[Sticky] USB-C Microphone (official topic)

@marcdraco I received the new THAT1510 from my parent's friends and... it works \o/

At first I thought it didn't work. I re-plugged everything, I did a new check for potential shortages and after I took all precautions to avoid ESD, then I did a record test with audacity resulting with no sound... but I took a big breath and realized that... I forgot to replace the old THAT with the new one  🤡

Anyway, I can now record understandable sound with correct sound level. There is a lot of noise, but I guess this is because not everything is correctly isolated.

All the circuit is inside of a metal biscuit box but I cannot fully close it due to outgoing cables. Also I badly wrapped the capsule into the brass mesh that was not sealed.

Thanks again @marcdraco for all your help debugging the electronic.

Next step will be to do all the mechanic stuff. I hope this will be enough to remove the noise, if not I may ask for your help again.

This step may take me some weeks, I'll start by making the capsule encasing. I'll post some photos and test records as soon as this part will be finished.

It's been an absolute pleasure dear boy. I hope you enjoy the rest of the project as is a true joy to see.

If I do a production run of "Varee" heads (which are higher performance than the capsule+FETs) I'll send you one with my compliments as a thank you for your detailed discussions to help us find the problem.

The vast majority of the noise is picked up at the capsule end. The circuit isn't immune by any measure, but that super-high impedance impedance converter at the capsule is like huge radio ariel.

 I'm delighted it was "just" the THAT - bit of a bind, but it was interesting to remotely diagnose a fault (we all hoped it's something like a duff joint and not an expensive chip).

If it makes you feel any better (and this includes the folk who are still following along) I could have bought a fairly serious professional mic and 48v supply for the money I've thrown at my passion project. 🙂
 
And I've loved every minute.

I still don't know (for sure) what sort of low voltage it will accept, but it should operate at 15 V (or 30 V) per Matt's original. Hopefully it will go all the way down to 5V to make it easier to power. Here's the final schematic as a PDF including the PNP output transistor that fell off the original rendering it totally useless. 

If that were an IC it would be tiny but then again...

It's a weird circuit that takes a fair bit to wrap your head around (at its core, I don't know who designed the kernel here with it's negative feedback loop (in developer terms it's a little like recursion but at a huge depth and speed that's effectively instantaneous). In real terms it happens at about 2/3 the speed of light but the two halves meet in the middle.

Bottom line (and you can try this in a circuit sim like LTSPice) is amazing performance.

The main impedance converter (Q202) feeds a second one (Q601) which then drives a class AB *power* amplifier stage (Q602 A and B) - exactly the sort of thing you would find in a high-fidelity amplifier.

Nothing too startling there - but the output of that pair is fed BACK to the main FET (actually a LSK170 or similar for low-voltage operation) at its Source.

The operation of base-emitter of Q202 (with about 0.65v drop) and the 330R resistor sets the FET current at 2 mA (give or take) while the action of the FET amplifier causes the drain current to change sending an audible signal to through Q601 force an opposite change at the source voltage which restores balance through... negative feedback! Quite brilliant coupled an exquisitely low distortion to a "pseudo" balanced output.

It's this circuit (or something very like it) is probably going to be found in the Rode NT01. 

The signal-free "cold" line in here normally carries the anti-phase signal of the balanced pair, but it's not that signal that matters, it's the impedance of the the wire connecting the mic to the amplifier. Not a massive issue on Matt's design, but on a run of a few metres or more-- the better that impedance is matched, the more effective the amplifier will be at removing stray noise (which, ideally is on both wires in equal amounts and that's why the impedance matters).

With PCBs (as tiny as these are) I've been able to do tricks that just aren't otherwise possible such as "guard" tracks around the high impedance nodes, board-level isolation and shielding and (I hope this time) enough of a solderable strip around the edge so that you can solder your mesh to the board if you're using that as your 0V rail.

Like the doofus I am, I totally forgot to add an output drive transistor to my last order so that one ate into me (I had to live on cold beans for a month) but I have no one else to blame. And I don't mind cold beans. It's the raw sausage and eggs that give me a problem. 🤢 

Hello everyone,

I took on this project and I'm at the final stages but I ran into a dreaded hum in my circuit, and I'm reaching out here for help.

I can say with 100% confidence that the source of the noise is the rotary switch, but I have no idea why it is causing a problem. I attached two audio samples: one with the switch, and one with the switch terminals jumped with a resistor. I have also tried three other rotary switches (two different brands), and jumping the line with a wire instead of a resistor, and the whine is ONLY present when ANY of the rotary switches are plugged in. 

Rotary Switch plugged in:

Rotary switch jumped:

I've also attached pictures of the mic and rotary switch(es).

An aside:

marcdraco your forum posts have been insanely helpful throughout this whole project, thank you for all the wonderful work you have shared with us all!

@marcdraco  Thank you for your reply, your project sounds pretty neat and I look forward to seeing the progress on here!  Did you have any thoughts on this post below?  If you have a chance to take a look that would be much appreciated!

"Hello, I am new to the forum and pretty much a beginner with electronics.  It seems you have contributed quite a bit to the forum.  I am looking to build this microphone and noticed this particular post when scrolling through and wondered if you could help me.  It seems some people have had a few issues with build, and from your post I could focus on 1) Replacing the 2u2 Electroytic Capacitors with Polyester Film to reduce noise; 2) Not use /or be very careful with Strip Board as it is very easy to short the components through "solder jumps" (Perhaps I could get a PCB printed to help reduce this issue?); 3) Add a RF snubber in there.  Did you mean RC Snubber (When I researched what a snubber was- (and my understanding is it is a device that deals with sudden voltage spikes - which cause electromagnetic interference - when there is an interruption in the flow of current) one of the options was RC snubber).  Would this be a resistor and a capacitor, and where would they go in the circuit?"

Thanks, Mike

Posted by: @justmike12

↑

@marcdraco  Thank you for your reply, your project sounds pretty neat and I look forward to seeing the progress on here!  Did you have any thoughts on this post below?  If you have a chance to take a look that would be much appreciated!

"Hello, I am new to the forum and pretty much a beginner with electronics.  It seems you have contributed quite a bit to the forum.  I am looking to build this microphone and noticed this particular post when scrolling through and wondered if you could help me.  It seems some people have had a few issues with build, and from your post I could focus on 1) Replacing the 2u2 Electroytic Capacitors with Polyester Film to reduce noise; 2) Not use /or be very careful with Strip Board as it is very easy to short the components through "solder jumps" (Perhaps I could get a PCB printed to help reduce this issue?); 3) Add a RF snubber in there.  Did you mean RC Snubber (When I researched what a snubber was- (and my understanding is it is a device that deals with sudden voltage spikes - which cause electromagnetic interference - when there is an interruption in the flow of current) one of the options was RC snubber).  Would this be a resistor and a capacitor, and where would they go in the circuit?"

Thanks, Mike

Hi Mike and welcome to our little corner of the web. 

RC (the circuit) snubbers and very simple low-pass filters to "short" radio frequency noise to ground before it gets to the amplifiers.

It sounds crazy at first because our initial thought is, "well I'm only interested in the audio, so how does that work."

Modern transistors (and when I say "modern" I mean even general purpose transistors going back to the 1960s that I know of personally) can operate at quite insane frequencies. These days, well into the 100s of Mhz.

Take a cheap and easily available common-or-garden part like the 2N2222 NPN device. It's can be operated up to around 300MHz after which it stops producing any usable gain. Such cut-offs are more usually due to device or Miller capacitance than the materials in use.

Now this is orders of magnitude greater than the 20 KHz theoretical max that we're interested in.

So what?

Well we only have a finite amount of power to drive the following stages with and if those stages are sensitive to (i.e. not filtered for HF, VHF and above) radio signals that we can't hear drain some of that power. This can cause weird problems like a circuit drawing large amounts of current (relative to its design) when it's supposed to be sitting idle (quiescent).

My (hopefully) last prototype Varee is literally arrived a few minutes ago so we'll hopefully have an update on that very soon. The latest version has a very simple snubber since it's designed to be operated in a "cage" that just uses a capacitor.

It's more usual to put a resistor and a capacitor in there so that the current flowing through the capacitor is limited by the resistor. In essence, you have a frequency dependant resistor with a minimum impedance at all frequencies of whatever the value of the resistor is. At low frequencies, the capacitor operates like an open circuit so no current flows. High (interference) signals "see" the capacitor as a short circuit but the excess power is dumped through the resistor and lost as heat. (If your dropping resistor gets hot though, you've got bigger problems.) 

Capacitive Reactance Calculator (omnicalculator.com)

This calculator helps to describe it. Note that at all frequencies for our "snubber" there is always a 1 K resistance across the output, but at 20 KHz the reactance (effectively the resistance) of the capacitor has dropped to around 800 ohms - so signals "see" a 1K8 resistor across the output (not really enough to notice at this stage since the output of impedance of a typical stage is much lower.

At 200KHz (AM radio) it's down to about 80 ohms and at 2MHz, it's down to 8. So you can see that as the frequency increases, the amount of signal drained away (prevented from going up our signal chain) is reduced.

Varee Rev. 1 doesn't have those resistors and effectively shorts everything to ground which is naughty of me but space on the board is tight so I really only included them as a half-hearted addition for completeness. Here's a 3D render, I'll put a pic of the real one when I have it working (cough, splutter).

My PCB has gone through several revisions (I've spun off several other projects as a result which is great but it slows things down for everyone, and that's my fault).

Electrolytics have a bad rap for any number of reasons. Usually it's because the buggers have a nasty habit of having exceptionally wide tolerances (+/- 20% of rated value is typical) and they can overheat in power supplies, especially switched mode ones if they are not "low ESR".

Basically, they are convenient ways of getting a large value in a "small" package (small is relative) but they do it by using a electro-chemical effect hence the weird behaviour.

ESR or equivalent series resistance makes them look like the snubber above - but the resistor is a "lumped" part of the component, an effect caused by the physics of electromagnetism (there's a tiny effective inductor in there too).

The small resistance has two effects, but the main one is to cause the component to get hot - the other is series resistance if the capacitor is used for DC blocking, I'll come to that in a second.

In a power supply the capacitor is used as a power reserve to remove the "ripple" caused by the effects of either the mains voltage fluctuating or (more usually these days) the switching noise from the switching IC. Switchers are everyday now, they're cheap, compact and reliable - but man are they noisy!

Now recall from the discussion of the snubber (the one that goes in parallel, across the supply which has an impedance governed by the frequency of the signal it has to snub and as I mentioned the resistor dumps the excess shorted via the capacitor as heat.

Same thing happens in an electrolytic and that "lumped" resistor which can be in the 10s of ohms  dissipates energy as heat. But heat shortens the life and eventually kills components. You'll see many repair videos on the web where an engineer will go through and replace all the leaking electrolytic capacitors (sometimes all of them as a matter of course). The heating effect causes the electrolyte to break down which, over time, causes the device to bulge. Modern electrolytics  always have a little cross on the top which is the weak point - designed to break open. Older ones (I had some as recently as the 1970s but I expect you might still encounter them) didn't always break and would go with a spectacular "KERACK" or a bang.

This should be the "uh oh" moment and you start to realise why ESR matters in power supplies.

We cannot avoid the heating effect so we have to minimise it by designing devices with very low "effective" resistance and just to make your head itch a bit more, ESR varies with frequency too.

I prefer to use low ESR types in serial connections (as Matt has here) too because they tend to be higher quality. It might be my perception of course. Some people swear by them, others (correctly) point out that the audio has already passed through any number of capacitors - many electrolytics - before it got to us so...

Electrolytics tend to be electrically "leaky" so they are less than ideal and should be avoided where practical. 

More recently there's been an explosion in SMD causing even more headaches for designers. Here we have the amazing multi-layer ceramic capacitor (MLCC) - they really are impressive too a 10 uF device can be had in 603 for pennies.

But there's a cost in quality. The larger value MLCCs (like X5R, X7R etc.) use ferrite in their construction and can't be used where RF might be an issue (assuming I read the paper correctly). There's a great intro to them here. 

basics-of-ceramic-chip-capacitors.pdf (johansondielectrics.com)

NP0 (the premium type) are only available in larger packages and smaller values to they tend to be reserve for the really tricky bits.

The very best capacitors are plastics like polythene and PTFE although there are some really exotic ones for specialist jobs.

Posted by: @swimwater

↑

Hello everyone,

I took on this project and I'm at the final stages but I ran into a dreaded hum in my circuit, and I'm reaching out here for help.

I can say with 100% confidence that the source of the noise is the rotary switch, but I have no idea why it is causing a problem. I attached two audio samples: one with the switch, and one with the switch terminals jumped with a resistor. I have also tried three other rotary switches (two different brands), and jumping the line with a wire instead of a resistor, and the whine is ONLY present when ANY of the rotary switches are plugged in. 

Thank you for such kind words, this is a terrific project but one that's caused a lot of issues because of the sensitivity of components, USB issues - you name it. It's also why it's taking me so long to come up with a design that just works. 

I must congratulate you on your work, that looks absolutely yummy and worthy of this project. Mine looks like a couple of rats got drunk and had a rave in the wiring box (my old lecturers will be spinning in their graves too.)

The rotary switch is a brilliant idea, in fact I've used it my own design but attached to a circuit by the amazing Peter Baxednall. Peter is responsible for this and the pre-digital tone (bass and treble) controls that are still in use today. I think it's discussed earlier in the thread but I'll go over it again when I publish the final design. Since I'm moving to a phantom powered "universal" head amplifier I need that bit working before I can produce a PCB.

The question still eats at me if I should do full through-hole (as we have here) or partial SMD. I'm inclined toward SMD although the boards work out slightly more costly (and there's a minimum of 2 full-built with five boards at JLC) once you add in the through-hole parts it makes the board larger (significantly so) and the cost more than the SMD parts do. I think I might have just answered my own question.

It's not the switch that's the issue, it's the cable and your sample gives a really good look at what's going on:

If you're not familiar, this is a Fourier series or a Fourier analysis. What we're looking for here are peaks in the wave. Now you'll notice right over on the left there is a big peak to start with. That's mains pickup at 60Hz in this view and the chart starts to go off the top at the very left - that's probably some left-over DC. So I'll take a stab and say you're not in the UK because we use 50 Hz. Of course I could be completely wrong.

Now change the size of the "window" and you get this:

See there's a fundamental at 1Khz and just look at those harmonics - each peak notice appears on a frequency multiple of the fundamental - there are odd and even harmonics which is usually characteristic of a sawtooth wave ("instant" rise time and sloped fall off) but looking closer I can see the signal riding on some DC.

Now it's obviously not as bad as this, I had to amplify that by a factor of about 40x to see what's going on. This DC issue is mentioned by THAT corp. and the the solution is to add a very large capacitor in series with the gain resistor such that its reactance (resistance to AC) is low at the highest gain and at frequencies below hearing. The highest gain (60dB) requires something like a 5 ohm resistor from memory and the capacitor is going to be >=6,800 uF according to THAT.

The other thing THAT mention (and to be fair this applies to all instrumentation amplifiers) is that you shouldn't really have the gain resistor on the panel because it's in a *very* sensitive part of the signal path.

I haven't built an original version per Matt's design as I already had something that did the job but I really need to test this stuff on my bench.

(Of course the cat picks now to start using my hand as a scratching post so I'm working with one finger hunting and pecking.)

The PCB looks very good, is that your own design or someone else? Excellent work.

TL;DR

The issue looks like noise coupling onto the wire connecting the gain resistor to the rotary switch. If you can run both wires through some de-soldering braid that would work. Alternatively some small-signal stereo cable will work. The idea is to keep the wires as short as practical up to the board but you solder a trailing wire to the screen (in your case probably at the board end) and fasten that to the grounding point. This should produce small faraday cage and may even work without the ground (I haven't tried but the theory sounds reasonable.)

More of a concern is where that 1 KHz whine is coming from. It looks more like a small square wave than the Fourier suggests and it could be a couple of square waves at different fundamentals. I ran into an issue with one of my early designs (I'm from an early analogue world, so well past my sell-by date) which used copper fill that effectively radiated all the crud the ground was picking up all over the board. 

The issue with such sensitive circuits is they pick up more crud than my fleece in a wood shop. OK, perhaps not an ideal metaphor but you get the idea I'm sure.

Hopefully (fingers, toes, eyes, etc. crossed) using a screened/shielded cable will fix it though. Come to think of it, twisting the wires together might help (still needs a screen if you go this way) and the wires should be the same length.

If that doesn't fix it, we'll take another look but that's the most obvious thing I can see.

Screenshots from Audacity, which is free and open source - useful even if you do have an oscilloscope and doubly so if you don't! 🙂

Marc

Hello:) 

i haven’t been here for more than a year.. I once read that there will be an updated version of the mic in 2023, but as far as I know, there is no video. I still really want to build this mic, and already started more than a year ago, but that didn’t quite work out, because I couldn’t get the right parts. Did someone post an updated guide in this forum by any chance??

thank you!!

Here's a non working prototype (it would have worked if I'd paid more attention to fixing something I did months ago... but hey ho.)

I'm close to achieving the dream of a capsule adaptor that can work from 5V all the way to P48 with very little re-configuration, just not quite there yet. 

Evicting the last couple of gremlins means building a prototype on stripboard.

Curiously (and I was today years old when this one sunk in) I finally realised why P48 (other similar systems) use two POSITIVE supplies and a ground. It's one of those "can't see the wood for the trees moments, at least that's my excuse and I intend to stick to it.)

It's genius really.

The mic is powered by two (effectively) identical impedances - that's the reactance of the cable run between the impedance converter (Varee) and the input stage of the THAT1512 in Matt's original.

It seems "obvious" that (especially if we have split supplies as is typical in audio) that each wire would carry a different voltage. Indeed in Matt's design, that's exactly what he did - using the front end of a classic "Schoeps" amp used in pretty much every cheap Chinese mic like the Scott Helmke version here.

Now I should stress this is not a criticism, the JFET in this case forms a part of a voltage controlled, voltage divider formed from two 2K2 resistors (left hand side of this schematic).

The voltage at the JFET's gate causes the voltage at the Drain terminal to move in the opposite direction to the voltage at the Source, splitting the signal into two equal signals that are in opposite phase.

Powering this circuit is fiendishly clever and requires more explanation than I have time for today, but the JFET is powered via a regulated supply using a simple Zener regulator and each output "leg" is driven by PNP transistors using the 6K8 resistors in the P48 circuit (at the pre-amp stage) as emitter loads, effectively as emitter followers in fact. The Zener/resistor combination isolates the signal wiggle imposed on the power supply lines from feeding back too the JFET and causing distortion.

I believe Shoeps was first to demonstrate this (but if not, it's as eponymous as the word Hoover is to vacuum cleaners; sorry James Dyson. Prior to that the circuit required a transformer at each end to feed the DC in and allow the AC (signal to pass). Again, this is something for another day but it's darned clever.

A did make up an early Varee with this precise circuit, which drives both "legs" and drives pretty much any P48 compatible equipment until I fell over the final internal design by accident. 

So what's the problem?

Sure I could have saved myself a world of hurt sticking with the tried and tested method but that wouldn't work with lower supply voltages down as low as 5V from USB but that limits us to professional grade pre-amps and that (potentially) means fairly expensive power supplies. This 50 ish volt standard seems to derive from the very early condenser mics (and still the best) that use an external "bias" voltage - sometimes as high as 80 or 90 volts. Electrets are made with a special material that doesn't lose its charge for (it's claimed) a century or more. So if you want to leave a microphone that'll outlive your grandchildren and still work, you probably need an unbiased capsule.

Anyway, Chinese mics and Schoeps. 

When it demonstrated the fully transistorised circuit similar to those above, PNP transistors were expensive and poorly specified compared to the more ubiquitous NPN designs. Something that's no longer really an issue thankfully.

This meant that (compared to the much more expensive) transformer powered versions there was a transistor distortion that some people (claimed) to be able to hear. I don't know as my hearing isn't that acute, but it can certainly been seen on proper test gear.

But it was cheap, simple and reliable - so the world and his uncle started making them.

But that's no where the story ends.

JFETs, while falling out of favour compared to bipolar and MOS transistors, became the preserve of high-impedance converters for devices like condenser mics and some medical transducers, for example.

Quality improved and manufacturers started to be able to improve performance far beyond the simple biasing scheme offered by Schoeps.

The *really* high-end mics are individually tuned, that is the JFET channel (drain-to-source) current is selected to be as close to the manufacturer's ideal as possible.

For the LSK170A (my preferred device here) that's 2 mA, but devices do vary - the LSK170 comes in no less than four different versions, but the A and B will operate comfortably here. In fact, many JFETs can be dropped in here, including (according to reports) some "switching" JFETs too. I think the J301 was mentioned in one discussion for example.

And there's the snag with the split 4k4 impedance. At 12V (which is typical for these circuits) total FET current, and it's the current that matters, is 2.7 mA and that applies irrespective of what FET you drop in there.

You don't have a lot of current to waste either - for all practical purposes P48 taps out at about 10 mA by which time your supply voltage has dropped to about 14 volts (due to the 2 x 6k8 resistors limiting the current out to the mic. These resistors are in parallel giving us:

10 mA * (6,800/2) = 34 v dropped across the limiting resistors.

The fascinating solution I chanced across uses the JFET as a Source Follower but as part of a feedback loop that controls the FET current by what amounts to a constant current source by raising the voltage at the Source with a low-impedance, class AB power amplifier output stage.

A single resistor programs the FET current (this can be variable but not on such a tiny board without resorting to very fragile & unreliable trimmers). Configured as a source follower like this linearises the FET dramatically thus reduces distortion.

If your brain is waving a white flag at this point, don't worry. I promise to do a full breakdown of everything when I have a device you can build at home.

I don't really want to do "two" or more variations of this unless I have to because a "universal" stage is going work out cheaper for everyone.

(I'll probably do one just to carry a JFET as I've done previously so you can convert any large diaphragm electret mic into a simple microphone. But the larger PCB is more versatile as it fits most 34mm and 25mm microphones. (I've got to see how it fits inside another classic, donor body, the MXL990 eventually).

The genius of P48 Power

Ok so this one flew right over my head making the noise like a horny bluebottle which shows just how deaf I am!

Many of you know that the beauty of "balanced" lines (again, per Matt's original) is they are almost immune to electrical interference. We're talking cable runs of 10s of meters without care, as they just get thrown around the studio where stray electrical noise is like a storm!

This is why we need two opposite "phases" (signals running a mirror image of each other). When those two phases are combined at the pre-amp, one is flipped at the the result is that the noise gets cancelled out. (That's why the impedance matching matters so much - so the noise pickup is equal.)

But what about that bluebottle?

Imagine this, you have 10 mV of rippling hum and even noise running up that "regulated" 48v supply. 

That ends up at the pre-amp too - BUT since it's driven by the same power supply any crap coming from your 48v regulated supply *also* gets cancelled out by the differential stage at the pre-amp with the rest of the noise.

Now that's not to say you can skimp on the regulation of course. Zener diodes (for example) are a only a perfect regulator in an ideal world so any ripple that gets to that has a chance of creeping onto your signal causing distortion and other unwanted artefacts. I'm using a TL432 IC regulator so I can get the voltage way down below what Zeners can manage (around 3 v) but the same applies.

The output stage is not unlike Schoeps original and uses a PNP transistor to "tug" (my word, modulate is more accurate) the supply which alters the drop across the 6K8 resistors.

This is a gremlin I clumsily stepped on like garden rake though.

Checking the specs of professional mics and you'll see the output impedance is at most 600 ohms and is often as little as 50 ohms in some cases.

Sounds reasonable doesn't it.

Until you strap a 50 ohm resistor between P48 and ground that is (even with a transistor controlling it). And there goes the farm. Each 50 ohm resistor drags the P48 voltage line to less than half a volt when the transistor turns on.

What works well in Schoeps (which is a different topology) won't work as well here, darn it. And this is where my sloppy simulation reared up and bit a chunk out of me.

TL;DR

I've ordered a bunch of through-hole components to test my theory on real parts to fix any unforeseen problems. Should only be another couple of weeks but I promise it should be worth the wait.

I know because I've heard it. 

Hi,

I made some progress so far. I started doing the mic casing. The first step was to make a first iteration fast enough to check that I can have an exploitable sound before buying remaining materials.

Here is the current result (of course it is in french):

This is not the best sound yet, but I did not finish to wrap every cables correctly. All the electronic is wrapped together in a metal case. The junction between the capsule's enameled cable and the recycled cable is not shielded, so I had to wrap everything in some brass mesh otherwise the mic would get way too much noise (cf first photo). Maybe that can explain part of the noise in my test record.

I saw in your previous post that I should also shield the "gain" cables, so this may be my next step.

Also, I'll restart the capsule case from scratch, I made it too long, so the noise reduction does not work well (at least, I hope this is the reason).

Now I should buy all the reaming parts (brass rods, tubes etc) and create the final case. I have some questions about this:

On the official video, at 9:24. I don't understand how it is possible to solder the rod and the tube together. Should I use classic soldering tin wire? How it is supposed to enter that small gap?

Also I never worked with brass before. Do you have any hint on what kind of polish to buy? Same question about sand paper, and/or fabrics used to smooth the surfaces and angles.

[ quote data-userid="1596" data-postid="4530"]

Hi,

I made some progress so far. I started doing the mic casing. The first step was to make a first iteration fast enough to check that I can have an exploitable sound before buying remaining materials.

You’re way ahead of me on that one. If I’d managed to do that in the first place I could have saved myself a world of pain but it’s been great learning new stuff. 🙂

Here’s a picture of the only extant prototype of the new (in production as of today) of the new Varee which does *appear* to work on the bench (my first SMD versions don’t because I overlooked the *obvious* problem that was a holdover from a Schoepes design. This is entirely my fault and I lost the bet I had with myself. LOL

I’ve spun that into two different forms. One for 48v unbiased mics and one for 5-15v supplies which is the one that’s off for prototyping. It should offer improved noise rejection at the capsule end than is easily achieved with Matt’s solution – which, ironically is how I got here! I simply lack the skills to bend the material. I'll get to that though.

Posted by: @ldoppea

↑

Here is the current result (of course it is in french):

This is not the best sound yet, but I did not finish to wrap every cables correctly. All the electronic is wrapped together in a metal case. The junction between the capsule's enameled cable and the recycled cable is not shielded, so I had to wrap everything in some brass mesh otherwise the mic would get way too much noise (cf first photo). Maybe that can explain part of the noise in my test record.

I can’t really *hear* a lot of noise in that recording but my hearing isn’t what it used to be. I have to rely on a combination of simulation (initially) and test gear (prototypes) to see how things are likely to perform. Considering the amount of exposed cable you had, that sounded pretty good so far. A nice long sample with the mic muffled is better and then we can look at that in Audacity (which is a passable oscilloscope for this use).

Posted by: @ldoppea

↑

I saw in your previous post that I should also shield the "gain" cables, so this may be my next step.

Yeah, the gain resistors are in a very sensitive part of a (typically) high gain circuit.

This isn’t a “classic” instrumentation amplifier design but you might be able to see what's going on her in the following. This is a drawing (not a simulation) but I find it quicker to use LTSpice when I just need to knock up a quick schematic of that's clearer than with all those protection diodes, etc in the way.

This is a variation on the class "long-tailed pair" differential stage that's found in many high-gain amplifiers. In the simplest case the two current sources (the circles in the emitter circuits) are replaced with a single resistor controlling both "legs.

The transistors are biased into their linear region using the current sources (the symbol in LTSpice is slightly different for practical reasons which leaves the bases open to receive a DC voltage. 

You can probably see from this that the gain resistor is connected between the emitters and without going into the details, the smaller that resistor is, the greater the overall gain. This appears a more convenient way to control gain vs. classic 3-amp design. 

BUT note this is also part the signal path - any noise on pins 1 and 8 (RG1 and RG2) will appear at your output. This is one of those "you probably shouldn't do" things. If you can guarantee low noise inside the case, it's fine; if not, you're attaching a huge radio aerial just as easily as if you didn't screen the microphone capsule.

Posted by: @ldoppea

↑

Also, I'll restart the capsule case from scratch, I made it too long, so the noise reduction does not work well (at least, I hope this is the reason).

Noise can come from a number of places (depends which noise you mean - there is mains hum (50/60Hz) and wideband white noise ("hiss") and then there's the real PITA - switcher noise. I'm literally considering two separate NMA0515s at the moment for this niggle alone.

Now I don't know what sort of crud comes out of Matt's computer 🙂 some of mine make noise like an unfiltered effluent outlet. And of course that gets on the rails and we're off.

There are a couple of more advanced techniques to combat some or all of this. But something I did notice when I bothered to pay attention (I often have many things in the air at once and that's not good) is the isolated ground.

The NMA05..., 09 and so on power converters provide an output that is 100% isolated from the voltage at the input (up to some 1000s of volts, nothing is immune).

Phantom power has the advantage that any noise picked up on one signal line is also picked up on the other at the same level, both "referenced" - and this is where it gets a little weird - to the 0V line. The crucial point for P48 phantom, etc is this:

Differential (pair) signals are "self referenced" so they don't need what we might think of as a 0V or GND connection. 

As a matter of interest for anyone unfamiliar, all USB data is transmitted via differential pairs and received by differential receivers for the same reason. 

Now if we use the USB 0V line as reference throughout the design that defeats the point of having the isolation. It doesn't hurt of course: but if we use that part of the circuit and call it our "high impedance, noise sensitive part" and only send the differential signal to a separately powered THAT1512 we are isolating the whole power NMA0515 and capsule from everything else.

This might be a decent way of getting the the benefit of a shielded and screened cable. Really fancy equipment puts returns some of the signal to the outer (aluminium RF shield). This requires rather expensive cable and the THAT doesn't provide that signalling anyway, but I include this for information.

It's something I'll experiment with when I have at least a powered capsule working to Matt's standards.

Posted by: @ldoppea

↑

Now I should buy all the reaming parts (brass rods, tubes etc) and create the final case. I have some questions about this:

On the official video, at 9:24. I don't understand how it is possible to solder the rod and the tube together. Should I use classic soldering tin wire? How it is supposed to enter that small gap?

Also I never worked with brass before. Do you have any hint on what kind of polish to buy? Same question about sand paper, and/or fabrics used to smooth the surfaces and angles.

Brass is a beautiful alloy. Give it a good polish and it's tough to distinguish from gold (although it tarnishes, unlike gold).

The thing with brass is that it takes a while to corrode (presumably due to the large copper content). That essentially means it takes solder just like a clean piece of copper does. It doesn't require a lot of heat. It's possible to use a large soldering iron but a butane powered chef's brûlée torch is perfect.

Brass should take any solder wire but IF and ONLY IF you have decent extraction facilities you can use leaded solder as that flows much better and generally cools slower so you don't get that horrible crystal structure that makes the joints look dull.

I can't stress this enough. Don't use leaded (Pb) solder unless you can guarantee you're not going to breathe the fumes. I've (really) signed my own destiny from decades of using the stuff and I know my likely future and I've lost a lot of brain cells as a result.

As for polish, your options are likely limited to what you can get locally. Brass is very soft (zinc and lead are both soft metals) so you won't need anything harsh. The strongest abrasive I would use is a "soap filled pad" which is steel wool with a little soap in it. We used to use them as pan scrubbers years ago.

But anything for polishing stainless steel, "Barman Keeper's Friend" do several grades which are good. There are also products like "Brasso" infused wads but those things are for removing the chemical dullness.

I've even used toothpaste but if you really want to get into the weeds there are things like Jeweller's Rouge (which is, I suspect, iron oxide) and another one in a green form that I forget the name of.

Posted by: @marcdraco

↑

A nice long sample with the mic muffled is better and then we can look at that in Audacity (which is a passable oscilloscope for this use).

I'm not sure what does "muffled" mean, but here is a sample with no voice. My room is not entirely silent, but the louder sound in the room is a wall clock that does big TIC sounds. The fact that I don't hear them in the record makes me think that the sound sound recorded by the mic is internal noise, isn't it?

But it is weird that the noise seems to fade on the second half. The mic did not move, and there was no change in the room.

Posted by: @marcdraco

↑

That essentially means it takes solder just like a clean piece of copper does. It doesn't require a lot of heat. It's possible to use a large soldering iron but a butane powered chef's brûlée torch is perfect.

 

But how does the brass get inside of the very small gap between the tube and the rod? Is this something like capillarity?

It goes from this

To this a few seconds later in the video

A a few seconds later, after some cuts

Looks like the soldering went deep enough so both parts appear as one single block over 2cm long.

I don't understand how this is made as liquid solder wire seems not liquid enough to travel though all of this (I did not try yet)

Posted by: @ldoppea

↑

Posted by: @ldoppea

↑

 

Also, I'll restart the capsule case from scratch, I made it too long, so the noise reduction does not work well (at least, I hope this is the reason).

 

 

Noise can come from a number of places (depends which noise you mean - there is mains hum (50/60Hz) and wideband white noise ("hiss") and then there's the real PITA - switcher noise. I'm literally considering two separate NMA0515s at the moment for this niggle alone.

Sorry I should use different words for different noise sources. Here, by "noise reduction" I was talking about reducing real noise around the mic like keyboard sounds.

Posted by: @marcdraco

↑

Here’s a picture of the only extant prototype of the new (in production as of today) of the new Varee which does *appear* to work on the bench (my first SMD versions don’t because I overlooked the *obvious* problem that was a holdover from a Schoepes design. This is entirely my fault and I lost the bet I had with myself. LOL

Did you forgot to add the picture? I don't see any

Posted by: @marcdraco

↑

You’re way ahead of me on that one. If I’d managed to do that in the first place I could have saved myself a world of pain but it’s been great learning new stuff. 🙂

 

I started the project doing this way because I knew I would struggle resolving potential electronic problems and I was not sure if I would find someone to help me on this. Hopefully you were here 😊

Hello. Tell. I took the project of one of the participants @lmarzen as a basis. The project was downloaded from https://github.com/axydavid/diy-microphone/tree/main. After I assembled it, such a problem appeared that 100Om resistors burned out (+15.-15 line). What to do, what to do?

Oh my .... I did forget that picture - and I apologise for my utter dog's breakfast. (I was trained at the premier electronics specialist and those poor lecturers must be spinning in their graves.)

Problem solving is a great way to learn new skills - that's working out for both of us! I've never diagnosed a faulty IC over the Internet before, so that was a new one. Of course, having the THAT schematic and your excellent tests to hand, I eventually made sense of what was going on.

SOLDER

OK, so let's talk about the amazing thing with solder. It's not just a metal there's also a lubricant in there called flux - the name comes from "flow" - more specifically from 14th Century Latin for something like goes everywhere. If you've ever spilled milk you've seen what the Romans called it. It's also word in your native language and it means the same there too.

But the alloy of tin and lead (the solder metal) is also fascinating stuff - as you'll see here the stuff melts at very low temperature and the hotter it gets, the more liquid it becomes and its surface tension changes.

Effect of Temperature on Wetting Angle (sjsu.edu)

In essence then, yes, solder is pulled up the thin gap mostly by a capillary action. You'll note from "wetting" - when you coat a part in a thin later - you can get a better joint.

NOISE

You were correct in calling it noise. Any unwanted signal that reaches the recording equipment is noise. I assumed you were referring to electrical noise such as hums, hisses, crackles, pops, whines and such.

External noises like keyboard clicking, cat's tearing your carpet up, the other half complaining you didn't do the dishes... those are more difficult.

Designers refer to a mic's design with a view from above to show where it is most and least sensitive. There are various patterns drawn on a polar (circular) graph so that the most sensitive "lobe" is at the top of the chart. All audio microphones (with the notable exception of ribbon mics) pick up sound from everywhere - the question is more of how much. Most graphs bear some resemblance to a classic heart shape - this open air design is a wide cardioid response with *some* "off axis" sound leaking in.

Off axis is a term to mean an imaginary line going straight from our face (or the sound source) directly to the microphone' Most microphones are least sensitive along this line from behind (where the two lobes of the heart connect. You'll find lots of discussion and pictures on the Interwebs.

The problem is that we tend to think of air as being an empty space - even though most of us know, "In space, no one can hear you scream." As the 1970s poster for Alien put it. Of course, everyone does here you scream through the microphone in your headset. But I joke.

You're probably aware that air is just a lot of gases all jumbled up and flying around randomly. And that's what causes the effect we know as pressure (in a container). On the macro scale air pressure is the physical mass of several miles of gas being held down by gravity.

What's not as obvious perhaps is that sound travels very much as light does - as waves. The air doesn't move very far in real terms, it's almost like a three-dimensional Newton's cradle or the tide washing up on the shore. Note how we see a mass of water moving in and out, not the bits that make it up. (This is why even those sub-1mm diameter holes can allow enough sound energy to pass on MEMS microphones.

Another strange effect we rarely notice is that we can't tell where a sound came from if the source is in another room. We might infer where it comes from by the sound (running water, a toilet flushing, the television). But an unfamiliar sound... that could come from anywhere. This effect is why professionals use booths or studios with strict noise controls in place. The nearest we get to "true" directionality are shotgun (supercardioid and hypercardioid) microphones but they have issues all of their own.

If you need a miniature studio at home that doesn't echo but also keeps the noise out, the poor man's (very cost effective) method is to surround all four walls and the ceiling with some old blankets or quilts. What matters is that it has lots of tiny, lightweight fibres. A down one would probably work very well. Sound waves hit the fibres and which absorb some of the energy and pretty soon the sound is "muffled" (in reference to your earlier question). 

A muffled sound has lost most or all of its high frequency information due to electronic filtering (a low-pass filter) or through via an absorbent material.

Note that to work, the fibres are lightweight. The diaphragm on your mic is also, I would say paper thin, but it's a lot thinner than that. It's incredibly light and that means it's easy to move. Easy to move means it "hears" things that you and I can't in terms of absolute volume and (in my case due to aging) the frequency response too.

Dynamic moving coil (*moving magnet?) microphones are a better choice for pure vocals, and especially singing, because of the sheer air pressure created the human voice is capable of overcoming the mass of a coil easily whereas off-axis keyboard clatter just doesn't have the energy. Off-axis rejection (actually lower sensitivity) is better and typically these mics need amplifying another 20dB (10x) for similar results at the mixer.

Capacitor mics with their incredible sensitivity and wide frequency response sound like a fool's errand for vocal use then, right? So why do we use them as here. Wouldn't it make more sense to just use a capsule from a vocal mic like the Sure SM58?

Simple answer is no. More complex answer is "it depends".

The finest microphone capsules in the world (in wide use) are all capacitor microphones. Nothing comes close to the sound they deliver in terms of distortion, frequency response and sensitivity. I would argue that ribbon microphones are superior in these regard but they are also hugely expensive and incredibly fragile. Ribbon mics also have a very curious response and a real dead zone off axis which is why they were used in the first talking pictures - as the early cameras rattled like box of spare bolts caught in a whirlwind. We can't hear them because they were arranged so that the mics always faced sideways to the camera.

In terms of quality and range of reproduction these ultra-fine materials are unbeatable (the lower the mass the better) but as larger diaphragms are also preferable, ribbons might just win out. The only way I could conceive of improving that is with light but I don't know if anyone has cracked that problem yet. At least not outside the realms of science fiction.

TL;DR

Won't blame anyone for jumping over that treatise on the benefits of microphone diaphragms. 

Capacitor capsules are tough, easy to make (tens of millions of electret mics are made every year) and they last. This is what makes them popular. The JLI2555 we're using here is an intermediate step between a pure condenser microphone (which needs a bias voltage of around 50 - 100 volts) and those little 10mm capsules that they almost give away on eBay because they are electret - pre-charged capacitors - with a JFET mounted directly inside the capsule. (There has been some talk of putting better quality JFET or even ASC chips inside these little blighters but I doubt there's a market for that as the pros have mostly moved to MEMS which are smaller still.)

So with the JLI2555, although I call it a bridge capsule, you get the best of both worlds. A pre-charged capacitor capsule so you don't have all that fiddling around with zapped fingers or even lips BUT that you get to chose the impedance convertor. This is what makes the difference between a "meh" design that works out of the box or push the design harder with better quality components.

I haven't seen the inside of the professional mic Matt found the JLI2555 in (the one in the video) but my guess is, if it's worth the sort of money they are demanding, it had be better than simple JFET source follower (or worse) a common source amplifier. A common source amp is the one they use in pretty much all cheap electret capsule. The gate goes to one side of the capacitor the source is connected to the case. 

When you put the 2k2 (typical value) on the main terminal and connect the ground to the other side that literally creates a little common source amplifier and the output that goes to your computer's digitiser is tapped off directly through a capacitor of a few uFs. Very simple with a performance chart that looks like a line of me on my way home from the pub after a couple of pints too many.

That's a lyrical way of saying it distorts it horribly.

The Schoepes method using the JFET as a phase splitter is far better (that's what Matt's original did) and also generates the two phases required for a balanced feed, ensuring more signal at lower noise, so win-win. But there is still some distortion from the JFET's non-linear response. It's only the high-impedance, low current noise from the JFET that keeps then in production, but there are very few decent ones remaining.

Hence I nicked the most powerful circuit I could find and adopted it for our version of phantom power at just 15 volts. It seems to work down to 5 volts but I'll believe that when I see it! 

I can't promise, but I suspect this circuit is capable of producing an excellent response even with some lower cost JFETs. It would be interesting to test this in future with the 2SK208 or even the J103. A tethered USB mic (i.e. one that is just plugged into the USB) can benefit from much higher current than the P48/P15 design I've done so that's another area where it could benefit although I honestly doubt it would change things audibly. 

The thing I'm really itching to measure properly is the screening built into the Varee PCB.

SILENCE?

The way your recording fades off sounds (to me) as if you have a very high-frequency noise on there that's driving your THAT1512 so hard it's unable to keep up. Probably need to look at that on an oscilloscope as it won't show up here due to the design of the A/D and filtering elsewhere. That noise is wideband "white" noise, mostly Johnson noise I would imagine. There will always be "something" (and the more we amplify the more that something becomes noticeable.

I didn't see a lot from the THAT1512 in my quick tests, so it's possible that white noise is coming from the digitiser. Not all A/Ds are as quiet as others. Some are dreadful.

Oh, yes. Beautiful work again. I wish I had that skill!

@g3nts hey, i noticed that you mentioned there is an error in this gerber file that you uploaded. But i have already finished the circuiting. Can you check this out and tell me what is the mistake so that i can fix it. It would be a great help and would save a lot of trouble 🙌🙏

@acidiconionas I just got there and Im stuck here too, what did you do eventually?

UPDATE: 04/04/2024

What has two thumbs and a red face? 👍😳 👍 

That's right but I just thought I'd just trust to luck and discovered that wasn't one of my better moves (I didn't have any TL431 regulators so I just assumed it worked). I know better but you know... I'm easily bored.

The TL431 is an integrated circuit voltage regulator that separates the supply/signal side of the circuit where the voltage varies by as much as several 100 mV from the FET pickup side which needs a rock solid supply because any noise on that causes distortion.

Lack of room on the board (this was originally 25 mm diameter) meant I kept it simple and (of course) I went too simple. So to my repeated shame, I've ordered some TO92 TL431s to make sure that the pass regulator works before I order the next lot.

Assuming, that is, I can find the room for that. 🙂

Hi, sorry long time no see,

I finally received all the brass part a few days ago and I did great progress this weekend 🙂 

A few weeks ago I tried to make the capsule casing like in the vidéo: cutting a flat sheet of brass and shaping it as a circle (I used a sheet of brass I got from and old Aliexpress order).

This went pretty badly, brass was too hard to be easily shaped, but not enough to resist the drilling process that bended it badly.

Hopefully, when searching for a french provider for all the brass parts, I found they did propose some 30mm tubes which was the perfect size for the mic.

This made the process wayyy more easier and way cleaner -> the result is a perfect circle

Here are the photos of the process. The wood part was to support the brass during drilling process, but I found later that this tube brass was solid enough so it was not necessary.

Here is the result (in the middle) compared to the first try (on the right), and the second try (on the left)

 Sadly I did not success to correctly align holes, even with the paper guide I used. But the result is good enough, restarting a fourth time would be pure waste of material.

(Seems like I reached some limits on attachments, I cannot add new ones for now. I don't know when I'll be able to post new ones. I'll try again tomorrow)

Regarding the tubes, I did try to solder them without and with some flux, and you were right, this thing does magic.

I succeeded to to the mic's arm. I'll post the photo soon.

Next steps will be to do the fixations, the case and finally the capsule support.

Posted by: @marcdraco

↑

Assuming, that is, I can find the room for that. 🙂

So sad those components cannot be compressed, life would be so easier 🙃 

Good news!

The latest batch of Varee boards arrived at around 9:30 this morning and they finally work as I had intended. (My sigh of relief scared the cat and when I ran the first test and played it back it knocked my socks clean off.)

I don't know if Matt (@diyperks) wants to do a video yet... but I've finally got something this community can stick it to Rode with.  (It should do as I cribbed the basic JFET loop from a presentation by Doug Ford their former chief designer. cheers Doug. Probably just as well he doesn't work there now or he wouldn't be much longer. 🤣 🤣 🤣)

P48 Phantom

Fully 48v phantom power compatible with noise levels so low I was unable to hear them on a Focusrite Scarlett Solo. Which means this head is fully compatible with all professional recording equipment if you don't want to make the base. (Personally, I think it looks amazing as a single unit, I got the Scarlett to make sure they worked before I tried to create that part.) My next step is to make a switched mode boost to get us the necessary 48V - with up to 15 mA. In reality it's more like 12V at 10 mA when you take account of the current limiting resistors though.

I found a booster on Amazon that claims to do 5V (USB C) to 48V but I'm suspicious so it's sat on the bench some of with my shattered nerves. What makes this doubly interesting/bloody difficult/terrifying is that I can't find a switcher chip capable of such a jump. I know it's doable (the Scarlett does) and I've designed the basics of something but this is all going to need proper prototypes. At least I can do this bit on the bench so turn around of prototypes is in terms of a few days (time allowing) not weeks - currently my turnaround is about 10-12 days from making a change to being able to test it as with these. 

How does it sound?

Sensitivity and quality are hard to judge without a proper screen so I'll try to mount one in one of those Chinese mic bodies as a screen. This is really now moving right into Matt's skills (cough, splutter, choke) and but it works with the 34 mm and 25 mm electret heads, but it needs a screen. I busted my JLI255 by soldering and desoldering it just too often so I've only tested it with a larger capsule. They seem to pick up mains hum like a huge radio tower so the head assembly will need some screening.

I THINK (and this is down to someone with more skill than I have in machining) that if the capsule is screened and the screen attached to the common at the rear, the device will operate (essentially) open backed as there is an internal screen for all the electronics sandwiched in the actual PCB.

The working prototype looks a bit odd (pictures to follow) because I've used professional grade cable to hook it up. Fine wires per Matt's original will look far better and sound just as good but you can run many yards/metres of this stuff without picking up any excess noise.

The LED does glow in use and looks really cool but (of course) it's around the back in normal use.

How much is it?

Small quantity price works out at about a £5 per board but there's a minimum order from most PCB shops so it's difficult to price it out in quantity. I hope someone on eBay (or even the Chinese) will pick this up and run off some low-cost ones. It's a trivial matter to replace the through-hole LSK170 with a slightly cheaper, if slightly poorer, JFET with an SMD 2SK208. A factory soldered pin header would be useful too as this thing is a bit tight (even in the 34 mm variant due to the extra parts like a pass regulator.)

I've done two versions (both should be OK with P48) but one uses a lower set point so it should work with Matt's original pre-amp with some minor modifications. Electrically speaking, Matt used a split power to get the JFET to deliver the differential signal. 

P48 doesn't allow that, so it will be necessary to just feed +15 into each signal/power lead and we'll likely need to change the current limiting resistor but that's about it.

Mounting 

The board has four holes ready to take the suspension (nylon fishing wire might work well) but there's nothing to say you can't just mount it in a handheld unit. These things only weigh a few grammes so it should work with Matt's original hardware (and JLI2555) without issue. The solder braid screen should look amazing and it's possible you could make this thing look like it was floating! You can go full steam punk with this thing.

Here are some 3D renders and the circuit from KiCad, photos to follow when I find my camera... and a sound test when I can find someone who doesn't sound like me. Since I don't play an instrument (except maybe my own trumpet) I'll have to find someone who does. Honesty you don't want to hear me, I've got a face for radio and a voice for the printed word.

Beta Testers

If any of the beta testers would like one, drop me an email and I'll send a complementary one with my gratitude for your help (even if you don't think you helped, I can assure you, everyone did).

I'll write up the circuit description in a separate post later in the "what are you working on" section as it's going to be complicated and I've already filled this thread with umpteen development updates. Details of how this deceptively simple circuit works (and why it's way more complex than Matt's) are beyond the scope of this thread.

Posted by: @ldoppea

↑

Hi, sorry long time no see,

 I finally received all the brass part a few days ago and I did great progress this weekend 🙂  

That looks amazing work. I feel ashamed, part of my training was in basic metalwork, so I should know what I'm doing, but I lack the tools these days and my Aldi pillar drill probably isn't up to the job. Well, it probably is but I'm a crumbly old goofball and I leave things like that to people who have the equipment and skills he says looking roughly in Matt's direction.

I love how this project has inspired so many people (myself included) and I've pushed myself, gone without beer money and learned some interesting new workflows. SMD was the only way to do this and the four layer board made a big difference. The final "head" unit is over engineered as result and it shows. 

I note you're having trouble aligning the holes. Did the drill skid (I see you used a centre punch to mark them)? Drilling into tubular parts is tough at the best of times. I'd have been tempted to use the drill to align them by going all the way through to the other side if that makes sense. The Varee design seems to work with an "open" back (I designed it with that intention) which should hopefully make screening easier. Per usual that would have to be confirmed by someone else.

Starting with sheet brass (and curving it to fit) I would probably drill the holes first. I think that's how Matt did it. But it's a truly beautiful material when it's finished and polished. 

Posted by: @marcdraco

↑

Did the drill skid (I see you used a centre punch to mark them)?

Yes I used a centre punch, but I'm not sure why I failed. I started drilling with a 1.5mm bit and it seems good on the paper sheet I attached to it, but when I removed it I realized the misalignment. Maybe the paper moved a bit when I centre punched it.

Posted by: @marcdraco

↑

I'd have been tempted to use the drill to align them by going all the way through to the other side if that makes sense.

I'm not sure this would be easier. First, because having 2 opposite holes not aligned would not be noticeable, but having 2 coincident holes not aligned would be way more visible. Second, because you'll have to be very precise on how you place the "brass circle". With a circle of 3cm diameter, each error of 1degree would shift the bottom hole by 0.5mm (which is big compared to the 4mm holes I made)

Posted by: @marcdraco

↑

Starting with sheet brass (and curving it to fit) I would probably drill the holes first. I think that's how Matt did it. But it's a truly beautiful material when it's finished and polished. 

This is what I tried for my first attempt. But the brass was too hard to be bent easily. I succeeded to give it a curve shape on the "plain" side, but the "hole" side would not bend between holes, resulting into a polygon.

During my research, I found that there are 2 types of brass, "écroui" and "recuit". Those are french terms and I don't know the english ones. But "écroui" means that the brass has been shaped while being cold and "recuit" means that it has been shaped when hot. My understanding is that first one is harder while the second one can be shaped more easily.

So I suspect that mine is the harder one and that Matt did have the "recuit" version (the way he bends the metal in the video seems to require less effort than I actually had to put on mine)

Posted by: @marcdraco

↑

The latest batch of Varee boards arrived at around 9:30 this morning and they finally work as I had intended. (My sigh of relief scared the cat and when I ran the first test and played it back it knocked my socks clean off.)

 

Congrats!

Posted by: @marcdraco

↑

I THINK (and this is down to someone with more skill than I have in machining) that if the capsule is screened and the screen attached to the common at the rear, the device will operate (essentially) open backed as there is an internal screen for all the electronics sandwiched in the actual PCB

Posted by: @marcdraco

↑

The board has four holes ready to take the suspension (nylon fishing wire might work well) but there's nothing to say you can't just mount it in a handheld unit

I'm not sure to understand, do you mean there is no need for a "faraday" casing around components?

Posted by: @marcdraco

↑

If any of the beta testers would like one, drop me an email and I'll send a complementary one with my gratitude for your help (even if you don't think you helped, I can assure you, everyone did).

Sadly I don't have any phantom compatible equipment, so I would be helpless.

Don't worry about the P48 Varee for now (it should work with Matt's amp too but you'll need to make a couple of very slight changes). I say should since I haven't tried it but I did consider that possibility when I designed it. There's something niggling me that I can hear on the Beyerdynamic headphones (albeit at volume) and that won't do so I'll check my figures and run another batch when I nail it down. That's the "fun" of iterative design.

I did a quick check and sure enough, brass (like stainless steel) is "work hardened". With iron (the only metal I've done this with) the hardness is determined by the crystal structure so heating and cooling at different rates affect how the crystals form. Crystals in metal - crazy eh? Even though I knew this I'd never seen it until I saw some chaps making a fancy knife and talking about this very effect.

So the thing with brass is this - and it's something you and I can exploit. Even if you got the hard stuff. It can be heat softened! When you work it (bend and shape it) the stuff hardens. Weird I know but this has affected production of the Cybertruck and it's going to make it a nightmare if you dent it. Ordinary metal will snap back or can be beaten. Stainless steel work hardens so when it's bent, it stays bent!

I don't have a lot of forming tools myself (I'm really a writer by profession, not that you'd notice) so it looks like it's a case of getting that brass nice and HOT - a chef's blowtorch should be enough for a small piece a blowtorch will be sufficient for even a big bit - I'll say this to protect everyone - let it cool before you touch it! I know you know this but ... someone is bound to expect it to cool like solder (almost instantly) and it doesn't. Brass retains a lot of heat due to the copper I expect.

It's pretty late here, I'm playing catch up with posts, but I'll try that myself with my bits. I have a propane torch that I use for the BBQ - that should get it plenty hot. There are probably videos on YouTube but I'll see if I can repeat the experiment and work harden some. In theory beating it with a hammer will harden it too. Probably need a big hammer though. And and anvil. I don't have either. I think my neighbours would complain. 😉

Since brass is soft (and this is how I was taught for what it's worth) you should mark up directly on the material. That way you can see what you're doing (I hit this issue trying to make a Red Special guitar with my daughter's B/F) The paper shrank - not much - but enough and we ended up with ... a lot of sawdust and a guitar shaped piece of firewood.

Matt does make it look easy though - TBH that's what always attracted me to his channel. The way he recycles and reuses so much stuff and his final work is just beautiful. He doesn't come across all superior either or seem to have a full machine shop at his disposal like some of the DIY channels. 

You can be the smartest guy in the machine shop but if you're only allowed basic tools it's still going to be difficult. (I spent two weeks in my first year filing bits of mild steel flat. I was very, very bad at it.) I did learn a hell of a lot though. I got to use lathes, mills, grinders... all full-scale machines and honestly it was a breeze even then. These days it's nearly all CNC like 3D printing so you don't even have to get your hands dirty.

It's all about budget. Throw enough money at a problem and you'll usually find a solution within the confines of physical laws that is. 

What you need (I think) for ease would be a "mangle" - a roller press actually. You feed the flat sheet in and it bends it for you over a roller so it's consistent. I made something like that with a 3D printer for my ribbon mic project - mine is a corrugator but the same principal applies. 

Thinking out loud - here - I wonder if you could use some suitable wheel bearings? A mate of mine is an retired machinist (we called the fitters and boy these guys were amazing). Very unassuming, I'll ask his wife to ask him what he thinks, she might still be awake painting. Wonderful people in their 80s and still going strong. She used to be a chemist and a teacher but boy that lady is an amazing artist across all sorts of paint and even embroidery. A true genius. She did a book of her work a few years ago (oooh 2009 - quite a few years ago). Embroidered Jewellery: Amazon.co.uk: Sherris, Shirley Anne: 9781906388119: Books

Time I got mine autographed actually. Kill two birds with one stone! 🙂

OK, quick email the some idiot box before bed methinks!

I'll get back to you as soon as I can and you're very welcome to a sample board from the current or a future batch. It performs better than cheap Chinese mics but it's still not as quiet as it should be and like Matt, I'm a bit of a perfectionist. I'd rather blow my pocket money and give everyone something top notch vs. everyone have to keep buying (and swapping out) my iterative designs. This one finally proves - on an SMD board, that mounts direct to the capsule isn't just possible but a real practical solution. It's good enough for podcasting and vocals but not good enough for me. That said I haven't actually tested it with the JLI capsule since that had an unfortunate accident with a screwdriver, de-soldering iron and an idiot in control of them.

Twice.

I snapped the case tabs making it impossible to mount properly.

Mounting the high-quality "pre-amp" board on the head allows more room in the donor bodies to mount a USB digitiser! Varee doesn't have to operate at 48 volts, it should work down to 3 or four actually but probably not in a phantom configuration. That said, if the whole thing is mounted in a donor body, it would be a fair bit less complex.

Anyway, I apologise for digressing. 

I'll nip and ask my pal for his input.

EDIT: He's written to ask what sort of thickness and so on which means it's a virtual certainty he can help. He's just had his 83 birthday, what a lovely guy! I'm going to see them next week so I'll either order some more from Amazon or try to find the piece I mangled into a twisted mess!

I'll update other bits at the end but although I've had something of a breakthrough, let's look at the brass problem.

I don't know if Matt put the length of the strip in there but I calculated it to 79mm (78.5 more accurately or 80mm to allow a bit of slack in.

Ken tells me (it's pouring down here so I can't safely check) that you can soften it by heating a cooling. He said you can rapidly cool it (that hardens iron/steel) or let it cool naturally. It has to be work hardened which will happen to some degree in the next steps. It doesn't need to be hard like a brass instrument though because it's not a mechanical (supporting) part.

You'll need two things a mould and a press. 

Which isn't as bad as it sounds. Ken said find a socket (from a socket and wrench set) or buy one of the peg if need to. A 1/4" drive should be dirt cheap but 1/2" isn't too bad for a single. Remember it's the outside diameter of the socket, not the size of the nut it was designed for. I can't be much help on that but metal rod would be an alternative- although likely rather expensive. Softwoods like pine would likely break doing what I'm about to describe due to uneven stresses and hardwood dowel is expensive.

A local mechanic might help you out if you don't have a suitable "inner ring mould".

Ken suggests we just gently work from the middle outwards and press the strip against the socket a little at a time until it follows the shape of the socket.

The outer ring (press) made me facepalm for not even spotting it - so easy. We call these things jubilee clips here, I'm unsure of the French but this in the thing. You can screw these things quite tight and swishing a piece of well-formed brass against a socket should be enough to bring it into final shape. I won't correct major "Draco-sized errors" but it will add a final bit of shaping.

Quick update on the Varee, I found the problem and went back to the simulation board, there was something that never sat right with me but I figured, well Doug Ford (etc.) and he drew it out all wrong. I've got to try the new one on a protoboard but the component count is reduced somewhat. It's just not really that good for P48 and what I'd done was the best I could. Not good enough. But this is iterative design and lack of training to some degree.

I'm now going to slap my self silly for not realising my mistake in Varee earlier - this extra space might, (emphasis might) give me enough room to fully balance the output - doubling sensitivity while keeping that JFET operating at its most efficient/linear.

UPDATE: 24 hours later and I've managed to crash the PC twice but I'm trying a whole new approach. If they pass JLCs testing I should have a FINAL one back next week. I've also done a design (untested) with an 2SK208 on board. The only advantage is that the JFET is attached (less soldering).

Now I have a busted keyboard thanks to the cat's big feet! Oh the humanity!

I feel more like Elmo every day. Difference is, I'll own up to my mistakes because I am the designer. 😉 But the good thing is almost every time I drop the ball, and I've been bit of a butterfingers recently, I find another one when I go to pick it up, so there's that.

I wonder if NASA are still giving money away? I've got a great idea for a new rocket engine.

Ok, self-fish-whipping aside, I totally forgot to check the Vce of the 2n3904/6 transistor which is a fantastic, low-noise jellybean up to (wait for it) 40 volts. P48 isn't quite enough to create Magic Smoke, but it's more than enough to make the transistors creak a bit. 😶 

Checked every other part for voltage and power handling except the most sensitive of them all. The hardest lessons are the ones that cost us time and money. And looking back over some of my earlier attempts, I should have been more careful to check what was going on rather than committing the cardinal sin of *assuming* that if it works in LTSpice it will work in the real world.

Do I feel like a prat? Sure. But while I could do an Elon Musk on y'all and not even document my assumptions and "Diagnosis: Got it wrong again", I won't. Development (esp. part-time in today's horrible world) can be slow. Things I would *and should* have checked are often overlooked. (I ran one of Doug Ford's designs for example on a single rail, as originally drawn, but when I took a step back, it was clearly not suitable for Phantom power.)  

Posted by: @marcdraco

↑

But while I could do an Elon Musk on y'all and not even document my assumptions and "Diagnosis: Got it wrong again", I won't.

I wouldn't say this is a failure... this is a success at verifying if the theory was true about the fact that this design choice would fail 😅

That's how one of my heroes put it too. “It doesn’t matter how beautiful your theory is; if it doesn’t agree with experiment, it’s wrong.” 

I really should get some better parts - I've been "breadboarding" on SMD which is expensive. 😉

I've got breadboards (the right way to test things) but I find that the cheap components I tend to get in bulk don't make great connections and stripboard can add problems of its own.

I've done a 5v op-amp (no voltage doublers, etc.) variation I just need to validate it works as it does in LTSpice before I run a bunch off. It's much simpler and should perform quite well.  

@perrytheplatypus @acidiconionas Odd, I missed those - they must have been stuck in the mod queue. Weren't my files so I can't be a lot of help - I'm keeping mine private until I've validated everything works: WELL. About half worked to some degree but none were up to snuff and I'm not going to have people spend money on my mistakes. 

I've ordered a small batch of a simple "universal" head-end adapters I made up recently and if that works there should be enough to give a bunch away to builders. They are based on a very early working design that was primarily meant to save soldering a JFET to the capsule with all the potential for mechanical errors that brings with it. I've got an experimental fully realised head solution in there too - 5V USB with a direct output for the digitiser and relatively inexpensive parts.

I can knock out a "working" one (based on Matt's stripboard original quite quickly) so I could always put it on the current order (provided I don't go over my shipping weight). I've got a bunch of much more advanced prototypes but they're not compatible with the original design (they work from +15V to +48V). This isn't ideal if you've bought the parts. Given that I need a break, I'll see about getting some done this afternoon. Would be handy for me to test how well it works. My other attempts have been noisier than expected due interference on the USB, and expect a four layer without lots of copper pour would work better.

HELLO!!! It's me again with my constant delays, almost an entire year and I haven't ordered my parts yet. 

Anyway, I have some quick questions about the breakout board. To my understanding, it just connects data positive, negative, power, and ground all connected to the audio card with also power being split into an XH 2pin. I wonder if I'll have any trouble with the USB breakout Host and non-host thingy. Will something like this breakout work??

https://www.amazon.com/ANMBEST-Connector-Receptacle-Adapter-Support/dp/B091CRLJM2/ref=sr_1_1_sspa?crid=9ISKT1FONVQ3&dib=eyJ2IjoiMSJ9.BLhd61AtY0A9b983TIXzIZMYG1Cw1X0bJn3mReSOQBHgDR1ROLHndU7ZrHpiOLun8pQCEhUK5lP6D6qJoBU5FY62QUREPGD0oQ-e77exQ8A8zL79HvSnnTBZqJIOUbhiKFa0PGfqR_bFjt1wPWDYKolRFvm29C4o_Fj4Hs5_1P92AILpjObv3k9kU1cvHkGgW634Tey_BGye6AlwUQpE_aLFiRGmNG84RGeWIOoQLGc.qcRngMX6xm_C-klIS-RU3Af4XSWV8r_-VF3aiWdB6cQ&dib_tag=se&keywords=10PCS%2BUSB%2B3.1%2BType%2BC%2BConnector%2B24&qid=1714292549&sprefix=10pcs%2Busb%2B3.1%2Btype%2Bc%2Bconnector%2B24%2Caps%2C188&sr=8-1-spons&sp_csd=d2lkZ2V0TmFtZT1zcF9hdGY&th=1  

That's a stupidly long link...(thanks amazon)
For the caps, I heard Electrolytic capacitors have a bad lifespan and aren't the best for audio use, due to fancy current magic voltage ripple stuff that's too advanced for me to understand. I feel I'm getting ahead of myself and the caps I've selected will work just fine. I did find an "audio grade" alternative to the 2200uF cap but something tells me it's just snake oil—I'll post what I've planned and what possible alternative I'm planning to use. https://www.mouser.com/ProductDetail/647-UKL1C222KHD / "audio grade" alt https://www.mouser.com/ProductDetail/Nichicon/UFW1C222MHD?qs=kArNe9LFxXloZwf1aCQpjw%3D%3D

Also don't understand the whole grounding thing for the mic capsule and the circuit in general. 
ALSO also, I don't know what to use as a rotary switch too many diverse options that I don't understand.

Almost forgot about to ask about the transistor I found a cheap Chinese one https://www.ebay.com/itm/392649330332?mkcid=16&mkevt=1&mkrid=711-127632-2357-0&ssspo=Otankvg1Qnq&sssrc=4429486&ssuid=&var=&widget_ver=artemis&media=COPY

only concern is the quality and it having Mot in the name like in motfet and I think it’s supposed to be jfet. Also found 2N4416A spooky A with “professional” in the title but have no way to decipher spec sheet.

This should be my final set of questions till I finally order the stuff and eventually have more.

BREAKOUT

Yup, that breakout (and don't you just hate Amazon links? Me too.

If you can hang on a couple of weeks (maybe not that long) I've got a new set of "beta" boards in production. The more mature ones make this job a lot easier.

JFET

The transistor is and isn't critical. You need one with a low Vgs and low noise. Most JFETs have very low current noise but they do exhibit other problems like capacitance at the gate and more. I will have to check with 2N3819 (a jellybean JFET) but at a guess it should work OK.

You're right that a MOSFET won't work in this configuration, they're not really suitable for the application anyway. MOSFETs are mainly used as very fast (and mean VERY fast) switches that can handle oodles of current. Everything that JFETs can't do.

However, a JFET designed for this sort of job is far better - the 2SK208 from Toshiba and the LSK170a are excellent parts. The 170 is widely used in higher end mics. I don't know what the top of the line guys use (they don't want me talking their stuff to bits).

The parts you've picked are similar to the one Matt used and they are primarily intended for RF work (radio) in mixers and oscillators where noise is less of an issue than in audio. The professional one you mention has a noise of 4nV/root-Hz, the LSK170 by comparison measures 0.9nV/root Hz, which is more than 4x better.

One of the capsule heads I'm waiting to get back can be ordered in a bunch of different configurations: including with a 2SK208 mounted ready to solder. All you add is some wires and solder the capsule in place. The gotcha is that if you're not confident with small parts and these ARE small, you'll have trouble with shorting some of the other optional parts.

Now you might be (reasonably) thinking why are the other ones he's talking about only have three pins and come in plastic cases? (Epoxy if I'm being pedantic).

Well, the extra pin is the screen which creates a shield around the JFET and the little leads that connect it to the outside world. We simply don't need it because the whole gubbins is enclosed in a shield to stop noise at the capsule. I've tried running mine experimentally without a shield and it's unbearable. Even taking measurements is difficult.

The shield on the metal cans is intended to screen the JFET alone when it's in a circuit and in a RF design there's going to be EMI all over the shop.

LSK170s are reasonably cost effective, perform excellently and are in production so you don't need to worry about supplies.

I tried a couple of others (mainly dual configurations) but it was really a case of just chasing a rainbow. The noise levels added by the digitiser easily swamp the improvement we get from a dual FET. Duals have a place but not at this stage. In fact, I doubt many people could hear the difference without top notch, unbiased capsules and so on.

CAPACITORS

Now this is a tricky one. The thing about electrolytics and lifespan can be deceiving. Over time they tend to "dry out" due to heating caused by the ripple they are exposed to but that's only "hurting" them when that ripple is high current too (for example, in an early stage of a power supply).

When used in audio (at the preamp stage) the signal has a piffling amount of current (a few milliamps) and most of that just passes right through. Capacitors, ideal ones, don't have any internal resistance or inductance so in a perfect world they only store and release charge. In the real world, they have a small amount of resistance (typically a couple of ohms for a cheap one, milliohms for the stuff you'd put in the PSU). A failed cap might have an internal resistance of 10 ohms or more. You need special gear to measure that as it has to be done with AC.

This little resistor (it's not a real one of course) forms part of the capacitor and when current passes through it gets warm. Over time, as the electrolyte ages (or just leaks) due to these temperature changes, the device will start to break down and eventually fail. That's the MTBF numbers of a few 1000s of hours you're seeing.

There are two schools of thought on audio-grade capacitors and most of that is determine by where they are in the signal path. ESR comes in here again because it affects the line impedance as seen by the THAT1512 (for example), especially with increasing frequency.

BUT (as one of my mentors put it, and I'll paraphrase him slightly) "every recording you listen to, no matter where [even in the headphones at the studio!] has already been through dozens or even hundreds of capacitors - many of which will be "cheap" electrolytics."

I've been caught out like that too.

I'm a returning to this field after a very long hiatus - in my day 2200 uF capacitors were BIG. The smallest electrolytics we could get were tantalum beads up to about 10 uF with very poor voltage handling. When I built my wiring examination piece, a stereo amp with 2N3055 transistors at the output (that'll tell you how long ago it was) I dropped a bead into one channel the wrong way around. I had that thing (still do actually) for many years and was always troubled by this damn crackle in the left channel. So one day I stopped grumbling and went looking for the fault - a tantalum bead DC blocking cap was the wrong way around and the negative end was exposed to a couple of volts from the previous stage. That made it crackle so reverse it and hey-presto all is well!

Also, 16V is really not enough (I know the board is supposed to run at 15V) but that's too close to the component tolerance. The NMA0515 boost converter isn't regulated so the output is often higher than 15V.

But check this out.

I've shown it here with some of the optional parts fitted. It's a four layer board with a LOT of copper hidden away in that sandwich. There's space for four different triode JFETs (2SK208, LSK170 and others). Configuration is done during the order but I've got to validate it works before I will open it up for people to download and order from your choice of board house. I use JLC as they are cheap and fast but you might have one more local. Many of the SMD parts can be discarded when used with Matt's design but it can equally be used to make a mic for other uses using any one of a number of decent quality capsules, including the 34mm electrets.

This is another one in the current prototype run - this is Matt's original on a PCB. I can't say this works right now because I might have missed something. I've checked both and the seem to be OK but I've made more than one error because I don't have someone checking my work (which is the normal way to do it if you're in industry).

I didn't really want to show this one yet but it's instructional. I've left the "huge" electrolytics on there because the damn things are expensive and people might have bought them already.

But the 22uF non-polarised capacitors?

Look down near the where it says "designed in". The little chunky brown fella down there? There are several of them actually. Those are 22 uF
non-polarised capacitors! I won't go into all the ins and outs of these things (they're imperfect too) but a better option that a chemical soup that makes up the big uns. The smaller brown ones are the 100 nF caps to smooth out any nasties on the supply lines. They have be mounted as close as possible to the chip to function properly. A little more info is on Wikipedia if you're interested. These are "MLCC" and come in several versions - C0G is the best for audio but they are only available in small capacitances so generally we use XL5 and similar where it doesn't matter.

Ceramic capacitor - Wikipedia

I did add a cheeky LED to it but otherwise this is exactly as Matt's original just scaled down to employ these parts. I've a feeling some are redundant, but since I can't be sure, I'll see how it works in practise.

ROTARY SWITCH

I feel your pain there. For a later version I'm dropping an improved volume control but right now we're working with the gain on the THAT1512 (you can also use a 1510 in this circuit with slightly different resistor values).

What you need is a single pole rotary switch with say 6 positions. You need one for each resistor. In essence the switch connects one pin of the THAT1512 through a selected resistor and then back to the THAT. I have a drawing of it if you need one that's clearer.

I'm tempted to use a digital gain control for V2, but that will happen when it happens. My priority has been getting something to work with a wider variety of equipment but also to work with the original. Parts (or lack of them) have stalled that progress until recently since I needed a transistor that could handle 50-60 volts. The lower voltage ones seem to cry a bit... LOL.

GROUNDING

OK this one is a very complex topic and I'm not going to launch into a full explanation because it's nightmarish because it looks simple but small resistances (a fraction of an ohm) can cause major headaches. 

I think what you're referring to though is the Faraday cage around the mic itself. This bit is critical because the ultra-high impedance of the capsule is primed to pick up every passing electromagnetic field. Most of what we experience is the mains signal (or at least the magnetic field it generates around the wires). 

A Faraday cage catches all that stray EMI and sends it to the ground rail, ultimately back to the earth or somewhere else where it's harmless. Although the screen is grounded, the cage works without it. There's more to this of course but that requires diving into the physics of radio waves. Suffice to say that in very sensitive equipment we will often "drive" the shield with a low impedance output to improve its performance. Doing that here would require a complete re-design and it's just OTT (not to mention more complicated).

Good luck with the ordering but I'd honestly look at the LSK170, even if you do just solder it direct to the capsule (which is fine) because it's a great little JFET that's ideal here.

Hi all, I have an odd problem I'm hoping someone can help me with. My microphone works (amazing tutorial and posts in here, btw!) but only if the traces under my THAT1512 are bridged. If the traces are cut, like they are supposed to be, then I get absolutely no output from the mic. I discovered this because I accidentally forgot to score them initially and the mic output a quiet but decidedly good quality sound, perhaps doing this damaged the chip? When I realised what I'd done and cut the traces the microphone stopped working which is why there are some ugly repairs bridging the cut copper.