[Sticky] USB-C Microphone (official topic)

Ok, so this is a render of the last (I hope!) batch of designs for the JLI 2555. Since they are so small (just over an inch in diameter) I've popped four completely different versions on the same order.  Unlike my previous work, which required separate 5-15V DC, these all work from a P15 (15v) version of more traditional P48 power. Even the op-amp one as the pre-amp should be able to supply about 10-12 mA without the voltage drop making the circuits unworkable.

One, as you can see, uses a traditional operational amplifier, so long as it's a BiFET or CMOS type, any low-power dual amp with power supplies up to +/- 15 volt should work. The other three are varying degrees of complexity to see what sort of balance is possible between noise and performance. At least one highly regarded design used in "professional" mic upgrades uses one of TI's OPA series of dual BiFET designs. I'm gonna see what it sounds like with a relatlively noisy TL072 because I have some in stock.

I've simplified them somewhat (mostly a cost thing) although the simplest one does use a LSK389 in paralell so it should be exceptionall quiet. *should*. At a guess, I'd suspect that this particular FET, especially in parallel as it is here, will far exceed the noise floor of the capsule and would find use with more up market capsules.

The other two use an SMD FET from Toshiba the 2SK208 which, while not as performant is the LSK170 and LSK389s in terms of noise, is still specified for condenser mic impedance matching - which is, in essence, all these things really do. One of them features a pair in a parellel connection using a "trick" to force them both to use the same drive current, in effect, making the more like the LSK389. They're not of course, and it's going to be interesting to get a batch of these back and pit them against each other!

I just hope the simulator isn't playing tricks on me ...

@marcdraco I made a template first and then used a router with a bearing cutter. A router is pretty essential. You're gonna need one anyway to cut the pickup cavities and control cavities and the truss rod channel in the neck. Doing it by hand would take forever, and is never going to look neat. People normally would use a bandsaw or a pin router to cut the outline of the body, but neither is cheap. A hand-held router works just fine. I used it to make those indents behind the bridge too. You might want to just buy a ready-made neck. Just buying a set of nut-slotting files, to cut the grooves in the nut, will cost more than a decent neck. That's why there's no neck on mine. It's in the guitar shop, getting the nut cut. I just can't justify buying the files. I own all the tools to level, crown and dress the frets, but those are things I'd use again when the frets eventually need work, or when someone else needs work done. If you're using a plastic nut, you can always do what I did last time...let some guitar strings get rusty, then use the string itself to cut the nut slots. 😀

 Technically, pretty much everything can be done with normal tools...technically...kinda. There are tons of videos on Youtube that can help.

If you have to buy any specialist tools, just avoid Stew Mac like the plague. Their stuff only makes sense if you're building hundreds of guitars. They'd stencil a picture of a guitar on a hammer and call it a "guitar hammer" and sell it for 5X what a hammer costs.

But what you need to buy, depends on how much you're doing yourself. Some people even make the pickups themselves. I didn't. Those are DiMarzio's straight from the factory. And NOBODY makes the bridge or machine heads themselves.

 What level of DIY-ness are you going for? Cus even buying a ready-made neck, usually means cutting your own fret slots and pressing the frets.

 But technically, it's possible to build a guitar as much from scratch as professional luthiers do, using very few tools

Hi @marcdraco

I saw the first recommendations you gave to improve Matt's project such as the type of capacitors to use and their values: the ones you linked cut at 18Hz? Do I need to change some resistors? I seem to understand that I should replace the non-polarized input ones, and output one? The polarized ones? 

How could I also insert a filter to cut the high frequencies?

Sorry for all these questions, I would ask you more about your project too but I would become annoying. 

Edit: the capacitors make a rc circuit so it should be a filter to cut the high frequencies, why then should the signal below 18Hz be clipped? I still wish I had a filter to cut the inaudible frequencies

Wow! Thank you that's a lot of great information. I have a router and we have the plans for Brian May's Red Special.

Wood isn't really an area where I excel. I can put shelf up but that's about my limit. When we made a 1:1 Proton Pack (Harold Ramis' from GB1) in 2021 we used 3D printing but that's far too expensive now. Not that I'd make a guitar from PLA ... but that's the limit of my skills really. My routing skills are limited to gouging stuff out of blockboard and even then I'm not 100% accurate.

I'll show him what you made just to make him jealous! 🙂

@ridian Did you ever figure out why this happened? I have the exact same issue.

@mrhempman69 sorry I missed this. I haven't shared the gerbils (sic) yet because I don't want people to pay for a sub-optimal or, (heaven forfend!) a board that simply doesn't work. It's been quite a long road and I've already had a couple of last minute "hold the front page" moments like day before yesterday when I managed to upload the WRONG gerber to JLC.

I had a couple of transistors upside down on the schematic but I'd been staring at it for that long that I'd completely missed the error.

I have a small batch of working boards that fit the JLI2555 but they require an external power source so they won't work with Matt's pre-amp directly.

It was spotting that cockup - serves me right for designing the head without looking at how Matt's worked - that forced me to re-design the whole blasted lot. DERP!

Some good has come out of this - the first production boards are on way back from China now - these have the new preamp, headphone amp and so on. As soon as I've tested them and made sure there aren't any.

I'm just asking JLC to cancel the order for the faulty ones now.

What a plank I am. I could warm the street, I'm so red-faced.

Now the upshot of this (and assuming they won't chew me a new one for putting several designs on a single order) is that I've come up with a bunch of brand-new designs - ranging from a single 8-pin DIL socket and a 3 pin connector which takes many low-power BiFet op amps such as the OPA1642 right through to one which while cheap (in real terms) potentially offers superb performance right out of the box.

IF you want to have the Gerber/BOM and order some at your own risk, I'm quite happy to share in a PM.

I'm using a lot of SMD parts now and many are very small (402) so they are really best order pre-manufactured unless you have a very steady hand and an oven or hotplate. Even then, 402s are so small I'd suggest letting the machines handle it.

@marcdraco No worries! The mods took a few days to approve my (first) post. When it was approved, it slid back onto page 14 - essentially back in time.

@marcdraco Thanks for the PM, but I already ordered the parts and PCB to build a ~1.5 version from one of the githubs listed earlier in the thread ( https://github.com/axydavid/diy-microphone ). 

Also, I can't reply to PMs until I make 5 posts... 

But your influence is not unnoticed! A copy of the third edition of The Art of Electronics arrives soon! Looking forward to truly understanding electronics on a much deeper level.

Posted by: @lordece

↑

Hi @marcdraco

I saw the first recommendations you gave to improve Matt's project such as the type of capacitors to use and their values: the ones you linked cut at 18Hz? Do I need to change some resistors? I seem to understand that I should replace the non-polarized input ones, and output one? The polarized ones? 

How could I also insert a filter to cut the high frequencies?

Sorry for all these questions, I would ask you more about your project too but I would become annoying. 

Edit: the capacitors make a rc circuit so it should be a filter to cut the high frequencies, why then should the signal below 18Hz be clipped? I still wish I had a filter to cut the inaudible frequencies

@marcdraco It took a while to post for mod review. I think you missed it. I extend the question to you @DIY Perks as well. 

thanks in advance for the reply

@mrhempman69 A great book indeed. My second edition is just about to give up, I have the Kindled version of the 3rd so I know how much better it is. Win (Winfield Hill) and I had a brief email exchange last year - what a NICE guy. Never mind being clever. Paul Horowitz is a terrific guy too.

I'm sure you'll enjoy it - they do have a some math in there but it doesn't go into the silly explanations that so many other texts (and Wikipedia) do. It's as if they have to show off how smart they are. 

Sometimes all we need is the formula (1/2*Pi*R*C) for example to get the cutoff of a simple filter. The derivation goes into all sorts of weird stuff like imaginary numbers and while that's interesting to know, if you just want to design a simple filter, you just need the formula to punch some numbers into.

Now it's true that AoE does have some fairly complex math but that's only when its necessary.

The V1.5 PCB is perfectly functional which is a good thing, David made a great job. My designs (for now) must be regarded as untested. That's part of the reason I'd sent six different variations to JLC as you saw. Now I think they all work and it's mostly a case of figuring out which one is quietest and work out a budget vs. quality.

I expect the weak point in this design is the digitiser which suffers with terrible channel separation - or rather lack of it. If it's that bad it doesn't bode well for the remainder of the design. At least Matt is only using a single channel.

I guess we'll all know in a few weeks but postal services permitting I'll have a few spares I can throw at people at cost (which is a couple of quid/bucks) a piece to experiment with.

What I DID forget (and this was naughty) was to make a version to take Matt's original FET although there is space on the parallel boards to fit the leads from a device recovered from being soldered direct to the head.

@theguy I've been busy doing this behind the scenes, although I've concentrated on large electrets like the 2555 Matt chose for this.

More by accident than design, I've found that my choice of FET (Matt and I did a lot of this in PMs) which is still in production, the LSK170 or LSK389 which is the dual variant, is quieter than the original 2N4416 by 6db! Put two of them in parallel on the same die and you can half that noise figure again!

Pair that up with a THAT1512 and you have an amazing chain but what lets it down? The JLI electret which has a self-noise of (from memory) around -60dB which is huge compared to the better unpolarised large condensers which have self-noise around -90dB or better. Now this is from memory and I know I should go Google that but there are only so many hours in the day and I'm sitting on my hands waiting on a bunch of experimental versions ranging from a BiFET (op amp) version which is dead easy to upgrade through paralleled LKS389 (or 489), dual and single LKS170s and even a couple that have their own integrated SMD FETs from Toshiba that, themselves, have pretty decent noise figures.

I've also re-designed the pre-amp board to include a better volume control, high and low cut filters and a fairly decent headphone amp (until I ran out of space). This isn't a criticism of Matt design, it's just meant to augment it. The board can be ordered with or without factory parts depending on what you require - but I haven't got the final prototypes back so I'm nervously awaiting them coming back from the factory for testing.

Assuming that they work, I'll work out some way for Matt to sell these (should he want to) or otherwise I'll put the working design and BOM into the public domain. There's nothing really clever in what I've done, it's all just simple electronic lego bricks in essence. 

The 48V version is on track though. I have to work out how best to bias the capsule and the FET for low noise. I tried a couple of 1N4148 diodes with a bootstrap (to replace the more usual 1G resistor) but I suspect they were a noise source and I've been unable (being in bed with an infection) to test that properly.

1Gig resistors tend to be expensive and hard to find - let alone noisy because Johnson noise increases with impedance.

Electrets (and I still don't understand why, although I have some ideas) work without the load resistor. This may be because they're DC coupled whereas the normal condensers have to AC coupled to isolate the 48V charge voltage. The capacitance (including the parasitics) seen at the input has a reactance of a megaohm or more at low frequencies so the resistive load needs to be much larger to overcome that.

TL;DR so I'm looking at boostrapping some more commonplace resistors to create a larger impedance with smaller values.

So I planned on building this and kept putting it off and ended up buying a Neewer NW-800. Its a good mic for a starting setup, but no where as nice as this ones sound quality. Much better than the built in mic than my corsair headset though. 

I've been thinking of replacing the sensor in the neewer mic with one DIY perks used to improve the quality. The only thing I didn't like was the mic's size. As I play fast FPS games I feel like it'd be odd to see in my line of sight. But if I combine the better sensor, and nicer housing/aesthetics of a manufactured setup it would give the best of both worlds. 

Actually looks like someone has already done this. One of the reviews on the sensor said they did this on a BM800, which has some other brand name on it on amazon yet looks the same. So I'll just order the sensor and see what I can do. I'll take pics and post them later though if thats something people are interested in! 

Almost all of these Chinese mics use the same body design and internals - copied from an early Neuman I expect. As I've replied elsewhere, they also use the Schoeps drive circuit which is pretty easy to modify. Someone on one of the specialised websites (about mics) remarked it's very monkey-see, monkey-do mentality. But with each iteration, as they copy from each other, it becomes literal Chinese Whispers. 

Mistakes creep in and over time it starts to show as successive mistakes create larger and larger problems at the sharp end.

I suspect the next "fix" for these will be to repair that awful 48V inverter which I'm pretty sure uses CMOS inverters rather than a pure AC feed and a Cockcroft-Walton type multiplier. You only need a four or five stage CWM to get from 12V (AC) to over 48V and after that a standard regulator can bring it back to heel. But that's more parts and parts cost money.

Perfectionists would have you believe that you can't use SMD components - mostly due to microphonics in the SMD capacitors and there's some truth in this. But they fact is, even Rode uses a LOT of MLCCs in its designs - most of the big boys do until you get up into the really expensive stuff when you see a switch to Poly--whatever ("Poly Put the Kettle On" as former head designer at Rode once remarked).

PTFE in particular, but also polypropylene and polystyrene capacitors are far superior in every metric except for price but until you're using a VERY good capsule (think £200-300 and up) and the sort of equipment that can hear the errors, you're not going to see a good return on the investment.

The really high-end stuff uses valves and transformers too. None of this transistor rubbish. Of course, the joke is on the people who buy into this claim. While I'm a big fan of audio transformers (a well-made one is going to give a Schoepes circuit an absolute paddling) they're expensive - very expensive - and again don't really represent any sort of value in terms of ROI.

Valves come with a haulage vessel's worth of problems. The obvious one being the rather hair-frazzling HT supply needed to convince electrons to flow but there are others that are less obvious. Aging is one that most people don't know about but unlike transistors that only drift a little with age, valves take entire detours.

And they get hot. Very hot.

It is possible to run a modern valve at low voltage and this is something I intend to try at some future time. Simply because valves do exhibit a soft clipping that's hard to achieve with modern transistors. It's possible (so I've read) but valves (tubes) do have a certain charm and if all we have to worry about is the heater making the case a little ouchy, and not actually killing some poor sap (i.e. yours truly) that's something worth looking at.

In theory, it's possible to use valve(s) as the front end to roll off the clipping so it doesn't produce all those nasty harmonics, and then amplify that nice clean signal with more modern parts.

That said, Matt and I have discussed making a mic that's effectively impossible to clip (until the actual capsule reaches its limit) by increasing the dynamic range using a sneaky little trick. I just need to find the time to write the software as the physical hardware isn't all that complicated.

@marcdraco was inspired to make my own diy mic after seeing the video for this. after reading through the forum a bit, i see that there is discussion of possible design changes. would you recommend waiting until these are finalised before starting? also what mod would you recommend for that chinese microphone that you have listed?

apologies if this has been answered or if this is posted in the wrong place, i’m not too good at navigating the forum :/

Not to worry. It's been a while coming but the final batch should be here later this week, FedEx and a following wind allowing as they're produced at JLC in China.

I've re-used the more expensive components on the main board and designed several versions of the head unit (electronics) so we can see what gives us best bang for the buck.

I expect we'll put the Gerbers/BOMs out as Open Source Hardware under a Sharealike N/C licence. It's all the same to me but obviously this is a business decision for Matt, should he wish to sell kits which are partly pre-made. The full kit has a stereo headphone amplifier too but that's just on it's first prototype so there will certainly be some kinks to work out.

The head units are all designed to work with the original stripboard layout but most (and perhaps all) exceed the performance of the single FET solution plus they use current parts. JFETs are quite esoteric these days since they are a PITA to design for/around and most applications work as well either with op amps, MOSFETs or modern bipolar transistors.

JFETs are pretty much reserved for specialist applications now where we need to interface a very high impendence source to some external circuitry - condenser mics are the only one we meet in the builder's zone, but some medical transducers and military applications also need them.

Matt and I trawled everywhere looking for devices that are still in production and we came up with a couple of options that are still widely available in "through hole" (another tech that's heading for the scrapheap) and some modern SMD versions too.

I tried the SMD versions (the performance is outstanding) but they are just not practical to solder by hand unless you're an expert and have a binocular soldering scope (or an SMD oven/hob). Obviously most of us don't, so through-hole it is!

The devices we've found to give stellar performance are the LSK170 (A/B/C/D), LSK389 (A..C) and LSK489 (A..C). The 170 is a three-pin, single and the x89 devices are both large die monolithic.

The LSK170 outperforms the original transistor by a clear -6dB (50% less noise) but you can twin them up (and get an even larger noise reduction). This is what I've done for the LSk389 version on the new designs. I've had to use a couple of tricks to make this work with the 170s because the thing that makes JFETs such a pain to design with is they way their performance is a bit of a shot in the dark. A BJT (normal tranny) isn't that predictable, but they are sufficiently so that we can design around the shortcomings quite easily.

JFETs... not so much so which is why they largely fell out of favour. I expect if anyone ever figured a way to make one that worked like a jellybean BJT, they would get a Nobel.

For this run, I've added in a version that takes a low-power BiFET amp like the ones used in the OPA Alice (you'll see that referenced all over the place). It's just a traditional FET design by Scott Hempsky (sp.?) but with a low-noise Burr-Brown/Texas Instruments fitted for the impedance matching stage.

On paper though, the original JFET Matt picked still outperforms that by a significant margin.

The most expensive parts - the JLI2555 mic capsule and the THAT1512 are completely compatible with the uprated design. I'd be tempted to use an LSK170A in place of the original FET as it's hard to get hold of now though and if you're careful, you can get it working and then put it on the improved board if you want to later on.

@vint Sorry, missed the bit about the BME800. It's really case of swapping out the capsule. The difficulty is that some come with electrets that have internal FETs and others come with ones that don't.

You can't really tell just by looking (although you can test for a bias voltage). Once you know which type you need, you can order a much better quality one from somewhere like MicBooster.

It's useful to have a testbed to test the capsule you find in there but that's a bit chicken/egg situation.

@vint Sorry, missed the bit about the BME800. It's really case of swapping out the capsule. The difficulty is that some come with electrets that have internal FETs and others come with ones that don't.

You can't really tell just by looking (although you can test for a bias voltage). Once you know which type you need, you can order a much better quality one from somewhere like MicBooster.

It's useful to have a testbed to test the capsule you find in there but that's a bit chicken/egg situation.

@diyperks

"They're heeeerreee..."

There's nothing quite like that feeling when the boards arrive.

There's also nothing quite like that feeling when you spot the first error within about 30 seconds of visual inspection 😳 . It's not a deal breaker (yet) but does mean a small change to the final artwork assuming nothing else turns out to have gone sideways.

It's not easily visible on this but the mic entry point (lower left) has a three-pin socket header - well, the holes for one... and somehow I managed to use the wrong size hole spacing, it's a standard header spacing but more suited to wave than hand soldering. 

The heads final batch of heads should be here soon so fingers crossed everyone can get this cracking project built even better performance that we thought possible - and the original was pretty sweet!

@marcdraco Wow, I’m so excited! I’m definitely going to be first in line if Matt decides to put out a partially assembled kit. Either way, I’m buying parts as soon as everything is in order for you guys. Thanks for all the hard work!

@lamechial I found a couple of bugs - nothing serious but one does affect the chain so I've had to fix that before we churn out the final board. 

It's rather forced my hand into figuring out NGSpice, which is a highly advanced electronics simulator built right in to KiCAD. I already used Spice simulations to make sure my designs worked ahead of time but since everything is done over several sheets, there was always the chance of a boo-boo that I wouldn't catch until they came back.

And sure enough... there's a boo-boo in the chain from the Baxendall volume control. It's not a huge deal-breaker for me (it only affects the headphone section) but obviously that's no good to you guys.

The issue is that NGSPice (as fitted) doesn't simulate op amps and this thing is brimming with them as you can see. Add in that when you get a Spice model for most amps, the NE5532 is used extensively here, it only simulates one amp in the package. So in effect, it's impossible to check the circuit as drawn over multiple sheet (there's six for this board alone since there are so many discrete parts in there).

There is also the issue of the amount of power the NME0515 power convertor can supply. These are quite greedy chips hence why they are powered separately from the main 15V rails and that, unfortunately, is where the error crept in. (This is another reason I really need a simulation, while I can estimate how much load the amps put on the NME0515, it's a balance and the simulator makes that easier to calculate rather than having all the back and forth.

Long story short, when I added the Baxendall circuit in the late stages I forgot to add the required DC blocking capacitor where it enters the 5V powered headphone circuit. I've also fixed the issue with the microphone in pin-header to make it easier to solder.

The hexagon "FEThead" pack came too. I'll run these through some tests over the next few days to see how they perform. They're all based on a mature design so I don't anticipate too many problems but ... 

I don't mind if people want to risk it but if you'd rather wait, that's probably the sensible thing to do.

OK folks, I've done some gremlin-hunting and fixed a couple of show-stopping cockups on the preamp prototype (no magic-smoke) but that's why we make prototypes.

Since the design (by arhum design) breaks out the pre-amp section to make it easy to use the board as either a mic amplifier with mono headphones or as a headphone amplifier, it's possible to drop other functions in there - albeit ones that need their own power, etc.

The one that struck me is a value (tube) simulator. This is actually surprisingly simple to do since it's a essentially just an op-amp with a couple of diodes. I'll post something to this effect in the What Are You Working on Forum, for anyone interested in the theory of it.

If anyone wants any more functions, drop me a note here as quick as you can so I can see if it's practical before I send the finals off to China to have them made up.

@marcdraco are y'all going to be selling boards from this run? I was originally really interested in building the OG design until I came across these amazing write-ups. This is looking like a great project. I know you are constantly improving things, but do you think you'll have a "final" BOM sometime soon.

Next one will be the final one. There was a goof (entirely mine) on the last one when I copied the schematic from LTSpice into KiCAD and also I've discovered another expensive mistake on my part (another schoolboy error, I'm afraid!) by not checking the datasheet for the NMA0515 sufficiently carefully. 😳 😳 😳 

Long story short, the 15/-15V rails we're using to power the THAT and the head unit just doesn't supply sufficient go-go juice to drive the rather power hungry 5532 op-amps. There are replacements that would do the job with the existing design but they are horribly expensive and apart from being very low power, they don't offer one iota of improvement. The SMD 5532 is a really nice little jellybean and given the number we need ... at £5 - £10 a pop for the alternatives, you don't need me to explain! Also, JLCPCB keep them in stock and will assemble them at their end. More on that in a moment.

I'd rather own up to these mistakes when people ask where it is (given all the delays due to the 6-8 week turnaround for prototypes) than let you all think it's vapourware.

The other mistake isn't so much a mistake as a facepalm on my part, again this comes froof m my experience of 1970s through-hole parts and using that to judge the size of 2023 (well, current) through hole parts. I know not everyone has my level of experience soldering so I wanted to keep it all as "DIY Perksified" as possible.

So this is what happened - KiCAD, the amazing free software I use for the design/layout work has an enviable list of parts. Many have appeared since I quit doing this professionally (and to my shame, I haven't kept my hand in). 

You'll appreciate that everything is getting smaller - the SMD 5532 is better in SMD for a bunch of reasons - other things are tougher to judge. The size of connectors being one of them.

The header pin connectors aren't available from JLC in SMD form, only through hole and that means extra costs in soldering them since they have to be done by hand.  Wave soldering imposes some strict requirements on design and I don't think they offer it anyway.

So what? Well, in my haze of Cat Scratch Fever (it's a thing, trust me... even though my locum thought it was C19 sans cough and I hadn't been out of the house... etc.) I picked some male headers, for through hole soldering that looked right. 0.1" (I thought, they'll do) - and so they would but in my haste I hit the wrong list and picked 0.05" (1.27mm).

Through hole is usually pretty easy to do once you have the hang of things and a 0.1" (2.54 mm) pitch is about as tight as you want to be before things get dicey. Can you see where this is going? Oh trust me, it gets worse...

So of course, do I check them? Well yeah, I do a 3D render (all looks fine and nice) and copy/paste these same connectors through the entire board everywhere that I needed something to either pop out to the volume control, the filters, power systems (this has dual power inputs right now but I'm revisiting that due to this cockup) --- everything. Including the revised designs for the mic head.

Tl;DR these things are an utter nightmare to solder. The ones on the mic heads are difficult enough but there's only actually three pins and only two are *required* to make it work (I should really fix that for the next batch) but even with a lot of experience, pretty nice iron with a superfine tip and even LEADED solder (which I recommend against, but it does flow better) trying to prevent bridges was enough to give me a serious need for copious amounts of alcohol or a ferry to Amsterdam (you get my drift). 

A short won't break anything as it's seriously current limited but it will stop it all from working and unless you have the right gear such as a stereoscopic soldering station, they are a bit of a swine to remove without damaging the board.

In "the good ol' days" of Radio Shack/Maplin (in the UK) I'd have hopped in the motor and gone and got some news bits and been home in time for cornflakes but now it's a case of double-triple and quadruple check the Gerbils (sic) before I send them off as I'm outa "new user" coupons which were subbing my attempts to get the head part right.

Life has gotten in the way of making a testing harness (which I've promised myself I will do this week) so I can test the new head designs without having to wait for the main board to come back. I've tried best I can to make everything re-usable from Matt's original. It was never my intent to ask people to throw everything away. Might seem daft, but I learned a lot using and re-using (over and over) the same parts because it was simply too expensive to go buy new (irrespective of my previous comment). I'm from Yorkshire originally, if you know the Monty Python sketch, if not it's worth looking out).

I didn't reuse the FET which I should have really, but Matt and I found a better one that is more widely available for people overseas with better performance and at a lower cost. It's not pin compatible but easily adapted to even the existing design (the LSK170 by Toshiba). It comes in three versions A, B and C but in this application any one will do. The dual version, the LSK389 appears in one of the 6 test designs I did but I've already done a dual LSK170 that works like a charm and is ready for us- with one important problem, that particular board requires external power and Matt's preamp is essentially "phantom" powered.

Many designers, myself included (I'm not in the same league as these guys) shrink at the thought of Phantom power outside of transformer-based deigns where it originated. So while that design (codenamed Cerberus) is tested and working, there are some changes needed that I don't want trouble people with. Esp. as the Vero/strip version is a PITA to work with.

The new one, codename Hydra (6 heads per factory unit) has a variety which I still need to test but they all are designed to work with a variation on Matt's 15/-15 phantom.

I'll let some guys with far more expertise than I explain why P48 is a major source of dead input stages on stage equipment: (PDF) The 48 Volt Phantom Menace Returns (researchgate.net)

The old system of transformers was robust in this regard since the input was protected from the phantom supply by virtue of transformer isolation. There wasn't any way for 48V to leak back into the input stages, but with the advent of transistors and ICs transformers (which are bloody expensive if you want decent audio ones, the *cheap* one for my ribbon mic design set me back almost £100 with shipping!) fell out of favour.

You might have gathered then that sticking 48v (even at reduced current of about 12 mA) down the input terminals of your expensive preamp which runs at 15/-15 and might have a common-mode range of just a couple of volts is going to cause the sort of magic smoke that not even the hardest of hard rockers want to deal with.

But getting back to the NE5532 problem, this one is more thorny, so I've had to make some major changes - including dropping the active low and high-pass filters (I've replaced them with passive versions which less ideal but don't load the 15v rails) and I'm running some tests on the Baxendall volume control which sets the output level for both the headphone amp and microphone output.
 
5V seems a little meagre but the swing on consumer headphones is in the order of a few hundred mV so that should be enough "ummph" because the power is in the current: which is the reason we had to add a second 5V input in the first place. Yours truly here never thought to check that there was enough power left from the inverter... 

I won't compromise on quality and I'll drop the designs for the valve/tube simulation when they are ready. I've had to temporarily shelve that development so I can give the available time to getting the meaty bits done. In an ideal world, this would be a small stack of boards and I still might try and slip that past the folks at JLC although they'll probably want me to pay for each one as a separate design.

I prefer the modular approach just in case I smeg up something (as here) because one dumb mistake in one part of the board (one section) won't bring the entire project to its knees as this has. The new designs are almost ready, I just need to get the single-supply Baxendall down on a breadboard - it works fine in LTSpice even with slightly random component values but I've been down that road before and missed the "Caution: bridge out" sign with predictable results.

I hope everyone will forgive my slow progress, but I want you all to enjoy these designs without having to resort to some horrfic mish-mash with wires spreading around the board like a rat's nest. One of my lecturers, many years ago, drilled us that a board that looks good, works well - he was right too. I don't fully understand why but there's no question that a board that's been carefully designed AND tested will give many years of trouble free circuit than the Heath Robinson pile on my desk

Valve/Tube Simulation

Oh, the valve simulator is something of a late-comer. It's really meant for vocalists using a modified version of the desk mic. If I go the way of a modular design (or do a V3) it will drop in there. It uses some simple electronics to simulate the "soft" overload characteristics of a valve amplifier. 

I'm not a fan of overloading, but it can be used for artistic effect (most notably "fuzz"). Transistors enter their overload zone - clip - at the peaks of the signal and do it very suddenly. Now if you've ever heard a pure square wave oscillator you'll know the awful discordant noise it makes. This is all the harmonics (extra notes) that a square wave makes and as our ears are trained to react to gently changing pressure, that's like fingernails down a blackboard.

Valves clip too - every amplifier does if you push it hard enough, even the microphone will if pushed hard enough, as do speakers!

But unlike transistors valves don't come to a full stop when the limits are reached, they run out of steam more gently and bend the peak over slightly long before they clip it off. All of this happens thousands of times a second but rest assured that our ears can hear tell the difference. This is part of what's called the "valve sound" that musicians love. There's another valve sound that is more difficult to simulate and that's the effect caused by limitations in the transformers required to interface the output "toobs" to the speakers. I'm not 100% sure if there's a direct (cost-effective) way to do this without actually using transformers.

Still, I've waffled on enough - I thought I'd reply quickly while the cat was on my lap for a quick scratch. But as is, the bugger has decided it was time for a nice long snooze so I'm trapped.

I'll keep everyone posted when there's any significant news.

Hello! I'm in the process of building the mic and I was wandering if it's possible to plug the mic output into the input of an Apple dongle DAC as the USB interface, I use one as my PC headphone out and it would be nice to have an all in one solution. Can anyone help me?

I'm planning to build this but I'm confused about some parts as I don't have any experience in circuits or building. 
for starters a lot of the items I don't know where to get like:

Brass strips, what sizes were used and where can I get them?

Brass mesh, I can only find small 1-inch meshes. there were sheets but seemed too thin.

Thin copper wire, are I supposed to break apart some random old toy for this wire?

Desoldering wire, Will any desoldering wire work?

Brass wires & rods, how many of each is needed and of what lengths?

Wood, What are the dimensions for the "AMP BOX" used in the video 

Tools, What tools are used in the build?

Also since I'm located in the US what are some ways I can get the parts for the stripboard and transistors? 
If someone could explain or drop a link for these items that would be great.

@matt269 It should be fine as the output has DC blocking so it's pretty agnostic when it comes to interfacing to a consumer DAC which will expect well under a volt. Just take care that you check it with a meter before you hook it up. While many DACs have an input blocking capactor, quite a few don't and that can cause problems.

@kirby All the metal parts should be available on Amazon. The mesh is quite thin as it's not stuctural and used as a Faraday cage. All the electronic parts are widely available, eBay is the best place to start.

Mouser.com will have anything hard to get hold of - in fact given the shipping costs, you might be as well to get everything from them. The THAT1510 is pin compatible as are the INA217 (Texas) but be aware that the gain equations for both parts are slightly different and that affects the choice of resistors.

There aren't any complex tools required but bending brass can take some practise. A hole saw will be needed to drill a plug to form the front/back grills.

Matt and I have found the LSK170A transistor is a good replacement for the FET which is no longer made and therefore troublesome to track down. It comes in a TO92 plastic package so doesn't feature the screening pin but that's not used. You just need to check the pin out to make sure you don't mix the gate with source or drain. FETs are pretty tolerant if you swap source and drain but they get awful fussy (i.e. break) if you put a positive voltage of more than a diode drop at the gate (it's intended to go from 0 to some negative voltage).

It also performs quite a bit better than the original part too.

Non-polar electrolytics aren't strictly necessary and can be a bit hard to find. The input ones are polarised with the positive side facing AWAY from the THAT1512 and the output one has the positive side towards it. The reason for this is complex but there can be a considerable amount of DC appearing at that output which you really don't want going downstream.

The non-polarised ones are specified by the THAT corporation for p48 (48 volt) phantom power in cases where the mic can be unplugged. Mind you, without protection resistors, the discharge of over an amp is enough to blow the delicate input circuitry anyway (THAT now say you should put a 10R resistor in each line to prevent that).

You can use ordinary electrolytics but wire them in series (so you need ones twice the capacitance) and wire them back to back. Looks horrible but it does the job.

EDIT:

I apologise for missing an important question here, yes the de-soldering braid does make a difference. It all should be a braided tube off very fine copper mesh but I've found (as have others) that some braids are just flat braids. How to tell? Probably buy from a large supplier or several small ones on eBay. Although the size is specified, there's nothing to say that they maker can't just say 2mm (wide) rather than 2mm diameter.

The copper wires you need are called ECW or enamelled copper wire. Unlike plain copper wire (PCW?) or tinned copper wire (TCW), ECW has a very fine insulating layer of very strong paint that has to be scraped away. You can get it from a quite a few places - I get most of mine smashing up old wall warts (the heavy ones) because they contain transformers that are inefficient compared to a switched mode supply. 

However you can also get it from any large electronics store like Mouser and of course, on eBay, but as said previously, with the shipping costs from places like Mouser, it's probably better to use one place as a one-stop shop as you'll either save on shipping or even get it for free!

@forrest The main board had an error that only showed up after I got them back which is why I can't order these in any good consicence.

The main "mic" head boards look to be OK but they are a bit tricky to solder. I don't know at this stage if we should make a small adapter board to covert the very small (but still hand solderable) connection or just give people instructions on how to do it.

It's taken me a couple of weeks (this is soul destroying work) to track the actual problem down since there is very little good information on these devices and even large-scale manufactuers have differnet ideas about what's best for a given solution.

What we have is a form of P15 power - which is a standard Matt's made up for this job. As a wise man remarked (in computing) the great thing about standards  is you have so many to chose from.

P48 is the main standard but we don't need it here since the plate is pre-charged at the factory. The P15 (or should that be P30 since it's +/- 15V) is completely different and while it seems to work well, there are a bundle of issues that can cause issues. The one that bit me right on the hiney was the slight imbalance between the + and - inputs which doesn't show up in a simulation but rears up immediately on a prototype.

Imagine there's a 1/100th volt difference between + and - at 60dB that appears on the output as 1V whcih can be blocked by a capacitor but it's less than ideal. The DC (common mode) should be 0 because we have DC blocking capacitors. Only... we sorta do and don't. Capacitors charge up by causing a voltage differential across the device - had me scratching my head for a while since all my instrumentation amplifier designs have had to deal with PURE A/C. But here we have a 15 or -15V volts on each input respectively and that causes quite a large differential DC to appear at the inputs which I hadn't allowed for. My oversight entirely.

There are two solutions to this, the active one as suggested by TI is elegant but complex. THAT's is a bit simpler but does mean the insertion of a fairly chunky electrolytic.

TI's is a servo which "watches" the output voltage for DC potential using an integrator and then corrects the output offset by adjusting the "ground" reference pin. Clever but means adding another op-amp into the mix.

THAT corporation, on the other hand, just places a capacitor in the gain circuit so any DC is negatively fed back 100%.

I'll have to write all this stuff up for other people who will stumble across this issue when they are using the THAT (and other instrumentation amps) for audio purposes.

The problem with the way THAT Corp. does it is this. It can roll off the low-end of your pre-amp, creating a high-pass filter in effect (which gives me an idea actually) so for maximum gain you need a VERY large capacitor to push the -3dB point down well below the 20Hz base end of the human hearing range.

How big? 6,800uF is its recommended one - for 60 db gain - and they are only available in nasty electrolytics or quite pricey audio grade ones. Not THAT costly but still... Its so physically large it won't fit on my current layout (not a major issue) and that's what I'm up to now. Re-routing around that and putting the volume control on a separate supply since the poor NMA0515 (power converter/inverter) can't supply enough juice. That niggle killed off the active high and low-pass filter too. Poor thing. 🙂

There's some math behind all of this stuff and this isn't the place for it - I've got copious design notes with my assumptions (and cock ups) that I'll be sharing after I hit the button to send the headphone section to China for final assembly.

I'll drop in a 4.5V supply for "normal" electrets if there's room since a few people have asked for that and even 15v is likely to kill the internal FETs PDQ. The JLI2555 are a little tricky to source but there are some stellar FETted electrets out there (JLI make one but that's reflected in the higher price).

There's no sound with this (the irony!) but no one wants to have every cat in the neighbourhood come running to their door or have the dog sing along. This is a silent update so everyone knows I am working on this, but as I've alluded to above, I ran into a set of buffers because I had a DC offset in the THAT section (and some less important errors) in the prototype, all now fixed that prevented it from working.

(Sorry about the whip-pan, there's nothing hidden in the cut, I promise) but you can see I'm using a nasty little JLI2555 copy from Ali so even if could win the X Factor, you it wouldn't give the design a fair shake.

It should be possible to see the THAT1510 (10, not 12) while I pull the camera around and you can see the utter chaos that passes for my workbench and the mic head unit which is in the bottom of the shot. 

The 1510 has a slightly different gain equation but for the sake of the discussion it's set t0 60dB or 1000x amplification which is sensitive enough with this gear to pick up the gnomes chatting up fairies at the bottom of next door's garden. I'm joking, fairies aren't real... not so sure about the gnomes. They look a bit shifty.

It's not clear here but at x1000 with my "yodelling" and nothing particularly clever going on, I was seeing a peak of 10V (5v each way 0r 3.5v RMS) so it looks like I'll have to throttle the THAT a bit unless we put a "booster" setting in there - this signal level, exceeds the input for consumer ADCs and even professional grade ones too and and that's without trying terribly hard. A 20dB cut switch should suffice actually since that would take 10V down to 1V which is just about right.

Condenser mics are very sensitive which is why they are so popular compared to dynamic mics which are great for speech or a rock-performer but not so hot for more subtle instruments like pianos, harps, etc. Even natural recording is far better $ for $ with a condenser because they have a more even frequency response.

It's not clear from the video (not a lot is) that this is one of the experimental heads that comes with its own FET out of the box. I've done six different variations - I'll find a victim, sorry *volunteer*, to help do some sound tests when I've wired the other five up. The one I'm most curious to test has an Op Amp in a socket so you can pick one that suits you provided it has relatively low power consumption and a JFET or MOSFET input stage. TI/Burr Brown's OPA series should work OK, but I don't believe in putting my mouth where your money  is so I'll test them first!

For the curious, this is the circuit under test. No prizes for spotting that I really ought to have put the ground pin in the centre. It doesn't make any difference for soldering (the ground is carried on the screen) but it looks bad. (I'll slap myself with a haddock after this).

The "workhorse" FETs used here are Toshiba 2SK208s [2SK208 (semicon-storage.com) picked because they are actually specified by Tosh for this application which means you won't need to source a particular device. The current orders come with all six versions but I can't say when they will be on sale or for how much but I don't expect them to retail for much more than the cost of a decent dual-monolithic audio FET. Right now, I'm sitting on my hands until Matt and I can make sure we can arrange a reliable supplier... I barely eBay so there's no point asking me. 

JLC PCB will make them for you (fully populated) but the price goes down dramatically the more you order so even though I will publish the Gerbils (poor little rodents) in time, it's not likely to be cost effective to order a set (30 boards in effect!) unless you have a lot of mics or just want to experiment. I did a six-pack of different designs so I could test out different variations on the same theme.

 Oh, and yes, it is fully (pseudo) differential too even though that's not strictly necessary. The vast majority of noise pickup is from the unscreened parts of the capsule and the PCB screens a good chunk of that by design.

EDIT: I should probably point out that this is running off a pair of 9v batteries with differing charges. I did this to make sure that this design isn't too picky about its supply. (I've borrowed some tricks from op amp design to achieve this which makes it fairly agnostic to supply ripple. I have to put this assembly into a Faraday cage so I can find out how much noise this - unscreened - cable run is. 

@marcdraco Thank you for all the help, but I am a complete newbie to this stuff. 
what is the LSK170A replacing in this context some "FET"?
some other stuff and the build needs more resistors or else it could blow?
Right now I want to focus on the build itself first like the measurements.
Starting with the brass strip I'm guessing a 1/2 width and around 1mm thickness for the mic band. And a 1-inch width of around 1mm thickness (maybe more) strip for the "AMP BOX"  but its just an eye-ball, also I have no idea what lengths the rods ended up to be.                                                                                                                                                      Some opinions on the lengths would be nice or do you think this is more of a loose template to follow?