Skip to content
Notifications
Clear all

New mic design.

13 Posts
3 Users
4 Reactions
428 Views
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 
IMG 4561

Anyone following along with my attempts (and multiple failures) to get the noise down on Matt's design (which was already very low noise) will know the lengths I've gone to. I haven't given up, but life got in the way.

So in order to try something new I've done a bunch of new designs; one works (Perks) with Matt's original, you just add a JFET and you're done, one (Sarah) comes complete with its own low-noise JFET so it's even easier.

The others are experimental at this stage and I need to pluck up the courage to sit down and carefully assemble them, but (if they work, and they do on paper) this could be interesting.

So to the design ideas:

Joshua: requires a 2SK170 (or compatible FET) and has about a 20dB gain at the output. Single rail supply of 12V and up. NE5523 configured for virtual earth and a single gain stage.

Ouroborus: Has about a 20dB gain at the output. Single rail supply of 12V and up using a 5532 wired as in a feedback to increase the effective input impedance, thus hopefully removing any FET noise, but we're still left with an op-amp.

Sarah: uncommitted JFET on the board. Might work with Matt's original but I don't have a working board to try it with (bad Marc, who's a naughty boy?!)

Perks: uncommitted JFET board for single or dual supply. 

Roman: dual FET (2SK170) design for a constant current to reduce noise. More details to follow assuming it works.

Phil: essentially a combination of the Joshua and Sarah designs. Single supply 12-20V.

Running out of names of mates, I've gone all Harry Potter on the last three:

Incremento: A daughter-board for some of the other designs with a 20dB gain and power pass through.

Redigo: Converts a 48V DC (phantom power) down to a lower voltage with an uncommitted socket for a zener of your choice. I've suggested a 15.6 volt one for 15V regulated out. It has a 10uF 65V capacitor on the output to block the DC from getting back to the signal side.

Auris: Now this one is the gold standard (IF it works) because I've had to special-order the part - a MEMS mic - and design the pad layout so as of this writing I'm absolutely crapping myself in case I did it wrong. 😉

Auris is driven by 3V MAX - so can easily work from a regulated LDO via the USB power or even a small battery for the IM68A130V01 (IM68A130V01 (infineon.com))

But it gets better - if my math is correct (cough, splutter) it should be possible to mount two of these back to back - and yes, I goofed and didn't think of this until this morning, so there are no mount or pass-thru holes -but due to the very small distances between the actual mics, this has the potential to emulate a cardioid or omni mic at the flick of a switch.

This also has the potential to be made into a very good shotgun mic by venting the the rear of the two units - like a "real" shotgun has. The idea is that sounds arrive at the rear of the mic (the antiphase side) at the same time as they do at the front and sounds from the front arrive later and don't get cancelled out. This makes it highly directional (you can't do this with an "omni"). 

But as I've said this is all experimental - once I've validated these work as they do on paper and I haven't dropped the ball somewhere (I've lost a lot of balls recently).

Rome wasn't built in a day... but props to JLCPCB who managed to turn these around (from my PC to my door) in under a week!

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 13/11/2023 4:40 pm
matthiasgalvin and Roman reacted
(@roman)
Posts: 5
Active Member
 

Hey, just saw your messages, but I need 5 posts to reply

 
Posted : 14/11/2023 7:51 pm
DIY Perks
(@diyperks)
Posts: 150
Estimable Member Admin
 

Thanks for all your work on this Marc! Great summary of the individual designs. I won't be able to play around with mic stuff until Jan but I'm looking forward to testing out your circuits!

 
Posted : 15/11/2023 5:46 pm
marcdraco reacted
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

@roman no worries mate. I’ll send you more when I have everything clamped down.

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 15/11/2023 9:32 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

@diyperks Auris works a treat. But as is (it was always experimental) it’s only a mic level output. 

I’ve tested some of the others (roughly) but they need a screen to really get the performance analysed. The drag of only having a couple of capsules to play with means I’m using cables rather than taking advantage of the actual screen on the board. 

Now the real killer is this - with the right op amp (TI make a couple) this entire head unit can produce line level output from just as 5V supply.

I was telling Roman earlier and this is what we spoke about back in Jan. The whole shebang without needing the 30v converter, and all the work is done by the head! And because the IC is standard, it’s a drop in replacement. What works with the 5532s, works even better with these recent very low power amps! 

One design (and idea I tried a while back) does away with the FET entirely!

But better yet, Infineon make dual output and even digital output devices that fit in the same space. Slightly different layout but in both cases I need to sort a way to get them to JLC as they need special gear to make them. 

The boards are quite low cost assembled with the passive components so anyone with a small SMD soldering oven could make the high-performance ones at home. 

I’m looking to see if we can use a socketed 8pin DIL for upgrade but SMD is better as the device sits closer to the shield. 

These experiments have been very useful  in developing several different mic designs but as I said, screening is an issue I have to resolve before I can finally make a proper test.

Not gonna lie - I’m having a ball! I’m just a bit ashamed it’s taken so long. 🤓

I think it’s a good bet I could get the Infineon up to line level on that 2” board with an amp. Single supply too. Which is better for everyone. I’ve had an idea to stack the boards - in fact this batch can do that quite well - if I got it right (IF) and I have to try this yet, the the “mic out” boards (except Auris) will piggyback into the second amp and give a better level out for the digitiser. 

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 15/11/2023 10:01 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

Good news! Looks like I can get the OPA2992IDGKR and make a head that runs at 5v! With up to 40dB gain at the head. I'll get them ordered soon although it takes a couple of weeks for them to get to JLC.

Now to see if we can get those Infineon MEMS mics! WOOT!

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 16/11/2023 4:59 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

Quick update - I haven't checked the other designs yet (poorly tum-tum, which is naturally what happens when I cook my own food).

But I'm *experimentally* (yes more) trying out a DIL version of these designs. It's difficult because the size of the socket is such that it goes over the centre hole where the "live" pin of the mic is soldered. It's not impossible just... er, cozy.

These won't cost me a boatload to produce if I only put passives on them and I'm quick enough to catch the Black Friday sale. At a PCB board house? Apparently, yes.

This will allow the mounting of a variety of different op amps to suit the mood (lower noise, maybe) lower voltage and standby current are definitely up there and a real issue with the NE5532 which is as greedy as dog that's been fed on MacDonalds for a week. 

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 16/11/2023 7:24 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

@roman @diyperks

Well finally a bit of good news (hey, it's only been nearly a year) 😉

Minor bummer - the Ouroboros is too noisy, even when screened so that's a failed experiment but it was always an edge case so I won't lose any sleep over that.

I've crammed a couple of the new designs into an old mic case and the results are better than I'd expected. A great relief after all this time.

Auris (for all it needs external parts which are special order) works a real treat and doesn't need any special screening. 3V in, mic level 400R output which will drive any MIC input - note, not a line level on a digitiser board, although some do have a "mic" level input for this reason. Professional pre-amps ones similarly have this.

Phil - which is currently set for 20db (x10) nearly blew my ears off it's that sensitive. However that won't be enough to given enough "oomph" to the digitiser so my next job (when people leave me alone long enough to take a breath "Dad, me car's broken down... again" ugh) is to hook it up to the "backamp" - Incremento which was designed specially for this task. Designed - but not yet tested! That's got a clear 20dB gain bringing the actual level 100x which should be enough to drive the digitiser. It is possible to do this with a single stage (x100 is pushing the NE5532 about as hard as I'd like given it's GBW* of 10Mhz).

Unlike Auris (which is RFI and hardened) it does require some amount of screening - but nothing special and it will be compatible with Matt's original head basket design. They all are but only "Perks" is designed to work with Matt's original pre-amp. It's nothing more than a pretty mount with some extra screening for but designed for a more modern and lower noise FET.

More details to follow - with some samples when I jury rig a recording device to the mixer. I've noticed my commercial (Peavy) mixer is quite noise in its own right though so I'll see if I can bypass that. 

----
* Quick guide: GBW or Gain Bandwidth is the maximum frequency the amplifier can operate in open loop or with a gain of unity (for unity stable amps). The quick and dirty method that they won't tell you at Wikipedia is to divide your GBW (10Mhz) by the gain of the amp, so for x100, the amp starts to roll off quite sharply from 100KHz which is well outside the audio band. x1000 brings that down to 10KHz so that's best done over multiple stages.  

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 17/11/2023 10:34 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 
54ECA619 292F 44B6 B03F 8F586D3C0753

The test rig pulled open in a a Newer NW-800 donor body. The head unit would normally go in the basket but that’s not a requirement

and yes, this one works, beautifully.

There is room for a digitiser in here but the system requires 10-12v to work, so cannot be powered from USB as per Matt’s original. 

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 20/11/2023 7:53 pm
(@roman)
Posts: 5
Active Member
 

Hi Marc, I don't think my email got to you, Need just one more post to reply to the messages on here, so il just do that when this post gets approved

Either way nice to see the progress going on, some of these boards definitely have potential, Hows Roman going? just curious on how my 3cm PCB twin is doing

Oh and this reminds me, Iv bought some of the parts way back to try and assemble Matts original design, Iv just gotten around to CNC'ing the PCB and soldered everything up. Let's just say I'v managed to mix up the source and drain polarity from coming from the NMA. So Drain ended up getting -15 volts and source ended up getting +15 volts. So I fed the THAT1512 pin 2 and 3 the wrong polarity. Was wondering if you reckon I killed the THAT or j-fet or anything else in the circuit. I get a interesting sound that gets higher and higher then stops and repeats. Though Iv managed to get a few seconds of (rather noisy) actual sound. The second picture pin 2 and 3 have the wrong polarity in the schematic and that's what I first soldered. Any insight would be much appreciated. Also thinking it might be the fact that I have a ground going from the 3.5mm jack and a ground from a USB, since I have a separate USB cable for power and a 3.5mm for audio. Matts had just one ground coming from a USB cable since he used a digitizer. Don't know if what I just spewed was nonsense.

 

 

Screenshot 2023 11 20 221014
Screenshot 2023 11 20 223118

 

 
Posted : 21/11/2023 6:39 am
marcdraco reacted
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

FETs are weird, but for the sake of this cockup (and lucky for both of us, as I just did it myself!) they are unipolar. Normal BJTs are bipolar (two polarities) so they can easily be destroyed by reversing the connections.

In effect a FET is (in simple terms) a variable resistor between the Source and the Drain terminals with a control lead, the gate, which controls the "resistance" if that makes sense. That's a crude way to describe it, but it's useful way to think of them. The reason they rarely see use in variable resistance applications is that manufacturing the damn things is difficult - well difficult to make them behave with a tight tolerance like we can with bipolar transistors.

In summary, I think you'll be fine - so much so that it's often not that important to get them in the right way around since most (not all) exhibit the same behaviour regardless. What  matters is the difference in voltage between the gate and one end of the channel.

Noise issues can be caused by lack of screening and can cause the amp to go into oscillation which will swamp it - and if that's up in the radio frequency range you won't hear it but the whole thing will appear to be "dead" but noisy.

Your little 2.6cm (what's 4mm between friends? 🤣) clone is one of the more "advanced" designs inasmuch as it uses dual FETs to linearize the output more effectively than the single FET design so it's output is quite feeble and needs a good 40dB gain stage to test it, but I'm also fresh out of LSK170s to test it - the current version has a pseudo-balanced output which is a bit fiddly to solder given that I changed the screening system on the others to have a separate power and analogue grounds as you can see in Phil - this virtual ground is used in several of the designs.

I'll post full descriptions when everything is ready. This Phil adds some RF suppression in the feedback loop to kill very high-frequency noise. Shouldn't matter if its screened properly, but this is belt and braces.

The design used a common-or-garden NE5532 which, it turns out, is a very good chip indeed according to Douglas Self who forgets more about audio over his morning cornflakes than I know...

It will even work on Veroboard if anyone wanted to put it together at home - swapping the 100K feedback resistor for something larger will increase gain (1M gives 40db or 100x) at the cost of more noise. Although with some other changes it's possible to use 1K and 100K in that loop. More on that later.

image

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 21/11/2023 1:38 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

Well I found a couple of LSK170s soldered to an old experimental board so 10 minutes of furious fighting with a solder sucker and lots of burny fingers, I had them into your little fella.

And he works like a bloody trooper! I didn't try him in "resistive" mode, rather going the whole hog with the pair of FETs for ultimate performance although, in all honesty, I couldn't really tell any difference after the Peavy mixer had run it through its (fairly noisy) electronics between that and the other ones that use the 5532 and 2SK208 - no question it will be better, it's case of how much and it's the only one that will drive a "proper" microphone input on high-end gear - balanced or single ended. I've put a unity-gain phase splitter on the next set of boards so it can be made to operate in fully differential mode for the rich and discerning. I mean, spend a $1000 on this thing it must sound better right (even though there's no difference on paper... cough, cough).

Like all the "plain FET" designs, it needs about 40dB of gain before wiring it to the digitiser and while I have enough bits to do that, I ran into the cat's sleepy time and he wanted HIS chair back, so I'm out of action until tomorrow now.

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 21/11/2023 9:22 pm
marcdraco
(@marcdraco)
Posts: 453
Member
Topic starter
 

Happy Christmas/Holidays everyone

Egad! It's been a month already!

Some of this post is a little technical, but there's plenty for everyone to enjoy and please, do feel free to laugh heartily at my expense. Trust me folks, I'd rather everyone learn by my mistakes because I make plenty of them and there's no reason it should cost you.

For anyone following, I have a load of new designs tested and working. A couple need bit of TLC, like the one where I removed and reinserted an Op Amp into the schematic and hadn't noticed they were the wrong way "up" (+ and - inputs reversed) so that went about as well as expected!

STEREO

Ok, so this one is one of the most requested features. While not useful for a desk mic (even if there are several of you, each one would have a mic) a there are several designs that use baffles and omnidirectional mics to arrange just two microphones (most stereo recordings use three to mix the centre stage). I can't find a picture but they do exist and while bulky (they're about the size of a typical bookshelf speaker) they are said to produce a sound similar to binaural recording but one that can be played back via speakers.

BINAURAL AUDIO

I've also done an omni design for insertion into several projects seen on the web such as the Jecklin Disk which is a simpler than the human head design for TRUE surround sound recordings - although you have to play them back through headphones which means it's not suitable for speakers (it sounds odd).

These boards also make decent "boundary" microphones for conference use and even find application in theatre if that's your thing.

BME-800 Upgrade

Upgrading one of these Neewer mics is now a doddle as you can add several of the new boards directly into the head basket! While Neewer's default capsule isn't in the same league as the JKI-2555 or one of the Primo units it has the advantage of coming with a stand so the driver board can be soldered direct to the mic and the mic pushed back into its holder.

The only gotcha with this is, by default, Neewers are designed to work with 48V phantom power (although I have ... "questions" over the ones they supply, it only has a single signal cable, not the two you'd normally find here).

Even so, these things kick out far more juice (by design) than a typical mic input is expecting and many run from 5V (again, this is by design) so you won't need a costly microphone level digitiser as these feed the DigitalLife USB 2 audio capture card (sold by Amazon) or pretty much any ADC you like. 

At the moment I'm working on a USB power cleaner because my computer seems to have a particularly bad case of crud getting back onto the power lines and many others are too. USB3 might be better but this is a belt/braces approach. 

Headphone Section

Hands up, this is where I goofed (again!) I mean who would have thunk it, in the words of Deep Thought (Hitchhiker's Guide to the Galaxy) "I checked it very thoroughly" and... much to my annoyance the LM324s powering the 5V version of this design aren't stable into a capacitive load like the NE5532 is. (The math is a bit beyond this discussion and WAY over my head as it's to do with poles and zeros and stuff that looks like hieroglyphics). But the bottom line is it works for about a minute and then slowly starts to oscillate at a very high frequency (outside of hearing range) and stops the headphone section operating. This, naturally, is what happens when you can't run a simulated IC and even if I could, most SPICE sims can't simulate this sort of crass instability. I should have breadboarded it but I just took the existing 5532 design and swapped in LM324s.

(The feint dripping noise you can hear is the sound of me crying into my beer... glass. Because I spent all my beer money on the last batch). 😉

At least the Baxendall section works (and it works a treat) so volume adjustment is a doddle and you can use a choice of a set of switched resistors or a linear pot to get a proper logarithmic response that our ears expect. And none of this fiddling around with weird values that THAT dumped on Matthew and other designers battling with this design which is technically "just" an instrumentation amplifier.

The THAT1512/1510

I've done a couple of versions with the THATs on but the traditional designs are, to my utter disbelief, as good in this application - most of the heavy lifting is done by the LSK170 FET. Being SMD they come soldered to the board and save a lot of space. Oh they're a great deal less expensive too. That's design is unquestionably superb and a great choice but it's also a PITA to drive - you'll note Matt's original design used an NMA0515 rail-splitter to get a split 30V supply from 5V in! I originally blamed the USB noise on that device (the noise leaps all over the board making it an absolute nightmare to pin down if you're not expecting it). I honestly admire Matt for sourcing these parts in the first place, let alone getting them to work with a 5V USB supply.

PSU Improvements

Apart from the improvement to the microphone's power supplies (the less USB crud that gets up that line the better as it's getting amplified by 40dB and that makes it very intrusive) I found a reference on USB 2.0's specs that the largest capacitor we can have is ... *drum roll* --- 10uF!

The problem is that USB2.0 is limited to half an ampere (500 mA) and the "inrush" current is quite spectacular as it charges up. The capacitor looks like a total short circuit at first and charges over time - and for a few milliseconds the draw can easily exceed that 500mA (in simulation, with one ohm internal resistance it peaks at 5A) which is potentially enough to cause a potential shutdown or damage over time. A lot depends on how fast the limiters can kick in many operating systems will flag a bus-powered device if it tries to draw excess current for a few seconds).

USB3 is far more developed "better, stronger, faster" (and if you know that reference you're showing your age) but it still has limits that we have to observe if we're going to be good. I've included the simplest solution in the latest designs. The new supply is safe for use on USB2.0 and above which is all that matters (and it won't kill a wall wart either).

image

 Not my idea but one that's done the rounds for ages. The capacitor charges through the 100 ohm resistor without looking like a direct short (it peaks at 500 mA but only for a few microseconds). If the downstream electronics suddenly demands a short boost of power (someone screaming into the mic for example) the capacitor discharges back through the diode (this weird looking diode has a very low forward voltage drop and is ideal for this problem).

For the love of Odin, I hope this last batch is the last of this project. I'm just trialling a couple of ideas for minor tweaks which I'll include in the newest versions. When I have a completed set of working designs (sans gremlins) they'll be on my GitHub page ready to load into KiCAD. While the electronics may be beyond some of you, there's plenty more you can do - particularly in designing new board outlines to your own specification. KiCAD has some (occasionally buggy) but generally pretty good features for designing and testing your own and as you've seen here it will render them in glorious 3D (even raytraced) so it's a breeze to change the mounting holes, etc. The "stereo" pair of MEMs boards (ones without amplifiers) could easily be altered into a boom microphone for a headset - and you're feeling really posh, it can even be put on a flexible PCB!

OK, so time to get back to it. But I hope you all have a great break and who knows... you might even get one of these in the New Year. I have a limited number to give away but no way to do that so I might have to nod in Matt's direction for that one... (sorry mate).

Take everything I say with a pinch of salt, I might be wrong!

(Cybertruck avatar is a riff on my inability to deliver my designs in reasonable time so far.)

 
Posted : 19/12/2023 7:40 pm