Sunday, 5 July 2020

Distributed computing.

Since May 1999, I contributed to the Seti@home project. It's subsequently gone into hibernation, and given the current COVID-19 pandemic, I though I'd switch to putting my unused processor cycles to a valuable project...

This project uses your spare CPU or GPU processing to "fold" proteins, and look for possible solutions to various diseases, and, in this case, COVID-19.

It works on Windows, Mac and various Linux distributions.

Go and give it a download, and use some of that spare computer power whilst looking at Doz' Blog, Facebook etc!

Join the Doz' Blog team if you'd like...

The team number is 266687.

See you there!

Saturday, 27 June 2020

Yamaha KX-580 fail...

It's the last of the famous George's kit to look at ...

It's a mint condition KX-580 with a nasty noise on the left hand channel.

When I say mint, I mean mint... it looks like it's never seen a cassette in it's life.

Sadly I rapidly trace the fault ... it looks like an open circuit head ...

... and sure enough it is ...

Game over. It's a special low inductance GX-37 head , and utterly unobtainium. I hate being beaten, but sometimes that's just the way it is....

Copland CSA-14 repairs

Jon what'sapped me ....

"Got a Copland here, previous owner accidentally dropped a bare wire across the speaker outputs... Care to take a look?"

He also sent a picture ...

Ah ... the magic smoke has escaped..

So, the thing is duly opened, and examined. 

It's quite obvious the unit has been modified by a previous owner. Caps replaced with "audiophool" grade parts. The speaker protection relay bridged out, because, you know, sound quality. Led hanging on the end of two bits of plumbers solder...

The quality of workmanship left a lot to be desired. Where some caps had been replaced, the leads hadn't been cropped off fully, and just bent under the board. The caps are of the wrong pitch, and don't sit snugly against the board... and one was fitted back to front, and had vented it's anger...

Cap in the wrong way!

... I mean ... what ?!?!
oh, and the fuse fairy has been in and liberated the fuse for the -42V rail to the right channel...

On the plus side, there's a pair of possibly the finest E88CC's ever made lurking in there, the Mullard CV2492.
OK, out with the output stage, which is unsurprisingly short circuit... both NPN and PNP transistors. The bias thermal compensation transistor is also OK, unlike the BD139 and BD140 in the speaker protection circuit, which I expect caught the brunt as they attempted to open the bypassed relay!
So we need a 2SA1302 and a 2SC3281... obsolete in 2000. But a reputable supplier has it listed, excellent. The two transistors are duly ordered along with a few replacement caps.

They duly turn up, and are fitted... but I can't set the bias correctly. The 2SA1302 is drawing far too much current. I check and check and check everything. Finally I swap the transistors out of the left channel, and everything is fine. I then google 2SA1302 ... oh dear. The world is full of fakes. My reputable supplier is contacted, and after an investigation, they remove the fakes from sale, and issue me with a refund. Reputation upheld!

Consulting the datasheet, I reckon an On-Semi 2SA1943OTU and 2SC5200OTU will fit the bill nicely. They're ordered from an official On-Semiconductor dealer. I'm not risking this again...

New transistors fit, bias correctly, and there's no difference I can measure (or hear) between the left and right channels. Why should there be? It's not a valve!

It's not a bad thing to work on, but for some reason, known only to Copland, the PCB is single sided, but has through-plated holes, just to make removal of a faulty part that little bit more tricky. The Metcal SP440 de-soldering tool worked a treat though.

It's given a LONG soak test, and is pronounced fit. 
Jon, it's done.... "Yeah, that Goldmund is playing up on one channel" ... this fills me with dread ... more later!

Later - other than a crap DC offset pot, the Goldmund was nothing serious, except I'd nicked a trace on the fragile front panel PCB, disabling an output LED, which meant a few hours work stripping it all down again! 

Saturday, 20 June 2020

NOAA Image site rescued.

It appears I didn't read the small print for the web hosts that up until yesterday were hosting the NOAA pictures website. It disappeared overnight.. I thought perhaps we'd gone over our bandwidth limit, and they wanted me to subscribe and send them some money, which is fair enough...

... but no, apparently they do not allow automated uploading. The politely shut my account and wished me luck. Fair enough.

Anyway, I put out a plea for some webspace on my facebook page, and had plenty of offers. Thanks to every one of you.

Matt (a regular reader of this rubbish I write, who is also building a NOAA receiver after seeing some of the images I've received) came to the rescue, as he has his own site which he hosts...

So he's organised a new bit of the web for me... ladies and gentlemen I present ...

Even the URL is better!

It's all up and running again after a bit of fiddling about with settings at Matt's end, and him spending hours fault finding with me. Expect Malvern to come on line in the next couple of days. Evesham is currently off-line as I'm not there at the moment, and Gloucester isn't quite ready for the off just yet...

Anyway, here's to Matt. Have a virtual Doz' Blog Medal for Outstanding Service to the cause.

Monday, 8 June 2020

Raspberry Pi Zero NOAA finishing touches.

I'm chuffed with the Pi Zero receiver in the last post, and I'm going to install it in my workshop in Evesham, to take over the job done by the laptop and my FT-817.

I needed to pretty it up a bit.

I found a suitable small case to house the Pi, the SDR and an old LNA (previously failed, but a new amp chip and all is well) 

I got to thinking.... 

Nice if we had some status LEDs on the front panel.

Power (obv.), READY (we've got a pass schedule, so ready to go when a bird pops up), AOS (Acquisition of Satellite) and PROC (processing images).

A front panel is created and printed on a piece of clear acetate, which is then stuck to the front panel with lacquer. It's OK if you don't look too closely...

To drive the LEDs, there are two bash scripts to alter ... and

both are found in ~weather/predict/ needs the following lines at the end..

#Switch READY led ON
sudo echo "24" > /sys/class/gpio/export
sudo echo "out" > /sys/class/gpio/gpio24/direction
sudo echo "1" > /sys/class/gpio/gpio24/value now looks like this..


# $1 = Satellite Name

# $2 = Frequency
# $3 = FileName base
# $4 = TLE File
# $5 = EPOC start time
# $6 = Time to capture

# reads and creates folder with current date / time (i.e 05-30-2019_07-48 *windows friendly*)

NOW=$(date +%m-%d-%Y_%H-%M)
mkdir /home/pi/weather/pics/Folder${NOW}
#switch on AOS led, switch off READY LED
sudo echo "22" > /sys/class/gpio/export
sudo echo "out" > /sys/class/gpio/gpio22/direction
sudo echo "1" > /sys/class/gpio/gpio22/value
sudo echo "0" > /sys/class/gpio/gpio24/value
sudo timeout $6 rtl_fm -f ${2}M -s 60k -g 45 -p 55 -E wav -E deemp -F 9 - | sox -t wav - $3.wav rate 11025
PassStart=`expr $5 + 90`

if [ -e $3.wav ]

#switch off AOS led, switch on PROC led
sudo echo "0" > /sys/class/gpio/gpio22/value 
sudo echo "23" > /sys/class/gpio/export
sudo echo "out" > /sys/class/gpio/gpio23/direction
sudo echo "1" > /sys/class/gpio/gpio23/value
/usr/local/bin/wxmap -T "${1}" -H $4 -p 0 -l 0 -o $PassStart ${3}-map.png
/usr/local/bin/wxtoimg -m ${3}-map.png -e ZA $3.wav ${3}.png
/usr/local/bin/wxtoimg -m ${3}-map.png -e CLASS $3.wav ${3}.CLASS.png
/usr/local/bin/wxtoimg -m ${3}-map.png -e MCIR $3.wav ${3}.MCIR.png
/usr/local/bin/wxtoimg -m ${3}-map.png -e MSA $3.wav ${3}.MSA.png
/usr/local/bin/wxtoimg -m ${3}-map.png -e MSA-precip $3.wav ${3}.MSAPRECIP.png
/usr/local/bin/wxtoimg -m ${3}-map.png -e MCIR-PRECIP $3.wav ${3}.MCIRPRECIP.png


# PROC LED off

sudo echo "0" > /sys/class/gpio/gpio23/value

# copies files to the new folder and deletes the original ones

cp /home/pi/weather/*.png /home/pi/weather/pics/Folder${NOW}/
rm /home/pi/weather/pics/Folder${NOW}/*-map.png
rm /home/pi/weather/*.png
rm /home/pi/weather/*.wav
wput -B -u --dont-continue --reupload --tries=5 --binary --verbose --reupload ftp://mywebspaceusername:mypassword@myurl ~/weather/pics/Folder*/*.*
# READY led on
sudo echo "1" > /sys/class/gpio/gpio24/value

You can see from the script that the GPIO pins 22,23 and 24 are used for the LEDs (The POWER LED is driven from the supply). Because only one LED is on at anytime, we can cheat and use only one 220 ohm current limiting resistor to a ground pin, and drive the LEDs directly from the GPIOs, a huge saving of ~3p. Every little helps...

GPIO pin 22 is the AOS LED drive, 23 is PROC, and 24 is READY.

The finished receiver...

In situ, waiting for a pass...


Saturday, 6 June 2020

More NOAA APT shenanigans , the £25 NOAA receiver

"Shenanigans" ... that probably doesn't translate well... Wiki says "silly or high-spirited behaviour; mischief."
... anyway ...

There's a bit of talk on the noaa and Raspberry Pi forums, that the new Pi Zero (W) hasn't got enough poke to run a NOAA receiver... so we shall see....

£9.80 is duly shelled out on a Pi Zero W from the lovely people at PiHut.

I thought I'd just clone the existing memory card ... no such luck, the drivers for the Zero's wifi aren't included in Stretch. 

So a new install of Raspberry OS lite is created, and set up. You can follow the instructions here.

You'll need to install wput and lsftp if you want to upload to your webspace.

sudo apt-get wput
sudo apt-get lsftp
You also need to create the pic directory

mkdir ~/weather/pics

Things didn't exactly work out....

That's no good at all...

Now, there are issues with predict on the newer OS. 

When you run predict, and enter your station "callsign", lat, long and height, quitting out leaves the screen in an odd condition, logout, and log back in again.

I had to run predict a number of times before I could get it to swallow all the data... keep running it until it finally shows all fields correctly filled in.

In the previous picture, the receiver had thought it was at 0 degrees long, and 0 degrees lat ... just off the coast of africa! 

After finally getting predict sorted ....  superb pictures are received.

So ... a £14 SDR receiver, and a £9.80 Pi Zero! well, nearly... You'll need a micro SD card and an antenna, but we've saved some money on the computer!


Now my good friend Ben is keen to implement a receiver in Gloucester, and is going this route... once he's up and running, I'll add another page on the NOAA extension to this website, so you can enjoy his pictures as well!

Thursday, 4 June 2020

Goldmund Mimesis 3 Amplifier repairs.


Jon messaged me.

"How's my Goldmund amp looking?"

erm ...

"What Goldmund amp?"

"The heavy 1U thing.."

"Nope, I've not got it Jon ... "

anyway, after a bit of head-scratching, it turns out Colin's got it... it duly arrives..

It's dead and has a noisy left channel.

Modern looking thing, very neat in it's 1U case. Razor sharp heatsinks ...

Smart. It doesn't want to power up...

Off with the top cover, and ... well, I'm shocked...

That's a lot of caps.... in a small space, and it's going to be hot in there...

Then I'm filled with nostalgia .... where have I seen these before ... ?

... oh yes ...

The old favorite Maplin 75W kit... a right of passage for any budding tech in the late 70's & 80's. A little bit of a challenge to get the thing to amplify rather than burst into oscillation, they weren't particularly stable, and required careful power supply considerations... but once running seemed bomb-proof.

Anyway, this appears to have had plenty of consideration to the power supply!

Under that black box in the middle there are 4 large toroidial transformers. A neat trick to get it all to fit in a 1U case.

Holding down both power buttons it's supposed to power up after a delay of a couple of mins... nah. Relays are clicking about, but nothing much doing.

 This is the soft start PCB, That 2.7K looks a bit cooked, but it reads fine. It's in close proximity to those caps...

After some disassembly, the caps are evicted and changed.  There are 2 x 4.7uF 100V and 2 x 22uF 63V

... and now the unit powers on fine :)

Running the unit at a humble 20 watts into a 8 ohm load, and those heatsinks are getting warm. The whole thing's rated to 200W, so 100W per channel, obviously not a constant tone though.

I have a word with Jon, and he wants all the caps changed to ensure longevity. That's a lot of work...

The parts are duly ordered and arrive next day.

First off 4x 15,000uF 63V...

Which, well, gives me a problem....

To remove the cap, we need to remove the PCB. To remove the PCB we need to remove the screw. To remove the screw we need to remove the cap.

It's schrodinger's capacitor...

It turns out it's not quite that bad, but still damned awkward.

There's some recessed screws which allows the front panel to come away...

But that still leaves the mounting bars covering a few of the 5 pins we need to desolder...

... right next to that LED too....

The trick is to desolder as much as you can from the back, and then get a long thin bit down between the cap and the board, where pins 1 and 5 are soldered to the print of that side, and work the cap out that way. It's not a nice job...

After several hours, and many words my mother wouldn't approve of, they're replaced...

There's a few smaller caps on the board to change too...

This on the left hand side of the power supply/front panel board, It's just tacked across a couple of tracks...

This cap, has had it's top damaged where it doesn't clear the Right hand amplifier's PCB...

... this cap has suffered the same fate, and it's rubbed it's value off... It's 4.7uF at 63V, and decouples the 24V rail...

To obviate the issue, I've laid the replacements on their side..

Next to re-cap the amplifiers themselves...  Access is at least easier this time... The thermal compound has dried out somewhat, and is cleaned off, ready for replacement once the amplifier is finished.

The print is not easy to work wth, it's double-sided, and has almost no thermal relief, so desoldering the caps where they meet the fat power rails is a real challenge, and clearing out the through holes a real pain.... it takes hours.

Halfway there ...

There are 4 caps on this back panel, which is also a right pain to get out, as the IEC socket must be carefully desoldered. 

Nothing is easy in here. The print is fairly fragile, and has no thermal relief

Finally it's all back together. Powered up, and after a few minutes to stabilise, the bias is checked. It needs no adjustment. I leave it on test for about an hour.... I'm in the lounge next door to the workshop... *Crack* and it goes quiet in the workshop ... 

Damn thing has blown 3 of the 4 supply fuses. Looking at the fuses, someone's been here before. They're not as spec'd in the manual. Two are T6.3A, and two are F8.8A (never seen 8.8A 20mm fuses in the wild before!) They are all supposed to be F6.3A. 

Cursory checks of the amp board shows our left channel to have a heavy load on it... sure enough one of the 2SJ49 FETs is now a short circuit. They're unobtainium (although equivalents are available, but this would require all the output FETs to be changed....) ... but guess who has a NOS spare !

It's fitted, the fuses replaced and powered up again. All is well. Bias is rock steady. I bet that's the cause of our noisy left channel too...

It's given a 12 hours soak test before it's pronounced fit.

All in all there's a lot of work gone into it... getting on for 50 hours... so hopefully it'll be good for another 30 years :)

The guilty parties.

Sunday, 17 May 2020

Video rack repairs and new video multiplexer upgrade!

5 years ago, I built a rack to house the standards converter and video source equipment in the workshop (you can see that post here)... it's time for a few repairs and a refresh.

It's got a few niggling faults.

1. The DVD player no longer ejects, and sometimes doesn't want to play.
2. There's a couple of channels which no longer work
3. There's quite a lot of crosstalk between channels.

Let's address a few of these...

1) The DVD ...

First off some disassembly of the rack case, which allows the DVD player to be removed.

Powering the player up, and there's nothing on it's output... until after a few minutes the unit powers up, and reports "Bad disc" on it's on-screen display...

A few screws later, and the lid's off, the tray front removed and the trapped disc released.

A known good DVD is inserted. At this point it's obvious why the thing won't eject. I suspected a failed tray belt, but that wasn't it... the tray is catching on the front escutcheon. I'll deal with that later. The disc attempts to spin up, fails miserably to get to speed, and then "Bad disc" appears once more.. On the second attempt the player seems to re-start it's firmware...

I check the outputs from the power supply with my multimeter... +5, -12 and +12.. +12 is varying between about 10.5 to 12.5, which may be acceptable. I switch to the 'scope and check the +5V again....

Woah! That's awful - 40V pk-pk of 50Hz crap riding on the output !

The power supply is removed, and a blanket change of all the electrolytic capacitors is performed, to ensure reliability.

The power supply is replaced, and on powering up, the disc spins up rapidly, and then reports "No Disc" ... hmmm ...

The laser pick up is given a gentle clean with some window cleaner (Don't use IPA on any CD or DVD, it damages the coating on the lens), and ....

Bingo, the test pattern DVD fires right up!

The guilty parties....

The sticking front tray is resolved by relieving the front a touch with a file... I suspect it's got distorted by the shelf above it in the rack for all those years...

Good, one fault fixed ... on with the next...

2 & 3 - Channels not working and crosstalk.

This is being caused by the mechanical switches failing on the front panel, there was always a bit of cross-talk (one channel seeming to "float" behind the video of another) as the switch does not provide any correct loading, or buffering between channels... 

What's needed is a proper video (and, of course audio) multiplexer.

I designed this a while ago in Eagle... thankfully Kicad reads eagle files :) KiCad, the gift that keeps giving :)

There are 8 audio and video inputs. 

Each video input is loaded by a 75 ohm resistor, and fed to two MAX4312ESE multiplexing ICs. One outputs the video to a UHF modulator, and one outputs the video to a 625/405 line standards converter. So, under the control of my favourite flavour of microcontroller, the ATMEGA328P, any of the video sources can be sent to either 625 or 405 sources.

The select line of the video multiplexer also feeds two ADG1408Y multiplexers, which switch the audio through to the relevant output device in the same manner. There's a 5532 buffer on the audio output, as there's none in the audio mux, unlike the video mux.

The audio mux requires a negative supply rail, and this is supplied from the microcontroller. A continuous squarewave is output on pin 5 of the micro, is floated by C9, and rectified by the double diode D1 (BAT54S) to provide a "-5V" supply, in reality it's more like -4V. 

The micro is controlled by a front panel , which has two push buttons, and a small OLED display.

A PCB was created... not modelled as it was Eagle, but Kicad has a go at it!

When Kicad imports an Eagle creation, it doesn't seem to cope with the ground plane or component models, but you get the idea... 

... anyway, the real thing turns up from PCBWay.

... and duly stuffed. I'll clean that flux up later ;)
The front panel was also designed. Simple.
and again, built up... not quite so simple though, as my new display has the power and ground pins reversed from the one I modelled :(

... so a minor bodge is deployed...

The software is very straight-forward. It acts conditionally on the actions of the select switches on the front panel to increment the channel address selected, and update the display. I used the U8glib library this time out, as it's got some nice fonts, and is quick, if a bit bulky.

The software can be downloaded from my git.

It will be noted that the hardware supports 8 inputs, and two outputs, but the software doesn't support that many, as I don't need them all! It's a simple task to enable more, if you wish.

Nice font...

So, back to the rack.

Out with the old...

Tidy up the RF wiring a bit. The black box in the bottom is the Aurora 405 line standards converter. The loop of co-ax you can see is a stub filter, as the output from the aurora is very harmonically rich, and the filter eliminates aome of the patterning caused by the harmonics effecting the UHF. What I really need is a low pass filter .... one day.

The UHF modulator (modified to run on DC, see here) is once again disassembled, so I can adjust the channel and sound carrier spacing via two buttons on the front panel. I've just soldered some wires on the existing switches, and made them available to the front panel.

The Cyclone media player, and HDMI converter are mounted on the shelf, and the multiplexer PCB behind that.

The power supply is a salvage 12V 5A supply. This feeds the multiplexer, and the aurora. It also feeds the modulator via a 42 ohm resistor (to give about 8V) , and an eBay 5V buck regulator, which feeds the Cyclone and HDMI converter.

The front panel is marked out...

... and the holes chewed out with my teeth. I do wish my metalwork was better, anyway, it's only for my own consumption!

And it's all wired up ... what a rat's nest!

The back's fitted, and it's all powered up and tested - No cross talk, all the inputs work - oh bliss!

Now, there's a rake of telly's here need fixing ....