[Sticky] USB-C Microphone (official topic)

Posted by: @ldoppea

↑

Posted by: @ldoppea

↑

And another question regarding the wires: I have no electric motor and I'm not sure that I want to buy a working motor just to dismount it. Are there some alternative where I can buy only some wire? Is there a dedicated name for this kind of wires? (so I can google it)

 

I realized that I have some old earbuds, maybe I can recycle their cable? Would it work?

 

Did you pop this in the wrong forum? Don’t remember anything about a motor but I might have missed it. 

Sorry my previous post got unnoticed because I'm a new member and my first posts took multiple days before being approved and so they were already lost behind more recent answers at the time they were accepted (you can find them here, here and here)

If your question was about the need for motor wire, this is about this part of the video (at 2:59) when matt suggest to get enameled copper wire on old electronic motor:

(that being said i just realized that matt did tell the technical name so my question from last day is answered 😋 )

Ah! Now I remember. (Senior moment I’m afraid.)

What you’re after is enamelled copper wire (ECW) which you can get on a spool or even rip some from an old “wall wart” transformer. Newer ones are no good as they use more efficient (and cheaper) switched mode regulators. 

Depending one where you live, you might be able to get it by the metre/yard from eBay etc. but it’s often easier to go “dumpster diving” or to charity shops, etc. and get an old radio or similar small electronic devices which have transformers in them. 

However this is potentially risky (it might be obvious but you’d be surprised the things proper do). 

headphone cable should work too but it won’t look as nice as Matt’s all copper design. 

Posted by: @marcdraco

↑

headphone cable should work too but it won’t look as nice as Matt’s all copper design. 

This shouldn't be visible, should it?

@ldoppea Ideally, no which is why Matt chose ECW with a de-solder braid as a screen - really quite brilliant actually. You won't see the wires well through the braid though - and alternative is to get some *braided* screened, stereo flex with 7-core inners. 

This stuff isn't that easy to find, many cheaper screened cables use "lapping" which is far cheaper because it's easier to make the cables and the machines will doubtless be cheaper too.

Braided gives a better screen that's more resilient to being tugged, bent and such - whereas lapped screens tend to develop small holes over time and use which may (emphasis on may) reduce the screening ability. In my experience that's never happened and with audio signals I doubt it would make a lot of difference. Modern lapped cables seem to often come with a lapped aluminium screen on top of the inner copper which probably makes them equally effective but still lower cost.

Braids probably have slightly lower impedance per unit length so don't load the circuit as much but again at audio frequency this isn't something we need to worry about.

But if you can find some braided stereo flex (eBay or one of the electronics specialists like DigiKey, Farnell etc.) all you'll have to do is remove the out plastic sleeve with a sharp knife and a LOT of care. I suggest splitting the top section for a couple of CMs (about an inch) and then peeling it a little like one might peel a banana so you don't either nick the braid or (worse) cut yourself!

I attempted this build recently & finally got everything put together, but I'm a bit confused as to what I've done wrong. If anyone could point me in the right direction that would be sincerely appreciated! Some things to note for my build are:

- I've connected the red audio connector from the USB audio capture device to the board, as I wasn't 100% sure it mattered which cable to use.

- I wasn't sure whether my capacitor orientation was correct.  

- Rather than using a rotary switch, volume control has been set aside in favor of a single 4.7ohm resistor.

- Plugging the device in only gives an extremely faint static, & all interaction with the microphone capsule seems to do nothing.

Thanks in advance.

It's times like this I feel ashamed the V2 is taking so long... It is coming (a very few people have received early versions but the most recent one isn't even at the PCB house yet).

Enough of that though - at a guess, I would imagine that >60dB is sending the THAT into serious oscillation; although it's sold as a microphone amplifier, it's actually a decent quality "instrumentation amplifier" with a 60dB bandwidth of 1.6Mhz which is well into the AM radio frequencies. Nothing in the V1 design does anything to "roll off" that so noise elements can creep in and cause feedback well beyond anything we can hear. It's something I've paid a lot more attention to in the Hydra Heads which form part of the V2.  

Oddly enough (and to my further shame) I just spotted an "obvious" error in the original design which I never actually thought to check. Which was daft of me. I figured, well if Matt's worked on stripboard...

I understand the error isn't obvious unless you've worked with Op Amps - they don't "like" driving capacitive loads and the maximum isn't always specified in the datasheet but it is for the 1512 and it's an astonishing 300pf which is quite decent as these things go. Driving a 22uF therefore is rather out of the question yet for reasons which aren't entirely clear ... it does seem to work. Explaining this effect requires maths of complex transfer function which I don't know well enough to give an understandable explanation.

In simplest possible terms a pole rolls off (reduces) a signal at a rate of 20dB per decade of frequency; where 20dB is 1/10th of the original signal as the frequency increases by 10 times. A zero is the opposite - the gain rises at 20dB per 10 times increase in frequency. 

While this is happening the phase margin is altered too and while the 1512 uses a resistive gain it still relies on feedback to keep distortion and noise low. You can see this loop formed by the final 5K resistor in the output stage. The other 5K or 10K resistors give the output a net gain of 1x or 0.5x.

When we add a capacitor into the mix, it creates a zero at some frequency which can cause the amp's gain to spike due to the phase margin to descend below 45 degrees (a safe margin). Designers sometimes use what's known as rate of convergence (which you usually do in a simulator like Tina or some other Spice. If the pole and the zero are too close (or overlap) you've got an oscillator.

I ran full-speed, face-first into this one with my last headphone amp by assuming (*smacked bottom*) that the 10 ohm "isolation" resistor that works with an NE5532 would work with a similar(ish) Op Amp, the LM324. Of course it didn't and the thing ran away oscillating in a matter of 30 seconds (sometimes it takes time for the feedback to break away). 

The net result of all of this is the output stage is totally saturated by uncontrolled oscillations which sounds a little like a gentle hiss on headphones as the signal fades away but looks like an utter mess on the oscilloscope.

While it's possible something is squiffy, the first thing I'd do is reduce the gain down to something a little more manageable - about 40dB (x100) is quite sufficient for these capsules. I've done this with all of the line-output Hydras (part of the V2 design) and it's more than sufficient. A typical electret mic operates around -40dB (10mV) for 1 pascal at 1KHz sine wave. The JLI2555 varies from -45 to -39dB. This is a common and easily comparable value.

That sounds like a lot (but these are the extremes) and can easily be compensated in the DAW. 

For comparison the world famous Sure SM58, one of the most used vocal mics in history only manages a piffling -56dB under the same conditions. (Dynamic mics have a lot more mass to drive so they need a lot more oomph.) This is why the THAT1512 goes to 60dB. This makes electrets more sensitive, making them ideal for instruments, wildlife and so on. Some can even be used to detect bats! (That's a project for a.n. other day.)

But they're also stupidly cheap to make compared with a dynamic capsule and can be produced on almost microscopic scales. Some of the new heads I've done use MEMs (Micro Electronic Mechanical System) which measure just mms and that includes the internal electronics with a "port" just a fraction of a mm wide.

TL;DR

If you skipped to this I really don't blame you. I sat scratching my head and chasing shadows for quite a while because sometimes you can't see the wood for the trees! 🙂

So first step is to reduce that gain by replacing that fixed resistor with something more manageable (gain wise) - about 50-60 ohms should be fine. The sheet is quite specific but that's for precision measurements and with audio we need low noise, not high precision. (And yes, I know what THAT claims about the noise but you can ignore that as a bit of an "elonmusk*".)

At 40dB noise levels are so low that any residual noise is more likely coming from the digitiser.

I'd also be tempted to check the DC offset voltage on the 1512 with the inputs grounded. Make sure you do this directly on the chip so you don't set fire to the NMA0515 inverter. 🙂

I forget what THAT said it was, but it's unlikely to be more than a few mV which won't harm the following stage. With the mic connected it's probably going to hit 13V - although the meter might not be able to pick that up as the oscillations are almost devoured by noise and it isn't an exact science (technically it is, but it's one for the simulators as there are too many nodes to figure).

Anyway, with the INPUTs at 0v, the output should be almost zero and that's safe to connect to the following stage - although (and I haven't done this) it might be wise to drop a small resistor in there to reduce interactions with the digitiser's input. 100-200 ohms is probably enough but I'm guessing since I don't have access to the design of the digitiser's input stage (any more than I know the output *resistance* of the 1512 (this isn't the same as it's impedance although it is related).

Anyway, hopefully that will either get you going or have you come back at me saying it didn't work! 🙂

Let me know how you get on.

 * Elonmusk: A huge fib that they hope never to get called on.

@marcdraco Thanks for looking! I've just tested it again, first using a 47ohm resistor and then again using a 100ohm resistor, and neither yielded any results. (Acknowledging that this isn't quite in the range you suggested, but is rather close.) I'm not quite sure how to measure DC offset as I've never done that before, but as I continue to look at this thing I suspect more and more that the noise I'm hearing is little other than a product of the audio capture device's unused inputs. I looked at the board under a thermal camera and can at least conclude that the circuit has power and that the THAT1512 is a fair bit warmer than the other components but... That's all I've got, really. What now?

Hi mate, sorry for the late reply - you need to have a few posts up otherwise they sit in the moderation queue until Matt can get to them.

The THAT1512 does run slightly warm under normal conditions maybe 30 C or so wouldn't worry me. Anything over 40 C (about 100 F) and I'd be concerned. I'm jealous that you have a thermal camera, lucky dog. 😉 I just use my finger or even upper lip [not even joking].

Those resistors would have yielded a result, they were well within acceptable limits to get a 40(ish) dB output. The 1512's biasing is quite fiddly but for audio you don't need an exact gain within a fraction of a percent. That's for instrumentation where we're measuring something - the 1512 is actually an instrumentation op amp, hence the high price and specification.

Rev 2 boards should be with JLC in the next week or two. I haven't taken a break over Christmas, rather I threw out my design books and started from first principals (not to mention nicking a few ideas from Rode).

When you say "noise" do you mean crackle and hiss or hum? Hiss is the broadband white noise that sounds like an untuned FM radio with the muting turned off. Or the sound a cat makes when you annoy it by staring at them too long without blinking. (I'm joking, cats are awesome - they only hiss when scared.)

The hiss from Matt's suggested digitiser is very minimal indeed - and while it's possible to raise the volume in the DAW (or Audacity, etc.) the signal should swamp that out so you can't hear it even in the quiet parts.

So... DC offset. That needs a picture.

I hope this makes a little more sense. Really what you're doing here is insuring the INPUTS to the 1512 see an equal 0V (that's a little trickier in reality but a simple shorting wire, connected to ground (that's important) means that the output from the 1512 should also be 0.

In practise it's going to be a few thousandths of a volt - 10 or 20 mV I would imagine - I don't have a working rig I can test this on, but THAT Corp's datasheet describes this.

If it's sitting at some odd, but significant DC voltage, then something is off.

If you have a scope, you can do the same test to see if any noise is creeping out there too - there will be a small amount (we don't live in an EM vacuum) but we're talking in terms of nV AC here - noise from the internal electronics and that's all.

If you break your reply over several posts, hopefully that will free you from the moderation queue and I'll get you answers faster. 

@fishoutofwater Hey there, how is it going with the pcb design? Did you get around to finishing it?

Photo #1: There are the brass mesh sizes categorized by a number. I have no reference on what these would look like IRl. What size should I choose?

Photo #2: I don't recognize these 2 items in the project I'm guessing they are for video aesthetics.

Photo #3,

$18 seems a little expensive for an "audio capture card" but i can't find any cheaper and don't want to deviate from the original too much.

Photo #4,5,

Confused about this part of the build, in the video it seems like a female USB-C was fitted to the rear of the AMP box but for simplicity, I want to use the original audio grabbers USB-A but it is male so will it just hang for the back of the AMP BOX for my project?

Adding on, the 5-volt output was connected to the breakout board in the video. I don't see how I can do that myself without also implementing a breakout board.

Photo #6 This shows the audio cable being plugged into the XH 3-pin connector. What I do not understand is where the 3rd wire comes from as I only remember Drain and Source being used by the Mic to the connector.

Other comments

• Can't find a rotary switch that matches the resistance as the one used.
• Heard Capacitors can be dangerous, what are some etiquette for working with them, and are they dangerous at this level?
• A lot of the circuits have a lifespan of 2000 hours which is low but I'm guessing that value is the "at least 2000 hours" in the worst conditions.
• Another photo of my mouser old list got deleted due to cookies. ReAsking this but is a .6w rated resistor enough even 3.9k resistance which I'm guessing is gonna generate some heat?

• 

• Also, the screws and standoffs used are too different to be bought on Amazon, what are some "screw and bolt shops" guessing home depot can work.

Back at this would have ordered the stuff sooner but college is kicking my ass. gonna cost like $200 bucks but looks good on paper for transfers.

hope this doesn't take half a page of forums with the screenshots.

So... I after loooong fight I managed to assemble everything with working microphone... Quality looks really decent, but I have no idea how to remove background noise. It does not change whenever mic is fully shielded or not. It does not change when I move it away from other devices like laptop ect. 

Every electronic part is ordered from Mouser (THAT and NMA0515 even twice, works the same - don't ask...)

I changed PCB design to fit in 5cm PCB. I tested it on the orginal design, and results were the same (before I burned NMA putting PCB directly on shielding) .

Here is a analysed frequency during silence:

Here are my recordings on max gain (lower gains just lovers the volume, noise is the same):

Here are some photos of assembly:

So... I have 2 guesses: 

1. It's digitizer that makes that noise... maybe I should try to power mic from external source?

2. My shielding might suck... Microphone capsule shielding is made from tea stiner - but I made sure that it is well connected to capsule and later - to the PCB ground. Case for other electronics is coated in very fine metal mesh I had lying around for years. Maybe it is too thin or something? I don't want to buy new metal mesh and rework case just to test it, can someone confirm, that this is fine for shielding?

PS: @diyperks please change description on your youtube movie about this microphone. Link is directing to old PCB design with wrong capacitor polarisation. Took me sooo much time to find out there is new version on your site.

Hi @kirby, let's see if we can alleviate your concerns. Let me know if I miss anything.

Mesh: I'm not sure which mesh Matt used for his design but it's largely an aesthetic choice as at the frequencies of interest (in terms of the Faraday shield) anything from a 30 upwards should be fine. Don't worry about air pressure passing through the shield - it's a common misconception (largely based on the fact that speakers tend to be large - and people forget that headphones are also speakers - that you need huge holes for the sound to pass through. Some of the microphone capsules I've worked with recently have a port hole of around 0.6mm across although they are at the extreme end of consumer devices. 

Many professional mics use a large hole diameter of perhaps a mm or so but you can literally stretch some nylon hose over the thing without dampening the sound appreciably. That's how things like "balloons" and "dead cats2 (windbreakers) work. A 20 or 30 mesh is rather nice in my experience.

The switch in photo (the red case with the silver lever) is just for power. It's not necessary. The other part is a ceramic capacitor which (given that plenty of people have made this mic) is probably not required either. Technically, there should be two of them but I don't want to go down that road or it'll just confuse matters because it really means getting into the weeds of electronic design that you just don't need to know as a maker. If you do want to know more about it, there are myriad examples of it on the web and videos explaining what they do. They're called "bypass" capacitors and although the big chunky ones are also bypass capacitors, the fact that the perfect capacitor is an impossible dream (thanks to physics) we often have to use several.

Capacitors

Capacitors are odd fish and I understand your concern. Can they hurt you? Yup. Can they kill you? Again, yes.

BUT not at this voltage.

Capacitors are buckets that hold electric charge. I'll save you the complex stuff about coulombs and just tell you why. 

The size of the bucket is determined by the size of the component as measured in Farads (after Michael Faraday of the Faraday cage I mentioned earlier). A Farad is so large that most of the time w work in thousandths or millionths of a Farad (and often even less).

Modern technology has given us "supercaps" which are in a league of their own but let's stick to what we have here. The danger in a capacitor is the voltage held on it when charged up - and you only tend to find dangerous ones in power supplies, television sets and so on where capacitors can be charged to >50V which is the tipping point between "ouch" and a trip to the hospital or even the morgue.

The devices here don't get beyond 15V which you can't even feel under normal circumstances. There are stories of kids who have died putting 9V batteries on their tongue thanks to hitting a nerve which supposedly connects to the heart so don't try that. Really. DON'T!

Even tiny capacitors can be quite alarmingly dangerous if you don't exercise care in HT circuits - often because of effect that (last I checked) wasn't fully understood called dielectric absorption which causes the component to re-charge itself out of thin air as if by magic. It's thought to be an electrochemical effect and is far worse in electrolytics (the larger capacitors) than in other forms but it's nothing to worry about in this sort of circuit.

Audio Capture

$18 isn't a bad price and I've tested quite a few of these. There is a CMI-based one floating around which is excellent value for money under $5 but the problem is there's no way to know (for sure) which chipset the manufactures have used and while the CMI chip is top-notch, the other ones are bargain basement... This one is pretty much guaranteed to deliver the quality and it's stereo too - something I've used to great effect in the V2 which might eventually arrive this year - it's only a year late as of this writing (sorry everyone), I spank myself with a haddock every night and will continue to flagellate myself until I get the finals out (that's engineer humor BTW).

There's no requirement to use the Pol... wossit USB-C breakout, it just makes things a little more modern. USB Audio devices only need USB2 speeds but more modern machines are eschewing the old standard because it's a PITA for any number of reasons. 

You can tap the +5v directly from the audio capture board - just follow the RED (+5v) and BLACK (0V) leads to the board and solder direct from there. Not the easiest of jobs for a beginner but dooable.

Photo 6:

I believe Matt used that as the ground to the Faraday shield. Now this isn't strictly necessary (don't even get me started on ground loops, you'll find engineers all over the web discussing grounding as some sort of black art and Faraday shields are no exception. The first time one of my lecturers told me to fix a ground to one end of the circuit I thought he'd take leave of his senses. (RIP Taffy!) But no, he was right on the button. Ground loops and grounding problems are the bane of sensitive audio circuits and if I had any hair left at the start of my re-design, it would have all gone by now. 😉 Getting this right is vital and there's no point me producing a new design that doesn't exceed Matt's original - which is a pretty high bar. Between us we've come up with something pretty sweet but thanks to time constraints and high costs getting production prototypes back from China has slowed things to a crawl. It's soul crushing to spend weeks on a design, wait weeks to get it back and then watch it all go up in smoke ("magic smoke") because I screwed up a component or misplaced a via - I come from days of very early PCBs and SMD wasn't something I'd done before I started this.

In the "auld days", when I were just nipper and we still carved stuff out on stone tablets, blowing a resistor wasn't a major issue (you just looked like an idiot with soot on your face and a cloud of foul smelling smoke appearing over the bench - but never a damn genie) because the parts were large and easy to replace. These days many common parts are little larger than a pin head and soldering them requires a special oven (or a microscope and a very, very small iron). One of the beta testers in Canada (sorry mate) hit the buffers on one of my last set trying to piggy-back two of the boards. He's probably lost his hair too by now. I'll send him a new one when I get some. 

Misc:

Rotary Switch: simple answer is you won't find one because they're not made. You have to get the resistors (which are awkward values, thanks THAT) although the precise values listed - again by the designers at THAT, not Matt - are not that crucial. It's something I tossed out first chance I got and replaced with an active volume control that uses either a linear track potentiometer OR just single value resistors on a rotary switch per Matt's idea - which is a great way to make sure your volume says where you set it.

Lifespan: do you mean MTBF - mean time before failure? If I was working with components with a life expectancy of 2000 hours before failure I'd call them consumables, not components. Used within tolerance and properly rated, everything except electrolytic capacitors should last many years of constant use. Decades even. 

Dissipation: 600mW (0.6) resistors are fine for this and they're not even going to register on a thermal camera. The only components that get a little warm (and at that, it's only 30 degrees or so) are the THAT when pushed hard and the NMA0515 inverter.

Wattage is simply a function of current * voltage and is most problematic at DC.

But how much current?

Well ohm's law comes in here. The 3K9 resistors aren't going to see much current at all since they are getting a few thousandths of a volt so that's not even an issue. The 100 ohm resistor gets a good-ol' kickin' for a few milliseconds as the capacitor charges (called the inrush current, and in fact, that's part of the what that resistor is for) up but after that, it "sees" a 2k2 resistor in series to ground with a 15 volt potential.

OK so we have 15 volts divided by 2,200 ohms:

= 15/2200

= 0.0068 amps 

= 6.8 mA

To get the wattage you multiple the current by the voltage :

= 0.0068 x 15

= 0.1 watts.

or 100 mW.

Even small, leaded resistors are good for 250 mW so that's not going to be an issue.

An old colleague of mine [a true genius] did a brilliant Javascript conversion of Paul Falstad's Java real-time circuit simulator and while it's not the most accurate of simulations, it's dead easy to use and really friendly too. The interface took me a little while to figure out (I'm not a genius) but I'm also so old I make antiques look new, so that's my excuse.

Circuit Simulator Applet (falstad.com)

I can't recommend this enough because it also has an animated current flow and other nice features like LEDs that actually light up when operating so you can get an idea of what works. No simulation is ever 100% accurate (that's why they are called simulations not realthingulations) a factor that bit me right on the backside recently as I pushed my theoretical circuits beyond what the simpler SPICE sims could manage and they started arguing over which was correct. LTSPice (my usual go-to) said something worked whereas PSpice (TINA-TI) another free program said it didn't. The best one of all if money is was object but now you're broke is Microcap which was an absolute fortune but is now in the public domain and can be downloaded from the Internet Archive. I don't recommend it for anyone without a college degree in electronics though as it's like learning to ride a push bike and then jumping on a F1 motorbike on a wet track the day after.

My days of pushing my luck are long gone now so I'm resorting to what I know works in theory and in both the simulators I can use. LTSpice is what most people use with TINA being a little faster to use once you have the circuit drawn - but it's horrid for schematic capture due to very little thought being put into wiring stuff up. LTSpice almost looks like the 1970s called and asked for its CAD back but in use, it's really, really fast and has a huge library of fairly decent models. TINA is more limited in that regard. 

Paul Falstad's web-based one is sufficient for many tasks and while simulations generally won't tell you when something is going to summon the invisible genie, you can see it right here as I've deliberately not used a resistor in this circuit (which is usually mandatory) so you can see it does the job! 

Spice simulations can give some hilarious results too. I made a mistake a few days back and my 15V supply managed to generate several thousand volts. Which is clearly impossible (OK, it's not, but not in the circuit I was designing) but it's due to the maths generating results that the program isn't designed to test for. It's not a bug, it's a feature.

Here's the same circuit with various resistor values so you can see how the simulator tries to show the LED dimming as less current is applied. The little yellow dots show the direction of (conventional) current flow, this is something that's really handy if you're trying to see what each component is doing making this an excellent learning tool.

So... I am afraid that my post was stuck between other conversation.

In the mean time, I tried what @robert did in page 17 of this topic. I used external power source to power pre-amp (I used 2nd USB cable from my laptop). It removed "some" of the noise, but not everything. It is still not usable 🙁 

Shielding or no shielding seems to not do anything. I even did this:

Not even a change in recording 😉

I am clueless. Currently, it looks like this:

(recorded on lower gain)

I really need help with that. I have only shallow knowledge about electronics and sound recording. I have a feeling that I am just small step before finish, but I have no idea where to look for fix. Can someone at least point me to the direction I should go? 

- Is it shielding issue?

- Is it issue with components?

- Should I connect "box" shielding to the ground of pre-amp like it is done with capsule shielding? 

- Is it possible that this noise is generated by bad quality resisitors?

- Or maybe something I am not aware of?

@robert from page 17 of this topic mentioned some small capacitors added between 2 pins of NMA, and "right after the USB connection". But there are no specifics, and I cannot send message to him. How big capacitors? Where exacly? What polarity? I am trying to follow what he said, because his initial noise was almost identical to mine (and separating power to pre-amp with another USB cable worked a bit, like he suggested).

However... This is all guess work now for me. I would really appreciate any feedback. I have almost working microphone since 3 weeks now... And I really would like to finally use it 🙂 

@marcdraco That makes sense I thought rotary switches came with resistors but I think I prefer a smooth potentiometer.

my question is in Matt's design he uses values up to 560 ohms in his design, I'm wondering if does is enough to fully block out noise and if the closest potentiometer of 500 ohms would work.

Also found confusing in the video:

 It seems like 1 wire is connected to all the resistors instead of a source of power and "exit" after has gone through a resistor. 

on the diagram, it seems like both XH connectors connect to the same row so now I'm wondering how it works.

taking a second look this could be where the 4 copper lines on the stripboard got cut. still would like some feedback.

Matt excluded that from the schematic - I guess when we're so close to these things we forget (particularly problematic area too moving from drawing to prototype(s). I know how it works so it's obvious to me.

I imagine a log potentiometer might work there but I doubt it. You'll need a 10k device but I would suspect it better to use a fixed gain around 40dB which is 50 ohms across the gain pins and put the pot across the output from output to ground with the slider to ground being your volume controlled signal.

The gain resistor is placed across Pin 1 and Pin 8 - the part you highlighted.

The V2 (which is coming, albeit slowly) uses an entirely different setup which is far easier and only requires linear resistors (fixed as a chain or potentiometer). Not a circuit you see very often but it's perfect because it's predictable without requiring strange resistor values.

I have a little problem, I switched the 2N4416 for a J113 JFET and the THAT 1512 for an OPA 1641 but now I seem to get WAY too high of a volume on my output. How could I fix this? 
I think it has something to do with the gain rotary switch that switches resistors between the RG1 and RG2 pins on the that 1512.

The OPA 1641 lacks this feature.

Welcome to the forums.

The JFET is less than ideal but it should work, the problem is that you can't swap out the THAT1512 for any old op-amp (not even a good one).

The 1512 is an instrumentation amplifier - we usually make them with three discrete op amps but these monolithic chips can benefit from superbeta-darlingtons on the input (which means very low input offsets with a bipolar noise (very low at low impedance) and the really important resistors can be laser trimmed in the factory to 0.1% or better.

I'm not surprised it's LOUD -- that's operating the amp without any negative feedback and the gain pins have no effect (NC = not connected) so it's operating flat out.

The trouble with 1512s (and the 1510) is that they are expensive and therefore subject to Chinese clones. There are a number of "instrumentation amps" available but they tend to use their own gain setting resistors - so you can't just swap one of those out for the 1512. Even the 1510 uses completely different resistors.

You *can* make a single-amp INA but the common-mode rejection is poor (probably good enough with Matt's screens). I'm currently working on a new input stage to remove the need for the THAT1512 and put simpler volume control on there but it's not quite ready for the big-time yet.  

Short answer - swap out the 1641 for a 1512 and you should be golden.

Welcome to the forums.

The JFET is less than ideal but it should work, the problem is that you can't swap out the THAT1512 for any old op-amp (not even a good one).

The 1512 is an instrumentation amplifier - we usually make them with three discrete op amps but these monolithic chips can benefit from superbeta-darlingtons on the input (which means very low input offsets with a bipolar noise (very low at low impedance) and the really important resistors can be laser trimmed in the factory to 0.1% or better.

I'm not surprised it's LOUD -- that's operating the amp without any negative feedback and the gain pins have no effect (NC = not connected) so it's operating flat out.

The trouble with 1512s (and the 1510) is that they are expensive and therefore subject to Chinese clones. There are a number of "instrumentation amps" available but they tend to use their own gain setting resistors - so you can't just swap one of those out for the 1512. Even the 1510 uses completely different resistors.

You *can* make a single-amp INA but the common-mode rejection is poor (probably good enough with Matt's screens). I'm currently working on a new input stage to remove the need for the THAT1512 and put simpler volume control on there but it's not quite ready for the big-time yet.  

Short answer - swap out the 1641 for a 1512 and you should be golden.

Here's a simulated (LTSpice) circuit from my own work which could use three small op amps like the one you have here to achieve the same effect as a THAT1512. Not as well, it has to be said!

And this is what comes out (top trace) vs. what goes in on the bottom two. That signal is so utterly mangled by "common mode noise" that it's time to give up and go home, but the magic of these two "differential" signals is that the signal is encoded out of phase so when one is subtracted from the other, only (almost only) the signal from the mic is left.

Clever isn't it? Resistor matching of the 10Ks is important. 1% to 0.1% difference in tolerance boots the noise rejection by 20dB! 1% to 0.001% (which are seriously expensive at this level of accuracy) boosts it to 40dB - and that's just the resistors. The CMMR (noise rejection) is given by (gain * 100) / (resistor tolerance).

So for a 10x gain overall, we have a CMMR (best case) of 1000/1 = 100 = 40dB.

For a 100x gain with 0.1% resistors that's: (100 x 100) / 0.1 = 10000/0.1 = 1000000 = 100dB which isn't too shabby and matches that of the THAT at the same gain. Not surprising since the circuits are very similar.

The catch? That's at DC... as soon as we start using proper signals that CMMR drops off quite a bit.

The THAT1512 has these resistors on the chip, all nicely laser trimmed for amazing accuracy. 

If that looks like magic (always does to me and I understand it) I'd encourage you to download LTSpice for free (and give it a go). The full circuit is above but be careful you match the values as shown. I've included my own here too - it's worth a play to see just how this clever circuit (not mine!) works. It's been around since ... almost forever and is used in all manner of high-end electronics and amplifiers.

Posted by: @dnamein

↑

So... I after loooong fight I managed to assemble everything with working microphone... Quality looks really decent, but I have no idea how to remove background noise. It does not change whenever mic is fully shielded or not. It does not change when I move it away from other devices like laptop ect. 

Every electronic part is ordered from Mouser (THAT and NMA0515 even twice, works the same - don't ask...)

I changed PCB design to fit in 5cm PCB. I tested it on the orginal design, and results were the same (before I burned NMA putting PCB directly on shielding) .

Here is a analysed frequency during silence:

-- attachment is not available --

 

Here are my recordings on max gain (lower gains just lovers the volume, noise is the same):

-- attachment is not available --

Here are some photos of assembly:

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

So... I have 2 guesses: 

1. It's digitizer that makes that noise... maybe I should try to power mic from external source?

2. My shielding might suck... Microphone capsule shielding is made from tea stiner - but I made sure that it is well connected to capsule and later - to the PCB ground. Case for other electronics is coated in very fine metal mesh I had lying around for years. Maybe it is too thin or something? I don't want to buy new metal mesh and rework case just to test it, can someone confirm, that this is fine for shielding?

-- attachment is not available --

PS: @diyperks please change description on your youtube movie about this microphone. Link is directing to old PCB design with wrong capacitor polarisation. Took me sooo much time to find out there is new version on your site.

 

Curious if you had any luck with your problems?

I had similar noise (mains AC harmonics?) which I worked around with the software Equalizer APO + reaeq-standalone notch filters.

Weirdly after not looking at it for months the problem seems massively reduced to the point the filters are not required and I have no idea why or what witchcraft is involved.

The only change in my setup since is a new keyboard, I don't know if that makes more sense than witchcraft 😶 

Yes... witchcraft. 😉

There's a lot of science to a decent mic preamp hence why I've dedicated much of my free time for the last year trying to balance excellent performance with low cost.

My final pre-amp designs are now taking shape with a headphone monitor for 5V - hopefully that whole thing will work on 5V USB 2 without triggering the limiter (it should but...)

The input is now based on a more traditional front-end - I've shown THAT the door in disgust. (Nothing wrong with the THAT but all this voltage flipping is a bind and the output drive is weedy.

The alternative which is just nearing completion (I sourcing parts we can actually get) is rather more exciting, if you want to whet you appetite, check out the Jensen 990 op amp. I've made some changes for modern components (the original Jensen appeared in 1980) and is at the heart of some of the world's premier mic pre-amps. It's that good.

The original Jensen used an LM324 which was 100 NPN transistors in two blocks of 50 giving a low-noise matched pair that is truly hard to beat. Impossible in the day.

The Mat12 (Analog Devices I think) is the closest I could find but there are some others and a couple are available as DIP devices which means I can have all the niggly stuff done in China and leave the few discrete components such as polyester capacitors and the input transistors for you to do at home.

I can't promise when it will be done, but performance in simulation and in the limited bench tests I've managed are mind blowing, far exceeding the performance of the everyday ADCs we're using. I've seen some snipping comments from "experts" - probably the same people who think transformers are the only mark of a good microphone - about using a THAT and I've also seen THAT's used in high-end machines.

It's easy to get into the weeds looking at 1512's datasheet and it is good, don't get me wrong, but a discrete solution is almost always going to be better simply because we're using more semiconductor material which, invariably, means lower noise. That's how the LM324 worked in fact.

My first prototypes will only use lower cost "intrinsically" matched pairs - that's transistors made from the same die, ideally adjacent to each other. A true matched pair is a bit more tricky and getting the noise down means making 'em BIG! However, that should give all of us an idea of just how loud (or not) these things are.

I've been beavering away at this for some time, flip-flopping on different ideas, but it struck me that while a "simple" and effective solution can be cheaply slapped together and still perform as well or better than Matt's original, they invariably mean more wires needing to be threaded in the shield and that's just ... wrong somehow. Also, for longer runs that means multi-core shielded cable which is getting into specialist stuff. 

So I'm sticking with mic cable for compatibility with Matt's original. I've modified one of the new pseudo-balanced head designs (a la Rode NT) so it should work with Matt's pre-amp. The impedance matching is more accurate than the simple JFET phase splitting solution so noise pickup is reduced considerably - particularly if you want to use the pre-amp with a longer cable. And why not? Although Matt intended this as a desktop mic, it's far more versatile. But some people in audio can be a mite snobby - indeed, I expect there are those who will pick nits off me for not sourcing LM324s or using paralleled MAT12s for even LOWER noise. 

Rod Elliot (ESP) has some amusing things to say about such folk. One of the best is people complaining that such and such design has capacitors in there and they want a pure DC signal path. Op amps are DC coupled which means they can deal in DC and all AC frequencies whereas capacitors invariably spoil that party.

As Rod rightly points out, every piece of music has been through a LOT of capacitors long before you download the lossless-FLAC audio that someone (cough) ripped off a CD. Same with the snippiness over the venerable 5532 audio amp. Classic albums from Beatles to Pink Floyd up to the 1980s tended to be mixed on 741s (or even valve amps) and similar op amps which are "crap" compared to a 5532 - and when the 5532 hit the audio mixers, it took over. So the vast majority of modern music be it hip-hop, house, jazz, orchestral - you name it - has already flowed through numerous "crap" op amps and ceramic or electrolytic capacitors.

All of this matters to us because capacitor microphones (the ones used most widely today, except for loud vocals and conference work) have very large output swing, perhaps 100mV or more. The really semiconductor and electronic low-noise (popcorn, Johnson, etc.) systems are meant for use cases where the signal is incredibly weedy at less than a microvolt - the sort of things that might be found in medical and highly specialised lab equipment. 

Hi,

A few days ago I finally received all needed parts and today I tried to solder/plug everything together.

Unfortunately, like some others in this thread, I receive barely no sound 😪 (I can guess my voice very very low with my computer's volume very very high)

My understanding from previous replies is that the two most common sources of failure are the THAT 1512 and the +/-15V power isolator. As I got both on them on Aliexpress, those are the first parts I want to check.

I did not test the THAT 1512 (although I'm not sure how to do it), but I did check voltage on its pins 4 and 7. Instead of +/-15V I get +/-12V.

Seems like bad news, because when reading THAT 1512 datasheet, my understanding is that it needs between 13V and 15V.

Do you confirm this is the problem? Do you see any other check that I can do to confirm that?

PS: I did not put any shield yet as I wanted to check the circuit first. But my understanding is that it should work without it but it would be very noisy?

Couple of points to note. First, cheap 1512s will probably work up to a point - or not at all. 

The 15/-15 is unregulated from the NMA0515 - so the measured (loaded) voltage might well be below +/- 15 but that's not an issue as the device is specified to operate all the way down to +/- 5V and probably works lower with reduced quality.

Running this thing without a shield is going to be a world of pain because the 1512 can amplify signals well outside of the range of our hearing - well up into the radio band in fact.  A little bit of noise can get amplified and "swamp" the device with signals you can't hear but that will show up on an oscilloscope if you have one or can get access to one. Get-you-by units are quite cheap now, £30-50 in UK but that's all they are. A decent one is still going to run you £300+ unless you grab an old CRT beastie. Don't thumb a nose at the old ones, a working one used can be had for a fraction of the original cost and many will give cheap digital scopes a reason to feel cheap. 😉

The fact that the 0515 is buckling slightly suggests to me this might be the case as the quiescent (standby) current drain is quite modest on the 1512 and shouldn't be too much to handle. That's a trap I fell into myself trying to run a bunch of NE5532s on the same device and finding out the hard way that I should check the datasheet first. Years of pouring over datasheets you'd think something would stick... but noooo...

Anyway - first thing to do is disconnect your mic from the preamp board and see what sort of thing shows up on the monitor (Audacity, etc.) on your PC.

Try picking up one of the 15V lines (that voltage won't hurt) and you should see a VERY large AC waveform appear. It's a little like a wonky sine wave - which it is because that's the noise of the AC coming into your home plus any crud that it picked up on the way.

Other methods are a little more involved so try that and see what you get, but it's vital to get those screens in place or nothing you do will be valid since this circuit is designed to be operated in a Faraday shielded environment.

Good luck, let's see if we can fix this together.

Wow that was quite a fast answer, thank you 😀

I stopped working on it for tonight but I will try this tomorrow.

I don't have any oscilloscope, but I may be able to use one on my local fab lab (unfortunately I won't be able to get there until mid-february)

but it's vital to get those screens in place or nothing you do will be valid since this circuit is designed to be operated in a Faraday shielded environment.

Ok, so I'll try to wrap everything in the brass mesh without having to do the entire casing build.

Anyway - first thing to do is disconnect your mic from the preamp board and see what sort of thing shows up on the monitor (Audacity, etc.) on your PC.

By "mic" do you mean the capsule + transistor parts?

Try picking up one of the 15V lines (that voltage won't hurt) and you should see a VERY large AC waveform appear

Just to be sure to understand (as noob in electronic I'm a bit afraid to do something bad), what you are asking me is to plug a wire between the 0515's +15V pin and the "mic+" pin?

Sorry, my bad it's difficult to get a handle on people's experience with this stuff - esp. with my head stuck into a "new" operational amplifier. We're all learning so don't worry about being a noob. The 

But yes, by "mic" I mean the JLI capsule (actually it works well with a huge swath of them) and the JFET. So you should have two wires and the screen to that. In Matt's design this is the "split" positive/negative supply with a reference to 0V via the screen. This is a variation on the conventional, tried and trusted method of phantom powering. You have to have the FET as close to the capsule as physically possible for reasons I won't discuss on this post. But the reason is one of noise rejection.

This may come as a shock (it made me smile)  but a common-or-garden metal biscuit tin or paint can is a very good Faraday shield - cheap too. I sometimes think this is a inside joke in the industry as I've even seen it used in demonstrations by companies like Texas Instruments (TI) for noise analysis.  

If you're using Audacity to record the signal, you'll see that it has a "monitor" on there which will allow you to listen to the signal live rather than have to record the "smush" and figure out what did what later on. Better to do this with speakers rather than headphones just in case you get a lot of loud noise. This will cause feedback (a screech) but it's better to do that when you don't have something close to your ears.

Now I know it's unlikely (but not impossible) to start a feedback oscillation in a headset (capacitor mics are super sensitive so this can be an issue) but it's far better to make a loud noise by accident through speakers than right into your ears - our hearing is far to precious.

Those 15/-15v lines I was talking about are the ones that supply the microphone power. The two that go up the inside of the de-soldering braid.

It's one of those bits of weirdness of phantom power that is totally counterintuitive because we spend time, effort and money into getting those DC lines to be a smooth and flat as possible and use a phantom circuit to dump an actual signal on there.

"Classic" phantom power uses +48V on both lines, which is enough to make your head spin the first time you see it. The advantage of the higher voltage is that it's easier to sneak a very low voltage AC signal on there (perhaps just a tenth of a volt) without the other electronics affecting it. Cheap condenser mics in headsets also use phantom power of a sort at just 1.5V - so the potential for cross-contamination from the power supplies is quite severe. In practice this isn't all that bad since the circuit is partly isolated by the load resistor (often 2K2) such as is shown on the JLI2555 specs. 

Some "self-powered" microphones (lavaliers - tie-clip mics - for example) use a small cell such as a AAA that can last for months without ever being turned off because FETs demand very little power even in use. This is the cleanest way to power a simple mic but it's not something we see in most professional gear (Rode and others do offer devices with both self and phantom power) because there's little call for it. Imagine going on stage and then finding out your mic battery was flat!       

Anyway, enough of this historical stuff.

If you had a signal generator (which is the proper way to do it) you'd inject a signal there and watch the output but most of us lack this sort of of gear and with such a ridiculously sensitive input as a mic (all medium-high impedance inputs do this) you can stick a pinkie end on there and use that to "inject" a nasty bit of mains hum. This is what happens with those awful "phono" plugs that have pervaded consumer audio for decades. If you've ever plugged one into a "live" input you'll have been greeted by that horrible "buzzzzzz" when the centre terminal (signal) makes contact before the outer leaves do. This is the complete opposite of how it should be done and an amazing piece of "what the hell were they thinking?" but it's what we have unfortunately.

I managed without a scope for years growing up and while they are an amazingly useful and fascinating piece of test gear you don't "need" one. I do a lot of my initial testing on the bench with my fingers just like that because it's often quicker, especially to make sure that everything is doing something before I hook up a full testing harness. Most of us (myself included) don't have the space to set up a full testing harness with scopes, meters, sig gens and so on. That's the reserve of chaps with money to burn and the space to spare. (I'm so jealous.)

If you can post a screenshot of what you're recording that might help. Audacity has some analysis tools that than help too. But let's take this one step at a time. 

@xane 

So... I managed to find out what was the biggest mistake, and, of course, it was my fault 😉 I forgot one connection on PCB board. After fixing that, microphone works really good. There still is some noise, but honestly, it's good enough for my use.

What I changed:

1. Digitizer is outside of the box and shielding. Moving it away from PCB fixed some problems with noiuse.
2. I have separate USB cable that powerup pre-amp. (only red and white cables)

I think I can make it a little bit betterm by reworking case and pre-amp shielding, but I am already using that mic... I don't really want to risk breaking it again;) It is, after all, really great for my use.

Here is graph takend when there is a silence in my room (some fan noise included):

And my raw recording without any filters on max gain:

And how it looks in reality?

 Ps: My PCB schematic posted in previous post is 100% working.

Can I directly connect the usb-c to any device and use it as microphone with any additional software.

Hi,

Sorry I didn't do the test yesterday as I said. But I did it today.

What I did was:

- Put the circuit in a metal biscuit case

- wrapped the capsule in the brass mesh (although is was not 100% sealed)

Nothing changed, still no sound (even worst that previous tests as I don't ear even low level noise anymore)

Then I tried to do what you asked for: I plugged a wire between +15V and the mic entry. Here is the signal I got by touching 3 times the "mic+" entry. The signal only happens at the exact moment I touch the pin, but nothing happens then stay in contact. This is why you can see 3 "noise" signal on the screen corresponding to the 3 times I put contact.

Also I found that the contact between the capsule's copper wire and the "classic" cable was easy to break (those thin enameled copper wires are very hard to solder), but for the first test with the capsule I double checked that the contact was good. But maybe this can be part of the problem?

PS : I'm not sure where is the Audacity feature to live visualize the signal so I ended up doing a classic recording.

This is positive. At least we have an idea of where things are going sideways. 

The little peaks you get when you tap the input pins are capacitive coupling, so that helps to isolate the problem.

Here's the part in the Audacity manual where you can monitor levels and such 

https://manual.audacityteam.org/man/meter_toolbar.html#recordinglevel

The most requested feature for a V2 has been the headphone monitor and I recently completed that part. I'm just tweaking the input stage now to perform better than the IC (THAT1512) solution which, amazingly, is possible using discrete parts. It's just fiddly.

Back to where you are though. You can use a heavier gauge of ECW in the braid if you want but during testing some simple, screened audio interconnect cable would be more reliable as it's designed for more mechanical strength. I have an idea to improve this part too but I have to get my other bits working first and it's been in R&D for over a year now so don't hold your breath. 😉

When you popped it into a screen, a lower noise is what we'd expect since the screen is working as a Faraday cage and eliminating it.

Next thing to try (and I'll try to check in often so you're not waiting) is to check the resistance between pin 1 and pin 8 on the THAT1512.

This is the gain control and it's crucial this part works correctly or nothing will. Being the old hack that I am, I'd probably just solder a resistor of perhaps 1K across them but that's really naughty or even leave it without one.

When set up this way with either (around) 20dB (x10) or even 0dB (x1) gain you could hook up a line-level source to the input and use that to check if anything is working. The larger gains the THAT is capable of are for microphones but nothing says it can't be used for other sources - including the headphone output from your mobile phone! 

This would be the next thing to do in fact. I would suggest you remove the gain setting resistor/switch completely and attach something that will drive a small headphone direct to either of the inputs. It's important to "inject" the signal via a capacitor though as both inputs have a DC potential on them.

If you need assistance in making a quick test harness let me know.