasting, acoustic guitar etc. And get that level back at the push of a button.

Does that make sense?

Hello! I watched DIY Perks video about this microphone when it first came out and really wanted to make one. I bought all the electronic parts and assembled it but it had no sound output.

As I was also busy with school, I put the project on hold and watched the forums for a situation similar to mine. I saw many mentions of fake THAT1512s so I bought another to test but no luck. I recently came back for China(The place of cheap electronics!) and bought another hoping that the problem was just that the previous were fakes. 

I first checked the voltages in the socket as there seemed to be a reference of what was normal and it seemed fine although the inputs voltages were a bit lower at around 14.5V(rest of the pins at zero). I thought that was probably fine so I inserted the new chip and still no audio. When I check the chip again I found it hot(prob over 40c maybe more) and quickly unplugged it. Just to check I plugged it back and quickly measured some of the voltages and found the inputs voltages to have decreased signficantly and by different amounts (~10V and ~8V I think? One is negative but don't remember which)and the output was about -4V.

I assume I fried that THAT chip(haha) and buying another from a more reputable seller would cost me ~30CAD which I would definitely not like to spend.

Also some tests with the old THAT chips that I followed from the forum:

Putting my finger on the mic wires did create a humming noise although I needed to amplify the sound in Audacity by quite alot. Things like tapping the capsule made spikes in the wave form and some looked like a discharging capacitor?(A large spike followed by what looks like a capacitor discharge curve)

And pics of course:

And just adding that I do have access to a multimeter, variable power supply, and a cheap oscilloscope that can also generate wave forms.

Looks like you've hit a buffer a lot of people ran into with this. Primarily that the design is on Vero/Stripboard which is why I've been beavering away trying to improve it a process that has proven rather difficult because, for all its failings (and Matt admits that it could be better) it's actually quite a solid design. When it works.

One of the advantages of doing this for my own education is that I don't care (all that much) when things go wrong and trust me, I've got a couple of kilograms of failed boards to prove just that. 😉

Apart from my first attempt (when I didn't think to check the NMA voltage inverter's specification, which was a lazy and expensive mistake) I've been troubled by noise of the sort that's very difficult to track down since (as it turned out) was coming up the USB lines. Now I knew they are noisy, that's the worst kept secret in USB, but I didn't really count on how bad it was. One of the Lenovo machine was so bad in this regard, it had to be run on battery (due to mains creeping through the unearthed switch-mode PSU).

Before we go any further though, you might have noticed I found a huge potential error in the design which can fry the audio grabber - for that reason I suggest you disconnect it for the moment.

Here's that problem.

The THAT151x chips are capable of output swings (on a +-15 volt supply) of 13.5v each way which is way more than the audio grabber is designed to accept. More than 12x overload. I blew two digitisers before I figured that out - it's one of those things you just don't think of (even though you should).

I've fixed this in the new PCB which will be available soon. I don't want to say when in case I jinx it. This has take a while since I have to design a board, have it made, test and debug it and then start over. I've had a "working" version of Matt's original for some time, but the later versions improve in several key areas. I've also made them "me" proof - most components are factory fitted and those that aren't are easy to handle, except for some experimental parts that might (or might not) make the final one.

One experimental board I've done replaces the (expensive, if excellent) THAT amplifier with a discrete design constructed from three operational amplifiers. Functionally it's almost identical except the for the gain which requires different resistor values. The basic circuit (minus the gain resistor) looks like this:

It's made from two dual operational amplifiers (a quad takes a little less space on Vero but the SMD design is more efficient as two duals.

The fourth amp is unused as you can see. If someone else has time to make this on some Vero, it can be made with long pins so that it fits into the socket for the THAT1512. Note the little markers on the circuit that show where the various pins connect to mimic the function of the original chip.

There are any number of explanations as to how this works on the web so I won't repeat that here. The circuit can be made with one or two Op Amps but this version has better control of gain and far better CMMR which is what you need for an "instrumentation amplifier". You sometimes see a "Sense" pin brought out of this design which "drives" the secondary screen but we don't have that luxury here and in fact, it's really only useful for very low voltage transducers such as you might find a medical setting. You could connect Sense to a metal case but that causes other problems with grounding so it's really an advanced function.

The main advantage of this is it can use cheap amplifiers like the TL072/TL072H which are tough and plentiful. You won't care so much if you blow a TL072 that cost less than a buck, right? Most dual Op Amps will work in this place in fact. More exotic ones that run at lower voltages cost more anyway, so no harm no foul.

Noise performance suffers of course and you'll want to replace those with either OPA2134, HiFi Op Amp from Texas which I've used in some other designs and is very quiet indeed. Nothing comes close to the THAT ultimately though - it's more a question of if you can hear it.

Oh No, Magic Smoke!

That THAT (sic) getting a bit toasty? That's actually normal when the device is being pushed to its gain limit (which is something in excess of 1000x) and those output drivers are swinging 13.5 volts into whatever load you've got... That produces a fair bit of heat but probably (I haven't tried this) drags the rather weedy NMA0515 close to shutdown so it's unlikely to overheat. Again, I haven't tried this.

But wait there's no sound, right? Have a look at this:

https://www.thatcorp.com/datashts/1500data.pdf

My instructors used to tell me that datasheets are your friend. So long as you can read them and there's a humorous story right there (cough, cough).

And right there, top the page, we all missed it - 13.3V swing on 15V supplies. Whoops!

Spanked botty for Marc then.

But right below that "Wide Bandwidth: 7MHz @ G=100"

G=100 is also 40dB and is roughly the gain you'll use with a typical microphone capsule of this type. Moving coil designs (like the famous Sure SM58) need about 1000x gain (60dB) - which is why the gains go so high.

The key here is 7MHz - and that's a 40dB. So this chip is capable of huge voltage gains orders of magnitude beyond what a human ear can perceive we tap out about 15-18KHz by our twenties. If this is an issue you should be able to see it on your scope although it might just look random. (My eBay/Ali handheld can manage it, modern instruments never cease to amaze me.)

It is possible to limit the bandwidth (in fact, we frequently do) but impractical on a design like this and I expect THAT figured the phase shifts etc. were too detrimental to the overall performance.

I've also discovered this design has a tendency to self-oscillate (worse on a PCB ironically) at very high frequencies which can also cause this mysterious heating. All of this assumes that you've put everything together correctly and set for 40 or 60dB gain.

TL;DR

Ok really it's time to stomp on a few bugs. I suggest you read all the steps here before you pull bits apart. You might decide to a different order.

IMPORTANT: with reference to the "bug" above it's absolutely vital the digitiser section is removed. There's high chance that you will push the THAT into overdrive and that will destroy the both channels of one section of the two stereo audio inputs or worse!   

Step 1

We'll start with the capsule and work from there. So first of all you should remove the FET from the capsule.

The capsule itself should measure roughly 30-50pF (0.03nF) if your meter has a capacitance range. Don't worry if not, these things are quite robust if handled properly. I took mine apart and to see how it worked, but I'm just naughty like that. Maybe we'll make one, one of these days - just for fun.

Testing 1-2-3

Next bit get ready to blow your mind. With care you can see the microphone working on your scope. You'll need to select the most sensitive range you have to begin with, but the input to any modern scope is a fancy dual JFET affair not unlike the FET you've soldered on it. Give the mic a gentle tap (or if you connect it well enough to hold) talk to it and you'll see the waveforms appear.

This is only possible with electret microphones like this because they have pre-charged plates and no el-cheapo JFET in the way. If you've read this thread you'll probably note my comments regarding 100 for 1c electret capsules that are produced by the zillion ever year. 🙂 

Step 2

Let's remove the THAT1515 (use a puller if you can, or a screwdriver at each end if not). Snapping a leg on these things isn't fatal to the chip but it's horrible to solder. I have my cat to thank for one of mine! He dug it out of its parts bin and ran off with it while I wasn't looking. For weeks I wondered where I'd left it until I found it with my foot one evening stumbling around in the dark. It's now soldered to an experimental pre-amp board.

Step 3

You can use a voltmeter for this or your scope set to DC. I find a voltmeter faster but scope probes are easier to, well probe, delicate parts. It's important to know that the NMA0515 isn't regulated. That means its output is likely to be several volts higher than specified on a modern instrument. This is normal and not of serious concern. So when I say 15 volts, I mean somewhere in the ball park a couple of volts over and up to 0.5 volts less..

Power up the board and measure the voltage with your black lead on Pin 5 measure the voltages at Pin 4 (-15) and Pin 7 (+15). You can also measure with black to Pin 4 and red to Pin 7 for 30 volts. The DC level should be 0 (referenced to Pin 5). Might be a good time to check those.

Step 4

Bit of a tricker one this. Measure the voltage from Pin 5 (which is the ground or 0V reference) on the Drain and Source pins of the JFET. You might find it easier to do this at the board than the capsule. I find the scope is useful here. Set to with 2V/cm you should see the move to show the voltages you measured at the power pins for the THAT1512.

Step 5

If you removed the JFET at Step 1 - which is optional - you can put it back on now. Don't put the THAT back at this stage.

Power the board and with your scope in AC mode, probe either Drain or Source terminal of the JFET.

Even without sticking a finger in it (oh, those were the days, before I stuck my finger in the back of a TV...) you should see a considerable amount of mains noise. It's a nasty looking sine wave, 60Hz where you are. Sticking a finger on the mic's centre tap (which is the JFET's gate) should cause that to ramp up quite a bit.

Step 6

This is strictly, Step 1 because we'd normally start in the centre of the circuit and work out (a technique called divide and conquer) but I find it's instructive to understand what the various components are doing in the chain of operation. I'll assume from now on that you know when to apply power to do tests and when not! And now you know why it's vital to read the instructions before you start pulling things to bits! You did read this first, right? 🙂

Measure the DC level from Ground (Pin 5) to Pin 2 and Pin 3. This should be 0V or very close (some residual charge may show on the decoupling capacitors but that goes away over time. (This can be another "gotcha" when you read the spec sheet fully but it on effects people who want to unplug their mics.)

Now measure for DC between Pin 1 and Pin 8. Again, this should be 0V. You can also check your gain resistor at this stage. If you haven't made the switch yet, I suggest popping 50 ohms in position which will give you x100 gain which is a good starting point. 50.3 ohms if you bought one but 40 - 60 ohms is close enough for testing.

Using your scope (AC mode) look for noise signals on Pin 2 and Pin 3. Expect to see the same sort of thing you saw at the Source and Drain terminals. (This ensures you have your decoupling capacitors in place and correctly functioning).

if you have enough fingers and at least a second channel you can validate that the phase splitter is working as it should by comparing the noise signal at Pin 2 and Pin 3. One is shifted one full half-cycle to the left/right - which is 180 degrees of phase shift.

Step 7

It's Miller THAT Time!

Power off (just to be safe), put the THAT back in and let's look at what's going on now. Deep breath now because this is where we get to find out if your THAT1512 has gone to silicon heaven or not.

Using a voltmeter check (with reference to Pin 5) voltage on Pin 4 (-13 -> 15); Pin 7 (+13 -> 15); Pin 2 and 3 - both 0  or very close.

Pin 6 (output) should sit at 0V but it can rise to a perhaps 5 mV - so the output capacitor isn't usually necessary. If you this terminal has a DC offset on it power down and temporarily short Pin 2 and Pin 3 to Pin 5. This gives the amplifier an input of 0V (very close anyway) so even at full gain (Pin 1 and Pin 8 shorted) it should not swing more than 5 mV.

Any DC at Pin 6 in these conditions means that (*bad joke incoming) the THAT1512 is now a WASA1512.

You can use your scope at this point. The signal at the output (Pin 6) should be a mirror of the signal at Pin 2 or Pin 3 only much larger.

The signal also appears at Pins 1 and 8 referenced to the Pin 5 and across those pins too if you care to measure it.

Step 8

This is about as far as we can go because if all those tests passed correctly, the microphone should work. It won't work WELL until you have everything screened. I can't stress this enough. The input (via the JFET) is incredibly sensitive and the unscreened capsule will generate a large amount of noise; mostly mains hum.

Let me know how you get on and we'll take it up from there. Good luck!

@marcdraco I'm trying to order Michelle-Through-Hole from JLCPCB and have them to assemble the parts. Do you have the file that includes that parts?

@kulmam92 Hi, Michelle is at an "experimental" stage.

I'll drop the current Michelle 1.4 which is vastly superior to the original but hasn't been field tested yet. It's a three or four board design now although you only need one. The other two provide optional functionality like filters.

The "Varee" capsule adaptor is working well and outperforms mics costing many times as much - however it doesn't work correctly with Michelle - there's a fault/feature in the power supply design (cough) which is giving me headaches. Matt worked around it with (very) large capacitors and even a regulated version isn't playing.

The issue is with the NMA0515s - I can see the inductive spike on my supplies so I need to suppress that better. (Murata has a recommended schematic for that so I'm sat waiting for parts).

Varee works faultlessly with devices like the Focusrite Solo with 48V phantom power.  

I've been pulling my hair out for months but I've tracked the problem down now (*it was time to get some better test gear and it's amazing how much easier it is to track down annoying faults ).

I EXPECT this version should work with Matt's original capsule design but I haven't made the time to check. I'll do this as a priority now. The new version has jumpers to configure the board in multiple different versions.

This is my working version - so some of the part numbers are incorrect or missing at present. The circuits are complete and I think the layout has everything. DRC will stop on errors over the plug-in boards.

You'll need KiCAD8 for your platform to view these files. Keep an eye on this forum for breaking news. I thought the issue with the Varee designs was in the capsule and it turns out I was examining the "wrong end of the cow" if you know what I mean!

@marcdraco, thank you for the prompt response. It would be fantastic if you post the below information.

1. Michelle
   1. description
   2. Stage / Version
   3. optional boards
      1. name and description
      2. part list for assembly
   4. part list for assembly
   5. design file - Gerber and BOM/CPL file
   6. Assembled picture like Nukebommer360's post above
      1. Microphone to usb
      2. with optional boards
2. Chimera
   1. description
   2. Stage/Version
   3. part list for assembly
   4. design file - Gerber and BOM/CPL file
   5. Assembled picture like Nukebommer360's post above

I'll get the parts (BOM) together today hopefully - I need to order the prototypes but there are issues with the power supply which I'm finding, er, troublesome to deal with. It's a nasty issue (causing the head to shut down) that may be down to screening but I'm not 100% happy as yet.

Varee is by far the best performing head and it does appear to work (so far) with "professional" gear like the Focusrite Scarlet Solo. P48 preamps are surprisingly cost-effective these days (not as cheap as DIY up to P12, necessarily) but Varee is designed to work from 12 to 48V without any real changes. One resistor might need adjustment but I can't be sure until I've tested it with some other P48 equipment.

All the complexity matches or betters that found in some top-end designs and self-adjusts with 100% negative feedback to keep the FET in check.

*Late breaking - I've added space for a 10G resistor for either SMD or a through hole (10M will work at a pinch). Seems a lot of the issues with these higher specced circuits is that they get very twitchy about the JFET biasing. (That's a yolk I've carried for the last two years... and now finally it bites me on the rear.)

So what about all those boards with Michelle?

The main board carries the main power supply and now an optional isolated supply for even better CM noise rejection.

This section also houses the THAT1512 (or similar) with an option for an SMD alternative as it's a standard instrumentation amplifier. Output is clipped hard at 600mV give or take to prevent overload destroying the downstream equipment.

Just below the power supplies on the main board is an (optional) driver for balanced line out.

"Quinn" the little daughter board is the really poor man's alternative to the THAT151x using three amps (from two duals) but with SMD pads to it can fit into the space normally used by the THAT151x. I'll have to run a set of calculations to get the correct gains for each as different manufacturers have their own internal resistor strings.

The larger daughter board ("Alexandra") houses a veritable feast of useful functions that are better applied at the input side.

* High pass filter (low cut) rolling off the subsonics and unwanted low frequencies with cuts of 12, 24 and 36dB per octave.

* High pass filter (proximity) a shallower, softer low cut to reduce that "proximity" effect on the low frequencies when speakers get very close to the mic.

* Low pass filter with a cutoff around 15KHz and an (optionally) adjustable peaking to shape the high end with a bit more "whoohoo!"

And a headphone amp of course.

The expansion port is for optional functions such as a Baxendall volume control which I've designed but not yet tested "in anger" so to speak.

This is an overview since these boards are not strictly official and have to be assume to be experimental until I say otherwise: and trust me, I will let everyone know with a proper release announcement. 🙂

@kulmam92 Re: Chimera. That's really intended for a different audience to this one since it's designed to adapt to a small USB-C digitiser and fit inside the body of a BME-800U. It's good, but people tend to demand discrete over op-amps and you can't really drive much of an op-amp with 4 mA to spare.

What you'll need is the Varee design OR (they're here somewhere) the Sarah/Ubuntu carriers which are meant for mounting a single JFET to those small capsules for simpler circuits.

@fishoutofwater I've just been helping my son create his version of Matt's amazing design

It's been great fun, but I don't have either of their technical genius - so if you have a PCB board design from JLCPCB, I'd love to see it!  Thanks 🙂

That's delicious!

I've perfected the head (capsule) board. The pre-amp is still experimental (a bevy of new functions). I hope we'll have an answer soon... I do have a working pre-amp design (somewhere) but I'm a perfectionist so it has to justify Matt's hardware. There are some earlier versions further up the thread but they are incomplete/untested. If you grab a copy of KiCAD you can open the designs to see what I'm up to. I've thrown the kitchen sink into that lot. 😉

I can "announce" that I've developed a simple Vu meter (not calibrated yet) that drives a low-cost WS2812B LED strip from Amazon. 

It's not quite ready yet - I need develop safety diodes etc. to protect the input to the microcontroller.