Wireless Microphone Troubleshooting

@marcdraco 

Thanks so much for taking the time to write all that out. I really learned a lot! I’m still trying to wrap my head around the whole virtual ground concept, but your explanation (especially the comparison to potential energy and the current and voltage readings across circuits with Falstad) helped me understand it a bit better. I probably just need more time to fully absorb everything haha.

I also had no idea about the different types of grounds and how they're represented in schematics, so I’ll start being more mindful about the symbols I use from now on.

From what I understand about the green arrow that Jules used, it essentially acts as a virtual ground that connects different parts of the circuit together. He didn’t just use schematic lines to connect them directly, as I assume that that is the proper way to design his schematic, which can be confusing for beginners like me. That’s why it’s really important to learn the basics and intermediate concepts before jumping into real-world schematic design and printing it.

As for the PCB, I didn’t choose the assembly service from JLCPCB. I followed Jules’s design (I assumed that I get his reason why), which had the resistor mounted vertically. I didn’t really know the downsides of that at first, but I’ve always felt it wasn’t ideal, even if I couldn’t explain why haha. I just noticed that 90 to 95 percent of the PCBs I’ve seen have resistors mounted flat, so there must be valid reasons for that.

I also ran into a fitting issue that I think is one reason Jules opted for SMD components instead of through-hole. I designed the board to fit inside a BM800 mic, and using standard THT footprints made all the components not fit to the board. I might try using the larger 1206 SMDs, since I don’t have a hot air station yet and would need to use components I can solder more easily. I’ll just check first if the specific values I need are readily available here in our country.

I’ll also look into TL431s and LDOs when I start improving the design. Since those have three pins, unlike a Zener which has two, I’ll need to learn where to connect each pin. I’ll probably give that a shot once I feel more confident (and once my wallet stops crying haha).

For capacitors, I used radial leaded ones, because I was into audio gear before, and I always saw brands like Nichicon being used in modding handheld headphone amps (like Zishan Z1, Little Bear BX-4, etc.), especially the Fine Gold and Muse BP series. I’ve always assumed they were ideal for audio circuits, unless someone tells me they’re no better than the cheaper brands (I’d be a little heartbroken haha). I’m also considering replacing the solid polymer cap with an MLCC.

Sadly, I already printed the PCB, but it only cost me around two dollars for five pieces. It has flaws in the schematic, but it taught me a good lesson haha. Hopefully my girlfriend understands why I’ve suddenly been spending money on this project haha.

Thanks again! I’ll keep refining this and post updates on the forum.

I remodeled my schematic and PCB, fixed the virtual ground issues, and arranged the resistors flat on the PCB. I managed to fit everything by extending the board outline, and I'm hoping my measurements are accurate haha.

 

Now THAT is orders of magnitude  better my friend. Consider giving the capacitors a little more room - perhaps a millimetre or two spacing - so you won't find the size tolerance to be an issue. And naming if for someone special is always a lovely thought, isn't it? 

There's a saying in PC development (which also applies at least equally to electronics), "good looking code works better".

It sounds a little strange (and I can't remember who came up with it, though I expect the electronics lads had it first) but there's actually a lot of truth hidden there. 

When you write clear code it's easier to maintain, when you draw a well thought out layout it's easier to assemble and it's much easier to find any errors (which you're less likely to make).

I find EasyEDA a bit fiddly myself (I'm a KiCAD guy) and in PCBNew, the PCB section, we can specify components with "hand soldering pads" which are slightly larger than the the usual ones for SMD so you have some space to heat the pad in place. 1206 parts are quite chunky in the grand scale so with care you shouldn't have too much of a problem.

Alternatively, you could swap out the axial leaded THT capacitors for SMD parts. They have quite large pads and are generally a good deal smaller than the similarly specified part. Through hole resistors are fine here and it makes the job a little easier. 

Best practice these days says you would be better off using four layers vs. two, and dedicate the two inner layers to circuit ground. It's not really necessary for such a small circuit, particularly one that's going inside the BM800 donor body (because that offers a neat and effective Faraday cage). So much so that you can, if you wish move the JFET onto the main board. Practically speaking there's no real advantage though. We mounted the JFET in the V1 (and V2) so we can have the capsule mounted far away from the low-impedance side. 

Grounding in Schematics

You're spot on there. We use the GND symbol to connect parts together on the schematic rather than having a rat's nest of wires running all over the place. As you gain more experience, you'll find that EDAs have other short-cuts to indicate parts of the board that connect together (net names and labels) which tell the EDA software how to connect everything without having to draw lines everywhere. Simple circuits don't really benefit from this but anything with a Virtual Ground should really have that rather than using a ground symbol.

Some very complex boards even have solid planes for power too - 5V, 3v3 and 1v8 for example plus grounds to contain the EMI. It all depends on what you're trying to achieve. I'm tying myself in knots over the last few traces I'm dealing with because the board is crying out for thinner traces (less than 0.2 mm) and I'm even avoiding 0.2 mm because that's the limit of what JLC and others will make on their low-cost boards. Four layer doesn't cost a great deal more and there's nothing to lose and everything to gain from having your grounds all link up "automatically" plus those grounds are much closer to the fields they're trying to capture so your board should work better, even at audio frequency. 

KICAD calls them zone fills, I haven't checked EasyEDA although it does import from KiCAD now so that's a bonus.

Capacitors

Caps are weird, particularly when we move from pure insulators (ceramic, polymer) to chemical ones which are made from a Swiss roll of two layers of foil separated by an electrolyte that acts as the insulator. MLCCs are a wee bit trickier as they come in a variety of temperature and X5R, C0G/NP0 and so on constructions. Unlike the more usual capacitors we see MLCCs have some very peculiar qualities than can really spoil you day.

Better quality audio grade caps generally have a tighter tolerance around the rating. Check the spec sheets where you'll find that budget electrolytics might have a tolerance of 10% or 20% beyond their rate value. In audio circuits that can serious degrade the performance of a filter and unbalance a stereo signal.

 

C0G are expensive and only available in very small sizes (of capacitance) physically an X5R will always be smaller than a C0G of the same capacitance and voltage but they don't have the weirdness of MLCCs which are:

1) Sensitivity to vibration. If you tap a board that uses MLCCs in the signal path (and most do, often for practical or budget concerns) you can see the effect on a scope or even hear it in a microphone! I believe this is because due to the nature of the construction when the part vibrates, the electrons (which are only held in place by the electric field) get shaken off the plates and cause a tiny amount of current to leak. 

2) Reduction in rated capacitance under DC voltage. Now this one is very weird on its face but I think (I haven't found a decent explanation) that at a physical level, these capacitors can saturate in much the same way as an inductor does.  Saturation simply means it's stored as much energy as it can so that means there's no room for more electrons to migrate to the opposite plate when a signal is applied. I could be totally mistaken but that's how I think of it. The trick is to use a larger voltage rating than you need wherever possible.

If you've got a circuit producing a 6V DC with a 100mV AC riding on that, then a 6v3 rated MLCC wouldn't have a lot of space left for that extra charge to move around, whereas the 12V (or larger would). In fact, much the same applies to electrolytic capacitors but that's because they tend to get hot if mistreated due to ESR.

ESR isn't unique to electrolytics, but it's invariably a lot worse with that form of construction. I won't deep dive into this, but when you see a capacitor on a schematic it looks like a couple of plates but as far as the current is concerned there's a deeper story.

We call the schematic components "lumped" because it hides a lot of the bear traps in the lump. Every component we have (and even the traces) has capacitance, resistance and inductance. They're usually so small that they can be ignored until you work at R/F (well above 100 KHz) but ESR can really spoil your day, mostly in power supplies where the man who ignores the ESR is a man whose board is going to break sooner rather than later! 🙂

Here's an example: a real capacitor (in the simplest form) looks like a capacitor with a small resistor in series. (It has inductance and a parallel resistance too but those are trifling and rarely need to be accounted for. That series resistor can be an issue though.

Resistors "slow" current down by impeding the flow of electrons through the material. As capacitors age this "internal" resistance gets larger - it starts out typically at less than one ohm but can rise into double figures and when it does that means the part is doomed.

If you put too much current through any wire (or a resistor) it gets hot - that's energy conversion (energy can't be created or destroyed, only changed). But heat is the enemy of components and the electrolyte inside those caps needs to stay cool.

This usually becomes an issue in a power supply where the bulk capacitors (like the ones that Matt used in the original) are there as low-pass filters - allowing the DC through while passing the "ripple" to ground. As that internal ESR gets larger, the effect of the resistance is to make the part heat up and over time it will fail completely. Most capacitors have (or should have) a cut - often an X shape in the top so rater than exploding (and that does happen!) when the device bursts, it pops open at the weakest point and "safely" sprays electrolyte upward. In many cases, that fails and the darn things spread all over the circuit board which can, in extreme cases even destroy the PCB tracks. 

This is one area where MLCCs can really help as they are available in quite large values now and by nature have ESRs measure in the milliohms! This makes them perfect for supply decoupling - particularly if you wire several in parallel which reduces the total ESR while raising the capacitance! 

The other "gotcha" and why the better electrolytics cost more is tolerance.

We often need to block DC from entering other parts of the circuit (op amps are DC coupled but we even need to add a capacitor to block incoming DC, such as the 48V phantom power.

Balanced lines (where we have two equal and opposite signals) MUST offer the same impedance to the circuit they're driving or the balance is lost to some degree and "common mode" noise and other rubbish gets through as a result.

Capacitors allow AC to pass as you know but they have a response called reactance which is determined by the frequency of the signal we're interested in and in the low audio band, we need a large capacitor or we'll end up with what amounts to a large resistor filtering out frequencies of interest.

The audio band goes from 20 Hz to 20 KHz (this drops off as we age but out brains fill in the blanks so we perceive higher registers that we can actually hear, esp. with familiar music. The speech "band" is 300 to 3 KHz (it's actually much wider than that but this comes from the early days of telephony) so we only notice that we're going deaf as it becomes increasingly difficult to understand other people.

But in music sounds go all the way down to 20Hz (sound effects in modern movies go even lower, perhaps as low as a few Hz but we don't hear that, we need special drivers to reproduce it and done right, we literally FEEL the sound).

But to my point here, the input (be that single ended or differential) will usually have a high-pass filter comprising of a capacitor and a resistor:

We can calculate the -3 dB point (that's a convention and where the audio has dropped to half of it's power) from the formula:

F -3dB = 1/2 * pi * R * C)

With C in Farads and R in Ohms

For a single ended (single signal) this is pretty easy to figure out because we know the input resistance (a few K for a single ended input and as low as a few hundred ohms for differential) so given that we can calculate the capacitor we need to pass frequencies of interest and attenuate the ones we don't want. The very low frequency - sub-sonic - stuff isn't useful in everyday audio and while it might seem prudent to pass everything, the power present in subsonics can saturate the amplifier to the point where it's not working for us. Generally, we set the high-pass F -dB at around 15 Hz so we don't cut too much of the useful audio. Many microphones can reproduce sounds that low anyway so a higher cut off is sometimes used.

OK, so let's assume we've got a single ended input with a DC potential (any electret capsule will have up to 10 V bias voltage, typically less with a signal that we need riding on it. So we place a high pass filter designed like this. Let's put that input resistor at 10K for simplicity and run the numbers:

Sites like Omnicalculator have these pre-made so you don't even need to faff around with pencil and paper, so for a single ended medium impedance input, that works out to 1 uF (using a preferred value). To be more exact, the impedance of 1 uF capacitor at 20 Hz is a little under 8K.

In effect the input "see" a voltage divider comprising of an 8K resistor in series with a 10K resistor... which is how it works.

Sounds simple enough. Now let's look at a differential line and where this can become a problem.

Our differential impedance is very low because that makes the input far less sensitive to stray magnetic fields from the mains. The opposite is true of a JFET which is why the capsule has to be screened.

The induced current is tiny but a tiny current appearing across a very large resistor gives a lot of voltage - and voltage is what we're amplifiying!

So by getting the effective resistance of the input as low as practical (radio systems typically use 50 ohms) we can almost entirely eliminate those signals but there's a catch. 

Recall that I said the high-pass filter is a voltage divider - or to be more accurate, a frequency dependent voltage divider. And it's frequency characteristics are controlled by the resistor (DC resistance) in combination with the impedance of the capacitor at the the frequency of interest.

For a 47 uF as you've used here works as a voltage divider with the combination of the line resistance (47 ohms per side) plus the input impedance of the circuit it's driving. 

Which seems a breeze on paper but here's the catch.

On a schematic a capacitor has a fixed (and perfect) value. In reality, electrolytics (and the cheaper ones are awful) have a value tolerance as much as 20% - when NEW - and that will likely get worse over time.

Since the balanced line impedance has to match (ideally perfectly) any deviation in the value - and 20% is a LOT - is going to change the impedance of the little network, knocking the impedance of our balanced line all to pot, perhaps as much as 40% between the two and that means any "common mode" noise (that's noise that appears equally on the two lines) is no longer "balanced" so the input driver (THAT151x etc.) can't remove it.

Again, a lot of this stuff is less than obvious in prose, it's something you really have to experience for yourself but there are lots of videos on YouTube that will show you these effects on a scope so you can see how common mode interference can really spoil your day; all for cheaping out on poor-quality components.

As usual, I salute Matt on his choice of parts. He seems to go out of his way to specify decent quality and it shows. Made my task more challenging to as he set the bar I have to beat but that's part of the fun and I'm having a ball! 
 

@marcdraco 

I’ll make sure to give the capacitors a bit more breathing room in my next layout. I’ll also be creating another board with two SOP8 to DIP8 adapters, since I saw that the OPA627AU is available at a local electronics salvage shop. Based on what I’ve read in forums, it’s considered a drop-in replacement for the OPA Alice build, and better than the stock op-amp, as long as it's adapted into a dual op-amp configuration. I really want to learn how to properly create a schematic when using other op-amps, since I know there are always better options out there. The only issue is that most of them aren’t drop-in replacements for our builds.

I appreciate the tip on hand-soldering pads. I’ve been trying to find the balance between compactness and solderability, especially since I don’t have a hot air station yet. I’ll probably give 1206s a go for now, and maybe switch over to SMD caps too since I’m already tight on space.

Your explanation about multilayer boards and ground planes really cleared up a lot. I’ll hold off on four-layer designs for now, since I’m not sure if they would increase the manufacturing cost, but I’ll definitely keep the idea in mind for my future projects.

I’m glad to hear I’m on the right track when it comes to grounding in schematics. I’ll be more careful not to use the ground symbol for virtual grounds. Maybe in the future, I can apply what I’ve learned about ground references to a PCB design that really puts my understanding to the test.

The part about MLCC behavior was a new knowledge to me. It’s similar to what we call “microphonics” in IEM cables. I didn’t know that vibration in MLCCs could cause audible effects. I always assumed a 10µF capacitor was always 10µF, so from now on, I’ll start paying more attention to voltage ratings and tolerances. Also, thank you for the detailed insight into ESR and how it affects performance, especially in power circuits. With the electrolytic caps heating up and failing, It reminded me of the story you shared before, about how large electrolytic capacitors used to blow up, which is why they have an X-shaped marking at the top, to release pressure safely if they fail.

For now, I’m still waiting for the new wireless lavalier microphone so I can continue with what this thread was originally about haha. I got sidetracked with building a decent recording microphone using the Alice schematic, but it was all a lot of fun. Thanks again for guiding me through it!

The difference between four layer and two layer is literally a couple of $s on JLC and the advantage you receive from the field containment is really false economy.

There are some excellent videos from Youtube.com/@RobertFeranec where he discusses PCB design with people like Eric Bogatin and Rick Hartley (both men also do talks for Altium).

The key is largely in the distance between layer 1 and 2 & layer 3 and layer 4 -where the plane forms a capacitor between the stripline (the signal or power trace). 

If you recall from my discussions on capacitors (and the range of the electric field) the ability of a capacitor to hold charge is determined by the size of the two conductive plates and the distance between them. 

The electric field pulls electrons from one plate and drives them (in a chain) around the circuit until they clump on the other plate and are attracted by the opposing electric field just like the opposing poles on a magnet. Because the electric field is very weak, the closer the and the larger plates are, the more electrons are trapped. (If you blow or pull an electrolytic cap apart (carefully, electrolyte is caustic) you will see they are like a swiss roll of two long strips of metal or metalised plastic film which allows for two very large plates to sit in a small physical space.

FR4 (the plastic used on low-cost boards) is quite chunky so the space between layer 1 and 4 (on a four layer board) is often similar distance as the that from layer 1 and 2 on a two layer board - around 1.6 mms. But layer 1-2 and layer 3-4 are much, much closer so they make much larger capacitors (albeit in the very low nano-farad range or less but that's enough for signals to couple to a return path.

Rick Hartley made a very interesting point about how difficult A/F design is in this regard since low-frequencies couple far less well to the return plane than they do at radio and very high frequencies. This was one of those "OM*G" moments where I started to realised how much we're taught was wrong. I won't bore you with the details of the new V2 (because it's quite secret until we can reveal a properly production ready board) but the manufacturer's datasheet goes into VERY detailed explanations about how we should separate the ground reference planes into digital/sheild/analogue and more and it's so far away from what best practise (based on sound, scientific principals) it's not even funny. Rick Hartely waxes lyrical at length about this but it's really difficult to unlearn stuff that was current even into the mid 1990s, (which I'd learned in the 1970s and 1980s) and switch tack into what seems ... wrong.

Back then a fast IC would have a rise or fall time (its "slew rate") measure in the low milliseconds to a few hundred nano seconds. So those boards and those techniques just worked.

A friend of Eric Bogatin, has done some amazing animations showing this visually.

I'll attempt to explain this in prose (but you should go look out Eric and Rick's talks).

We're taught in basic physics that a when we connect a battery to a a lamp (it would be an LED these days) that the current starts flowing from the positive terminal to the negative one (conventional current flow) although electrons actually move on the opposite direction. That's why it's convention to draw a schematic with Vcc/Vdd (most positive) at the top and Vee/Vss at the bottom. Ground reference often sits in the middle of the power section but, as you've seen it's used all over the place so it's generally marked with that little arrow. Vcc - meaning "voltage at the collectors" (for a bipolar design) and Vee means voltage to the emitters. Vdd and Vss are the same for a FET design or a FET chip - which pretty much all digital electronics now.

But that story about the little electrons moving quickly through the copper is so far from the truth it's nuts.

What actually happens is that when we connect a supply to some circuit, a electric field start to emerge (in FR4) at about 1/2 light speed from BOTH the positive and the negative (or ground) reference which starts to move the free electrons and those in turn create the H (magnetic) field which follows immediately behind. 

This was predicted by James Clerke Maxwell in his famous equations from the 19th century but not proven until much later. These days we have things like 2D field solvers that can predict how current flows, including things like the "skin" effect which is how currents at high frequency only actually move electrons around in the outer layer of copper (the ones nearest to the electric field). A DC that field sinks all the way into the copper (or other conductor) but as switching frequency rises above a few KHz, the fields change so fast the the electrons deeper in the material aren't affected. This effect also causes the effective conductivity of copper to fall away, which is also why we have Litz wires (a specially wound bundle of thin copper wires) in induction cookers. If we were to use a solid block of copper the wire would get very hot and that would lose most of the magnetic field that we're trying to create.

So the point is that the current doesn't magically go into one end of a circuit and come out the other - it travels as the electric and magnetic fields through the plastic or even the air. Remember that even light is just electro-magnetic energy - moving through space. This is one of those concepts that makes no sense to many people - but without it radio, television.. everything we take for granted wouldn't be possible.

But a demo that make my noodle itch (and still does a bit) is that of a rectangular wave guide. At DC (and even at modest frequencies) a wave guide is a short circuit because it's a cross sectional tube of solid metal. But the frequency of interest, something almost magical happens. The width of the tube is set to 1/2 the wavelength of "light" at the signal and ONLY that precise frequency just fits so snugly inside the tube that it passes right thorough as it were radiating into thin air - but inside the tube hence the name "wave guide" because we can steer and deliver it somewhere else and use the guide as a very high Q (very steep sided) notch fillter.

This is also possible with waves of physical energy - like sound - something used to great effect in SAW filters - Surface Accoustic Waves - that work a bit like tuning fork only at frequencies massively higher than we can hear. The difference is the this is a moving pressure wave vs. a moving EM field.  

At real audio frequencies our problems are worse because the capacitors formed between the various layers of our PCB are so small (they have a high impedance to low frequencies) that the fields spread out much further and can start to energise parts of the circuits that they have no business interacting with. Electric fields always follow the path of lowest impedance and that's either the plastic (inner layers) of the AIR around our board. This is something we have to design for rather than running wires and hoping everything will work. Pouring copper on a double sided board does help to capture the fields on the opposite side but it also causes currents to spread around too and that rears up and bites me on the bum as audible noise.

The idea of separate digital and analogue grounds derives from this misconception. and can even make matters much, much worse. Because the magnetic field has a much lager range than the electrical field, devices that cause inductance (everything some some degree) and but mostly the wires and inductors can leak fields all over the place. Key here is that the field only causes a problem when the device (switches). This is expressed mathmetically as dI/dt - delta-current/ delta time. It's these edges where the change in current is large (milliamperes or more) vs the switching time (millionths of a second or less) that determine the size of the radiated fields.

Once again, the great Rick Hartley recalls a time where a lift company had been making the same board with the same ICs for years without issue until one day and entire batch failed. 

It took Rick to figure out what was going on and in fact, it was nothing more than a simple die shrinkage. As we've got better at making chips smaller with smaller transistors (CPUs being an excellent example but all ICs are affected to some degree) the physical distances between the millions of transistors gets smaller. And as that happens (since the electrons have less distance to travel) the switching time (slew rate) gets shorter so dI/dT the ratio is bigger... and we get MUCH faster edges.

It's entirely possible to slew limit signals (the ASIC at the centre of V2) employs slew-rate limiting to limit transients on some I/O lines but that's unusual. 

Die shrinkage save the manufacturers considerable amounts of money and it also makes the chips faster (which is usually a good thing) but in this case the transitions getting shorter, created much larger field bursts which upset the board's logic. That's pretty wild but Rick has loads of similar stories from his long and varied career so he's a must watch for anyone starting out in board design so we don't make the mistakes that generations of the old guard insist as still valid.

Which is a bit like saying that we should use drum brakes on a car because they work despite being so much less efficient than a disk and beyond a certain point are almost useless.

But it's great to do this for yourself, there's no replacement for experience.

@marcdraco 

Now I will consider that small cost difference, if the 4-layer board offers overall improved field containment and return path control. Thank you for the detailed explanation!

I also just learned that I have to manually place the power and ground planes on the inner layers of a 4-layer PCB. I might get lost when it comes to which components should connect to the inner layers, so I’ll try to study it more carefully.

As for the main topic of the wireless microphone, here are my major takeaways (for now):

If you want to skip the hassle, strictly stick with the components Matt used in the video.

Reasons why I said that:

I’ve used this kind of wireless lavalier,

(yeah I messed up the mic in pads of this one, main reason why I bought a 2-in-1 haha)

and also the square-type lavalier before. What I encountered with these types of wireless mics is that they sometimes use an SMD electret mic and don’t provide any pad for soldering an external electret capsule.

The major downside of not sticking with the exact mic Matt used is that most of the time, there’s no 5V bypass pad on the mic board, which is crucial for powering the LM386 preamp.

Some of the mic boards I opened had a labeled 5V pad, but it only works when the device is plugged in. I even tried removing the battery and powering it via the charging port, but the board doesn’t turn on without the stock battery.

That’s why I’m still stuck using a long USB cable connected from my PC to the power input terminals of the LM386 preamp, which I suspect is affecting the recorded audio.

 

Unfortunately, I didn’t take a picture of the full setup, but basically, I placed the LM386 preamp and mic PCB inside the BM800 mic body without touching any of its metal parts. I also removed the XLR pin at the bottom of the mic to route the USB cable. I triple-checked everything before recording, and here are the audio samples.

 

In this audio, I’m pointing out that the noise gets louder when I’m holding the mic, which isn’t ideal, since I want the mic to be used handheld.

 

This is the specific noise that my DIY mic is mainly producing.

 

This part is just me getting frustrated with the mic noise HAHAHA.

 

The noise was noticeably lower when I plugged the capsule and LM386 preamp board into the Blue Yeti PCB, compared to when I used a wireless lavalier.

 

Pardon me as I’m speaking in my native language in most parts of the recordings, but these are the audio I’m capturing. I’m not sure whether the issue is caused by

• the USB cable running out of the mic body
• the BM800 mic body not acting as a complete shield for the JLI-2555 (needs some sanding/post processing)
• or still something within my circuit (capsule and lavalier aren't sync with each other).

That’s why as of now, I recommend using the exact same capsule and mic lavalier Matt used.