Profile cover photo
Profile photo
Michael Kukat

Component failures in electronics

I just had the idea to put together a list of component failures i observed during the last decades of repairing lots of different electronic gear. The list might not be complete yet, but i tried to put together what comes to my mind at the moment. The age of the components also is interesting, especially with capacitors, the lowest amount of problems with them is with 1970s gear, while all 1990s gear, when they started with those SMD aluminium electrolytic caps, has leaking capacitor problems.

Most observations described here are valid for normal operating conditions. Bad design, especially thermal problems, increases failure chances significantly. Especially for capacitors next to heatsinks, but also semiconductors like linear regulators or transistors.

Those are my personal observations and might not exactly match the observations of others. I didn't add any problems i just heard about, no urban legends here, just my own experience.

Tubes (mostly up to 1960s)
- Normal aging, fail slowly, degrading current flow, rarely current leakage a while after power-up or even shorts
- Applies to video pickup tubes (until early 1980s) and CRT display tubes (until early 2000s) also
- Rare vacuum loss (detectable by getter mirror getting white)

- Rarely fail
- In tube radios, resistance might increase with age (up to 1960s)
- Found open resistors from time to time, even ones with close to zero power dissipation (even with 2000s gear)

- Tube gear - oil/paper capacitors (up to 1960s), there are many lists of bad types out there
- Pre-1980s electrolytic caps rarely fail, but might need reconditioning if not powered up for many years, can overheat and explode otherwise
- Supression caps with plastic housing cracking open and exploding (1960s to 1990s)
- Ceramic caps shorting (rarely, 1970s)
- Tantalum caps shorting (especially 1970s-1980s), often exploding or frying protection resistors (which were there for a reason back then)
- Some types of electrolytic caps leaking (1980s)
- SMD aluminum electrolytic caps nearly always leaking (1990s)
- Electrolytic caps in switchmode power supplies drying out, (late 1990s until today)
- Cheap electrolytic caps losing capacity, sometimes dead already after 3 years (today)
- Metal film capacitors losing capacity, especially in capacitor power supplies for RC power outlets and similar stuff (today)
- Metal film capacitors shorting (rarely, 1970s)

- Selenium rectifiers, aging, detectable by smell (1960s and earlier)
- Silicon rectifiers, partial failure (diodes open) - especially 1980s
- Linear voltage regulators, rarely fail, but can short and lead to overvoltage, 1980s-1990s
- Power transistors in switchmode regulators - fatal in step-down converters (overvoltage)
- Small signal transistors, germanium and silicon (1960s until today) rarely fail
- 40xx CMOS logic, especially analog switches like 4066, 405x (1980s)
- Cheap modern semiconductors more prone to failure, possibly due to counterfeits
- LEDs (today, no issue before 2000s)

- Leaking NiCd/NiMH batteries, destroying contacts and/or PCBs, even traveling through cables, lithium non-rechargable batteries usually no problem
- Early lithium rechargable batteries often dead after several years, lithium polymer cells seem to be more durable, but can swell and develop other problems (2000s until today)
- In general the charging electronics and correct handling of LiPo batteries is essential for their long life, not all cheap devices do this right, if such a cell is discharged below 2.5-3V, it can no longer be safely used.

- Relays, contacts sticking together (RC outlets) (2000s until today, rarely with older ones)
- Switches, potentiometers, sliders - dust, dirt and spring pressure degrading can lead to contact problems (all ages)

- Drive belts wearing out or dissolving
- BGA solder joints falling apart (baking PCBs is a modern trend)
- Highly integrated chips delaminating (Nvidia knows this very well...)
- THT solder joints in thermally stressed areas cracking

Post has attachment
Panasonic NV-180EG teardown

The cute portable VHS recorder that came with my Olympus VX-301 needed some more work, because during a first test, something blocked the mechanics.

While having it open for a first inspection before powering it up for the first time, i noticed a small plastic part that fell out of it. After searching for a while, i found where it belongs - the cover of the machine has 4 tabs to lock into the base, and 3 of them are glued to this cover instead of being a fixed part of it, for whatever reason. I just found one of them, so i assumed the others are lost.

But now, i need to inspect the mechanics. After removing the cover and tried some pushing while ejecting the cassette, everything suddenly worked like nothing ever happened. Okay. Time to take it apart. With the help of the service manual and some educated guessing (the SM isn't really accurate describing the disassembly procedure), i managed to dissect it into many PCBs and screws. And some mechanics. I found another enclosure tab, but the third one is still lost.

During reassembly, i decided to take some photos of the process of putting it back together.

I'm really impressed about the maintainability of this machine. You need some care to not break something or forget a connector, but besides this, it's very easy to take it apart and put it back together. What a fine vintage device. While it looks like it's hard to reach for example the second board behind the front panel to measure something while operating it, i don't think it's impossible. Sure, it needs some care, but the front panel itself is not required for operation, the next board can be flipped down while still being connected, so the second board is reachable. You might not want to really come into this situation, but seeing how compact this machine is built, it's not that bad.
The color-coded connectors also make it easy to get everything together without writing down where everything belongs during disassembly.

And according to a quick test with a VHS cassette, it also works fine - after more than 3 decades, it's built around 1985. So no need to repair more here.

The camera has some white balancing problems, but i'm still trying to find a service manual, it has way too many trimpots in it for blind experiments. Besides this, it also works fine (i had to replace a leaking capacitor in the EVF), but if i'm bored one day, i might also take it apart and make some detail photos of it. I'm especially interested in finding this green/cyan/white stripe filter in front of the vidicon. And learning how this works on the electronics side, i only read about some theory, which sounds plausible, but i need to understand the details. Not that it's something i'll ever need in my life, just curiosity :)

In the last photo, i put my current full HD Panasonic camcorder next to the others - just for a size comparison. Same purpose as the large luggable stuff from back then, but fits in your pocket. Not talking about what's in every smartphone today, besides the optical zoom, even this stuff does the same as the good old 1980s luggables.
15 Photos - View album

Post has attachment
Roland Juno-60 restoration

Yawn. Another synth repair. Okay, i didn't make photos right from the start because i didn't want to bore you with the same routine maintenance over and over again :)

But on the other side. Vintage synth guts!

A friend asked me to replace the battery because it doesn't store sounds any longer. And while i'm on it, i could check the PSU also. So it really started as routine maintenance. But then there was this massive dent in the rear of the metal panel. And the volume knob was not original, very ugly. Okay, a bit more work.

After opening it, i decided to give it a full restoration because it was in pretty bad shape. Dirt everywhere, the foam rubber parts around the switches and sliders being brittle from all this dirt and age already, it just deserved some more work. While i'm not such a big fan of those single DCO Roland synths (i have a MKS 50, don't need more of them), they still sound nice. For sure the chorus takes a big part in that. And i don't know why, but i really like working on the Juno-6/60. They have some special kind of build quality that makes working on them enjoyable. Too many connectors and cables, like some Korg synths, but i've never seen a Roland get into trouble due to this - on the other side i've never seen a Poly-61 not being in trouble with the connectors. The leaked battery for sure makes the difference. Roland started using lithium batteries very early, and they nearly never leak. Many others used NiCd rechargable batteries. At the first glance, that's a great idea, because if you can recharge them, you never need to replace them, while a lithium battery will be drained some day and you need to open the synth and in most cases solder in a new battery. That's the theory. In practive, those lithium batteries deliver enough power for 30 years, while after 30 years a NiCd battery leaked in nearly 100% of the cases and destroyed the PCB it was soldered to. But in engineering, there are decisions that look bright in the first place but prove to be not so great decades later.

But okay. Let's get back to this one. First, i completely took it apart to get the cover free from anything because this needs some less delicate handling with the hammer. I'm not a specialist for this, but the result is good enough for my taste. You didn't see it before, the spot with the dents was at least 3cm bent inwards, i have no idea why the wooden base isn't completely destroyed. I wonder how this happened. Okay, With some hammer action, it's good enough again, there is not much room for more work without damaging the top side of the panel from the inside. So better keep the visible parts intact and accept some imperfections in a spot that's barely visible during normal use.

For the battery, i replaced the CR-1/3N by a CR2032 holder, i needed a good amount of hot glue here because it's soldered to the copper side of this single-sided board, it wouldn't make sense on the other side. This is a bit dangerous for the PCB traces if someone in the future isn't careful enough when replacing the battery.
And i decided to convert this to CR2032 because those batteries usually have capacities of more than 200mAh while the CR-1/3Ns i found are just 170mAh. So you don't lose anything, but win easier replacements in the future. Like in 20-30 years, when it's time again :)

The PSU got new rectifiers (i always do this in Roland PSUs meanwhile because i diagnosed half-dead rectifiers in around 50% of the Roland gear i already had in my fingers), new capacitors and a new +5V regulator. The 10mm snap-in capacitors are a bit hard to obtain, i didn't find a dealer having both the types i need, so i decided to use normal 7.5mm Panasonic caps as replacement. Some careful bending of the pins and some hot glue, problem solved.

And the rest was just boring cleaning of everything. The ultrasonic cleaner didn't do it's job right, so i had to improve the result with a brush afterwards, the rest was just toothbrush versus PCBs. And some surface cleaning while all components are removed from the front panel.

The only interesting thing here were the foam rubber parts Roland always added around the sliders and switches. Maybe as dust seal, maybe for cosmetic reasons because it looks nicer, anyway - they were no longer usable. I tried some materials and from conventional felt, modeling felt and doam rubber, i decided to use foam rummer again, 2.0mm thick. It might not live forever, but it somehow felt right to use what they originally used there. But i'm pretty sure i've already seen Rolands with those parts being felt. Anyway - instead of using glue to attach them to the sliders (that was evil to remove, really.), i used double-sided adhesive tape. Much cleaner. And the result looks really nice.

Sure, i cleaned and lubricated all sliders, potentiometers and switches before.

And with all those wires, it was a great feeling that it didn't immediately catch fire after putting it back together but worked on the second try. I really took care, but 2 connectors were swapped, so it didn't work on the first try. Anyway, nothing happened, everything worked after correcting the connections, added the remaining tons of cable ties.
And yes, i checked the PSU before connecting the rest to it. Ah, by the way - i converted the PSU to 240V also, you should do this today in 230V times, devices get more into overheating when set to 220V than into any trouble when being set to 240V.

With the parts i ordered, i also ordered a small selection of suitable knobs after some pre-selection, because the one installed (in the close-up of the knobs, it was the upper left one) really looked misplaced there. I have chosen the one i liked best and the owner of the synth also liks this one best.

Nearly done. The last thing was a full adjustment of everything, there might be a bug in the adjustment guide in the VCF LFO gain area, so i left this value where it was - most of the adjustment was still okay anyway. After loading the factory sounds and wiping off the whole synth with a soft cloth, it's done.

I think i have it some days around to play a bit with it.
21 Photos - View album

Post has attachment
CEM5508 / PD508 replacement - mission accomplished

Finally - the project completed. While i posted the steps inbetween privately, this is the public post as a follow-up to to tell you how this worked out at the end.

With the theory described in the first post, i implemented the control logic using a small STM32 development board, optimized the firmware and checked if everything works as expected. The timing is not as tight as with the original CEM5508, but on one side, it's more than good enough for the microWAVE, and seen from an external point of view, it even meets the signal timing from the datasheet - just the sampling period moved some nanoseconds internally.

Next was completing the prototype to replace the half-dead CEM5508 in the microWAVE and check if there are any sonic oddities. I tested many of the sounds in the unit and voices 1+2 with the substitute sounded identical to the remaining voices still using the original CEM5508. So this is the way to go.

The circuit was not that difficult, it's straightforward with the goal to use only cheap standard components and not special stuff like the SMP08, which costs 10 times as much as the classic 4051/TL074 combination with C0G capacitors (PP would even better, but you don't get them in SMT) i use for the S&H stages. For the switchable impedance, i didn't even look for something simpler, even if there was an 8-channel analog switch with addressable latch out there (i'm sure Maxim has one, they have everything :), it would add another expensive special chip that might be unobtainable in the future. So i use 2x4066, HCT259 and 8 resistors for this. The control logic is implemented using a STM32 because it's fast enough, cheaper and easier to handle than a CPLD.

So the PCB layout was the biggest effort, as usual. With those holes for the trimpots and capacitors, it was a bit complicated and took 4 or 5 days to complete. I might not have needed the holes for the capacitors, but they are close enough to the board that i would need to keep away components from those areas anyway, so it doesn't hurt, i can use the capacitors to fix the board in place with hot glue later. Sure, rising the board by another 2mm with another socket inbetween would also work, but that would decrease the stability of the whole thing even more.

The first PCB run failed, i missed my own documentation and botched the exposure time, but it was good to re-check if everything fits as intended, i optimized the layout a little bit after i found some critical spots.

Yesterday, i made the final PCB. The process from film to etched PCB took just half an hour, as usual, i gave it a tin bath for an hour then and started soldering. Around 4 hours of soldering (plus some breaks inbetweek), everything was put together and electrically checked, no shorts, i just had one broken trace, so the overall quality of this PCB is very good, the process meanwhile is close to perfect. To avoid that this stuff rots away with the years (or maybe decades?), i coated it with a plastic spray, which also gives it a very nice finish.

Today was the big day - i adapted the firmware to the final layout, which is necessary because i assigned channel numbers and GPIO lines in a way that simplifies the PCB layout, so the firmware needs to compensate this, which is not a performance penalty, everything is basen on LUTs. Performed a last dry test to not damage the microWAVE if there is a bug, and flashed the firmware. Also no big deal, no further problems. 26mA drawn from the +5V digital power supply, the analog supply is also not too heavily loaded with all those components.

Then that great moment - board installed in the microWAVE, the cable for power and reset soldered to the CPU, finger on the power switch, adrenaline level goes up to extremely unhealthy levels.

Flip the switch - nothing.

Okay, now let's connect the mains plug to a power outlet.

Next try - the usual popping sound on power up, no strange sounds. Hit some keys - IT WORKS!. I can't describe this feeling when you work on something like this for weeks, power it up for the first time and it just works. This mixture of excitement and panic... I still suffer from adrenaline overdose, it's so awesome, even if it's just a logical consequence of the work performed before :)

So this project is finally completed after several months i have the machine around already. There is another big project in the background and this challenge blocked it all the time, so i'm happy this finally worked out as expected and i can give the friend back his microWAVE, maybe the only one on the planet that will never again have problems with dying CEM5508s.

I recorded a bunch of sounds with the CEMs before, the half-dead one always worked fine when cold, so this is a valid reference, i can now compare those sounds with the voltage control prosthesis in place, will play with some parameters to check if the switchable impedance stuff works for envelopes as expected, but for now, everything sounds as it should. I also need to stress test the whole thing, closing the enclosure, bury the unit under some blankets to make sure it has the required long-time stability.

This design would also work with the microWAVE Rev. 2 analog board, but it would need a new PCB layout, which is most of the work. In theory, it would also work for the WAVE, but the space limitations in that machine are a real problem for this large amount of stuff you need for the substitute. But at least there is a way to save those microWAVEs, which get ripped apart to get the CEM5508s to repair WAVEs. And the solution for around 20€ component cost (without PCB, assembly and development costs) replaces 4 CEM5508s, which sometimes appear for 50-60€ each in the web. And they are selling used or NOS chips, that will very likely fail in the near future, so the substitute is the only long-term solution for this problem.

Let's hope the other CEMs and the ASIC in the machine a more durable...
10 Photos - View album

Post has attachment
CEM5508 / PD508 replacement

From time to time, when you fix old gear, you run into a common problem - you need spare parts that are no longer obtainable. That's a well-known fact among the collectors and users of vintage synthesizers, but you have the same problem in other areas of electronics also, just take the 1990s oscilloscopes with all their custom chips.

So today (or better say since several months), i have to care for a Waldorf microWAVE. It didn't take long to find out that one of the CEM5508s in it has a thermal problem.

Looking for spares didn't reveal something usable. There are some offers in the 50-60€ range (per chip!), but those offers don't really suggest that you definitely get what you paid for. And what if you get a used or maybe NOS chip that fails 2 weeks later? There are some replacements, also very expensive, but they also are not an option - more on that later. WAVE users currently tear apart microWAVEs to get their chips, which is a shame, but i understand the problem...

So i had a closer look and talked to the owner of this synth and we finally decided that i'll try to build a replacement for this chip. It totally doesn't make sense on the economic side because it will take me days, maybe weeks, to complete this. And technically, it's not even my problem. It's just the usual DIY aspect - an interesting challenge that needs to be solved, money and time doesn't matter :)
And maybe the final design can be useful elsewhere in the future.

But what is the CEM5508?

It's an 8 channel sample&hold. That's it.

Wait. You mean, like an SMP08? That's simple.

Well... there is more about it. At the first glance, especially when you happen to look at the 3-page preliminary datasheet, you might get the impression that this is really all you need to know. But it isn't. You need to either find the 7-page preliminary datasheet of the CEM5508 or simply use the PD508 datasheet, it's the same chip. And then, you find some more details. So let's start to find out how it can be recreated with off-the-shelf components.

8ch sample&hold - yes, i'm doing this all the time, 1x4051, 2xTL074, 8xPP capacitors, done. The 5508 is even 4051 pin compatible.

3 ICs, 8 capacitors so far.

But then, the 5508 has switchable output resistance. This is especially useful for synthesizers (any maybe due to not being useful anywhere else, the 5508 is no longer made), when you have an 8bit DAC and software envelopes and want to slowly sweep your filter with resonance fully cranked up. This usually ends up in unpleasant stepping. So you use the high impedance (50kΩ) there. And when you want a very quick attack phase, you simply switch to the low impedance (several 100Ω). Followed by another capacitor, this gives you a great smoothing filter for modulations.

So, let's add 2x4066 and 8x50kΩ resistors. And an 8bit addressable latch to control the analog switches.

6 ICs, 8 capacitors, 8 resistors so far.

But now things get complicated. If you know the 4051, you know that you need to open the gate (lower the inhibit signal) for the whole sampling period. The 5508 also has a feature to simplify interfacing. You just throw a short pulse at the inhibit signal and it does the whole sampling job on it's own. Sure, you need to hold the analog input, but the sampling period of 2µs is handled automatically. You may even change the address lines inbetween, because the idea is that you program your DAC and then just perform a quick read or write access from the CPU to trigger the sampling, with the analog channel number directly hooked up to the CPUs address bus. Obviously, you need some more logic for this. So add a monostable multivibrator and a latch to keep the address. Not going deeper into details here, but if we continue this way, add 2 more ICs. And a resistor and capacitor for the timing.

This switchable output impedance. How do you control it if the chip is 4051 compatible and has no additional input for this? Well, it's as simple as "write to the same channel twice, the second write is for the impedance switch". Hm, okay. So you sample a channel's analog output and then sample the same channel once again, using the analog input as a digital input to turn on/off the corresponsing impedance switch. Great idea. But we want to recreate the chip and this adds a 3 bit comparator and possibly some more logic. Add 2 ICs.

So at the moment, we have roughly 10 ICs, 9 capacitors, 9 resistors.

Now let's care about the voltages and how all this goes together with the signal levels. Roughly said, the 5508 is nearly compatible to the 4051 when it comes to power supply and signal levels. So it is able to handle something like - 3V Vee, +10V vdd, GND inbetween and the control signals are TTL compatible. Horrible. The 4051 has all this built-in, but you don't directly interface the 4051, you have other stuff inbetween. So my first thought was to build everything with 40xx chips, and i found a suitable level shifter which would solve all my problems.

Then i had a look at the timing. The 5508 is damn quick. Trigger it, it starts to sample within some 100ns and is done after 2.5µs. If you build such a complex circuit with 40xx CMOS stuff, the whole thing doesn't even consider starting to work within 2.5µs if your power supply is low enough. So this won't work.

74xx logic, fast enough, but then you add more power supply domains and need even more level shifters.

And you don't get around this. You have an analog part, the 6 ICs, 8 capacitors, 8 resistors, and you need a control part for the rest. And i thought very long about how this is done best, ending up with a microcontroller for several reasons.

74xx logic would be nice, but i need a larger number of ICs for the control part and that makes PCB layout even more complex than it already will be.

CPLDs would be an obvious choice, but they also need clock for the timing and i didn't find a source where i can get a small number of them for cheap prices, besides this, i would need to learn all this stuff first. I have this on my schedule, so it would have been a good opportunity, but the whole project is incredibly large already, no need to add this here.

Microcontroller. Like the STM32. I mean, the microWAVE runs a 68000 with 8MHz, why not put another 32bit computer with at least 8MHz right next to it to control the stupid sample&holds? It's a very cheap and flexible approach and might even simplify the PCB layout a bit because i'm not forced to use specific pins - at least not to a degree i would have with 74xx circuitry. I can compensate a "stupid" GPIO layout in software and already did this successful in other projects. So i had a closer look at this option.

You still don't get rid of the level shifting issues, and to get the required timing, you need to run the STM32 full speed, which eats lots of power (well, 30mA or so). Having the whole circuit per 5508, it would not be a very nice approach. So while the idea in general is universal enough to replace 5508s in every device, i decided to concentrate on exactly this microWAVE with a Rev. 1 analog board and make a PCB replacing all 4 of them. Because i only need 1 micro then and i'll just feed it from the digital part's 5V, not from the +8V line of the 5508s. And then i can design the circuit for exactly this purpose and simplify some things because i know that i only have TTL levels at the control inputs, that the analog input swings between 0V and 5V, everything easy enough to put this thing together with fast 74HC4051, 74HC4066, a 74HCT259 as addressable latch, 74HCT138 as address decoder, i can put everything together this way, powering the 74HCTs from the digital +5V, the 74HCs from 6V derived from the 8V, the TL074s powered by the - 5V/+8V and finally the STM32 from 3.3V derived from the digital 5V. This combination of chips completely hides the required level shifters within the chips i need anyway. STM32 (3.3V CMOS compatible) > 74HCT (5V TTL compatible inputs, CMOS compatible outputs) > 74HC (6V CMOS compatible inputs). All the input/output low/high levels match.

Phew. What a horribly complex thing. We're still talking about just replacing 4 simple, identical chips :)

So far for the hardware, which was finally defined this way during the last 2 days. Not yet tested, but i'm pretty sure it works. Too bad the microWAVE meanwhile "exploded", i need to order new capacitors for the PSU first, before i can continue on the hardware. But better the PSU dies now than when i'm done and the unit is put back into the owner's rack.

On the software side, i think i have everything in place also. That was a nice challenge to squeeze out the last bits of performance from this. Basically, it's a very simple loop consisting of just a handful of lines of code plus 2 large lookup tables.

The main loop waits for any inhibit signal becoming activated, it does this by waiting for an event (WFE instruction), to remove the big jitter in timing. It then immediately reads the GPIO with the inhibit signals, channel number and impedance data input (the analog input connected to GPIO). Using this 8 bit value, it uses a LUT to get an address of a 32bit integer. It then masks out the inhibit signals, reads the 32bit integer and ors it with the value, resulting in an index into a second LUT, which results in a control value for the output port. This is then immedialy written to set the channel and mux number or impedance latch, not yet activating them. Now, it is checked if the value is 0. If it is, the whole thing is aborted, it waits for the inhibit signals being deactivated and goes back to the main loop.

If not, it masks out the bit for gating the mux/latch, starting the sampling process. It then updates the "previous channel" value in the 32 bit integer mentioned above and then turns some rounds of updating the impedance switch latch output with the analog input value. Just to waste some time. After this, the sampling period ends, it waits for inhibit deactivation (usually happened long ago) and goes back to the main loop.

So the 2 LUTs have the following purposes:
- The first one basically converts the inhibit signals into a pointer to one of 4 32bit integers - or a 5th one in case of an invalid combination
- The 32bit integer holds the previous channel number, the chip address, and another bit to indicate an invalid bit combination
- Combined with the analog input and the current channel number, i have all information i need to control the output port with the right bits, this is what the second LUT is for. All entries for invalid combinations are just 0, so the port is not changed and the loop aborts, all valid entries perform the full sampling cycle

For performance optimization, i tried several things and found out:
- Running the code in SRAM instead of FLASH doesn't improve performance, it sometimes even degrades it
- Having the LUTs in SRAM improves performance significantly because the FLASH prefetch buffer might be nice for code that executes linearily, but not for random access - so i suffer from the 2 waitstates in FLASH
- Both things together help the micro to optimize the pipelining, so data and instruction fetches from the same source don't collide
- Best optimization level in g++ was O1. Os and O2 were identical, O0 much slower, O3 resulted in non-working code
- Counting clock cycles like i did on the 6502 back then is no longer helpful :)

And as a result of all this, i now need a minimum pulse width of a bit more than 160ns on the inhibit inputs (PD508: 200ns), no further address hold time (PD508: 300ns), sampling starts a bit less than 500ns after the inhibit pulse (here i'm worse, PD508: 200ns) and with the duration of the sampling, i'm relaxed, plenty of time, i set it to the 2µs of the PD508, which results in the given 2.5µs overall time of the PD508 datasheet.

The PCB layout will still be a bit challenging, but the whole thing will be a drop-in circuit board to replace all those 4 CEM5508s in the machine, so the owner can even sell them on eBay, 3 are still working fine. And there are people buying them. Maybe i'm making a board for the Rev. 2 also, if i get one in my hands that needs this fix, but for now, it will only happen for the Rev. 1. Maybe some day, there will even be a replacement board for the WAVE.

And this is where we come to an end for today - a second post will hopefully show the completed, working solution. Just some personal thoughts about the whole situation with this chip.

- People tear apart less valuable, perfectly working synths to fix more valuable ones with their parts. But does it really make sense to replace old, failed parts with old, soon failing parts, especially if you need to sacrifice a vintage synth for it?
- There are only very few "replacements" for this chip, which is not really rocket science. But those replacements usually are nothing more than a SMP08 adapted to the right pinout. Without the impedance control handling and automated sampling handling, this simply can't work at all in a Waldorf synth. Maybe it's okay for others which just use the 5508 as a 4051 on steroids, but definitely not for WAVE/microWAVE owners.
- How can it be that since many years, nobody does something against the problem. Why is it always me who needs to care about such trivial stuff? I really can't fix the whole planet :) be continued

Post has attachment
Planned obsolescence? Let's check

A friend gave me this nice kitchen scale yesterday and asked me if i can find out if there is some counter or so that renders it dead after warranty expired, because this was nearly exactly the time it failed. Completely dead, no display.

I had a quick look into it this morning, found out that something still switches power to some circuitry on and off as intended when operating the capacitive power button. The main CPU is unreachable because it's bonded to the PCB, but there is a TM1620 next to it. Checked the datasheet, it's an 6x8 LED driver with serial input. Hooked up the scope and the input looks plausible, the data fields change when pushing on the sensors, but the outputs of the chip stay off, not driving the LED.
Okay, doesn't look like planned obsolescence, but still leaves the possibility, maybe the thing is tricky enough to simply not send a chip enable command when it's time has come.

But i just had another idea - hot air. So i grabbed my hot air pencil and heated up the TM1620 while the scale was turned on and i suddenly noticed it drawing more power (had it upside down to reach the chip, so i couldn't see the display, in such situations, the amperemeter of the lab PSU is very handy). Turned it around - voila, it works!

Okay, no permanent solution, as soon as the chip cools down again, the LEDs go off. Operating it with that 80W hot air pencil would not be very energy efficient, but i told the owner that he can get some TM1620 (you get them in 10-packs for $3 from China) and i'll replace it, if he wants to.

But at least it's proven - not planned obsolescence here, at least none that's not forced by the consumer asking for very cheap electronics and getting exactly the quality he asks for. While 5 years for a 36€ scale isn't even that bad.

Post has attachment
Selenium rectifier inside

Why throw away broken stuff when you can take it apart and learn something?

This is the selenium rectifier i replaced in a tube radio today. You can easily see 8 small stacks of selenium diodes and some pieces of metal to connect them. Many diodes for a single rectifier, but selenium diodes can't handle large voltages, so this is necessary. If you ever cracked open an old HV rectifier diode from a TV (i did decades ago), you find many hundreds small chips in it.

Interesting. And i think i found the problem of this rectifier. The second cell in the upper row and the corresponding chip doesn't look so healthy.

Post has attachment
Selenium goes silicon

Well, it didn't work to put this Nordmende Carmen into the corner to complete the repair a different day. After the problem wasn't an open resistor, i checked the anode voltage (well, i always tell people to check supply voltages first, so why should i do it myself? :). 160V instead of 260V. That's a bit out of tolerances, maybe.

As my isolation transformer showed just a small current consumption of 160-180mA (the smaller Blaupunkt radio needed 220mA), it likely isn't overloaded, but the rectifier is broken.

It's an old selenium rectifier and in Germany, they are also called "gleich riecht er" (will smell in a moment) for a good reason - at end of life, their voltage drop increases, the power loss across it thus also increases, it gets hot and finally gives up with a very evil smell. Good that i didn't reach this point, i read it's hard to get that smell out of the room again...

As selenium rectifiers are less than optimal when it comes to voltage drop (back then, it didn't hurt to lose 20V at the rectifier), you can't simply throw in a silicon rectifier, the voltages would be too high due to the lower voltage drop. Besides this, silicon diodes have much harder transients when switching, which can result in enough distortion by overtones that the low frequency AM bands get unusable.

So a common trick is to use 1N4007, add some 4.7-22nF capacitors across them and add some resistors to simulate the higher resistance of a selenium rectifier. I never spent so much effort for just a simple rectifier, but it works great. I had some 15nF Y2 capacitors around, i had 150Ω 2W resistors, which sum up to 600Ω, which should be a useable value for this radio (other sources tell around 470Ω-1KΩ for similar types). And it looks nice. You could easily sell it as an audiophile bridge rectifier :)

After carefully putting this thing together and installing it in the radio, i first measured the anode voltage. 250V instead of 260V. Perfectly okay, i have set the radio to 240V and run it off our official 230V power grid.

And the most important thing - it works now. No fluctuations any more, clean FM reception, great sound and the EM84 is bright enough that i decided to keep it.

So i can finally archive that topic :)
3 Photos - View album

Post has attachment
Nordmende Carmen restoration

A bit more to do on this one, but i got it very cheap, 10€ or so. After working on this radio, i know that restoring tube radios will not get a new hobby here. Sure, it's interesting, the result is extremely satisfying, but there are also some very annoying moments around this. Big respect for all out there who frequently spend all the time and effort to keep those old radios alive.

But let's start. First problem was a FM tuning capacitor that was stuck, you could hardly rotate it at all. I tried some Kontakt 701 externally, but it wasn't enough, so i had to remove the cover of the FM tuner to get to the innards of the capacitor. There, a very small dose of 701 fixed the problem. Not sure if it's a permanent solution or if the grease re-hardens with the time, but at the moment, it works fine.

After this, i could test it a bit more. Fluctuations. Everywhere. The reception quality on FM was coming and going, the whole magic band flashing, but in general, it wasn't that dead. I already checked the main filter caps some days ago, they were good enough. But after seeing all those infamous ERO 100 capacitors, some of them in a very questionable condition, i decided for a bigger recap job.
I ordered replacements for the capacitors, most are Mallory 150, just the ones i didn't get the required values are TAD Mustard, Orange Drops 715p and one silver mica capacitor. Add another nearly 30€ for those parts...

Yesterday evening, i could finally start to replace all those capacitors. It was no fun on the PCB because reaching all the necessary spots on the solder side without removing the whole PCB is a bit difficult, especially if you need to take care all the time to not touch the wiring and burn it.
Replacing the capacitors underneath the chasiss also was no fun, because this radio is built to last, not to be repaired. If those capacitors would live forever, the whole radio likely would also do so, but well. Capacitors. So it took a while to replace them all, being confused here and there if i connected it to the correct place, constantly verifying with the schematics, but finally, all the capacitors in the radio were replaced.
There was one shielded capacitor (with 3 wires), those are unobtainium today, so i used a standard unshielded one for now. If i would need to shield it, i could just wrap it in some aluminum foil and connect this to gound. But i don't notice audible problems at the moment.
And while working on this, i destroyed two resisotrs, they easily fall apart.

Another careful check of all the places i worked on - you don't want to make mistakes in a device running off several 100V.

Let's continue with the visual appearance. Just some dirt here and there, i have a nice set of small nozzles for the vacuum cleaner, so removing the dust was easy. The plastic parts were cleaned with glass cleaner, in some places an old toothbrush helped a bit.

The dial glass backlight diffuser is just a sheet of paper wrapped around a cord. The lower end was getting loose, so i glued it back in place, didn't more on this it's still good enough.

Old foam strips that totally dissolved over the decades were replaced by some self-adhesive felt strips.

The dial glass was a bit of a problem. Nordmende attached several self-adhesive things to the painted back and the paint suffered under them. I hope the chemical process meanwhile ended, but i left those stickers in place, i fear removing them also removes the paint. It's not nice, but a new paint job would be really too much, i have to live with this. The painted side was just wiped with a damp cloth, the front side got the usual treatment with glass cleaner (finally it's used for what it's intended for :).

After putting everything together, it just worked on the first try. Sort of.

On FM, i still have fluctuations, and it gets even more crazy. When it's connected directly to the wall outlet, FM nearly doesn't work at all. On the isolation transformer, it mostly works. And it also depends on the orientation of the radio, which is not that surprising with the type of the antenna, but the effect is very dramatic.
I assume the gain control doesn't work, maybe an open resistor. The magic band gives plausible readings, but maybe the control voltage doesn't reach it's destination. The AM bands all work fine.
But that should be a solvable problem. I think i'll give it a new EM84, it doesn't look so healthy, but i'll dive a bit deeper into the schematics first. Not today. I have to move forward nearly 6 decades in technology now :)

But the sound. Wow. This radio has 3 speakers and with this really large box around them, the sound is really awesome. This is how i recall the old tube radios from my childhood.
26 Photos - View album

Post has attachment
Blaupunkt Verona ECL86/PCL86 tube conversion

During repair of this nice tube radio, i found a faulty ECL86. As those are hard to obtain those days and i have plenty of PCL86 around, i decided to consider a conversion.

The ECL86 and PCL86 are nearly the same tubes, just the heater differs. E tubes are commonly used in radios and heated by 6.3V, all heaters are connected in parallel, while P tubes are commonly used in TVs and heated by 300mA, all heaters are connected in series and their heater voltages added to determine the heater winding on the transformer (resulting in 300mA through the whole string).
There are some special cases like the ECC83/12AX7, which provide several possibilities to heat them, like 6.3V, 12.6V and 300mA.

Not the PCL86, this one wants 300mA and to make this current flow, it needs 13-14.5V.

But first, i wanted to try if my diagnosis is correct at all. The PCB layout in this radio allows to easily disconnect the heaters of the up to two (for stereo) ECL86s at least on one side. After some soldering (and using heat shrink tube to connect the unsoldered heater wires for the original 6.3V circuitry), i had a separate heater input for the output tubes. With the lab PSU, i could check if everything works - yes, the problem was gone, so it's really a faulty ECL86.

Now it's time to decide about a way to fix this. There are several options.
- Delon voltage doubler, this would make cutting PCB traces necessary
- Separate transformer just for the PCL86, this would remove 700mA load from the original heater winding, possibly increasing the heater voltage for the other tubes
- Additional transformer in series with the existing heater winding, this would remove just 400mA from the original heater winding, a good compromise

And after digging through my transformers, i found a nice small 2x3V/300mA transformer that looks good for the purpose. With a provisoric setup, i determined 13V heater voltage for the PCL86 at 220V mains voltage, that's the lower end of the range, but enough. And even if it's just rated for 300mA, it doesn't get too warm, so this transformer is usable for the purpose.

While thinking about how to install it in the radio, i had some more complex setup with a PCB and 2 mounting brackets in mind, but then i found this single transformer mounting bracket that perfectly fits the transformer used in the radio. Double-sided adhesive tape and a robust cable tie to fix the small transformer to it, problem solved. I usually don't like such unstable solutions, but a tube radio isn't something you shake too much, so it's robust enough this way.

Wired up everything with the original transformer (it's important to take care about the winding directions of the transformer to add the voltages, not subtract them), i had this new output for the PCL86 heater. Now i just need to bring this to the PCB.
There are 2 free pins on the connector, but they are gaps between primary and secondaries/capacitors of the transformer, so i kept them unconnected. Two pins are connected together at the radio end (one end of the heater and GND), i also left it this way to not introduce hum when i move this connection away from the electronics. So i needed another single connection for the new heater. Not as elegant as the connector for the power supply, but as you don't take this apart frequently, a luster terminal is good enough here.

After putting it together, everything works as expected. The PCL86 needs a bit longer to heat up, a bit more power would have been nice, but for now, it's fine, the radio works nice and i will never again have a problem with the ECL86 getting unobtainable with the time.
3 Photos - View album
Wait while more posts are being loaded