Skip to content
Notifications
Clear all

[Sticky] USB-C Microphone (official topic)

728 Posts
139 Users
138 Reactions
54 K Views
marcdraco
(@marcdraco)
Posts: 476
Member
 

Maddening isn't it? I love Australia, beautiful place to live. Crazy that shipping to Norway is so expensive.

I guess I'd better get back to this artwork. 🙂


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

 
Posted : 15/05/2024 4:45 pm
eiriksb reacted
marcdraco
(@marcdraco)
Posts: 476
Member
 
image

Didn't really want to give this any huge fanfare as it's a bit boring (in a functional way) when compared to some of my other designs. This one is designed to simplify the process and get you going faster and for less money than the original as it doesn't need a THAT151x or a voltage doubler to make it work.

This is (perhaps) the ultimate upgrade for anyone who bought one of those "Neewer" USB mics with the digitiser than sounds like the wind. Any of that sort of microphone actually.

While this is a variation on the Ultimate USB Microphone, it's meant for a different audience.

I won't bore you with the details (that can go in build logs) but the IC is a standard DIP (not shown here) OPA2134 which is a professional grade stereo audio op amp, but still affordable. 

Most of the components are SMD to keep it compact but there's space for an LSK170 (or others with the same pinout) OR a 2SK208 for this revision. The '208 is a good alternative to the '170 if you're having these boards made up because it keeps the board tighter, esp. if you're using a SOIC-8 OPA2134 for the drive.

Solder the IC into place by forming the pins under the body. Not really ideal but it saves money over the cost of a SMD version, at least via JLC. Other houses may be different, Aisler for instance, is worried about the sliver of copper around the edge. For ultimate cheapness I've done this version as a two layer board but the extra two ground planes which should (or might not) improve performance might by worth considering.

Tenting the vias is important though as smaller 25mm capsules will short them this is marked in the artwork now but I've also move the offending vias out of the way so there's less chance of that.

Give it 5V and ground and it spits out an exact copy (single ended) version of the signal from the FET x100 (40dB) ready to drive the digitiser without anything else - and since this part fits in to the back of the capsule, the digitiser can be fixed (3D print anyone?) inside the body. Or whatever you can imagine. The out cable doesn't even need to be screened for the short distances from the head to the digitizer so it's possible to take one of those nasty bodies and turn into a desktop microphone that should perform with the best.

The final quality depends upon the capsule - which is why I suggest sticking with the JLI2555, but the cheap electrets I've tried are surprisingly good. I just need to find something to record now.

Although the circuit elements are correct, be sure to check the part numbers against JLC's parts library. Don't end up with a 1R resistor when it's supposed to be 1K. 

For the brave (or if you have a small SMD workstation like this Miniware MHP30 Mini Hot Plate Preheater) who fancy a stab at soldering 603 sized parts, here's a link to some Aisler artwork.

https://aisler.net/p/SSRUKJFB

For the more sane among us, this is the KiCAD 8 project file and completed artwork, positions and BOM ready to produce this at JLC PCB. Order it with the '208 with 

JLC's systems will correct the part placement for you, other houses may need different position files.

Below is everything you need, including the four layer Gerber artwork, it costs a little more but is worth the extra for the level of screening and return paths provided.

For five boards expect to pay a little under £20 to the UK with the '208 fitted. You'll still need a suitable op-amp and this needs to be the OPA2134 or better because it has to work with a split 2.5V/-2.5V rail. This is done on board, so you can just supply power from the USB (ideally after some filtering if possible). Chips like the NE5532 just won't operate reliably (or at all) down that low. The TL072H (not the original) will operate at that level but is still a good bit noisier (18nV/Hz) vs. 8nV/√Hz for the '2134.

For most users the '2134 will be superior and easier to source. Price difference for short runs is, again, less likely to be of consequence I can't see this thing taking off. 😉

TL;DR

  • Does it work with the original Perks design?
    • Yes but note that does require a couple of tweaks to the wiring.
  • Isn't single-ended bad?
    • Usually, and esp. in electrically noisy environments, but for a short run of less than a metre, it's not going to make audible difference.
  • How do I get digital out?
    • This is really up to the constructor. My personal advice would be to solder stereo cable (two conductors plus screen) to the board, the screen is soldered to the outer ring. Red to +5V (Vcc) and white/blue/black to the signal. This can be tapped off to a Phono connector and plugged direct into the digitiser Matt picked for the original which is, to date, the best I've tried that takes Line Audio and it's stereo remember so you can have TWO of these mics, mixed right in front of you!
This post was modified 4 weeks ago 2 times by marcdraco


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

 
Posted : 16/05/2024 10:15 pm
(@carontes)
Posts: 3
New Member
 

Hi everybody. I watched the video the first time it has been posted but never had the chance to build a DIY microphone. Now I would like to build my own microphone for various purposes but I'm a noob aboit the argument. I checked online diferrent capsules and I was wondering if a DUAL Diaphragm like the TSC-2 would worth the price difference compared to the JLI-2555BXZ3. Also since it's a dual diaphragm if it needs a different circuit board or I could use one of the ones posted here, also I've never ordered a PCB online so I don't know if they send me only the PCB and then I have to solder the components or if they do it for you as well. Since I got a 3D printer I could desing and print the case for it. Thanks to anyone who can helps me

 
Posted : 17/05/2024 1:25 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

These boards are specifically for single diaphragms, the circuits work (later ones since I’ve goofed quite a few times) but the working ones have the files.

I’ll glom the whole lot together with details but let’s look at the problem. 

The duals need to be able to “hear” both sides of the mic at the the same time and we can’t really restrict the airflow without affecting the performance. 

I’ll mock up a “stereo” version when I get a chance but that means an extra wire going through the braided tube. 

I have an idea on how to do this for Matt’s desktop design so I’ll see how practical that is. 

EDIT:

When you order PCBs you can get them partly or fully assembled or plain copper and plastic with holes for you to assemble everything. Several here, Chimera is the most recent, will work with this setup, but it does require a small amount of soldering which isn't difficult but it does need a steady hand (and some desoldering wire to hand!)

This is a prototype (I've added the option to have the SMD transistor added at the factory now as this can work out cheaper depending on where you are in the world. Also the vents have had a little work. You can see from here that I've just soldered a chip but pin 8 (upper right) to hold the chip in place. The pic isn't how you're forming the legs, that's because the cat decided to play hide and seek. He hides and my bear bare feet seek. Much to the cat's amusement.

image
image

The zip with project files above has the three files you need so you can try JLCPCB to see how the process works. I've made it as painless as possible by uploading the actual ones I've sent over so we know they're as they should be!

The Chimera board is so called because it's a mixture of two technologies, and it sounds cool(ish) if you say to your mates, this is powered by DIY Perks Chimera. 😉

This one runs on a single supply from 5V to about 30V so it's essentially a stand alone unit. I'm working on something to filter the USB, it's often a noisy power source and unless some sort of isolation/filtering is done some of that noise feeds back into the mic and sounds "m'eh". You can use a small battery pack of course, 6-9v would be ideal and it the current draw is modest.

Matt's preamp has a partially filtered supply so that can be used with some modifications. Complete isolation (which means separating of everything including the ground!) is quite costly which is why I'm looking at a better filter. Cost is relative of course but let's not go there. 🙂

There is another gotcha with the TSC-2 too. It's pure condenser which means you'll need a stable 48 volt supply to bias each side of the capsule. That adds more complexity to the circuit because (traditionally) we use two 1 GIGohm resistors and some specialised capacitors, because the FET's gate has to be isolated from the DC bias.

I do have some 48V units here but they are designed for electrets (pre-biased) and are intended for professional equipment which doesn't care what's coming out so long as it works with 48V in.

And finally... 🙂

Mounting a TSC-2 in such a tiny head is going to be tricky - shorting either "live" side to the Faraday shield (the mesh bit) won't destroy the 48V kit but it might make a mess of the capsule as those gold sputtered mica films are very thin.

If you're up for making something like this, I would suggest you look at the Neewer BME-800 which is available from Ali Express from a few dollars plus shipping although you can get it from Amazon.

When you mount a capsule in a metal body like this it's not necessary to mount a board right up against the head (that is primarily for these "floating" designs and ones that might look a bit space age/cyberpunk).

The other advantage is you have LOOOOOADS more room to play with. 3D printing the body is possible but the sensitive bits have to be enclosed in a shield and it's usual to use the wires from the capsule as they are - unshielded but protected by the body. That's not to say it' can't be done but that's more up our host's street than mine. I'm just a writer who can wield a soldering iron - en garde, alez, alez!

Because we have more space it's quite practical to double everything up and a pattern switch to derive the various pattern from the capsule. Even put a low-cut or a 20dB cut in there.

I haven't done a 48V voltage multiplier yet but professional equipment does supply it. It's something I need to do next. Since there's very little current (we're only charging a capacitor of a few 10s of picofarads) this can be done with a MOS Schmitt inverter and some other gubbins. Ideally a switched mode supply is the order of the day but I've been unable to find one "off the peg" that converts the rather feeble 5V into 48 volts; however a laptop power supply (typical 19V) will be easier so that's probably the way I'll go unless someone magics on up from China.

You can bias the "plates" with a lower voltage but the results will be poor at best. Too high and the film gets too taught (I'm told) but some mics go as high as 80V. That's typically in the spec sheet or if not, it's safe to assume 48V. These capsules are a good deal more sensitive than "electrets" as a general rule too.

I find it rather amusing to note that the Neewer look the part (P48 adaptor, canon plugs the works) but they are only single ended! This probably won't make a scrap of difference in a home studio but anywhere there's a lot of noise about, you need differential. The other thing with a differential is you get two signals so twice as much power. The cables are cheap too. It all is but you get what you pay for and a $5 plus shipping from Ali... consider this a nicely made donor body and nothing else. 🙂

This is probably a lot to take in, fire away if anything isn't clear though.

This post was modified 4 weeks ago 2 times by marcdraco
This post was modified 3 weeks ago by marcdraco


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

 
Posted : 17/05/2024 1:47 pm
Carontes reacted
(@carontes)
Posts: 3
New Member
 

Thank you marcdraco for your explanation. I'm trying to familiarise with the JLCPCB website, but I don't understand which files are the BOM and CPL files. Also is there a tutorial or schematic on which wires to solder and where? Thank you again and sorry if I'm making silly questions

 
Posted : 17/05/2024 10:31 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

You’re very welcome this has been quite the adventure for me  I hadn’t touched a soldering iron in anger for over 30 years before I started this. So much new stuff to learn about surface mount technology It’s been worth every penny.

Of course, I made too many assumptions in my answer I’ll write up a tutorial when I’m more awake. It’s really easy when you’ve done it a few times, but yes it’s a process. 


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

 
Posted : 17/05/2024 10:44 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

@carontes

Posted by: @carontes

I'm trying to familiarise with the JLCPCB website, but I don't understand which files are the BOM and CPL files. Also is there a tutorial or schematic on which wires to solder and where? Thank you again and sorry if I'm making silly questions

OK, so here's a detailed post on how to go from 0 to a PCB in a couple of weeks, post allowing.

Ordering OSH from a board house – General Electronics & Circuit Design – DIY Perks Forum

Chimera is one of my more complicated boards in some regards and simpler in others - it's specifically designed as an adaptor for BME-800 bodies so we can put the digitizer in the body but the actual work of doing this does require more work. 

So this one is soldered direct to the back of either a 25 or 34 mm electret capsule and provides a signal suitable for a "line" input that is compatible with recording equipment, including the digitizer Matt picked for his Ultimate USB C Microphone project.

As with most, you only need three connections to this one.

0v (ground) - traditionally the black wire - and that goes to the outer ring. This is a design choice for ease in case you want to solder the screened braid to the ring. 

V+ (5 - 30V): While this board is specifically designed for a low-power IC (OPA2134) you can optionally power it from a battery or other stable DC supply and cheaper op amp like the NE5532 although it's worth noting that older designs like the '5532 are quite greedy so they will drain the battery much faster than a more modern design. Others like the TL072 will be much noisier, but you have the choice of what you want to use for your application.

Out (singing cat): Line out - the signal from your microphone to the digitizer or PA amplifier for example. The isn't the easiest way to make a microphone for a PC but it's one of the easiest ways to get a great quality, low-distortion output.

The *easiest* way to make a microphone compatible with a PC sound card is with another design suggested by a member, is Minerva (I think I've mentioned that I really need to do a build log with each board). This one carries a single JFET wired specifically to operate in common source mode. 

This can be wired directly to a 3.5 mm "mono" jack - all the extra electronics are inside the PC. While this works over a very wide voltage range, it can work all the way down to perhaps a volt and a bit. That's relevant to PCs because the internal supplies are very noisy and that can be reduced by a regulator.

I should, perhaps, have put a regulator on the Minerva/Chimera boards to power the FET and I didn't to keep the cost down, but some noise does still come through on a noisy USB. I'll test this at some point on a breadboard if there is any demand for a better version. 

Ultimately though, I lean towards cleaning the 5V supply before it gets that far up the chain with some passive filtering. This helps far more people than putting a regulator on here as even regulators will pass some noise through, especially the higher frequencies from the PC.

image

 

 

This post was modified 4 weeks ago 2 times by marcdraco


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

 
Posted : 18/05/2024 1:02 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

Just a head's up, I've found a new digitizer buried in the 1.3 revision of the DigitalLife dual stereo audio capture card in the wild. Externally it looks exactly the same, but internally there's a custom chip with far fewer external components. 

image

Fortunately it's basically the same size so nothing to worry about unduly.


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

 
Posted : 19/05/2024 10:06 am
xenon
(@xenon)
Posts: 10
Active Member
 

Update:

Some components started to arrive this week, I choose to use the largest and more expensive mic capsule. In the image you can see that I soldered the case and the case (ground) pin of the transistor to the mic ground to hopefully eliminate more interferences.

IMG 20240519 132607

I also received a brass mesh I ordered on eBay but I think it might be to coarse. Should I put two layers on top of eachother to make smaller holes?

17161189314719201803518750798987

I searched everywhere on eBay for a brass sheet, but I am worried that might be to thick. What's the thickness of the sheet used in the video?

By far the main structure looks good but I still haven't found the bigger tubes and rods to make the rest:

IMG 20240401 191632
IMG 20240401 191603
IMG 20240401 191641

About the other electronic components I've got everything on the way.

 
Posted : 19/05/2024 11:53 am
marcdraco
(@marcdraco)
Posts: 476
Member
 

The mesh will be fine, you're making a Faraday cage and at the relevant wavelengths are easily captured by that.

I've tried 0.5 mm and 0.3 mm brass sheet. The 0.5 mm stuff feels very inflexible but it will move with a little heat. The 0.3 mm is far easier to work with but won't take as much abuse.

JFET looks well mounted so you're off and away there but that part of the capsule has to be surrounded in either mesh or solid sheet, there's no escaping that I'm afraid and trust me, I've tried. 🙂 


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

 
Posted : 19/05/2024 12:02 pm
xenon
(@xenon)
Posts: 10
Active Member
 

@marcdraco 

Yeah, I'll enclose everything in the sheet with the mesh. Thanks for the suggestion on the thickness, I think I'll go for a 0.3mm.

 
Posted : 19/05/2024 3:45 pm
marcdraco reacted
(@carontes)
Posts: 3
New Member
 

@xenon

Great update! I hope it'll work as you expect. Is it possible for you to take pictures of all the wires connections once you receive all the components? The first picture with the capsule is a bit out of focus. Thank you

 
Posted : 20/05/2024 1:02 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

By request:

image

This is a fully through-hole (therefore MUCH easier to create/repair) version of my mixed-type (SMD/THT) "Michelle" PCB for the original project. This uses almost all the same parts with the exception of the three TDK FA28C0G2A220JNU0 22uF MLCC capacitors. These aren't expensive and perform better than electrolytics while taking up considerably less board space.

This is a full set of Gerbers - to save cost this is a dual-sided board, but for ease of reference the component values are printed on the back.

image

 


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

 
Posted : 24/05/2024 12:05 pm
(@gilex)
Posts: 5
Active Member
 

Hello and thanks diyperks, @marcdraco and all other posters in this thread for putting time and work into this!

I'm new to pcb design & audio stuff and have some questions about the Chimera design:

  1. Is my assumption right that only Q1 XOR Q2 is placed? Which component have you placed on Q2 (capacitor from source to drain?) in the picture with the 8 pin IC (the OPA2134?) mounted?
  2. Where (if possible) would I connect the rotary switch and what resistor values should I use?
  3. How would I add a mute switch (can I just disconnect/bridge MK1 or will this result in noise)?
  4. I'm assuming that MK1 is the bare microphone because the 2N4416 (or something similar) is Q1/Q2. Does the Unidirectional in the JLI-2555BXZ3-GP's description mean that polarity does not matter?
  5. Will I be able to add the USB Noise suppressor circuit to the current design or will I have to order a new pcb?
  6. How long should the rubber suspension be? Could I suspend the capsule inside a traditional microphone design with short rubber from the mesh?

I'm sorry if that is a bit much and I missed some information but the thread is quite long and there is no option to load & search all 24 pages at once to my knowledge.

Thanks in Advance!

This post was modified 3 weeks ago by gilex
 
Posted : 27/05/2024 8:30 am
marcdraco
(@marcdraco)
Posts: 476
Member
 

@gilex

  1. I sense a developer, not many people outside the dev word would know what XOR is. 🙂

    Yes it's either Q1 (LSK170) through hole or Q2 (2SK208 surface mount) you *can* use both at the same time but the results are undefined as each transistor will operate a slightly different gains so that would cause distortion which could be significant.

  2. Chimera is stand alone (it doesn't need pre-amp stage like the original design) because it's primarily designed as an upgrade for Neewer BME-800 (and similar mics) that are based on the the classic condenser designs from the 1950-60s.

    Could you add a sensitivity control? Yes. Although that means calculating out a resistor string with log-points which is a bit fiddly although not impossible, we tend to only meet this sort of thing is a log track rather than a resistor string. A better solution (and one which I personally prefer) was invented by Peter Baxendall in the 1950s. Peter is better known for his bass/treble equalizer (1952) which is world standard to this day. The Baxendall volume circuit, which uses linear resistance steps is part of the V2 design for this reason. You can have three steps or three hundred, but it will always follow a very good approximation of a logarithmic curve and that's important because so does our hearing.

    image
  3. Muting is definitely doable. It's not something I've integrated (it never came up) but it's simply a matter of connecting the input side of the digitizer (that's the most practical position) to ground or using the Baxendall with the "0" position on the Baxendall (BaxIn -> BaxOut) pictures above, is essentially a mute position because 100% inverted output is fed back to the input.
  4. The Unidirectional description of the JLI255 (and similar) relates to the directionality of the capsule itself - how it hears if you like. Pressure differential mics like the 2555 are most sensitive at the front. It's more usually called "cardioid" (from heart) pattern when viewed from above - sounds coming from directly behind are almost entirely blocked. The idea comes from the the ability to hear a speaker (sat in front) but to be "deaf" to the sound of an audience. Omnidirectionals pick up sound from everywhere at roughly the same level.
  5. USB noise reduction has to be designed for the load (which is why there are no actual values for the board as yet since I need to measure the current in use and the finals prototypes in China). Chimera is USB powered directly and you can put the noise suppressor direct in series with the power. The original design does draw significantly more current which affects the values significantly. That's not to say it won't work, it does, but the NMA0515 does pull a lot of current compared to a couple of small op amps. Without dipping into complex theory, you can intuit what's going on by understanding that the efficiency of the noise reduction relies on the storage and SLOW release of energy. The harder the load (that's the preamp board here) yanks on the energy store (the capacitor/inductor) the more buckles the ability of those components to smooth out the spikes. Under ideal conditions, each rising spike is "absorbed" by the inductor.

    As the spike falls off the magnetic field starts to collapse and the falling voltage is replaced by the energy stored in the capacitor. The harder these components are loaded the less time they will be able to satisfy demand - and therefore, more noise will creep in. This probably warrants a more detailed explanation but if you're a math nerd (I'm not) you'll find the equations all over the web.

    I've built a couple of functions into the suppressor - the first part is for sudden huge demands when the voltage drops more than 0.2V below median, this works due to the function of the Schottky diode. You don't *need* a circuit board if you're only using the filtering parts, you can put these sections anywhere convenient, even wired "point to point" or on stripboard. Several can be chained - but remember that each one loads the one prior to it.

  6. Any "soft" rubber suspension can be used to support the capsule. Matthew just made his really steampunk and totally rad! If you have the right sort of 3D printer (direct feed) you can get very soft plastics and print your own "buffers". Once again, there are quite a few teardowns of professional mics you can see on the web to see how the pros like Rode do this. The compliance of the material is what matters most as it needs to absorb very low frequency "thumps and bumps" but not wobble so much that the capsule bangs on the head. Cheap mics generally come with a hard plastic "saddle" that is hopeless in this regard. 

Returning to this noise issue (I've split this off for the original design) the NMA0515SC requires 256 mA to deliver 2 x 33 mA at the output. So we can assume that we need to "buffer" at least that in the noise suppresser with an input impedance of about 10 ohms (from USB 2), we'll use the Marki Microwave tool:

LC Filter Design Tool (markimicrowave.com)

Call it 250 mA at 5V - that's a load of 5 / 0.25 = 20 ohms.

Let's pick a Chebyshev, low-pass (we're trying pass DC, but block all that HF rubbish). The Chebyshev function "ripples" a little in the passband but what's more important to us is the steepness of its falloff - and that offers the best

Picking the exact cut-off frequency is a something of an art because it depends on the actual noise that's appearing at the output. In the case of my laptop there was a nasty whine coming through around 2-3 Khz, it's unlikely to be much lower and will often be higher. The problem we're facing is that there are multiple frequencies switching currents (often measured in amps) all over the place. And when you add multiple hard switched waves together you end up with a non-descript wave which in a Fourier analysis appears to show peaks all over the place due to myriad interfering harmonics. 

Let's push that cutoff back to 1 KHz and feed that into the tool - here's a fourth order (2 section) filter. I've asked the tool to use values we should be able to buy (you can wind inductors, but that's a massive bind and it's far better to let someone else make it for you). Power inductors tend to be modest in maximum size (100 mH is typical) but tools like this will allow you to find the ones that suit a given problem.

image

By 2 KHz, any harmonics have been reduced by a factor of 10 (20 dB) and will continue to fall off as frequency increases. 

image

But why can't we just stick a couple of huge capacitors in there? The reason is practical capacitors are far from ideal and are much more complex in reality, particularly at high frequencies above a few 100s of hertz where their lead inductance and internal resistance start to work against them. (Low ESR caps are best in this regard, and for "large" values 10-100 uF we can now get ceramic (MLCC) ones which are far better than classic electrolytics). I suggest using an MLCC in the first stage of Heychia for this reason. Although I didn't detail this, you can leave the resistor out (short it) and don't fit the diode. 10 uF is the MAX allowed by the USB 2.0 standard although many people have gone way beyond this.

The point is it's not strictly supported if you follow "the book" - but I've seen it work with 4,400 uF as evidenced by Matt's original. While we shouldn't do this, I expect most modern controllers have a foldback current limit so they shouldn't break. If you exceed the max draw on a port, Windows/Linux/MacOS should normally shut the port down and complain. Not something you really want to try on an expensive machine or the magic smoke Genie might visit.

The main reason we think of low ESR (equivalent series resistance) in a practical (electrolytic) capacitor - the ones we use for bulk store - is that it can be a source of heating. The larger the "resistor" is the more heat is generated by the fluctuations on the DC which, over time, dries out the electrolyte and eventually the component fails. The cross in the top of modern devices allows the pressure to dissipate when the component fails so it doesn't explode. (They still do sometimes and other times they just leak electrolyte all over your nice board.) 

The other way to look at it is as a low-pass RC filter with the equivalent series resistance being the R in the circuit. That resistance affects how fast the theoretical capacitor can charge/discharge compared to an ideal component.

image

 (Full discussion here: Why Low ESR Matters in Capacitor Design (passive-components.eu))

An ideal capacitor looks like an instant short circuit when power is turned on, a real one looks like an ideal capacitor in series with its equivalent series resistance. A large ESR can actually prevent circuits from going "pop" ironically. 😉 Not that it's something we should rely on - and that's why there's a series resistor position in Heychia in case you want to use a slightly larger value but retain a safe margin. 10 ohms is enough, but slightly larger values help prevent that margin from eating up the max current - because there's a larger load further downstream (our preamp in this case).

Designers are taught that while you can get away with working right up to the bone (a 16V capacitor in a circuit designed for 15V) we *should* take the next largest value to give ourselves some margin for error or component variations. The cost saving from 16 to 25 volts is negligible until you start order 1000s of these things. There's also a matter of size - the larger operating voltage parts tend to be physically larger. MLCCs are tiny in comparison which is is why they are slowly replacing electrolytics in where small(ish) values are required for bulk decoupling. (When I was training back in the dark ages, a 100 nF capacitor was about the same size as a 10 uF *through hole* MLCC is today (100x more capacity). We still use 100 nanofarad caps for local energy storage - you'll spot them dotted around very close to the power entries on all ICs, these days they're just very, very small indeed.

But why "as close as possible" to the IC? Because the tracks on our PCB also have resistance & inductance. Even the bonding wires INSIDE the integrated circuit have some. Now it's tiny, to be sure, but modern digital electronics switch from low-to-high and high-to-low at ridiculous speeds now and that tiny inductance/resistance can really ruin your day. That's a bit more than you need to know at this stage but I've included it because it's something that's rarely discussed outside of dusty electronic reference manuals and at degree level but it's good to know.

Rick Hartley and Eric Bogatin (both on various YouTube channels) have some excellent demonstrations on this sort of stuff. Rick will teach you "the energy is in the fields" - that, to me, was the most important thing I ever learned in PCB design. It's as basic as Ohm's law and yet (at least in my day) fields were just something that happened due to magnetism. 

Inductor theory tells of a magnetic field - and it's taught in high school physics because it's easy to demonstrate. But the energy in a capacitor is also a due to a field (we're taught that it's charges moving from plate to plate) but that's a literal view - it's a field that stops them just moving away when the power is removed! We can't demonstrate that field with a magnetic compass but it's there.

Fields are everywhere though. Mains "hum" (50/60 Hz + plus harmonics 100/120, 200/240, 400/480, etc.) is caused by the magnetic field that forms around all the wires in our homes.

Moreover, the idea that electrons bounce off the end of a wire (called a reflection) makes very little intuitive sense and worse, hard turns on a PCB wire also cause reflections which. The moment you can imagine a "field" (and this will bake your noodle) which is in the BOARD, not the copper - things make a hell of a lot more sense. It's the fields that move the electrons around or hold them in position and fields move at the speed of light in a vacuum. Demonstrations of reflections (and this is usually only discussed in RF, especially in antenna theory) often talk about the electrons getting reflected but it's actually the field around the conductor that's getting reflected - when it hits the end it has nowhere else to go due to the "infinite" resistance. Radio aerials are "tuned" a specific length according to their operating frequency so a wave (or a fractional part of one) can fit exactly into it's length - called a "standing wave ratio".

Eric and Rick are masters in this area so I won't try and explain more than they could so I'll leave this discussion there. Robert Fenerac and Altium Academy have both hosted these men and they are among the word's best. If they say I'm wrong, then trust me, I'm wrong. 🙂


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

 
Posted : 27/05/2024 2:20 pm
(@gilex)
Posts: 5
Active Member
 

Thanks @marcdraco for your quick and detailed reply.

> script-kid caught in action 😄

I've now decided to order the pcb from jlc but received an email from their support because some of the holes in both inner layers are misaligned with the top/bottom layers.
Is this intended/negligible or have I messed something up while placing the order?
I used Gerbers.zip from Project_Files.zip, but all other gerber-files you uploaded show the same mismatch in the jlc pcb viewer.

Picture & text from support:

... As shown below,hole can't match the inner layers, ...

downloadMessageFile

Picture from their pcb viewer:

Screenshot 2024 05 30 at 11 34 40 JLCPCB PCBA Viewer

 I also can't find the inner layers in the KiCad project to generate new gerber-files (not that I would know how to properly fix the wrong positions 😅).

 
Posted : 30/05/2024 10:02 am
marcdraco
(@marcdraco)
Posts: 476
Member
 

Morning @gilex

There are a couple of different versions of that board, JLC are excellent at spotting cockups though so I'm truly thankful for that. I wish they'd spotted a more recent one... Ouch. The offender and the one I hope I fixed are due back today in a couple of hours.

What's probably happened is I uploaded the wrong one by mistake (always best to check with an Open Source Hardware). I only have my other half to check and they're not the best [cough, cough]. 🙂

The inner two layers are just grounds to improve the EMI rejection but it's negligible.

That carefully milled hole I "dug" around turned out to be a bit too much for other PCB houses (JLC managed it without a hitch). Quite how I managed to upload the wrong one...  Err...

Actually I do know... this popped up after I'd posted that - I managed to leave some of the wrong Gerbers in the production folder... That's a gotcha when I'm doing this manually (KiCAD 8 doesn't have a JLC automation as of my last order).  

Idiot caught in action. 😉

I found that the earlier Minerva/Chimera boards picked up a bit of "hash" from the 5V supply (which is why I designed "Hesychia" for USB noise reduction).

Each "head" or device has to be analysed for the correct component values - a couple of random components won't be optimal. I'll get the closest ones when I have working prototypes to test.

This is (I hope) the last version of Minerva, now re-titled Chimera because there are so many changes. Costwise there's nothing in it unless you use the 2SK208 processed at source that might work out cheaper than getting the LSK170a. 

image

This one will be back today (with notably simpler drilling) and, I hope, better noise rejection due to a regulator in the most sensitive part. I'll test these first and upload all the project files here if I've stomped on all the bugs.

I'm sending you a PM later.


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

 
Posted : 30/05/2024 11:21 am
xenon
(@xenon)
Posts: 10
Active Member
 

While waiting for the capacitors to get delievered I was looking a bit into the circuitry...
I was wondering, why choosing a rotary switch instead of a potentiometer as described here? https://thatcorp.com/datashts/dn138.pdf
There is also a capacitor in series which makes me wonder if it's really necessary.

C1, at 6800 uF, avoids changes in dc output offset with gain.

And should I connect the ground from the USB to the ground of the rest of the circuit or just to the NMA? If I understand right it's already connected trough the audio part of the capture board right?
And as the last question for now... Is any type of shielded cable good for the part that goes from the desoldering wick or do I need an audio grade cable?

 
Posted : 01/06/2024 6:12 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

Posted by: @xenon

While waiting for the capacitors to get delievered I was looking a bit into the circuitry...
I was wondering, why choosing a rotary switch instead of a potentiometer as described here? https://thatcorp.com/datashts/dn138.pdf
There is also a capacitor in series which makes me wonder if it's really necessary.

C1, at 6800 uF, avoids changes in dc output offset with gain.

First of all THAT Corp. is very naughty suggesting a volume control on the gain stage. It's necessary in some applications (like this) but it's in a very sensitive part of the signal path.

The 151x chips are "instrumentation amps" that have been optimised for audio. The cap has to be enormous because it has to match the impedance of the smallest resistor in use (five ohms or thereabouts) at the lowest frequency of interest for the 1512. If it's too small low frequencies will be lost because of lower gain. At higher frequencies the capacitor looks like a short circuit so the gain behaves according to the equation.

So why is there a capacitor there? If not, a small differential DC voltage at the inputs will by amplified by whatever the gain is and appear at the output. THAT1512s (like other instrumentation amps) in professional designs won't have a series capacitor at the output because they can cause oscillations. Matt's design has an output capacitor so the input one isn't necessary. (It's a virtual certainty that the the digitser has a capacitor on its input too.)

Matt did this successfully but by the book, if you do this you're supposed to use a Faraday cage around the signal wires - screened cable in other words. The switch gives reliable and repeatable "volume" although in reality most users will set it and forget it.

Switching gains at this stage can cause pops so for V2 I've put an an active gain stage in that doesn't need a capacitor or weird resistor values. Although I haven't done it (time and budget constraints) this circuit can also employ a "digital potentiometer" for superb performance by computer control. (The V2 boards are ready but I'm still perfecting parts of the original.)

Short answer: No. It's not 100% necessary.

Posted by: @xenon

And should I connect the ground from the USB to the ground of the rest of the circuit or just to the NMA? If I understand right it's already connected trough the audio part of the capture board right?

This is an interesting one that really deserves a post of its own. The NMA0515SC produces a two 15V outputs referenced to an internal ground that is isolated from the circuit ground. The THAT151x has a ground reference pin (Ref) that tells the output where ground is in relation to the power supply.

Sounds obvious, but differential circuits have "their own idea" of ground and in some applications (again, not here) "ground" at one location might be 100s of volts different to ground where we are. This is where the Ref pin comes in - in fact, it's theoretically possible to drive the 151x chips from a single 10-30 volt supply by lifting the Ref pin to mid supply. This isn't as easy as adding a couple of resistors though as the pin needs to "see" an impedance of less than one ohm to reference (virtual ground) which, practically speaking, means more electronics. 

Short answer: Yes. Tie all the grounds together as close as possible so everything is referenced to USB's ground. The only practical "issue" doing this is we loose high voltage isolation and, perhaps, ground noise rejection. 

Posted by: @xenon

And as the last question for now... Is any type of shielded cable good for the part that goes from the desoldering wick or do I need an audio grade cable?

Audio grade cable will do the job very well. For a short run, differential pair like this the screen is almost unnecessary. The vast majority of noise pickup comes from the capsule due to the exposed JFET. Differential signals have this neat property of cancelling out noise picked up on the line and in fact, that's one of the main reasons we use them.

Version 2, which is now in "beta", despite me promising myself I'd call this one "done" has an improved head design, suggested by Matt, has much lower impedance (50 ohms vs. 2K) and lower distortion, although it's not really noticeable as is. V2 supports both the the original capsule design and the newer phantom power compatible ones (so you can use the JLI2555 or one of the larger types) and mount it direct to the circuit, add just three wires (two if we don't include ground/screen) and case you have a full, professional grade, P48 phantom powered microphone.

All the design files are available on request but I'll ask Matt if we can fork a different discussion for that design because it really needs a one of his videos to explain how it all goes together. V2 heads are also available with a factory fitted JFET (2SK208) because the soldering for the TO92 LSK170 is a bit tight, to put it mildly.

I have three complete kits to give away at some point (through hole V2 pre-amp) and a selection of heads. (They're all just gathering dust right now), but it's intended for educational use, perhaps a class project. I'll post details of the giveaway when time allows.

I will have a small number of he various capsule adaptors on eBay eventually, but they will be sold at cost plus postage in the UK which is about £6 for the head unit and maybe £12 for the V2 pre-amp prototypes. JLCPCB will make a set up of 5 for around £40 for the small ones and £60 for the larger ones with expedited shipping.

The designs are pretty simple and if you can use KiCAD you can order the board of your choice and just order the parts you need assembled. Making the SMD versions at home (from a bare PCB) is not for the feint of heart. 

Short answer: Any audio grade cable will do, the soldering braid is part of the design aesthetic. Rather ingenious idea actually but some of us have been caught out by Chinese solder wick that's look like a flat tube but is woven that way.
 

 


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

 
Posted : 01/06/2024 10:11 pm
xenon reacted
xenon
(@xenon)
Posts: 10
Active Member
 

Thanks @marcdraco for your time in the detailed and really interesting reply! I thougt I was trippin as soon as I saw the huge capacitor on the gain pins lol.

I think I'll go for a big potentiometer because imo it has a better feeling even if it's more "inaccurate".
(500 or 600 ohms plus a 5 or 6 ohm in series to prevent goin below this value)

image 2024 06 02 005051012

 

 
Posted : 01/06/2024 10:55 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

Yeah it is a bit of a "HOW BIG!? Is that misprint?" moment isn't it. They have to set the low frequency sufficiently low that it doesn't adversely affect the very low frequencies. Just for interest for people who don't know, this is how it works

The calculation is (if you know C)

R = 1/2*Pi * F * C

Where:

  • R = effective resistance of capacitor in ohms at frequency (F) = (impedance).
  • F = Lowest frequency of interest.
  • C = Capacitor in Farads

So:

  1 / 2 * 3.1415 * 20 Hz * (68000 e-6)

Comes out to about 1 ohm so at 20Hz (negligible) and at 40 Hz it's dropped to 0.6 ohms. Even with this fairly beefy, and potentially costly, capacitor the gain is affected, albeit slightly which is why it's better avoided. Omnicalculator can do this a lot faster than fiddling around with paper pencil and log tables 1 a pocket calculator. 

Capacitive Reactance Calculator (omnicalculator.com)

A more traditional way to do this is to fix the gain (say at 40 dB) and tap the output via a logarithmic potentiometer). The trick with this method is to get the THAT's gain so it's just clipping at the loudest you expect to speak into the mic and then use a potentiometer (or if you want to be flash) an active control like I've popped into V2.

You can use a pot at the gain pins too, remembering that it's a sensitive part of the signal path so it requires screening (unless the whole case is screened). You'll probably find that the DC resistance of your jumpers adds a couple of ohms, so keep that in mind when picking the lowest value for gain. That's another little "gotcha" that analogue electronics will do to spoil your day, although it's mostly on the grounds where the resistance of a poorly sited copper trace can create a ground loop and ruin your entire week.

What I love about DIY (and audio frequency stuff has always been my area of interest) is that we have a choice over what works best for us. I've lost count of the number of microphones I've made over the years, including a ribbon mic. None exist any more because I'm constantly improving them and I did the 48V version of the head (capsule adaptor) primarily because my video camera has 48V phantom power. 

 


1. Yes, I'm old enough to have done this on paper and it's a long-winded exercise.


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

 
Posted : 02/06/2024 10:21 am
xenon reacted
xenon
(@xenon)
Posts: 10
Active Member
 

Really hard keeping it in place while soldering the joint. I still have to put on the hooks and the mesh.

A hook has to go right over the joint but I prepared a clamp that should keep it tight.

The holes were made after bending it (whoopsies 🤪) but they aren't that bad.

I might rebuild it but I'll see how it goes. 

IMG 20240602 105002
IMG 20240602 104900
IMG 20240602 104900
IMG 20240602 104908
 
Posted : 02/06/2024 5:40 pm
marcdraco reacted
marcdraco
(@marcdraco)
Posts: 476
Member
 

Absolute swine isn't it? And Matt makes this stuff look so easy.

In fact, this is the problem that I initially set out to solve with a simple board that just carries a JFET that's soldered directly to the back of the capsule. About two years back, When I started mine, I managed to snap a lead right off the LSK170 because I tugged a bit too hard. And that got me thinking..

Kinda went a wee bit overboard... but we all win as a result. The V2 capsule adaptors have a nice ring to solder to - of course, as I've said previously we'll need a DIY Perks video to help everyone put it together.

I have to say that looks really rather good, esp. with all things considered. Wish mine was 1/2 as good. The electronics I understand (mostly) but without the help of a machine shop, metalwork is a bit outside of my abilities.


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

 
Posted : 02/06/2024 5:50 pm
xenon reacted
xenon
(@xenon)
Posts: 10
Active Member
 

Can someone tell me which cables of the capture board go where? I can't attach pictures because I accidentally uploaded some but didn't use.

black

red

black

red

none

white -->

black --> 

red --> audio positive I guess

 

I'll upload the a picture of the circuitry tomorrow because I think it looks very good.

Thanks.

 
Posted : 06/06/2024 4:16 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

Assuming we have the same board (there are at least three different) ones.

There are four connections on there, comprising two stereo pairs. We don't worry about which "side" here (although you can connect both) at the same time.

Black: ground

Red: right 

White: left 

Basically, all the wires that aren't black are audio inputs. 

 

This post was modified 1 week ago by marcdraco


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

 
Posted : 06/06/2024 5:02 pm
xenon reacted
xenon
(@xenon)
Posts: 10
Active Member
 

@marcdraco should I short the two red and white?

 
Posted : 06/06/2024 5:38 pm
marcdraco
(@marcdraco)
Posts: 476
Member
 

No reason why not. From your diagram you could just "short" that white and red. The others will be connected (internally) together. I've a feeling there's a mixer on the board but I've never checked.


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

 
Posted : 06/06/2024 5:41 pm
xenon reacted
xenon
(@xenon)
Posts: 10
Active Member
 

@marcdraco cool, il let the other two alone for the moment, just in case. Thanks 👍

 
Posted : 06/06/2024 6:01 pm
marcdraco reacted
marcdraco
(@marcdraco)
Posts: 476
Member
 

All go. Generally speaking you can "short" inputs (provided you know what is connected to it, we do here) but never short outputs.


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

 
Posted : 06/06/2024 6:07 pm
xenon reacted
xenon
(@xenon)
Posts: 10
Active Member
 

Well, it's cooked... 😐 
I wired it up w/o the shielding to test it but something probably went wrong...
The PC recognizes the device but it says "Device needs more power than the port can supply." and this was the first warning sign...
The second red flag was the THAT becoming pretty hot...
The pinout is as it follows:

1--__--8
2              7
3              6
4-------5

with 1 and 8 being connected with a temporary 200 ohm resistor,
2 --> -in,
3 --> +in,
4 --> -15V (passing trough 100 ohm),
5 --> GND,
6 --> 22uF cap non pol (output),
7 --> +15V (passing trough 100 ohm)

I think it's bad. Atm I don't have a scope or multimeter to check the voltages. 

This post was modified 1 week ago 2 times by xenon
 
Posted : 06/06/2024 7:51 pm
Page 24 / 25