Richard Mudhar

it’s always the capacitors – Logitech SB3 one channel low audio fix

It’s always the capacitors¹ causing trouble these days, and if it’s not them it’s the connectors 😉

When I started work decades back, in Test, on units that had a Z80 microprocessor I kept a Z80 spare. When a faulty unit came in, I always tried swapping the microprocessor. Never in my time there did that clear the fault, it was always something else.

ICs are surprisingly resilient, and after some decades of experience trying to fix things, there’s one suspect that seems to be the #1 troublemaker – the humble electrolytic capacitor, and particularly in its SMD form. I’ve repaired gear from the 1970s and 80s and often the fault is an overstressed transistor, they were more rubbish in those days. Often the electrolytic capacitors are still OK after decades, it’s the more recent ones that seem to fail earlier. Nowadays we just don’t seem to know how to make electrolytic capacitors that last. Or nobody expects modern electronics to be in service for more than five years 🙁

It’s not always the connectors and capacitors, but they seem to be the bad guys most of the time!

First suspect connectors, then electrolytic capacitors

I had a Logitech SB3 network player, and when I powered it up after a year it sounded tinny with low output on one side. I blamed the tinniness on the cheap and nasty headphones I used, but fiddling with the 3.5mm jack didn’t change anything. Tried the headphones on something else, they were OK, switched headphones, fault persists.

The usual suspect would be the connector, these are often mechanical in some way and prone to damage. However, there was actually something coming out of the low channel. Connector faults are usually honest all or nothing intermittency, or nothing at all.

A couple of months in the laboratory can frequently save a couple of hours in the library², so before I get this on the bench I enquired of Google, which came up with this thread, and this one . Both mention capacitors as the likely culprit. I took this SB3 apart

The offending region, caps C35 and C32. These may be marked different on a non 20050630 revision board, but you can buzz out the -ve pads to the headphone socket without unsoldering anything

I identified the output capacitors by buzzing out continuity between the jack connector and the negative sides of C35 and C32. If you look really hard at C32 you can see some gunk, but it wasn’t obvious at first glance, and there is some funky hand-soldering on the board connectors anyway. I determined that this board was made after the end of June 2005, by the original Slim Devices startup that Logitech bought out. So it’s 20 years old at most.

SMD caps were bastards to desolder – you can’t really use solder tweezers on them, but let’s hear it from Mr Carlson, just twist them off

This worked fine for me³ – the secret is in the construction of a SMD capacitor, which you can see from the wreckage

What you can also see is that C32 has leaked a lot. I was tempted to solder to the remaining posts after cleaning the gunk a little bit, but they didn’t take solder. I guess the clue is in the name – aluminium electrolytic, and ally doesn’t take solder.

Now you can see the mess from leaky cap even better. Isopropyl alcohol was my friend to get this off

What’s standing up is some sort of riveted connection to the leads/SMD contacts, so I snipped these as flush as I could with sidecutters to lift the plastic to show the full horror of the leak.

This construction seems to be a particular weakness in SMD capacitors. Look at the top of the replacement leaded electrolytics I fitted, and you’ll see they had to score the can, so if you fit these backwards or otherwise cause leakage it can break out the top of the cap. That implies the seal at the bottom is pretty damn good. While I’ve seem electroytics detonate when you wired them in backwards on a power supply, it it true, sometimes the top breaches explosively.

There was a problem with electrolytics made in the late Nineties/early 2000s and the Wikipedia article shows leaded electrolytics bulging at the top. SMDs just seem to leak at the base.

I had some 220uF 16V electrolytics I had bought at some radio rally. From what I know now I wouldn’t buy NOS electrolytics again, but I tested the replacements with a Chinese tester widget

OK that’s 20% low for 220uF but they were all 180 or 181, can’t really moan too much about the 0.4Ω ESR in this application

I decided I can live with that, and soldered a couple in place, forming the leads to make contact with the pads.

Not so fast. Reassembled unit and listened. Nothing on one channel. Eh? Then I observed the headphone jack was recessed – C35 was standing proud and fouled the curved back of the case somewhat. Solution was to replace it and lay it down.

Great. Sorted – back fits on OK and the socket is flush. Socket is slightly cruddy so it needs a clean, but signal quality is fine. I did take a look over the other electrolytics particularly the 220uF 16V ones with a magnifier, but they looked OK. I didn’t want to recap everything given I don’t have the right SMD caps, so I took the First Do No Harm approach.

There’s gonzo startup-ery written throughout the weird construction of this, from the uncleared flux on the hand soldered connectors and board interpass connection, though the main PCB looks properly made. And what’s up with that whopping great big 3300 uF 16V Nichicon cap jumped across at the top of the board, and another one patched in closer to the connectors?

In all fairness, however, their work has stood the test of time in that it is still supported, and this unit now works OK. Lyrion Music Server is the server end, which you can run on a Raspberry Pi using PiCorePlayer (which has a server offering). I couldn’t get NFS to work on that but a large-ish USB stick on the pi itself works fine. You can scp files to that to add new music.

More specifically electrolytic capacitors, unlike in days gone by non-polarised caps seem to last fine ↩︎
Frank Westheimer ↩︎
This guy pushes back on the twist idea and says hot air is a better way. Twisting worked for me this time, I may try it on some scrap gear to see if it is a general solution or only works soemtimes ↩︎

Røde AI-Micro USB-C phone interface

I am writing this from a field recordist’s point of view, not a podcaster POV. Podcasting seems to be the design intended use case.

There’s nothing more commonplace in the modern urban setting than somebody with earbuds glommed onto the delights within their a smartphone. Pair some binaural in-ear microphones like Soundman OKMIIs with a smartphone and you potentially have an inconspicuous and high performance street field recording rig. You’ll also look like you’re poor and have a cheap phone, because everyone haz bluetooth wireless earbuds, right?

measuring frequency with a 16F628A PIC

After disappointing results measuring frequency with an Arduino due to the Arduino system interrupts stealing clock cycles, I tried a 16F628A PIC. This is a much more primitive chip. In my system it has only one job, to spit out a count of 4MHz main system clock cycles that have passed between the onboard comparator detecting changes due to a 4.8kHz squarewave derived from a crystal oscillator.

That 4.8kHz is derived from a 2.4576MHz oscillator block, which I divided by 512 using a CD4060 to get 4.8kHz in the same way as last time.

histogram of counts using the PIC crystal oscillator

The spread of 6 in 64612 is 92ppm. I had expected even better and don’t understand the asymmetric spread. But it’s a darn sight better than the Arduino, where I was eating a spread of ~ 3745 ppm. That’s not a criticism of the Atmel chip used in the Uno – in the end if I am going to use the Arduino overlay to run the chip then I am not working on the bare metal¹ in the same way as the PIC.

If I look at the stream I can see that the low values come in bursts

Which makes me wonder if the serial port is quite as self-contained as it seems from the datasheet. But the whole thing is on a breadboard

You don’t really get the right to precision timing with this sort of construction

and with that standard of construction maybe 100ppm is good enough. I did have a decoupling cap jumped across the PIC when taking the readings. But the two oscillators could be beating with each other across the power supply, or the great big white wire carrying the comparator in could be picking up a mobile phone somewhere.

I need better short-term precision for a PCSM, building this properly would probably help, as the histogram shows the vast majority of the results are 64612 or 64613

Code – this is messy and the comments are more a stream of consciousness in development. The comments may be inaccurate 😉 Every half second this spits out the number of 1µs clock cycles that have passed since the PIC counted 620 transitions of the 4.8kHz. That should be

(1/4800)*620/2*1000000 µs = 64583 µs

I chose 620 to keep the number of counts within the 16-bit length of TMR1

#define _XTAL_FREQ 4000000


// CONFIG
#pragma config FOSC = XT  // Oscillator Selection bits (INTOSC oscillator: I/O function on RA6/OSC2/CLKOUT pin, I/O function on RA7/OSC1/CLKIN)
#pragma config WDTE = OFF       // Watchdog Timer Enable bit (WDT disabled)
#pragma config PWRTE = OFF      // Power-up Timer Enable bit (PWRT disabled)
#pragma config MCLRE = ON       // RA5/MCLR/VPP Pin Function Select bit (RA5/MCLR/VPP pin function is MCLR)
#pragma config BOREN = OFF      // Brown-out Detect Enable bit (BOD disabled)
#pragma config LVP = OFF        // Low-Voltage Programming Enable bit (RB4/PGM pin has digital I/O function, HV on MCLR must be used for programming)
#pragma config CPD = OFF        // Data EE Memory Code Protection bit (Data memory code protection off)
#pragma config CP = OFF         // Flash Program Memory Code Protection bit (Code protection off)



#define _XTAL_FREQ 4000000


#include <stdio.h>
#include <stdlib.h>
#include <stdbool.h>
#include <xc.h>


char int_temp; // throwaya reg for ISR
volatile uint16_t int_counts = 0; 
volatile __bit init_flag;

void putch(char data)
{
 while(!TXIF);    // wait until TXREG is available
 TXREG = data;
}
 

/********************** end UART functions **************************/

void __interrupt() ISR(void)
{
  if (PIR1bits.CMIF && PIE1bits.CMIE)
  {
    int_counts++;
    if (int_counts == 1 && init_flag == 0 ) {
        // this is the first time an edge has passed
        TMR1=0;
        TMR1ON=1; // start the system clock counter
    }
    if (int_counts == 621) { // should this be odd due to 1 start
        TMR1ON=0; // stop the counter
        int_counts = 0;
        init_flag=1;
        PIE1bits.CMIE=0; // stop interrupts (there are no others enabled))
    }
    int_temp = CMCON;  // read this, discard, so that
    PIR1bits.CMIF=0;   // this is cleared
  }
}

int main ( void ) {

    
    TRISA=0b11100111; // A2 is Vref out that the comparator uses
    TRISB=0b00000000; // all out for now
    VRCON=0b11001000 ; // enable, to pin A2, Hi range, VCC/2
    CMCON=0b11000101 ; // One comparators C2
    OPTION_REG=0b01110000 ; // tmr0 ext input prescaler off (count 256 cycles, let it overflow) RB pullups on. 
    T1CON=0b00000000 ; // set timer 1 running internal osc 1x prescaled, stopped (clear and enable TMR1ON when ready)

    
    TXSTA=0b00100100; // high speed BRGH=1
    
    RCSTA=0b10000000; // enable port async
    SPBRG=25; // from datasheet for 9.6k
    
 

    PIE1=0;     // disable all stuff here, will enable CMIE separately
    INTCON=0 ; // disallow all interrupts
    INTCONbits.PEIE=1;
    ei(); // enable all interrupts
    
    for ( ;; ) {
       TMR1ON=0; // stop this (should be already). It counts main system clock cycles
        //TMR1=0; // clear this, stopped so non-atomicity is OK
        
        // clear the low freq counter
        int_counts = 0;
        // we are all set now
        int_temp = CMCON;  // read this, discard, so that
        PIR1bits.CMIF=0;   // this is cleared
        init_flag = 0;
        PIE1bits.CMIE=1; // permit the low frequency comparator changes to run the ISR
        // now sit and wait until init_flag is set high in the ISR, result can be read;
        while ( init_flag == 0) {
             //do nothing. This may lock up if LF osc is too low or absent
        }

        printf("%u\r\n",TMR1); // TMR1 should be stopped here %u because TMR1 is unsigned

          __delay_ms(500);  // wait 1 second
	}

    return (EXIT_SUCCESS);
}

I don’t have experience of programming the Atmel directly, which is why I bottled this and went to the PIC where I do ↩︎

Arduino millis() interrupt frequency measurement blues

A cautionary tale for those trying to measure frequency using Arduino

For the paramagnetism project, I need to measure the frequency of the oscillator accurately. I did that with a 16F628 PIC last time, but I am after an easier life, and the obvious choice here is Arduino. I get serial output, a way to use I2C OLED displays rather than bit-banging a Hitachi character display 1980’s style, and all that extra wiring, what’s not to like?

The lack of consistency, that’s what. I count the number of 16MHz Arduino UNO R3 clock cycles that have elapsed with every rising edge on D2 in an interrupt service routine – the code is at the end of the post. Once I have seen 16 rising edges pass, I copy that value over, set TCNT1 to zero and start over. The first result will be garbage, after that it should be OK. Every half a second I print a copy of the last result out. This measures the period, in 1/16MHz segments. I collect 2300 results, of which the first 10 are like so

freq,cnt
4793.92, 53398
4794.01, 53399
4795.35, 53400
4794.37, 53396
4794.01, 53400
4794.01, 53400
4794.01, 53400
4794.01, 53400
4793.92, 53401
4793.83, 53402

and you can see right from the get-go that there is trouble in Paradise. These are two crystal oscillators. One is the 16MHz in the Arduino board, the other is a SEI 2.4576MHz oscillator block, which I divided by 512 using a CD4060 to get 4.8kHz

Okay, so the oscillator¹ is 24 years old. Just to eliminate this going bad I ran a scope on the 1/512 output, in persistence mode

It was OK, even on many more cycles in persistence mode.

Paramagnetism revisited

Been nearly ten years since I last looked at this. Most people’s interest in paramagnetism seemed to be soil testing with the Phil Callahan Soil Meter (PCSM), mine was investigating megalithic sites. I have Fabrice to thank for this and much useful information about the PCSM. I had unsatisfactory results in the field, lacking sensitivity. This was probably due to a bad choice of coil design. Looking at Bartington’s MS2D surface scanning probe, it’s clear I should have designed this more like a metal detector search coil.

Bartington’s MS2D surface scanning probe shows how it’s done

Which explains why I shelved this project after dismal results at Gors-Fawr I recall.

Gors Fawr measurement fail. I expected this value to be a lot higher

Sadly Philip Callahan passed in November 2017, it seems interest has dropped away without such an eloquent protagonist. For the first time i though I would look at soil, which you can enclose in a solenoid, which takes away a lot of the handwaving. I bought some rock dust from Allgood Farm on ebay and thought I’d investigate. Before we go on I should emphasise that Allgood don’t say anything about paramagnetism in their description –

Volcanic rock dust is rich in essential minerals and trace elements like magnesium, iron, and calcium that plants need to grow.
Slow Release: These minerals are released slowly over time, providing a long-term nutrient supply.

so this is about mineralisation, not paramagnetism. The last volcanic activity in Britain was 50 million years ago, so volcanic rock from the UK won’t be highly paramagnetic, because much will have eroded over 50mn years 😉

the sample was tested inside the coil, separated for clarity in the picture

I used a discrete component Franklin oscillator. I am also testing a LM358 oscillator, on the same breadboard

opamp and discrete Franklin oscillators on a solderless breadboard

It’s tasteless as hell to make these oscillators on a breadboard, but the discrete component oscillator had a tendency to squegging at these low frequencies so it helped debug that. I have this on DL4YHF’s Spectrum Lab

and you can see there’s some scintillation, whether this can be improved by building this on Veroboard remains to be seen. Using Spectrum Lab I measured the frequency f2 at 5.4886 kHz and f1 with sample at 5.4861 kHz.

Using χ = (f1²-f2²)/(f1² × 4 π) from last time¹ I infer the susceptibility is

(5488.6² – 5486.1²)/(5488.6² × 4 π) = 7.25E-05 = 72 µCGS. Probably +/- 5µCGS given the scintillation in frequency. Definitely paramagnetic, probably not stupendously so. I need to find some powdered materials to calibrate this with. And test the Franklin discrete component oscillator and the LM358 version for stability when soldered, rather than bodged on a breadboard 😉

My coil was ~1.2mH and resonated at ~5.5 kHz with a 1 µF mylar capacitor and the stray coupling capacitors of the Franklin oscillator. To be continued

Note the χ_m terminology then is wrong, it should be χ_v, the dimensionless volume susceptibility ↩

CAJOE ebay geiger counter kit

I bought this noname kit, from somewhere in China. I may want two of these in the end, to compare radiation near one site compared with the general background, so getting cost down¹ is worth having, and I can’t moan about the price, delivered for < £25 including tube. But I can moan about the total absence of documentation! You pays yer money, you takes yer choice. I looked at some of the offerings from Banggood to see if I could match it. Maybe I should have paid an extra £5 and got someone in China to assemble it for me but the board looks the same. Bangood’s board says RadiationD v1.1 (CAJOE).

It’s a tough life in Chinese electronics, even the copiers get copied. The write up is on Github – it’s a little bit obscure, the PDF overall schematic isn’t that useful but the more detailed schematic in blocks is.

Not a kit for beginners

If you are new to electronics, this kit is not for you due to the amount of guesswork and component ID required, buy it built, the difference isn’t that much. There’s nothing that special or earth-shattering in the parts however, all are easily available, and the board is a reasonably decent construction and easy to solder. There was some satisfaction in second-guessing our Chinese friends but really, how hard would it be to provide the link in a slip of paper or even on the original ad? Investigation of the original designer CAJOE with the Wayback machine at the URL indicated on the PDF didn’t deliver much enlightenment, but there’s enough in Github to give it a go. I used the block schematic with values, thankfully the screenprinting reference numbers on the board match the components in the PDF, which the exception of C5 which is MIA on the schematic, at a guess a 100nF decoupling cap.

The kit prep is a bit slapdash. The classic way to drive a mechanic bonkers is tossing some spare screws in his tin of screws taken out of the engine, and here they lob in some spare components – on populating the resistors I was left with a few over, same for the caps and a couple of transistors. Continue reading “CAJOE ebay geiger counter kit”

Running a Raspberry Pi off a LiPo battery

I wanted to take an IR photo at dawn. IR photos are easy enough these days, you can either butcher the camera you already have taking out the IR filter, or you can use a Raspberry Pi NoIR camera where they have taken the filter out/not fitted it at the factory.

The dawn part, however, is a problem. I am night owl 😉 I don’t do dawn, if I can help it, and since the target is static in my case, and the Raspberry Pi is the camera of choice, it seems a nice idea to get the Pi to do all the work of getting up early. Why keep a dog if you have to bark yourself?

There are no end of shims and gizmos that will up the 3.7 to 3.2V of a LiPo battery to 5V, for putting into the Pi. I’ve used a wide input shim to power Pi’s off a CCTV power supply rather than have one mains PSU per Pi.

In this case I favoured the piZero variant, and didn’t have aims to be connected to t’internet in the field. So I need a real-time clock, and all of a sudden a simple requirement has turned into dongle hell. This is where I wanted to be:

and at first considered a 5V LiPo plus RTC device like this only to discover it won’t start the Pi on a schedule. I then considered using a 16F628A PIC with one of the DS3231 dongles, a Chinese noname clone of this. It turns out the stock Raspberry Pi driver can support one single wakeup alarm on the DS3231 – the gory details are here. That will pull down Pin 3 on the alarm, although pin 3 isn’t brought out on the connector it’s easy enough to tack a wire onto that. Some Pis have a setting where you can pull a wire to ground to start the board; fortunately the DS3231 pin3 is open-drain so ti will work with that. The PiZero (not W, mine is a 1.3) draws about 30-40mA powered down, that’s much better than a real Pi but still a bit much for a battery. Continue reading “Running a Raspberry Pi off a LiPo battery”

piezo contact microphone preamp for plug-in-power

This is for a piezo contact mic that is powered off a field recorder’s plug-in-power supply. That is for small recorders like Olympus’s LS10, LS-5, LS14 series, which are typically unbalanced audio inputs, stereo on a 3.5mm jack, not for a P48 phantom power from a mixer or pro recorder¹. You also get an opportunity to drink beer.

Batteries are a pain in field recording. One battery to rule ’em all is best in my book, the fewer things with a battery required to make a system work the better. The recorder generally has to have a battery, and Sony agreed with that many years ago, back in the days of Minidisc, and came up with plug in power (PiP). to eliminate the battery in the microphone. An electret mic doesn’t need much power so battery life is weeks not hours, but it has a switch you often forgot to switch off.

Plug-in-power

Plug in power is about 3-5v supplied through a series 6.8kΩ resistor, with the output developed across the 6.8kΩ resistor. It so happens that electret microphones were coming in at that time, and typically had a FET to buffer the high source impedance of the capacitive electret mic, an audio source through a few tens of pF. The 6k8 resistor is the drain resistor of the electret mic. You can read more about plug in power here.

Ideally a contact mic would work with this just like an electret mic, after all the problem and the solution are the same – a high source impedance ² requiring a FET to buffer it.

there’s not much power available in plug-in-power

The trouble is that both the FET piezo amplifier and the opamp version need a 9V battery, which is a pain in the backside. An Olympus LS10 needs two AA cells and will run hours, and then the piezo contact mic wants a box with a 9V battery all to itself. That’s not good, and you tie yourself up in wiring.

The opamp version will never run on plug in power. The basic FET buffer is optimised for 9V, so it won’t run off PiP either.

The design

Initially I tried this with R2 omitted (replaced with a short) but there is not enough current to get the drain off the 0V rail. That’s fair enough – the IDSS is specified as 2-20mA and the most current you are going to get from a 3V supply through a 6k8 resistor is a little bit shy of 0.5mA. I bought a job lot of 2n3819s and experimenting with the value of R2 showed 470Ω put the drain voltage about 1.7V, with the source voltage about 0.5V, dropping about 1.2V across the transistor and 1.3 across the 6k8 drain resistor, which is good enough. Surprisingly, though they were from Ebay, nominally Fairchild parts, all five I tried biased about right with 470Ω up to 680Ω. Ordinarily you’d bypass R2 with a capacitor to maximise gain, but you are rarely short of signal level with a piezo mic, so I omitted that, it will slightly reduce gain, and also limit the current slightly if the piezo puts too much through the gate/source junction if I drop it. I have decided to take a chance with this omitting the diodes of the original FET version, because they’re hard to fit in. I expect peaks to be lower because I have mechanically damped the mic a lot. Time will tell if I get to regret that and the mic goes noisy or fails in service if I drop it once too often.

Your optimal value of R2 may be different due to the spread in FET Vgs- put a multimeter on the drain and aim for about half the open-circuit voltage³ ±10%. Mine is chosen purely for the supply through the left hand channel. If you parallel the inputs you may need a different value. I have hopes of using two to record stereo, else for mono paralleling the inputs makes for easier monitoring in the field.

Plug-in-power contact mic construction

First open a bottle of beer, trying to preserve the shape of the crown seal rather than bend it in the middle cavalierly. Obviously you can’t leave an open bottle of beer in the lab, so drink it. A 2cm piezo disc fits in a typical bottle cap, as shown

Now pick out any plastic seal in the bottle cap, and then clear the residue. I used a Dremel and the face of a small stone attachment to get this

You are going to solder to that surface, it will be the ground, and the cap provides shielding on that side, the brass disc shields the other side. This is a good time to drill a 3mm hole in the side of a cap, I used a small modelling pillar drill. If you don’t have a pillar drill you are probably going to give blood and turn the air blue with cursing. So better file a U shaped slot with a round needle file instead 😉 A hole locates the cable better in the later glue stage, but it’s not essential.

Now assemble the amplifier Manhattan style over the ground plane of the bottle cap, which also takes the shield of the cable. Use relatively thin cable for the output signal, because else handling noise will drive you bonkers. Some cable protection against the sharp metal would be wise, heatshrink in my case. Solder the piezo to the relevant place. You don’t want too much of the red and black cable going to the piezo element because you will have to lose this in a load of glue. I had too much.

Now would seem a good time to test this with your recorder, handling the piezo should be easily audible.

I wanted to keep a continuous solid interface to both sides of the piezo. Air is quite high impedance mechanically, and a contact mic detects signal in low-impedance solids. The sharp resonance of piezo discs is because they are a solid in air on both sides, and I wanted to damp this and match the disc to the solid medium. I used glue for that.This takes out a lot of the dreadful 2kHz honk in the sound, which is the natural resonance of the disc in air. In an ideal world you would put the contact mic in a solid medium with the same impedance as the ceramic either side to pick up the longitudinal vibrations, which would get rid of that resonance, but getting rid of the piezo-air interface on both sides helps a lot, even if epoxy and hot-melt glue aren’t a great match for the ceramic – they’re much better than air! The cap and glue mass-load the rear of the disc somewhat, which usually improves the sound.

In the final device, you want the disc to sit slightly proud of the bottle cap so the brass disc is in contact with the sound source, but the bottle cap isn’t. For that you can use hot melt glue all the way, though probably in two stages.

I wanted to glue a neodymium magnet to the other side of the piezo disc. Many things that are interesting with a contact mic are steel surfaces, and a magnet saves you trying to clamp or stick things to such a surface.

I discovered the hard way that hot melt glue is above the Curie point for neodymium magnets, and the loss of magnetism is permanent, turning it into an expensive metal disc. So I used epoxy resin to stick the magnet to the back of the piezo disc. I also got to resolder the red lead to the friable silvering on the back – best done quickly.

It pays to manipulate the magnet with non-ferrous tools, the best way to hold it in place is put it on the disc over a steel surface covered by masking tape. The next stage is glue this to the cold hot-melt glue encapsulated amplifier in the cap – the depth of the glue plus the magnet should set the brass disc proud of the cap. I also had to lose all the red and black wire in the gap and fill this with epoxy. I did this glue operation is two stages, because otherwise it was going to be a very messy job.

first the magnet and disk are glued to the hot melt glue surface

This stage is used to prove the brass disc is clear of the wires and the bottle cap. Only just in the case of the wires.

The second stage infills the gap between the disc and the cap, by mixing the glue on a piece of paper and using a fold in the paper to pour the epoxy into the gap.

You will get to scrape some epoxy off the disc. Aim to make nothing stand proud of the brass disc, which will then be held on steel surfaces with minimal gap.

Field testing – what does it sound like?

This is the mic attached to galvanized farm gate, the excitation is entirely the wind. I could not hear this sound standing next to the gate, it was a strong breeze, not a hurricane. I found it was very easy to overload, because you can’t really hear those low frequencies and metering isn’t fast enough for the transients. I record the main output on the LH channel, and put a series 15k resistor to the RH channel, which gives me about 10dB attenuation, but also more noise. I had to take this recording from the RH channel because I made a mess of the recording level.

Experimenting with stereo

I made another one of these, to see what stereo sounds like in a contact mic. Using a Y cable that connected the tips of two TRS jacks to tip and ring on another, I can combine these. This was on a farm gate, tapping the gate to test the sound

this was recorded on the opposite sides of the middle vertical strut of the fence below

this was recorded on adjacent struts

and this final one was recorded as shown below

You can hear a bit of the wind impacting the cables, perhaps I will have to clamp them to the gate to get rid of that effect. The speed of sound in steel at nearly 6km/s is much higher than the speed of sound in air at 330m/s o although the spacing of the third sample (pictured) would be outrageously wide for a spaced pair of microphones in air, the stereo effect is clearer for me on that than on the first two.

typically presented on three-pin XLR, or five-pin for stereo. ↩
The piezo disc doesn’t need anywhere near as high a source impedance as the electret. 1MΩ is fine for the piezo, the electret wants nearer 1GΩ ↩
That’s o/c without the FET present, typically 3-5V on a PiP input ↩

HP Stream 11 lives again with Xubuntu

I bought this must’ve been 2016. It was a bad move from the get-go, because the hard disk is only 32Gb. And it had Windows 10, and 32Gb is only just enough to get Windows on. Pretty soon I had to use an outboard hard drive to be able to update windows, and by about 2019 even that didn’t work. It’s a shame, because it’s otherwise serviceable, but totally non-upgradeable – the ‘hard drive’ is an eMMC soldered to the board. It lasted me three years. I did like the light weight and silent operation, but the overall gutless performance and slower and slower startup was bad.

I could use a linux laptop

I had been tinkering with a Raspberry Pi4 for amateur radio field use, but wrangling a Pi in the field for things like SOTA is a mess, because a Pi plus all the odds and sods you need to make it work is a collection of parts flying in loose formation, and unlike a DC3 they don’t always work well together. It’s bad enough connecting the computer to the radio via analogue audio connectors ¹, having to connect the Pi plus screen to a Bluetooth keyboard plus some sort of battery to USB-C power contraption gets a bit much in the field although it all works fine on the bench. I had already run the FLdigi and WSJT-X software on the HP Stream in Windows so I knew it was capable of decent performance, better than the Pi4 which struggles a bit to decode WSJT-X in a reasonable time.

However, I had heard bad things about trying Linux on the HP stream, because the Wifi card is very proprietary. The Ubuntu drivers seem to have fixed that now

It was surprisingly easy to load once I quit trying to install on the UEFI BIOS. I uses Xubuntu LTS 20.04, downloading the iso and putting this onto a USB stick using balenaEtcher. I found the install instructions for Xubuntu hard to find and sketchy, but they are good enough to feature this hint

If you don’t already know how to install Xubuntu, then please read this great tutorial, which applies as much to Xubuntu as to Ubuntu.

which is indeed great, and took me from there. But first I had to switch off the UEFI BIOS. It’s not that obvious to me what advantage UEFI gives me with a machine with a whopping 32Gb of disk space which is far from the 2Tb limit UEFI is supposed to fix, so legacy is fine with me.

A disadvantage of linux on a laptop, apart from the general gangly geeky oddballness of linux on the desktop as opposed to on the server is battery life is not optimised so well.

Continue reading “HP Stream 11 lives again with Xubuntu”

Chinese HW-131 breadboard power supply, clone of YwRobot MB U2

I bought this Chinese HW-131 breadboard power supply, which seems to be a Chinese clone of another Chinese product, the YwRobot MB U2. There are reports of unreliability with that device run off 12V, the suggestion is to run it of less than 12V if you are drawing notable power from it because heatsinking is marginal, using the small board. And definitely test all the output voltages before wiring this to something valuable.

It uses a AMS1117 3.3 and 5V regulator, and the minimum input voltage is 6.5V, due to the regulator dropout of 1.3V, so I will look around for something more suited to this. A breadboard tends to get shorts easily, and I could see the AMS1117 getting shirty trying to dissipate 12W into a short at 1A current limit 😉 The schematic matches this Addicore one

Absolute maximum junction temp is stated at 125C, and their example calculations say a good board will get thermal resistance of 45C/W, so if we start at 20C the chip can dissipate 2W max. I favour 12V since I can split the 12V off an existing LED lamp power supply, and sockets are a premium around the computer.

It’s hard to ignore some of the dire warnings on the Net about this, it always seems to be the 5V regulator that gives itself to the cause, failing short, which is bad in a series pass device, though typical. So before I put this into service I tested its resilience.

I could reduce the short-circuit voltage with a power resistor or a PTC current limiter. I only had a 12 ohm power resistor, so I used that. Voltage drops to 2.5V into the AMS1117-5 when I short the 5V rail, but it recovers OK, and more importantly the 5V rail recovers to 5V. However, when I replaced the 12 ohm resistor with a PTC 550mA current limiter the series pass device failed short and a wisp of the magic smoke escaped. There’s not much of it. I can see where these get their bad rap from – show me the experimenter who never shorts the power supply rail with the scope probe 😉 As Bob Widlar said in his paper on the design of the LM109 voltage regulator “Actually, the dominant failure mechanism of solid-state regulators is excessive heating of the semiconductors.” And it happens pretty damn quick.

Magic smoke output port. Single use only

So I ordered five of the AMS1117 +5V which set me back £2¹ and set about the spare board with a Dremel to break the + connection from the pin on the socket, to insert the resistor. There’s no need for the PTC 550mA device because it clearly heats up slower than the AMS1117. I shorted the 5V rail, and while you can smell the overheating resistor the +5V rail came back fine, and I diddled about with the short to see if I could catch it out. After about 20 seconds and a wisp of smoke rises from the resistor, although the picture shows it doesn’t discolour too much, and everything is still serviceable. I measure about 10V across it into a shorted 5V rail, across 12 ohms that is about 10W, pushing what is probably a ~2-5W resistor a bit. Resistors tend to fail open, not short, which is good in a series pass device. It was time to introduce this to the old National Semiconductor method of testing voltage regulator short circuit tolerance, dragging the 5V rail over a grounded rough file. As Bob Pease used to say

Don’t just see if it oscillates — see how it RINGS when you tickle it with a pulse of current. In other words, BANG ON IT.

output of 5V dragged along grounded file

Analogue shorted load test equipment

That has this sort of effect on the rail going into the AM1117 5V regulator. Since the DrDAQ I am feeding into the PicoScope only has one channel it is a different instance

12V input to AMS1117-5 with output dragged against shorted file

Modded board – you cut the +12V centre contact of the barrel jack and wire the resistor in its place

It’s not subtle, and it’s not clever, but it seems to save the board from swift destruction, at the cost of limiting power drain to 500mA. This one has taken a fair bit of abuse now. I am still tempted to desolder the great big USB socket and bridge a 5.6V 5W Zener across the 5V rail as a better use of the space.

It beats using a big lab power supply on my desk. My lab bench is in a garage, and it gets cold in winter, so for some PIC development I wanted to run a solderless breadboard on my desk. Two reasons – it’s warm in the house, and secondly the lab computer is an old Sony Vaio 32-bit machine. Microchip’s MPLAB X IDE demands 64-bit, because it is shocking bloatware built on Netbeans and Java under the hood. Until recently I’ve used MPLab 8 and assembler, but, well, they ain’t making any more time and assembler is time consuming as complexity goes up.

Tinkering with PIC programming on a breadboard tends to involve a couple of LEDs or perhaps an LCD display, so a few hundred mA is good enough for that sort of thing. If the project needs less than 30mA a decent alternative is to use the PicKit 3 programmer to power the board, but 30mA won’t power many LEDs.

Chinese electronics seems pretty hit and miss. I find buying components like power transistors there is a crapshoot, because these things are terribly easy to relabel², which is probably what did for my SkyTec power amplifier, over and above the rotten thermal design 😉 I’ve had better luck with their modules, which are often the only way to tackle some SMD devices. There seems to be a mix of pick and place and hand soldering – looking at the reverse of the board

It’s a tad on the scruffy side. The jumper wires I got with the kit are OK, I normally use stripped Ethernet cable for jumpers so more colours are welcome. The breadboard is disgusting. The grip on wires is weaker brand new than my 20 year old main lab unit. Life is too short to muck around with janky solderless breadboards, so I will throw it way to save my future self hours of debugging it.

Update – I replaced the faulty 5V regulator and fitted the resistor mod, and that board now survives the 5V rail being shorted. To change the 5V reg snip the three SMD legs of the old one, which lets you lever up the chip and desolder the tab, and then clear the snipped residue. It wasn’t too painful, but I am glad I found this failure mode out before losing something valuable. ↩
I couldn’t find an ISO9001 supplier of the 2SA1941 and 2SC5198 transistors. Digikey carry the latter, on a quote required basis, so you probably have to buy 10,000. So I bought my transistors from ebay, and they probably got them from a Chinese relabeller. I should have put the effort in to identify a suitable substitute that I could get from a reliable supplier ↩