Notifications
Clear all
Hey, no sweat man! This is a complicated build (the next one should be easier) but it's gorgeous so well worth it and I'm happy to help any way I can.
Don't worry if your measurements are off a few percent - even 10% is quite normal and FETs are notorious for being a bit weird. 🙂
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 02/03/2023 8:07 pm
OK @diyperks here you are - the full schematic for an lower noise, lower distortion, lower voltage and (woo hoo) cheaper mic amplifier for Fetless condensers like the JLI2555.
Missing from here are the supply decoupling capacitors - might need some at the head too but I'll know that when I have the board designed (I'll do that tomorrow as the cat is getting fidgety).
As shown, this outputs around 500mV for 1mV in. This should be enough to drive a USB digitiser although the capsule board's output is quite sufficient on its own to connect to a microphone input.
I plan to add a 20dB pad and a high-pass filter (at 60-80Hz) to the final design but this is fully functional as is if you want to experiment. The Op Amps listed here are probably not ideal at such a low voltage so a low-voltage BiFet should be chosen the TL07H is a good modern version. The JFET shown is in current production but this design will accept a number of replacements. Key is a low pinch-off voltage and the ability to run around 1mA.
Some JFETs “prefer” to be run closer to the middle of their rated Idss but adjusting that is a simple matter of adjusting R2 to a smaller value since that sets the collector current based on the voltage at the base (say 1.8V) minus a diode drop (say 0.7V) giving a base-emitter voltage of 1.1 volts. With a 1K resistor that gives you 1.1mA – simples!
The capsule board will (just) fit into a 25mm capsule with the gate connected as close to the centre pin as possible and the ground shielding connected to the edges.
I've lashed together a head unit but I'm waiting on the FETs arriving from eBay so I can make a working prototype. It even works with your existing wiring too, which is fantastic, I LOVE that project.
Using a better BJT than the 2N2222 that I've specified here will lower the noise floor but I don't have the equipment to check just where it sits. Low-noise, small-signal amplifier BJTs are easily sourced, but I had a couple of hundred 2N2222s sitting in a drawer!
This post was modified 3 years ago by marcdraco
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 02/03/2023 10:51 pm
DIY Perks reacted
1. do you think that a laptops internal sound card would produce better sound than a usb sound card?
2. Wouldn't this be useless since theoretically the soundcard would be what limits the quality of the sound?
3. Can't you integrate in your design a usb Audio DAC chip that fulfills 32bit/192 kHz?
This post was modified 3 years ago 2 times by axy_david
Posted : 03/03/2023 7:55 am
@axy_david Julian Krauss has an excellent breakdown PC sound cards
The upshot (and it's clear Julian has an excellent grasp of the electronics) is that PC sound is "acceptable" for most uses outside of Hi-Fi. What I've done here is separate the microphone from its pre-amplifier which means you can get a high-quality capsule, give it some power (say a 9V battery) and plug that directly into the microphone input of your choice of external digitiser.
The pre-amp, which really needs (well doesn't *need* but would be useful) is a 20dBV pad and something to roll-off the very low infrasound which can drain the signal somewhat. I'll add those today when I've finished testing the head design and brushed up on my KiCAD skills!
You're right that it's the sound card, usually referred to as a "codec" from "encoder/decoder", that controls how the eventual sound will come out.
Julian measures stuff that I wouldn't usually consider and yes, he does work in 192KHz, 32 bit audio - which is practically impossible to "clip" because it uses floating point math.
But as I've said previously, we need to consider the whole chain rather than a single part in isolation. I've bumped the head design to something that's more versatile than the original Alice using techniques I cribbed from Rode plus a few other stunts that only a few seem privy too.
This means it has very low self-noise and very low distortion, even with a cheaper FET and an everyday low-noise BJT in the signal path. Much of the distortion is due to non-linearities in the FET due to manufacturing tolerance, so this circuit circumvents that by adjusting the IDSS automatically as set by the designer, and that puts the FET into saturation where it's most stable.
Similarly the input impedance "seen" by the gate is pushed up by a bootstrap to the two diodes so the main impedance is limited by the impedance of the FET itself which is in the 100s of megohms. It won't load the capsule at all and you'll get the full detail from pretty much any electret condenser. This design isn't suitable for capsules that need to be charged (although a simple capacitor will sort that).
But the output from this is "mic" level - it's essentially the sort of value you'd get from an everyday WM61a with a Linkwitz mod because it's a source follower. The usual way to load a small capsule is to give it a 2k2 (thereabouts) to the gate and you get a little bit of gain so these mics sound more sensitive.
The second circuit is the optional pre-amp which I'm using TL072Hs for. These are not professional grade instrumentation op amps like the the THAT1512 but they are more than sufficient to drive a PC sound codec (or even an external one) because the majority of the noise will come from that part of the circuit and we don't have control of it. You could bump these to pretty much any low-voltage, low-noise op amp you want but it's unlikely anyone could hear a difference. Many master recordings of famous bands including the Beatles and Pink Floyd were mastered on equipment fitted with 741s! Latterly, many professional mixers employ NE5532s but that's a bit of a current hog so not idea for a USB powered circuit.
As to a 192KHz codec... that's a bit above my pay grade. The ADCs in these things are monsters and only come as SMD components i.e. https://www.ti.com/lit/ds/symlink/pcm3010.pdf?ts=1677773485465
That's an awful lot for a single designer to work on and I really don't have the time to spare. In fact, you can already buy loads of pre-made ones that would work out far cheaper than making it yourself - even if it came as a kit.
There are loads on Amazon like this and you can plug my circuit directly into one so you have much more choice.
https://www.amazon.co.uk/Creative-External-Multi-Channel-Discrete-Optical-out/dp/B0953LL5R6/
Then there are the Trotters Independent Trading sound cards like this P.O.S.
I've got one and I've even used it to test my circuits. Sure it works but the quality is about what you'd expect for a device that cost me under a fiver! So don't buy on price, look to someone like Julian to get an in-depth review before you give these guys your money!
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 03/03/2023 2:15 pm
Topic starter
@marcdraco Amazing work! I'm extatic about this as it's such a simple design, and looks like it'll perform significantly better than the original.
Are you able to walk through the basics of the Capsule Board section? My impression is that J1 is the impedance buffer, with Q1 performing amplification, but as you know circuitry is not my strong point. If that's the case I can imagine a good combination of the two will provide an extremely low noise floor.
I'm just wrapping up a project for the channel at the moment but as soon as I'm done I'm going to get out my mic stuff and build what you've designed and test it out - incredibly exciting!
Posted : 03/03/2023 4:47 pm
I've (rushed) something in KiCAD - but I'm not great at it and the components I've used are tricky to find in their footprint libraries - more my fault than KiCAD's. I'll do a better one [all Open Source] shortly. The idea is to mount this directly to the capsule so the central pin pops up and solders to the gate with the shortest possible distance. The rear of the board is a ground plane and I'll fill the upper bit too. Two wires and a screen and we're off! 🙂
The head unit is the crown of this thing and while it's possible to build it on strip (Vero) or perfboard, I recommend against it as you can see from my mess up there! That was to make sure it worked as advertised and it does - when I sort a decent screen (faraday cage) for this I'll send you an audio file.
So let's take it from left to right to see how this strange-looking thing works.
The 40pf capacitor is just there to represent your capsule and its connections. It's not actually a small discrete capacitor. Signal enters the gate across a bootstrapped gate load formed from the two reverse biased diodes and R4. The function here isn't obvious but it's usual to have a very large load resistor in his part of the circuit, typically about 1G although I've used a 1M with acceptable results. The problem is that you lose signal level because the 1M resistor forms a voltage divider in series with the capsule's "internal resistance" - that's a bit of a misnomer but even the best capacitors have a little bit of leakage at some level of magnification (think pico amperes!)
The diodes & 1 Meg resistor, bootstrapped by the emitter of Q1 deliver a very high effective input impedance for the JFET in the stead of that very large gate resistor. You can use a 1G resistor but I've priced through hole ones on eBay and they are eye-watering. SMD ones are under 50p by comparison. (Doug Ford's discussion of a 4-leaded SMD JFET with internal diodes gave me the idea. Apparently it's not made any more.)
The "clever" bit of this circuit is formed from a version of a constant current source which acts as a very high (but variable) impedance. Current sources are often drawn like this:
Depending on the circuit you can use anything from a resistor all the way to an op amp. You can even buy current regulator diodes which (internally) are just JFETs with the source tied to the ground! This causes the device to go into full saturation and deliver its maximum current, but unlike BJTs they don't go into thermal runaway and blow up.
The advantage of these devices is (for all of their expense, typically a few quid each) is they do save a LOT of board space. Redrawn with a current sink (ideal) as might be provided by a current diode reduces component count AND distortion (it's hard to know by how much as an ideal current source can't be made.)
Another way to do this (with temperature correction) is to use a dual JFET (I've just stuck the same FET in here twice but performance will suffer a little like this. The advantage of a dual is you have two transistors in a package with tightly controlled temperature coefficients. However, they're unlikely to be "matched" despite being on the same die!
The current source I've chosen for this employs the voltage drop across and LED although you can use anything from a couple of resistors, to a resistors and a diode, a zener diode etc. The point is to put a voltage on the base a few tenths of a volt above the Vbe drop of 0.6-0.8 volts.
This puts the voltage at the emitter resistor at the base voltage minus the base voltage. For ease, let's say we have a 0.6 drop across Vbe and we manage to set the base voltage at 1.6 volts (you can do this but it's rarely necessary). This gives us Ve - Vb = 1V at the emitter. Put 1V across 1K and (drum roll) you get 1mA of current through the transistor. It's actually a little more than that because the base-emitter junction conducts but that's usually only a small amount and remains constant.
The really neat thing with these circuits (they are an essential building block) is that the current stays reasonably steady even with variations in supply voltage and that means that wobbles are ironed out - something a resistor can't do because it's a passive device.
But the function for biasing a JFET is really clever. Now this made my brain itch, but because JFETs are depletion mode devices, the drain-source channel conducts when the gate is at 0V and only starts to close down as the gate goes more negative. It might help to think of it as "pulling" the gate shut rather than pushing it shut (going more positive) per enhancement mode devices like many MOSFETs.
JFETs are often self-biased by putting a resistor in their source-to-ground terminal and grounding their gate through a resistor. This works because as the device turns on, current flows and a voltage develops on the source resistor which pushes the voltage at the source up to a level preset by the value of that resistor. You'll often see this split into two, sometimes with an adjustable trimmer and often with a tap bootstrapped back to the gate.
But the point to all to all of this is when the voltage at the source is slightly positive with respect to ground, that means that the gate terminal has a negative voltage with respect to the source - so the transistor is now ready to work.
You calculate ("guess-ulate") the value of the source resistor from the predicted load lines which is a faff on at the best of times and even then manufacturing spread means it's all over the darn place so it's a world of hurt.
The current sink (I call it a often source out of habit, but technically it's sinking current) works in much the same way as a resistor with the added advantage that you know how much bias current you're going to derive from the source terminal: 1mA in the example.
I'm using a 2SK170 in my proto-hash which has a IDSS spread at 10V VDS from 2.6 to 10mA with the gate grounded (channel in full conduction) so picking a source resistor isn't easy! Given that it's easy to see why a current source (sink) is the preferred way to bias the little beggars.
TL;DR
All this malarkey is more about linearising the JFET's performance while providing a very high impedance to the capsule. The drive current is a tradeoff between distortion and power gain (output impedance). I've set it around 600uA to get a balance between output impedance and distortion. A lower output impedance increases distortion a little. Not that we're driving anything hard! The next stage (I'm using low-voltage BiFet amps) is biased at 50% supply voltage by two 100K resistors giving an input impedance of 50K. This keeps the current low but is still usable if you want to switch to a bipolar amp, say, an NE5532. Although the supply voltage would need to be raised to a minimum of 10V.
Note that C1 is only 0.47 uF? That's because it's feeding into 50K which gives you a cutoff of less than 10Hz!
On reflection though, the buffer was there for an earlier version of this circuit so I've done away with that and just dropped an inverting amp in the output. 22 uF into a 1K load - this is an inverting amplifier - (which the 2N2222 stage drives easily) has a cut-off off around 8Hz. So all it needs now is a high-pass filter (I expect a passive one will be fine) and a pad and we're done!
I've only tested the head so far, tinkering with the simulation is far too tempting, but it's a pretty simple circuit that's well described all over the place. I don't even think too much about it these days. 🙂
And 5 minutes of tinkering later...
It's not obvious here (I don't think Spice does bits of hardware like switches) but the if you take the 22 uF C1 out of circuit the low-pass cut rises to around 80 Hz albeit with a gentle slope. Similarly, shorting the 470k resistor R6 out of circuit will cut the gain from 500 to about 50 giving you that huge pad. Putting physical switches in a signal path is a bit of a no-no though. MOSFETs might work but the cat is getting fractious as it's his bedtime.
I'm sure we can improve this more I'm having far too much fun though! 🙂
This post was modified 3 years ago by marcdraco
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 03/03/2023 9:03 pm
@marcdraco I used the ugreen one and once opened it had a decent chip I suppose, decent if you like 48 kHz 16bit mono input that is, however it is still cheaper than an behringer audio 2 pc and you can solder it into the mic since behringer mic 2 use seems to not work if powered trough wires, bought 2 of them and same issue. I have not found any higher quality usb powered ones that have a small enough factor to fit into the mic.
And there still is the question if there will be any significant benefits.
Is there any specific reason why you decided to split the board into 2? I mean can't we power our JBL capsule with 5V and some something like THAT1512? Or you have another capsule in mind?
Posted : 04/03/2023 12:20 pm
I've only split the board (actually the schematic) into two so we have the primary impedance conversion at the head so it goes on a little PCB that will sit at the head (in Matt's original design) and drive any pre-amplifier. The right hand side is the pre-amp with a gain of about 500x.
If you wanted to put the whole thing inside a microphone case (Neewer 800, etc.) there's no need to split it and the whole job lot can be put on a piece of strip board. This is what I've done myself but it's such a pig's ear after being made in two parts an then rammed inside a Neewer 800 donor (so I can isolate the mains noise) that it's a complete mess.
The THAT1512 is an excellent op amp designed for instrumentation use but the circuit would need small modifications. The easy way to do that is to drive one input from the head via a small value resistor - 500 ohms or more and ground the other input with the same value resistor to match the input impedance on both inputs.
I would suggest you don't build this yet because noise floor performance is unacceptable and I haven't isolated the problem to a single component. I'm only using parts I have lying around (save for the JFETs) and I need to figure out if the noise is coming from the mic (and where on the mic) or the digitiser in my iMac.
The condenser itself shouldn't have noticeable self-noise so now I have to figure out what's doing what. On paper the noise level should be well below human hearing but it's not... of course it could be the difference between real parts and simulated ones. I've popped a 5532 in there for further testing but I had a phone call before I could do any proper testing.
Looks like I'll be on to Amazon to get some strip board (I hate the stuff but it's what we got). 😉
EDIT: Sorry, I missed your question there about capsules.
This will work with ANY capsule that doesn't have an integrated FET and doesn't need a 48V charge. The JLI 2555 is just one of the better quality ones but I'm using a copy of the Neuman one for my own tests. (It's rubbish of course but I used these to get a baseline.)
As for digitising, 48KHz is fine for speech and less demanding things such as vocals which is what these mics tend to get used for anyway. If you were micing up something subtle and demanding with complex harmonics like an acoustic guitar then I wouldn't use a USB interface at any price.
EDIT #2: As I'd anticipated it's the diodes I've used in the gate bias that's causing the vast majority of the self-noise. It works without them of course but the linearity isn't as exceptional. Looks like its another quick visit to the simulator. The mains hum I was dealing with on the desk was drowning out the self-noise. Now it's not "bad" but it's nowhere near acceptable. Lower noise diodes would be better but I expect for the application it's a step too far. Sometimes simpler is better, per Occam's Razor.
Gotta love development. At least I had chance to check a 5532 (audio op amp) vs. a jellybean JFET op amp like a TL071/2/4. And to my utter amazement, the TL072 was as effectively quite as the specialised one. It takes less power too.
However, using a TL072-type amp changes quite a bit more if we're building inside a
donor body since the FET head amp isn't required. The reason I separated them originally was so we could get a small circuit in the head basket similar to Matt's original. In a donor mic all you actually need is a follower like the one shown here (the first amplifier with its output wired to the inverting input) and the capsule can drive the non-inverting input directly.
We need to ensure that the op amps are correctly biased since there's a x500 amp in there, so we can't have any DC flying around the place so this sort of thing will drive a capsule directly. Not that the OP07 is not suitable for this purpose but a general purpose JFET capable of running at 2.5V supplies should be fine. I'm including this for completeness so you can see what we can "chuck" if we're making something in a body that's just going to drive a USB interface.
The first op amp does the job of the JFET in the earlier designs. There are loads that fill this job if you have those +/-15V supplies. The second op amp can be replaced by the THAT1512 provided you observe the input impedance on both + and -. 1k - 10k should be fine, you ground one input (doesn't matter which) and feed the output from the follower into the other one. The gain is controlled by the external resistor per Matt's design.
This ONLY works in totally screened case or it will hum like the Devil trying to write music!
Also I don't recommend building anything I'm working on until I've verified that I haven't dropped the ball...
I've found a clever way to make a very large resistor which should be low noise and quite effective.
EDIT: the previous version of this schematic lacked a blocking capacitor between the follower and the gain stage. Sloppy on my part and although this was never meant to be built, I know better so I'll now go and slap myself senseless with a wet mackerel.
This post was modified 3 years ago 6 times by marcdraco
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 04/03/2023 3:38 pm
Quick update for anyone following my mad experiments...
I've discovered (my to my chagrin) that most of the noise floor is was/is seems to be caused by the S & H in the ADC. So after pulling the entire thing apart and putting it back together on a fresh board... Well, I'd say I was pulling my hair out but I don't have any, so I've superglued a fluffy duster to my scalp instead.
You should see the looks I get!
Full build log when I get it tested with an audio amp and not an ADC since that's the only fair way to test it apparently.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 05/03/2023 9:30 pm
DIY Perks reacted
I've been meaning to break in my new smd/soldering workbench. Maybe making an ADC board would be a good way to do that. I'm looking at the PCM 182X series of ADC's because if I'm going to do it myself, I might as well go completely overboard with 32 bit 192 KHz (and give myself the hardest SMD job possible). What sort of considerations would I need to make on the electrical side of things? Once the signal is digital I'm pretty confident I can hand the i2s signal off to a microcontroller to convert it to USB audio. But analogue electronics has never been a strength of mine. As for the digital end of things, I *think* the rp2040 should be able to handle it. The Programmable IO makes the thing super flexible, they're easy to find, and they have excellent documentation. They're a little underpowered but as long as I don't have to buffer more than ~50ms of audio (which is an eternity in microchip time), it shouldn't be memory constrained. Not sure about being CPU constrained, but my intuition says it should work, especially with the help of the PIO.
Of course this is all just in my head right now, whether or not I actually pick this project up is another matter entirely. I'm just wondering if this is feasible at all.
Posted : 06/03/2023 1:43 am
The PCM182x consists of the following blocks and features:
Stereo-channel, multibit, high-performance delta-sigma (ΔΣ) ADC
Differential audio inputs with a 2-VRMS full-scale signal
Hardware pin control operation to select the device features
Audio bus serial interface master or slave select option
Audio bus serial interface format select option
Slave mode supports the audio bus serial interface up to 192 kHz sampling
Slave mode supports a dynamic range enhancer (DRE) with 123-dB dynamic range for the PCM1820
Slave mode supports decimation filters with linear-phase or low-latency filter selection
Master mode operation supported using a system clock of 256 × fS or 512 × fS
Power-down function by means of halting the audio clocks
Integrated high-pass filter (HPF) that removes the DC component of the input signal
Integrated low-jitter phase-locked loop (PLL) supporting a wide range of system clocks
Integrated digital and analog voltage regulators to support single-supply, 3.3-V operation
Natty little chip that. I'm intrigued
As a quick time-saver for others browsing, here's the schematic for a typical application (using a pair of MEMS microphones). For our application you don't need any more than the microphone head end using an impedance convertor like a single JFET or a single low-voltage Op Amp.
It does support a mono differential input directly so you could use Matt's original design and offer the differential output signal to the microphone inputs directly. They're safe up to 100 mV so that won't be an issue and the integrated capacitor (*high pass filter) saves the need to block the DC.
CMMR [balanced line noise rejection] is improved with the PCM1821 from 40 to 60dB which isn't amazing but over a short run is more than enough. In a shielded case, it's far more than you need so that's over the top so I would imagine the 1820 is sufficient.
Here's TI's layout recommendation for low-noise. Note how the capacitors are placed near to the chip and the extended ground shielding. This is the sort of thing where PCBs excel of course, but it does put us amateurs at something of a disadvantage.
I think I'd be tempted to see if I could get an evaluation board from TI rather than jumping in with both feet and finding the water is a bit deeper than I expected though. 😉
What were you thinking in terms of MCU? The STM series are excellent of course although the IMXRT chip used on the Teensy 4.x blows everything out the water and then plays footsie with the debris.
And Paul's boards are designed so the can be mounted as a daughterboard too so you won't need to faff around designing even more support electronics for the I/O, power bus and such.
Paul did an audio board with, I *think*, an RTL PC audio codec on there which is an option for people who want to go this route but lack the hot-plate/oven to cook SMD components. He's also done all the programming grunt work too although the audio quality is "only" CD levels.
This post was modified 3 years ago by marcdraco
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 06/03/2023 12:09 pm
@marcdraco I agree that an evaluation board would be great, but at $140 from TI and almost $200 from mouser, it seems like those boards were made for commericial situations where there's more money than time. Unfortunately I'm not being paid by the hour (or at all) to toy around with electronics. I think I'd be better off buying 10x what I need in $3 components and messing up until I get it right. For the Microcontroller Unit (MCU) I was thinking of the Raspberry Pi Pico (which uses an rp2040 chip). Reason being I already have it, it's easy to get more of, and the programmable IO should help with handling the I2S signal.
If I actually get anywhere with this project I could look at integrating the rp2040 chip onto the same board, for what should be a pretty neat package. If I get past any kind of proof of concept it might be worth picking different components that are available for assembly at places like JLC for those of us without a hot plate/reflow oven/heat gun for smd work. But to be completely realistic I probably won't get past a proof of concept (I have a tendency to pick up overly ambitious projects).
The programmable IO is a pretty neat thing. When handling a signal in an MCU you usually have two options:
1) To use general purpose IO (GPIO) and "bit bang" the correct values from/onto the input/output lines and memory. This process wastes a lot of CPU time doing simple operations, but can handle pretty much anything if you have the cycles for it.
2) To use dedicated hardware, where a physical portion of the chip is built for interfacing between the microcontroller and the signal. The benefit of this is that you aren't wasting any cycles on the conversion/processing. But the downside is these things take up physical space on the chip and hence increase cost, all for a feature that might not be used.
Programmable IO (PIO) sits somewhere in between. It lets you establish the rules of conversion beforehand and hands it off to larger but more flexible dedicated IO circuitry. In my opinion it's the killer feature of the Pi Pico, as it allows it to act as a jack of all trades and interface with all sorts of obscure protocols.
If I were to use another microcontroller it would probably need dedicated I2S circuitry. I know some of the audio focused ESP32's have this, but I'm not sure what else does (but I haven't looked very much). Pi Pico is easy to get and I already have 2 on hand so I might as well stick with that. I do also have a teensy on hand that I think supports hardware I2S, I'll probably fall back on that if the pico doesn't work out.
This post was modified 3 years ago by ttamttam
Posted : 06/03/2023 5:34 pm
@pygmalion I agree that the device we're connecting to SHOULD have DC blocking but I tend to be overcautious when I don't know what I'm driving. We don't even know the input impedance but it's fair to assume that it's quite large.
I'm a little confused by your calculations though. This is what I get:
The source impedance as seen by the output pin of the 1512 (input into this circuit) equals two 10K resistors in parallel which is 5 K.
The "wiper" (PW0) picks a developed voltage of at any given point but the load seen from the THAT1512 is always just 5 K. This never changes although it IS loaded slightly by the following stage. I'm assuming that it's of sufficiently high impedance not worry about. I'd probably double check with a voltmeter just to make sure no bias voltage is appearing at the input, just to be safe.
OK, so the -3 dB point is given by 1/(2*pi * r * c)
= 1/ (6.28 * 5 e3 * 2.2 e-6)
= 1/ (6.28 * 5000 * 0.0000022)
= 14Hz (give or take)
You don't have to fiddle with all these numbers though, the magic of the Internet gives us sites like this:
http://www.learningaboutelectronics.com/Articles/High-pass-filter-calculator.php
If this is confusing (and bloody hell it wrapped me in knots the first time I saw one) you can think of a filter as being a voltage divider.
RC High-pass filters always start with a capacitor (because we're blocking infinitely low frequency down to DC) and low-pass filters always start with a resistor.
So the absolute simplest voltage divider looks like this. With two equal resistance values you just half the supply voltage.
The centre tap in this example is sitting at 2.5 volts but it's not very "stiff" (oo er misses) because anything connected from the tap to the ground "loads" the bottom resistor. That's why you don't see circuits like this used to split a power rail. It's not that they don't work, it's that they are mind-bogglingly inefficient.
Ok, so let's look at at high-pass filter arranged like this ladder.
The reactance (X) of the 10 uF cap is (give or take) 1K at 15 Hz. which means a sine wave at this frequency is divided into two just as in the resistive divider above. This doesn't account for phase shifts, but we don't need to worry about those here.
If we drop the f in to say 1.5 Hz, the reactance of the capacitor rises to around about 11K which produces a resistive ladder that looks like this:
The 11 k resistor I've shown here represents the reactance of the capacitor and you can see that the peak voltage is only 416 mV - 832 mV peak to peak. So if the AC (this example is at 5 V peak so 10 V full swing.
That gives us an attenuation of:
0.832 / 10 =
0.0832
(Guess that doesn't take a genius to work out - but I used a calculator so that tells you all you need to know):
To get the dB (peak-to-peak) attenuation we do:
20 * log(0.0832)
= -21 dB
For my own curiosity I did a quick spreadsheet with all the various operations including the current at a given frequency and the voltage drops across the resistor and capacitor. It charts like this with frequency increasing on a LOG scale across the top and dB of gain down the left.
I can't share the spreadsheet here but if you wanted it I'm sure we can find a way.
Marc,
thank you for your answer.
Firstly, I do not understand what is the input side and what is the output side for you. But I can tell you that my calculation is identical to the calculation of the CG capacitor in the THAT1512 manual on page 5. For example, a 6800uF capacitor at a frequency of 5Hz gives a reactance of 5Ω. And if you connect a 5Ω reactance in series with a 5Ω resistor, you get a halving of the gain, which of course corresponds to -3dB.
Secondly, I understand frequency filters, but I am not sure if the elements between the THAT1512 RG pins have the function of a frequency filter. As far as I understand, you did not use the R-C combination to filter frequencies, but to squeeze AC signal on the MCP41010 potentiometer into an acceptable voltage range. And even if this combination worked as a frequency filter, your calculation is flawed. After all, the 1kΩ of the potentiometer is its maximum value, while the minimum value is 0kΩ. This is exactly why I put a 1kΩ resistor between the potentiometer and ground to ensure that their combined resistance is never less than 1kΩ.
Finally, I firmly believe that the whole circuit is faulty at low frequencies. In front of the THAT1512 input, Matt has placed two 22uF capacitors, each of which has an reactance of 360Ω at 20 Hz, far from a negligible value. I think both the circuit and the THAT1512 chip are designed for audio frequencies and are simply not suitable for measuring infrasound
To conclude, I will try to solve the problem with accelerometers, which I suspect are indeed designed for lower frequencies. Let us consider the matter closed, there is no point in kicking this can down the road any more.
Marko
Posted : 06/03/2023 9:12 pm
marcdraco reacted
All good Marko, hope it works out for you!
@axy_david and @diy_perks, this one actually works as advertised. I've made a fair few changes from my earlier attempts partly because I trusted the simulator too much and partly because I took my eye off the ball... While they did work... let's just say they weren't up to snuff.
Links X1 - X3 show were the two cables and the screen (to ground) join between the constant current circuit which should be in the low-impedance part of the 2n2222 circuit not as I had it, in the collector load. (Spanky bot-bot time, naughty Marc!)
Pretty much any GP red LED should work here, same with a low-noise GP bipolar NPN BJT (BC109, BC179, BC239, etc.)
The 2SK170s were from Loadartronics on eBay in the UK but they're a popular and current JFET for low-noise audio applications so I expect they should be easy enough to source.
The combination of a GP diode (I actually used an 1N4001 because it's what I had, but they are a bit chunky) and the 1Meg resistor give a decent impedance to ground and increase sensitivity without the need for a bootstrap.
These components really do need mounting on a small PCB (see my post about slug tape, yes really, for a possible way around this). Matt is better at that side of things though although I'll do a board in KiCAD as soon as I get a few moments.
I've moved the LED away from the head (just to be on the safe side) but I expect they will all fit into a tightly packed board, especially if you're using one of those 34 mm"Neuman-like" capsule copies from Ali/eBay.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 06/03/2023 10:16 pm
I'm in the middle of this build. My experience so far was not so great because of the fake chinese THAT amp that i've bought, hours later in this forum and I've bought a real one that eliminate all my problems. I'm planning on 3d print the entire body. So as i'm planning to 3D print everything I want to ask if is there a way to put a mute button somewhere in the circuit, maybe with an LED that helps understand when it's muted.
Posted : 07/03/2023 10:46 pm
I'm in the middle of this build. My experience so far was not so great because of the fake chinese THAT amp that i've bought, hours later in this forum and I've bought a real one that eliminate all my problems. I'm planning on 3d print the entire body. So as i'm planning to 3D print everything I want to ask if is there a way to put a mute button somewhere in the circuit, maybe with an LED that helps understand when it's muted.
I HATE that. The Chinese will copy anything (including stuff their countrymen design). I have a siggen here based on an XR2206 which only works properly at lower voltages. It's badged XR2206 and works (broadly) like the XR2206 but it quacks like a pigeon.
A mute is definitely possible - but it's usually done with another JFET to stop the "pop" by buffering the switch transition for a few microseconds. Rod Elliot (Elliot Sound Products) has some examples I believe.
I'll put something suitable in the V2 circuit (above) if I get time.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 08/03/2023 1:38 pm
@marcdraco this is the complete circuit?
what's with the PCM182x ?
This is the microphone "level" output - essentially an improved FET impedance matcher. This will (and does) drive a microphone input on an existing USB audio codec (sound card). Quality on the one I tried was horrible though but it did cost less than a fiver so I shouldn't grumble.
The rest of the schematic needs transferring to KiCAD but it's very simple:
Just a x100 AC amplifier based on a single OP07 - but it DOES need a split supply as it won't work down at 5V USB. I've used an ICL7660 but the one Matt used is probably quieter as it's a monolithic with what looks like an EMI shield! The 7660 is bleeding noise onto my Vero prototype, so I'll need to work out some better decoupling. Output is via C6, R25 here is the internal load on the converter.
There's nothing inherently wrong with the PCM182x ADCs, it's just that the prototyping is beyond what I have to hand as it needs someone with proper etching equipment and SMD oven in house: and more experience than I have with this particular area. Usually a professional house is the only place that would attempt something like this. Turnaround time is in the order of 3-4 weeks for each batch of boards and that's assuming you don't fudge it up the first time. As development is incremental (I've been through four different versions of this one as I tweak bits) you can see the time adds up pretty quickly.)
Then there's software support too - it's not the sort of project I'd take on since there are already plenty of "affordable" - a term defined by your disposable income - ones on the market starting at around £120 if I recall correctly. It's not really practical for me to design one although I'd obviously support anyone who wanted to.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 08/03/2023 1:55 pm
If anyone is wondering about the PCM 182x and the posts I've made, I wouldn't worry about it. The part I'm talking about would be a DIY version of the USB audio dongle in the video. Thing is, if you want great audio at high cost you can buy a professional interface made by people who actually understand audio and emi. If you want good audio at low cost you can use one of those dongles like in the video. The thing I'm working on is (hopefully) good audio at high cost (i.e. a waste of time). I want an excuse to work with microcontrollers, audio, and to branch into hardware design. Any onlookers can safely ignore anything I say.
I wouldn't say this needs a professional shop, but you would need at least a hot air gun and solder paste as well as be comfortable ordering PCBs from fab shops like jlcpcb or pcbway.
I'm not very familiar with any of this, but there have been quite a few times where I wanted to do some project but couldn't because of that SMD hurdle. So I finally bit the bullet and got the equipment needed for amature SMD work. I don't want anyone to mistake me for someone who knows what they're doing. My education is in computers. I'm confident enough in writing software to think I can do this on the software side of things, but the physical development side is an unknown to me.
Posted : 08/03/2023 4:19 pm
marcdraco reacted
@marcdraco I'm pretty new to electronics for some exeption in 3D printing and drone stuff, so I can solder everything but I don't know what a JFET is. By the way, this is my circuit so far, I was planning to put the button between the 5v cable that goes from the usb-c to the pre-amp board. I notice however a super loud "pop" when I switch on or off. Maybe the JFET you are talking about is used to buffer this problem.
Posted : 08/03/2023 4:30 pm
I'm not very familiar with any of this, but there have been quite a few times where I wanted to do some project but couldn't because of that SMD hurdle. So I finally bit the bullet and got the equipment needed for amature SMD work. I don't want anyone to mistake me for someone who knows what they're doing. My education is in computers. I'm confident enough in writing software to think I can do this on the software side of things, but the physical development side is an unknown to me.
The problem is in designing such a complex SMD board - this isn't a simple task even for someone like me (although my training was in through-hole long before SMD was a thing). I did one for a replica proton pack (Ghostbusters 1984 version) based on the Arduino framework (AtMega328) but I did most of that through hole and what was SMD was ... a challenge.
If you REALLY want to try SMD, I can do an SMD version of the "Ella" mic and pre-amp (that's the code name for the V2 electronics I've done to complement Matt's work). The schematics will all be Open Source Hardware to compliment Matt's work. This is all you need to drive the "mic" input on the PCM 182x series.
The "through-hole" stuff is pretty straightforward. This is the prototype in KiCAD - now I know the electronics works but there's always the question of if I fudged something else. Here's the through-hole layout and a schematic you can copy into KiCAD. I've got issues with the pre-amp board because I don't have some of the parts in the library, specifically the 2N2222 which is a common(ish) NPN BJT that I'll need to nick from LTSSpice or re-define myself. Or given the damn thing comes in TWO different lead layouts (!) I might swap it out for an alternative as it's not that critical.
There are only four actual parts here but you'd be amazed the number of different sizes you can get. I know I was. SMD is the future of course so it's better we all get used to it or get left behind.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 08/03/2023 8:44 pm
@marcdraco I'm pretty new to electronics for some exeption in 3D printing and drone stuff, so I can solder everything but I don't know what a JFET is. By the way, this is my circuit so far, I was planning to put the button between the 5v cable that goes from the usb-c to the pre-amp board. I notice however a super loud "pop" when I switch on or off. Maybe the JFET you are talking about is used to buffer this problem.
Nice work that. You should see MY prototypes... or maybe you shouldn't... 😉
A JFET is a type of transistor, related to the MOSFET - both are "field effect" transistors that use a voltage on the input to drive a current on the output.
BJTs (the ones that were popular prior to the explosion of computers) use a current IN to get a current OUT.
FETs are useful for a bunch of reasons - largely they are easy to make by the billion... yes, billion on a wafer whereas BJTs are way more difficult. Hundreds yes, but more than that and you start to hit the limits of power dissipation because they aren't that efficient. MOSFETs switch at ridiculous speeds and when they switch "on" they present very little impedance (resistance) to the circuit so they don't get hot like BJTs have a tendency to do. They still generate some heat but it's not in the same league.
You can get Class-D amplifiers now that work by "chopping" an analogue signal into little time slices (quantisation) at speeds that are well beyond what we can hear.
This allows the output transistors to switch ON/OFF very quickly switching "huge" loads (often tens of amps in the case of IGBJTs which are related but way outside the scope for this discussion, dem tings be scary!) meaning you can get a very powerful audio amplifier that seems to lack much in the way of the huge heatsinks that we made "when I were a nipper".
JFETs are "junction field effect transistors" and are a bit of an odd one compared to the others. In the case of MOS or BJT we put a positive voltage (or current) on the control terminal - gate or base respectively and current flows in the channel - drain-source or collector-emitter.
There's a lot more to it than that but that's enough to get you going.
JFETs are usually depletion mode which means (unlike enhancement mode MOS or BJTs) they are turned ON by default and your circuit has to turn them off! We do this by pulling the gate negative with respect to the source.
Which is damn confusing when you have a single (say a battery) supply. But voltages are all relative to some point in a circuit, so if you "push" the source more positive than the gate, the gate effectively becomes negative with respect to that. And you do that (in the simplest terms) by putting a resistor in the source path. Current flows when the circuit powers up, causing a voltage drop on the source resistor and if chosen correctly, that drives the FET into its operating region. Turning the beggars off usually needs a negative supply though.
Still with me? Don't worry if this sounds confusing, the hours I've had to put in to unlearn my bad habits and learn how to do this isn't right.
All transistors can be made to operate as switches but MOSFETs and JFETs are more suitable for audio muting for a bunch of reasons and I suspect that the JFET is superior in that regard to since it's easier to make it "switch" over a few milliseconds using a simple CR circuit and let it slowly open or close the channel. It's the sudden "on/off" that causes the pop but with a JFET we can turn the volume down (or up) over a time that sounds instantaneous but is long enough to prevent the pop.
Rod Elliot has something on these ... he forgets more over a Vegemite sandwich than I know and if you want to study someone who really knows their onions... head over there.
https://sound-au.com/projects.htm
Rod has some detailed JFET work here and section 10 has most of what you need to reduce the volume.
https://sound-au.com/articles/muting.html
You'd put the mute between the output of the THAT1512 and the input to the USB ADC board in your application.
For more about JFETs if you're really in the mood for study, here's Rod again. He's a good guy and if he and I disagree, he's right!
https://sound-au.com/articles/jfet-design.htm
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 08/03/2023 9:12 pm
I was going to attempt to make one of the original ones, even went as far as having everything in my cart but it seems like I should wait it out and see if there are any more upcoming updates. I was planning to buy a cheap microphone for a donor shell or a casing off eBay, I would still prefer this route as I do not have the tools needed to create my own.
Posted : 11/03/2023 2:06 am
Hello, i recently Build the Microphone exactly as shown in the video. But it doesnt work at all. Because i couldnt find a place where i could buy the 2N4416 Transistor (only on ebay but the estimate delivery would be in 1-2 Months) i used this one bf245a. Could this already be the reason why the mic isnt doing anything? I also didnt want to use the rotary Gain switch so i just soldered a 100 ohm resistor where the gain connector would go. Since im not very experienced with electronics i cant figure out where the Problem is. It would be greate if someone here could help me or give some advise what i can do to identfy my Problem.
This post was modified 3 years ago by edding752
Posted : 13/03/2023 4:41 pm
Yeah, you might want to hold on there. The new one is mostly designed and the PCBs are ready to go to the fab shop in China. I'm trying not to outpace poor Matt as this is, ultimately, his design. Sorry for the slow reply, I've only just got the notification. I expect your post was in the mod queue.
I've done (am doing) several versions so people have a choice of which one they go with. There's an improved original that needs SMD parts so it's a bit tricky to solder and I've done (but not fully tested) a larger version which uses "standard" parts for what of a better work. Through hole stuff - but it's larger so you'll have to use a larger capsule (it's designed to fit those 34mm jobs).
Sound quality is excellent and the pre-amp runs pretty flat from 10 Hz to over 20 KHz but I've included a 20 dB pad and a low cut switch to reduce low end pickup (not that there should be any, but "belt and braces").
This is the SMD head unit so you can see it is designed but I'm holding off until I get all the PCBs on a single panel so we can have some spares. There are issues with SMD amps using multiple different pin outs (I've found three so far) which makes it difficult to pick a jellybean from somewhere.
The IC (an op amp) is cheaper than an FET and much easier to bias. The DIP version below can take pretty much any single JFET input op amp, but I think I might revise it for duals as they cost about the same and are easier to find. My reference is the Fet-CMOS TL071H which is a high-performance version of the TL071. Both mount directly to the capsule and (optionally) feature an LED power indicator. I can't wait to see what Matt does with this! I bet it's going to be amazing.
and this is the 34mm version for larger capsules. It's far too large for the JLI capsule though so you'd have substitute with a larger one (ALI express about a tenner, eBay about £20).
I can't really vouch for the quality of these capsules but they sound OK to me... Your mileage may vary of course.
Here's the preamp performance - first with the full-range:
And now with the 80Hz (actually it's about 76 for 3dB) cut. You can see how the plot slices right into the low frequencies. The three curves show the voltage gain vs. frequency for each stage.
This design doesn't use the THAT1512 but I have done an alternative that does. It's all Open Source hardware though so you'll be able to hack it as you desire.
EDIT: updated the 3D renders with the rev 2. boards.
This post was modified 3 years ago by marcdraco
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 13/03/2023 5:20 pm
Hey I've recently bought all the components for this build but there is one thing I cannot figure out no matter what I do. How do I uncompress desoldering braid/wire? I've tried everything I can think of and nothing seems to be able to open it up.
Thanks
Posted : 13/03/2023 7:26 pm
There are solder braids and there are solder braids. I hit this exact same issue myself. I've collected a few over the years (mostly from eBay) and several of them didn't want to play - and I picked at them until I lost my voice from all the cussing.
Chances are you got some like I did. I haven't examined it microscopically (I probably should have) but although it looks like a flattened tube, I suspect the braid is just flat. Presumably some clever person in China figured out how to make it look like a good braid but saved a few Yen (??) by making it like that.
The one that worked for me was a 2.5mm roll, the 2mm stuff wasn't having it at any price. I wish I could remember where I got it...
A hack *and it is a hack* to get you going would be to twist three pieces of the ECW together quite tightly and use one of them for the screen. Now this sounds mad, I know, but the way Matt designed the original with a THAT1541 the braid is not that important. It should be said that it's better WITH the braid than without, but the "common mode" rejection on the amp is so high that it will reject most of the noise the short run picks up anyway. Mains hum is the worst and this is one of the easiest to deal with on a balanced line. Separately, the braid gives the whole job some physical strength which it wouldn't have otherwise, so it provides a double function. Another way would be to get braided, twisted pair (and strip away the outer plastic cover) but that's not going to give you the low-mass Matt designed as the inner cores will be plastic wrapped.
Let me know how you get on, this is truly an original and quite beautiful project. One of Matt's best of this type.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 14/03/2023 12:05 am
Hello, i recently Build the Microphone exactly as shown in the video. But it doesnt work at all. Because i couldnt find a place where i could buy the 2N4416 Transistor (only on ebay but the estimate delivery would be in 1-2 Months) i used this one bf245a. Could this already be the reason why the mic isnt doing anything? I also didnt want to use the rotary Gain switch so i just soldered a 100 ohm resistor where the gain connector would go. Since im not very experienced with electronics i cant figure out where the Problem is. It would be greate if someone here could help me or give some advise what i can do to identfy my Problem.
Welcome - I'm just a user like you so this isn't official advice. The reason there's so much of my chatter in here about the "new" design is due to this exact problem. Matt and I are microphone crazy so this is a passion project for me.
The 2N4416 is out of production now, as are many JFETs, so I've re-done the project for a V2 (with Matt's approval, naturally) with op amps. There are two versions - the smaller one requires SMD parts and while they are solderable, they are tricky to handle so I've done a full through-hole design too for larger capsules. It's not clear at this junction when these will be available, but I'm open sourcing all my designs when I'm happy they work in practise, not just in a simulator! That's caught me out more than once.
Looks like the issue (and this is a guess) is with the pinch-off voltage Vgs (off). The BF245a is rated from -0.4 to -2.2 (JFETs are maddening) and the 2N4416 is -2.5v to -6v so that might be it. FETs are challenging to simulate because the software usually picks an ideal Vgs off and that can lead you up the garden path right off the edge of a cliff.
It's much easier to use an op-amp with JFET inputs that is guaranteed to work - and that's the way I've gone for my modification.
I did an earlier mod - that I got wrong several times - that "self-biased" the JFET (I used a 2SK170) but that's really quite naughty, even though it's a common trick you see all over the place. As a matter of interest, the 2SK170 has also been replaced by a tryptic of devices which *claim* to be drop in replacements with better performance. In the sim though, two of them fail to work at all. That's JFETs for you. Much as I love the things they are a mutha to get right.
OK, gain resistor looks solid. 100R gives you 40 dB which is 100x voltage gain. That should give you *something* particularly if you remove the mic from its shield and sit the thing on a desk - mains pickup is everywhere. You should be able to hear it if you put a finger on the capsule's centre terminal.
Of course, there's plenty more that *could* have gone wrong. If you don't get any sound by injecting a signal (e.g. sticking your finger on the one of the input terminals) then the JFET isn't the only issue. In this case, it might not be the issue at all.
When I learned troubleshooting, I was taught split the problem into two (roughly) equal parts and figure out which side it's in. Then divide that part into two and so on over and over until you finally end up with just the faulty part. There are cases where RFI makes this task impossible but in audio you should be OK.
Feel free to tag me if you need more help.
Take everything I say with a pinch of salt, I might be wrong and it's a very *expensive* way to learn!
Posted : 14/03/2023 12:31 am
@marcdraco I started testing the mic on a teamspeak server. When i started the test in Teamspeak it picked up something (there is a little bar that jumps up when the mic picks something up) but its only a very quiet sound like a coil whine from a gpu. This "tone" starts and then stops again but not in a rythm just randomly. Otherwise the NMA Converter gets very hot is that normal? I also put my fnger on the main Terminal with no effect. Would you suggest switching out the BF245a? Otherwise i dont know what i can do, i triple checked if i made any mistakes with my soldering and it looks like i did it right.
Posted : 14/03/2023 11:56 am
Here's a little update so far. I've got everything working. Thanks to @marcdraco i've added a mute button between the pre-amp output and audio interface input. I've also added a 1k potentiometer as gain control with a fixed minimum resistance of 100ohm (the fixed resistor is there because of a strange behavior when the potentiometer goes to 0, it seems to create some kind of ground loop, high pitched noise, dunno, not a problem with the resistance). Also added a 3.5mm jack output recicled from a laptop, in case I maybe need for some reason (always be sure to have the options).
Down some WIP on the main circuit.
Posted : 14/03/2023 5:42 pm
marcdraco reacted