Power

Good to go with a reassembled power test. First, the PSU outputs under load:

Not what I’d expected. I checked again with another meter, same results. May need to think about that a bit . . . 

Once everything has warmed up (half an hour, say) the dividers are getting +15·0V at pin 1, which is not what I’d expected but fine. The ‘decoupling’ capacitors are at +14·5 on the positive side, but alternate +1V, 0, +1V, 0, +1V, 0, +1V, 0, +1V, left to right, on the other. Which means they may not be decoupling . . . Leakage somewhere? I’m not sure why this would be a design decision but I’m not sure how a failure could do this unless they’re not on a single bus. I should check the caps’ conditions, but alternation of failures like that would be a bit much of a coincidence? So I’ll need to get the board back out to check the circuit.

Audio

So, initial audio check? It would be marginally surprising if everything was working here. Annnd . . . this is where we need to get our teeth out and get them into it. There’s so much wrong, it’s difficult to decide what it is.

• It’s NOISY and seems to be receiving a fair bit of AM Radio, but that’s with the top panel still off; we’ll see what happens later.
• IT’S LOUD (general output) but that’s probably a result of not having a volume pedal connected.
• The decay controls aren’t operating and all the strings are probably at maximum decay [IN18: identify and fix decay fault]. This is more pronounced on the bass than the treble side, except that B3 Cello has no decay at all.
• Pianos and bass aren’t sounding at all [IN19]. (Not even from their individual outputs.) They do have a lot of all-pitch oscillator noise feeding through, but that’s not unexpected if there’s a fault there somewhere.
• And Cello C5–D5 and to a lesser extent E♭5–F6 are producing a lot of noise, with very little tone in C5–D5. [IN20: Cello high notes noise.] But the oscillator and TOG are clearly working.
• The Ensemble setting, weirdly, prevents the treble side from playing at all (maybe not on pianos; we’ll see). I’m not sure if that’s intended behaviour but it seems odd. (Later: It is as intended.)
• The chorus effect may not be working. The strings sounds are smoother and more organlike than I’d expect. Usually with string machines you hear a bit more wobble. I mean, shimmering, that’s what they call it. Not that the existing sounds aren’t quite nice, and it does sound like something is happening with the Bass-side Cello but that might just be dissonance from the extreme decay.

This is probably normal, but should be checked: In Ensemble mode the trebles are active with only the faders, but the bass voices need their switches on. All four can be mixed separately. (n.b. as far as I’m aware Ensemble only applies to the strings sounds not the pianos.)

Division

Around this point, I checked the dividers in more detail, partly because I wasn’t sure how they compared to the dividers in the K1s. As far as I can tell, they’re similar, though they may have different power limits and do have different connections. (They appear to be equivalent to the AY‑1‑5050, for which more information is available. ^[2]) Here nine (rather than seven) of them are used to generate an extra octave as compared to the K1s so the connections are a bit different. The extra octave is F♯6–F7 (IPN) — produced directly by the TOG; this allows two registers of strings, and it looks as though it also allows the upper octave in the mix in the pianos’ keys F♯5–F6 which wasn’t available in the K1s. For this, the oscillator should run as standard at 667,040Hz, unless retuned in use. ^[3] (It was a bit sharp. I managed to get it pretty close but the trimmer isn’t behaving well, jumping about a bit, and wouldn’t be suitable for tuning in a live environment.) The dividers are actually producing 63 pitches below the TOG’s 12; I expect B♭0 and E♭1 are not used. ^[4]

[IN20:] All the pitches are being produced by the dividers, though E♭1 (divider 2 pin 5) is highly distorted, and can’t be automatically measured on my digital ’scope. Probably it isn’t connected to anything. All the other signals seem clean, so this is not the source of the Cello distortions for keys C5–F6.

A bit more probing to confirm my understanding of the voicing boards, and as far as I can tell the piano contacts aren’t sending a signal to the board sockets. That implies a fault in the supply to the pianos busbars? Yes, looks like it. The strings busbars have +20·8V (except in Ensemble Mode where the right bar is zeroed) but the pianos off-bar has only +0·64V and the on-bar has 0V, with the bass-range bar having a sine wave at about 0·8Vp–p 50Hz, which is presumably getting through from the mains. That makes it sound like an easier fix.

There are two small blue wires from the strings and pianos-off bars into the board at the left end, which may have been slightly obstructing the F1 contact springs, but I bent them back a bit. The strings (bass range) bar connects through one to a yellow wire direct off PSU out pin 2, and the strings (treble range) bar through the pink wire at the right end of the board, which goes to K2-14. The bass-range bar connects through the other to a green wire that goes up to the switching board — switching power, not in Bass Mode, but in Ensemble Mode. (So it’s an ensemble bar.) The pianos-off bar is the white wire at the right end, which should be negative, the pianos-on bar is grounded by the black wire at the right end.

Decay

[IN18:] While fiddling and probing some more I decided to give the sliders another clean, so took K2-13 off — and under a strong light, noticed a bend in the board between the decay sliders and the proximal board connector. Checking the underside under a magnifying glass it looks as though several of the traces are cracked, and may not all still be making reasonable contact. That may not explain the complete absence of decay control but it should be fixed. So I’ve bridged the traces with solder. In the process of scraping off lacquer (actual blood loss this time) I found another crack further along the board. The cracks mostly run from the slider screw holes — which aren’t used. It looks as though they are an outcome of a) vertical pressure — or possibly impact — on the slider caps, and b) that the only rear support on the board is the pin connectors. This may be a common fault with this model.

Giving it a test run, it has sorted the decays but I don’t hear any other difference.

I glued the board cracks with cyanoacrylate, and added 3mm rubber pads at the rear of each potentiometer to prevent a recurrence — although we might note that pressure on these will be transferred to the K2-14 board, and arguably another set of spacers, 10mm thick, under there might be advisable. But its mounting seems much stronger, and it isn’t supported by its connector pins. [IN18 closed.]

(Later: When I came to put the slider caps back on I noticed it takes quite a bit of pressure. I’d suggest always fitting them with the sliders at the front end, to avoid exacerbating this problem.)

That’s good but it did nothing for the pianos sounds. So I turned to the switching board (K2-13) to see if there was anything else going on. No cracks or anything, so I considered the circuit. Which meant drawing it.

There doesn’t seem to be a single point of failure for all five pianos voices here . . . but as far as I can see the switches’ action grounds whatever signal they’re carrying when in the off position. There are multiple ground pins here, so there’s unlikely to be a failure? (Later: Actually once I’d fully worked out the circuits I realised there is a single point for at least all four regular pianos voices, and optionally for Bass, in the output amp anyway, but that wasn’t it. The drawing has been through several iterations of increased comprehension.) Anyhow I mostly wanted to work out where the signals should be. So far I’m not getting very far with that. So I powered up again and checked all the connector pins with an oscilloscope, finding that only three have anything other than noise, and that depends on the position of the Ensemble switch. Not enough to be useful, but it indicates that there’s a fault upstream of this, as surely this board should be receiving or sending more than that.

Negativity

(Later: Those who don’t want to go through the tortuous fault-finding maze I got lost in might want to skip to the unexpected resolution here.)

Thinking some more about this I think there are some significant issues which boil down to the probability that somehow the negative rails are going positive, which may mean a short of some kind? Whether this is affecting the pianos voices I don’t know (fault in common?), but since the TCA350Zs may require a significant negative voltage (I’m not certain since I only have a datasheet for the TCA350Y), the chorus effect is probably affected. So testing the PSU outputs with everything disconnected . . . again it takes about ten minutes, but eventually the pin 8 and 9 voltages are −14·4 and −14·8. (The PSU negative capacitors might have been affected by a small positive voltage on their negative side, without the outcome being obvious? Not explosive anyway.) They might rise higher with time but that’s probably good enough to show the suspected effect.

With all the others out, just inserting the PSU connector makes no difference. I suspected the problem may lie in the chorus effect boards, so connected the M175 first. And that does it. The pin 9 voltage immediately goes positive. So I tried removing its daughterboards one at a time, only to find that even with them all off it’s still happening. That’s about the least likely possibility, as there’s nothing on it other than connection traces for the other boards. And I don’t see any extra solder blobs or anything . . . And it’s up on spacers to keep it from shorting to the grounded foil underneath . . . so, what? Well, reconnecting it while holding it in the air does not cause the voltage to go negative. So it’s shorting to ground? I tried putting it back in place and screwing in, testing again after each screw. The voltage started dropping after the left rear screw, though very slowly, so it might have been happening earlier.

So I’m trying it again with a sheet of polythene under the board. That seemed to cure the effect, though I’d prefer prevention. I reinstalled all the other boards and the negativity kept rising, but when I put the M174 in it all disappeared again, so this isn’t a simple fault.

(Later: This extra insulation was probably unnecessary, but I left it in even after I finally identified the underlying fault as I didn’t want to have to take the board off again. Can’t hurt.)

Over time the pin 9 output got to −6·4V without the M174 in place, and with only the M175 connected. And there is no other connector pin anywhere that has the negative voltages. On the M175, only the M174 seems to have a negative power input. All a bit confusing. Also, it turns out the clock signal in the TCA350Zs is locally generated on the K2-20-2s; the three boards do c.56KHz, 58KHz, and 55KHz (non-adjustable). But the modulation timing should be in common, presumably from the M174. I decided to leave it off for an hour or two to see what would happen.

Switching on again, the negative voltages started rebuilding reasonably fast, and I tried reconnecting the other boards to see the result. The top boards made no difference, but the left end connector on the K2-2 did cause the voltage to go positive. Leaving it to recover again . . . and I note that this is the other board where there’s a sheet of foil underneath. After leaving it another hour or so to rebuild the negativity (reached −15·16V) I connected the right-end connectors, which had no effect. Now to thoroughly test this I’ll need to pull all the K2-1s again . . . but I suppose I’ll need to get them off to get the board out anyway, and I already need to do that for reasons noted somewhere but currently forgotten. (It can be lifted slightly at the front but not pulled forward without untying the wiring harness.) Where’s that biscuit tin?

With all the voicing boards and the OS-5 off, starting up produces the increasingly negative voltage; and connecting the left-end connector turns it immediately positive. Hm.

I’ve been reluctant to do a whole wiring diagram but seems I really need to know this, so here goes.

Things I learned in the course of that:

• The little blue bodge wire with resistors running from the K2-2 left connector to the M175 connector is from the negative supply . . . is that how inserting that connector is causing the negativity to disappear?
• Messy though these more-or-less-handmade connectors be, they are seriously maintainable, as you can desolder and resolder on to them without having to fiddle with crimps or replace them. (In the course of these manipulations the said blue bodge wire broke off the resistors; I’d noticed that it was very weak already; and the joint was covered with old sticky insulating tape which had fallen back from the joint, potentially allowing a short. Saved me the bother of desoldering to put some heatshrink on it.)
• The Piano/Harp Output is self-switching, so without a plug in, it connects directly to the general output. The Bass Output is also self-switching but returns the signal to K2-14.
• The General Volume pedal takes power from the grey wire from PSU pin 5, currently +16·3V. In the absence of a pedal I don’t know what the final voltage should be.
• The white wire to the pianos-off bus bar is from pin 9, −15V. Which implies that the pianos envelopes are never getting power, which may mean that fixing the voltage is all it needs.
• The General Output amplifier is on the M175 somewhere; either K2-6 or K2-7. (Later: Both are preamplifiers but K2-7 is the final stage.)
• The tuning control wire is shielded and the shield is not ground, but a positive voltage for the trimmer to adjust (rather than a ground for a voltage to be adjusted against). (K2-2 terminal 13.)
• There are thirteen wires on the ground tag strip, but without desoldering or cutting the lacing I can’t tell which is which. All the tags are commoned with a small soldered wire.

• There are two almost separate ground systems, but they’re not separate signal and chassis ground:

  • One (hereafter called left ground) connects most of the black wires to the ground tags and the PSU, and to the earth terminal, and the majority of the signal wire shields through ground traces on the K2-2 and K2-14 (left end) boards. This includes the ground wire on the M174 connector, pin 6.
  • The other (hereafter, right ground) connects the remainder of the signal wire shields at K2-14 (right end) and the foil shields under the K2-2 and M175 boards, through the M175 connector pin 11.

  While otherwise separate they are then connected by the ground traces on the M175 board. A rather odd design? It may have something to do with the effect of connecting the M174 board on the power negativity, although that may not be normal operation?

  (Later:  both sides of this ground system are also connected in K2-110, but in different ways according to the switch state, for non-obvious reasons. ^[5])

Modulator

Time to examine the M174. First thing to note is that all the chips on it are UA4558TCs. Shunting a voltage from negative to positive is a possible failure mode, so I’ll want to look at that. I also notice a few scratches on the underside, which on close examination look like someone has cut through some solder bridges on the chip pins, or some others that may have been suspected but may not always have been there. Slightly rough manufacturing.

Checking it out and powered with ±15V from a bench supply. I started with a 20mA current limit on each side, but the voltages were highly unstable until I increased to 25mA. With the inputs stable, all eight op amp outputs were unstable on a DC meter. Which, if they’re timing signals for the delays, is to be expected. Using an oscilloscope . . . They seem good, actually; each of them is producing an approximate sine wave at both outputs; from left to right they’re about 5·8Hz, 0·578Hz, 5·8Hz, 0·578Hz; the ratio seems deliberate. These waveforms are also present on the two trimmers.

Somehow they’re being combined later on; the actual output pins give the superposition of the two waves, left and right pin at lower amplitude than the centre, and the left pin is slightly delayed from the other two. In summary, this board seems to be working perfectly well. At least — with 30V between the power pins, it is. Which it isn’t getting at present. Setting the negative voltage to +1·2 to represent current conditions, the outputs are flat. Lower negative voltages? Starting at −1 (it wanted a bit more current too; stable at about 30mA) and turning up slowly, we see the tiniest beginnings of the output waveform at about −1·5, reaching normal around −9. In conclusion, whatever else is happening, the BBDs won’t work right without a significant negative voltage maintained on this rail.

Main Board

Back to the K2-2 then. I tried putting a sheet of polythene under it too but it made no difference. Having a closer look at the underside, I see that the frontmost pin on the connector, the (currently?) −14·5 supply, is where the dividers get their negative power from. And it is also connected to every second 100n capacitor by the dividers, and to the OS-5 socket. The other negative voltage, which goes to the seventh connector pin, then goes through a resistor to the voicing boards’ pin 6, which is in the pianos section, implying that the wrong voltage there could be preventing the pianos envelopes from operating normally.

I’ve been reluctant to attempt this because these sockets aren’t always in good shape, and they seem to be functioning, but time has come to pull the divider chips. At least they’re socketed. And without them, the negativity returns. So does one of them have a hidden flaw? Eh well, actually replacing any one of the nine causes the voltage to go back to +1 and a bit.

A nice detail I noticed is that the pin 1 position for each of the dividers is marked with a dot in the copper on both sides. Unfortunately the sockets are also marked for pin 1 with a little bevel on the interior, and whoever assembled this for soldering seems to have taken the copper as a reason not to bother orienting the sockets correctly. It makes no practical difference, but is confusing. To correct that would involve unnecessary desoldering, so I’m just going to mark them for confirmation.

I doubt that the flaw here is the dividers any more than it’s the M174, because even under present conditions they appear to be working. And it would probably take something major to affect them all. More likely all these are in contact with the problem and with them out, it doesn’t get through.

So where is it? I checked the remainder of the K2-2 components, which are some diodes, and some resistors including the smallest power resistor I’ve ever seen, 0R56 3W, at the back right corner of the board (measured 0·7Ω, but that’s probably in tolerance to the level of accuracy I’m going to get).

Regarding that point about dividers or boards connecting something, it may be worth noting that the M174 has only power inputs, so the connection could be straight between the positive and negative rails. Which implies that there may be a failed component there. However, it does have outputs to the K2-20-2 boards so it could be there. Still, I went over the M174 in detail again. The only component which was in any sense even marginally suspect was the 100µ electrolytic, which I took out and did a full leakage test on just in case. It’s good. I also checked the overall resistance between the rails and rails to ground on the board; nothing wrong.

I checked the voltages again, getting the feeling that I’m back into faffing-about mode. The PSU with nothing connected gives −15·16V at pin 9. Inserting the connector with nothing else connected makes no difference so it’s unlikely to be a short in the wiring. Connecting the M175 (still with no M174 board) makes no difference. Connecting the K2-2, left end only, with no boards or dividers, does produce a fall in negativity — down to −9V in the time it took to measure, and still falling, but only slowly. I unscrewed the hardboard strip and checked again, then lifted the board and propped it up on insulators and checked again.

If anything, it’s worse; the fall in negativity is faster than it was. I also notice that I’ve been measuring at pin 9 but not always at pin 8; but currently that doesn’t seem to be affected; it remains around −14. Unexpected. Pin 9 also seems to be taking longer to reach −15 again.

I tried some resistance measurements across the K2-2 100n capacitors. (All connectors, chips & boards still out.) Nothing alarming. Every first has over-limit resistance; every second is in the 100–200KΩ range, though it seems to be falling across all of them the longer I try it. Which might indicate a capacitor charging from the meter but unless it had a transistor connected I don’t see how this would happen. The remaining circuit of course includes the K2-3 board linked through 146 contact springs (none of which should be making contact), and there’s a 470n capacitor on that as well as on the K2-2. A resistance check between the white and black wire solder points, which are directly connected to the cap, shows a low but steadily rising resistance like you’d expect from a functioning capacitor and quite a lot of capacitance in the vicinity, but with the PSU disconnected it’s over limit in resistance and right on target in capacitance. There are also a lot of diodes on K2-3 that I haven’t tested, but if one had failed the resistance there would be low? So . . . conductive dirt on the boards? Again if there was that much conductance there should be a low resistance path here to show for it. I’m faffing about aren’t I?

Resistance check on the PSU pin 9 to ground with nothing connected, just in case. Starts low and rises to 1·7Ω . . . exactly what you should get with a large capacitor over the negative and ground side of a full-wave bridge. A point to note is that any positive voltage on the negative lines will be limited anyway by the PSU’s DZ2 and R6. About +1·5V is about right; immediate serious damage to the capacitors is unlikely but dielectric absorption is going to happen.

When in doubt, draw things out. I spent a few days doing little else. This became a fairly ugly schematic because I didn’t have time or energy to summarise any of its repetitions and it’s easier to draw with repeating elements first.

There are a variety of things happening here which take some working out. The one I kept forgetting and returning to is that while most traces from the divider outputs come up to the B (top) side by the voicing card sockets, those for the lower octaves for sockets 17–19, 23, 25–27, 29, 30, 32, 33, 35, 36, 38–61 don’t. So it’s not always possible to check signal frequency at the solder points beside sockets; you have to put a small probe in the socket. (And checking frequencies on the underside while switched on is . . . a bit difficult.)

After much studying diagram and poking about, I’ve noticed that one of the diodes on K2-2 I tested earlier is between the negative and ground rails under the voicing boards . . . and it’s cathode to negative. Which unless I forget everything (which admittedly I’m prone to do) would basically allow that voltage through to ground. Perhaps explaining what I’m seeing, but . . . with some more examination . . . this doesn’t show signs of having been altered. To confirm my point, and check that it isn’t a Zener or something, with everything disconnected I ran 15V forward and backwards through it and found that, cathode to anode, it perfectly blocks +15V but passes −14·14 out of −15V. Which is about what I’d expect. So is this by design?

If so, how would you keep a voltage on the pianos-off bus bar? Well . . . there is a 560Ω resistor between the input pin and the envelope sockets, so that should have some effect, and the voltage on the K2-1’s pins should never fall below the turn-on voltage of the diode . . . trying a little mock-up circuit . . . −0·8V? That might be enough to be useful?

I tried measuring resistances on the K2-2 with all plugs out, and got a slightly odd result from that diode . . . around 4·3MΩ, where I’d expect it to be over limits. Repeated attempts got something similar, so I took it out. It’s a BA130, and OoC it has 3·7MΩ resistance. A bit leaky? That still shouldn’t be a real problem? I’m not sure, but I’m going to get a replacement and see what happens. (BA130s are available but they’re old parts at silly prices so I’ll go with a 1N4454 which is apparently a more recent equivalent.)

Testing the old diode in that mock-up circuit makes no difference. The new one has over-limit resistance.

With the new diode in place . . . well, starting at −15·2V on its pins with nothing connected, inserting the K2-left connector does start with around −0·8V at the cathode side of the diode but it gradually declines to −0·025V, wavering a little over time. Is that enough to be useful? I’m uncertain, but if — as it seems — this is just balancing the voltage on the PHB audio out signal, then it might work as long as it’s not actually positive. Only one way to find out and that’s to connect things up a bit more.

The result is unimpressive. With a couple of K2-1 boards in, and everything else except divider chips or the M174 in, the voltage actually becomes a few milliVolts more negative, but add a chip or the M174 and it goes positive again fastish. So back to plan A, except I didn’t really have a plan A yet.

Still working on the assumption that the unreplaced capacitors on the positive side are unlikely to be the issue here (since if there was an effect it should also be showing up on the positive outputs) — I noted before that this is a TIP32C as compared to the positive side’s TIP41A; they are (and I think ought to be) handling different collector voltages, but if the TIP41A has been replaced there’s no guarantee that it’s original spec. . . . and I’m beginning to suspect the TIP32C may be comparatively underpowered, at least for long-term use.

Around this point I finally got round to checking the voltage rating of the electrolytic on the K2-2 positive rail . . . difficult since whoever had put it in had done so with the rating directly down to the board . . . and it’s 16V, a bit lower than I’d prefer, if the working voltage here should be +14·5. It seemed good when I first tested it, but I’m going to check it again since it’s had several hours active. 1·02mF, 0·12DF; not much change, almost perfect capacitance and seemingly in good shape, but there could be some leakage affecting things here? Out of Circuit, then, it’s reasonable at low voltages; a little more leaky at +16V, but not enough to cause the effect found. I replaced it anyway, with a new 50V-rated radial capacitor which I had in, since I didn’t have a 25V one that would fit.

At this point I remembered that the positive voltages are adjustable. What if I just turn everything down? Well, with it down as far as possible we have around +18·6V rather than around +26V, and the negatives are affected too, producing −12V at pin 9. Connecting the wiring makes no real difference but connecting K2-2 at the left end (still no boards or chips in) causes the negativity to drop slowly, though after about three minutes it stabilises at −0·72V. Connecting the M175 board with M174 in causes it to go positive fast; taking M174 out allows it to regain negativity. In conclusion, the trimmer is not enough to do any good, but this supports the idea that the positive side is proportionately too high.

Another point here though is that if the negatives are affected, there is a possibility that the caps on the positive side are also affecting it. So I’m going to have to take them out and do leakage tests. Given the general warm-up slowness, I wouldn’t be surprised if they weren’t very good, but perhaps they’re a bit worse than this indicates?

All Power Corrupts, but bench power supplies are just too useful, sorry.

However, before trying to hack the PSU I’d like to see if there’s a voltage level at which everything just works . . . or even at which more things work, which might indicate something. So I’m replacing the PSU with external power supplies — six separate voltages. This is complicated, and I had to buy another bench supply to do it.

Meanwhile another thing to check is the zener voltage and identity code of DZ1. I haven’t done that hitherto, since it involves desoldering; but I clearly need to know now. So — it’s 11·2V and BZX 79C11V. No problem there.

Now replete with more working power supplies than I’d usually know what to do with, I’m going to test the whole thing reassembled but for the PSU. Starting with all six voltages at ±5, as I doubt anything here should be lower than that. There were no significant voltage changes, which seems to confirm that it’s the PSU at fault. I gradually increased those that should be higher, until I decided to take the remaining 5s up too. It started working, and for now I’ve settled on:

We have a piano and harpsichord voice, and the chorus effect is working, and very tasty. This seems like a much more sensible range of voltages, so it looks as if the PSU needs more work.

However, there are some keys amiss in the piano voices; C♯2 is too short, F♯3 is too loud. Also B1, B2 and B3 are silent on both voices. [IN21: Fix pianos notes.] And there is no Bass voice yet. [IN22: Find Bass voice fault.]

While the negative voltages are working like this, I’d expect them to end up at the current −15 and −14·5 again. Apart from anything else, I’d more or less expect the pianos-off busbar to operate at a similar voltage level to the strings-on bars, even if inversed. But that’s speculative. There is also still quite a bit of noise, which may result from the lid being off but may also result from the power being out of kilter somewhere, or a dysfunctional part.

Always Question Power

On to hacking the PSU then.

In summary, none of them are leaking enough to create this effect. Since there’s no way to balance the positive and negative sides with the trimmer, the remaining options start with transistor replacement.

That doesn’t seem to line up with the test results, as it would only affect the pin 2 voltage. Still, I got a new TIP41A — which measures h_FE 39.

Reassembling is quite difficult with the tag strip in. I’ve decided to drill the rivets out and bolt it back in when done. I also noticed a soldering burn in the wire to C8, though it doesn’t seem to affect continuity; so I’ve put some heatshrink over it. (Don’t think that was me.)

We Have Contact

At this point I was contacted by Sebastian in Germany who was repairing his K2 — the ‘Silvertop String-Orchestra’ I’ve mentioned elsewhere. In the course of the conversation he sent me a set of ELEX K4 schematics, which it had never occurred to me to look at, but when I did I noticed that many of the diagrams showed K2-numbered boards, which seem to have been re-used from the K2r1. Which means I was able to revise some of my schematic numbering in accord with the K4 at least, and get a better idea of the intent.

The K4 schematic also confirms that Q5 should be a 2N3055, but gives BC286b for Q4 and BC160b for Q1. Sebastian has sent some photos which show metal-cased transistors rather than TIPs there. Which doesn’t mean this and other r3s didn’t originally have TIPs.

In this schematic the K2-0 terminal 9 (my PSU pin 9) output is marked as −26V — significantly higher than I’m getting. (Compared with +27 at terminal 7 (my PSU pin 1), where I was originally getting +26·5.) So, given that nearly everything else has tested good or been replaced, and the direct voltage after the negative bridge seems to be only +15, worst case, could there be a fault in the transformer’s negative-side secondary winding?

Checking that output again with one wire desoldered to avoid any influence from the board, I get 19·8V AC; hardly any different from what I was getting before. And the post-bridge DC voltage from the positive winding’s output of 33·4V is +44; so with the same diodes, if functioning right, you’d expect 19·8V AC to produce (44 ÷ 33·4 × 19·8 =) almost exactly −26V, as specified for the K4. So, not the transformer.

With the new TIP41A in, the outputs stabilise (while disconnected, and with the trimmer readjusted to original position) at 1: +26·5, 2: +20·4, 3: +26·5, 5: +26·5, 8: −1·1, 9: −1·2. The negatives are clearly no good, so why is this happening? A bridge diode fault?

But they all seem good, checked out of circuit. While they’re out, I’ll try powering the board from an external DC supply to the negative side. The result is, whatever we put in, we get out at output 9, and about half a Volt less at output 8. This is more or less what was happening with the diodes in — but it indicates that the full input voltage from the bridge was not being reached. So I still suspect these diodes.

This also indicates that there may be two problems with the board as the K4 schematic implies that we should be aiming at −26 and −15 outputs. Sebastian measures his at −21·5 and −5·3 as compared to +39 as the incoming voltage on the positive side. That’s an even greater difference than on the K4 schematics.

So, working through each diode, they’re fairly uniform. They all block all but 110-124µV DC in the reverse direction and all have a rather high initial voltage in the forward direction, starting as low as I can measure with a simple test, passing 50 out of 50mV and gradually falling off to about 640mV at 20V. This isn’t how a normal silicon diode behaves, but I don’t think it would cause the failure I’m seeing. Perhaps one or more has too slow a response for a 50Hz AC input? To check, I’ll need to rig them as a bridge, give them AC, and look on the ’scope. (I could have done this to begin with without taking them out but the smoothing caps would have affected it.)

After several rounds of doubt, unexpected ground loops, and rechecking, I get a neat full wave at around the expected voltage. I don’t think the diodes are the problem. But since I have them out and they’re not quite right I’ll replace them anyway.

That leaves the resistors, DZ2, or the capacitors. All of these have tested good in the past, and the negative side capacitors have been replaced. It also leaves doing resistance checks on the board itself, noting that a cold joint might not have been noticeable.

Powering the negative side again, I measured the voltage behind DZ2, which ought to stabilise at about −15V, but it doesn’t; it stops around −8·3V. Even with C1 on the circuit I don’t think that should be happening. It may not be a complete explanation, but it’s about the only thing I haven’t replaced, so it’s probably necessary.

Another few days later I have a replacement. With that in (a pad lifted in the course of this), the voltage there rises to −14·57 and stays there up to −26V input. Everything sorted? We’ll see. First, solder in the new bridge diodes.

With that done (another pad lifted here) and things reassembled, with mains power, the negative outputs are significantly improved, and stabilise quite fast at −13·4 and −14, but I tweaked the trimmer to get +27V at output 1 and now get 2: +20·7, 8: −14·0, 9: −14·4. That’s still not the −26/−15 proposed by the K4 schematic, but I’ll see what happens once things are connected.

And the result’s a no, again. As soon as I pressed a key I could hear that the chorus wasn’t working. I measured the voltage on the M174’s negative input pin as +0·8. The negative outputs at the PSU are both around −1·2 and a bit wavery. Back to the start.

(I’m seriously considering just building a whole new power supply at this point; it might be less hassle.)

I tried again with the M174 board out, with similar results.

We Have Bad
⁁Contact

Just as I was looking at it and thinking of switching off, wondering what to do next, a rather unusual thing happened. I might want to emphasise that. A rather unusual thing. Happened. Which is, I saw some tiny sparks between the F2 (negative side) fuseholder and the fuse cap. I watched for a few minutes, trying to absorb what I was seeing; arcing seemed to be happening intermittently with gaps of around twenty seconds. Now, I was thinking of replacing these fuseholders once (if) I got the thing working anyway, because though I’d wirebrushed the visible corrosion out they still didn’t look good for the long term. This seems to confirm that it’s a good idea. Even if it isn’t . . . well, don’t want to say anything about that yet.

So, fuseholders replaced. [IN13 closed.]

I connected the PSU and switched on. Nothing. No voices at all. I had a look at the fuses; F2 is blown. Well, maybe that had happened already? I replaced it and switched on. This time I saw it blow. What then? I disconnected the power output from the PSU, replaced the fuse, switched on. No sparks. Test the power at the output pins. 1: +26·8, 9: −25·8. Almost exactly what the K4 schematic suggests.

This shows, I’ll say it now . . . it is the problem. Or part of it. I could have probably got this far weeks ago if I’d just replaced the fuseholders when I first thought of it. And now there’s a new problem — it’s drawing too much power somewhere, which may not be related to the original lack of negativity. But it’s conceivable this is inrush from the capacitors that had reverse voltage recently?

Since the M174 seemed to affect things previously, I’ll try with it out. Results — good, more or less what I had with the bench supplies, only, of course, the chorus effect isn’t working.

Next? The easiest check is whether it was the normalisation of the capacitors or some other temporary condition which should have been resolved by now. At worst, it just wastes another fuse. So, switch on again . . . 

. . . well, what d’you know? It’s working.

That is, almost everything is working. That certainly resolves the underlying fault with the pianos voices. [IN19 closed] With some experiment, the strings sustain is a bit shorter than I’d like, perhaps, and I tried an optical volume pedal I’d picked up recently, but it didn’t work, so I’ll examine that later. Other things that still aren’t working, as before, are pianos B1, B2, B3, and the sustain length in C♯2 and F♯3 [IN21], and the bass voice [IN22]. Conversely the noise in the upper keys’ Cello seems to have diminished notably. [IN20 closed] And there is no significant warm-up delay any more. Voltages stabilise almost as soon as switched on. And the AM radio effect seems to have gone.

In summary, while some of the components I’ve looked at weren’t in great shape and the instrument may benefit from their replacement in the long term, and DZ2 almost certainly needed replacing, the underlying issue was that the remaining corrosion on the fuseholders was allowing continuity, but losing me ten Volts. I’ve never seen that happen before, but I’m in no danger of forgetting it.