The Prophet 5, What a lovely Synth, when it works!

The number of Prophet 5’s coming into the workshop over the last few months has grown to three Rev 3.0’s, to say they come direct to the Bench would be untrue, they come into storage awaiting the time and the right moment to get some attention.

Over the last several months they have all gone into Triage, where they spend a day on the bench while we assess what is wrong and what it will take to put right, standard stuff for most synth repairs you would think (without the several month bit) but all of these have been broken for a long time and been in storage for a decade so we don’t feel too guilty about that.

The Rev 3.0 was the first of the P5’s to use the Curtis chipset and was a complete redesign of the synth compared to a Rev 2, but it did inherit a couple of (in hindsight) bogey factors from the Rev 2 era, firstly those ancient 2708 EPROM’s that require 3 different power supply rails, and those ancient and unreliable 1k X 1 6508 RAM chips that store the programs, that is 1 Kilobit for each chip and eight of those to make 1 kilobyte.. Two out of the three P5’s here have bad memory backup problems, the IC’s are almost unobtainium and very expensive if gettable, and old unreliable chips even if bought again have no guarantee of lasting very long.


The Sequential service manuals of the time are very good and well written and even informative, but with over thirty years under their belts, the old Prophets need a better system of voice assignment indication than “put your fingers on the VCO chip and if that voice goes out of tune, then that it the one being triggered”, That just isn’t good enough for a synth with multiple problems, especially oscillator tuning problems. The above picture shows our latest creation, a clip on voice indicator for Rev 3’s. It uses a 16 pin sprung IC test clip with a small stripboard with some high efficiency LED’s and is attached to (or over) IC 340. That chip is the latch that drives all the Gates and it made diagnosing a P5 with multiple voice problems considerably easier, and there were multiple problems on the first one to be finished.

The first one to go had to have…
The Keyboard Rebushed and totally cleaned, and all the contacts cleaned, realigned to make them all work.
All the the Pot’s taken out of the front panel boards, stripped, cleaned and relubricated and of course tested.

Everything was starting to look better when after another switch on, all Analogue functions of the CPU board were dead, no CV’s no Sample and Holds, no tuning up, nothing. After a couple of hours of careful Scoping it looked like IC 319 had died suddenly (the Prophet ailment yet again) and replacing it confirmed that it had. And then we get to the Voice card where there was one CEM3340 dead on osc1 on Voice 3 and very strange filter behaviour on Voices 2 and 3. A lot of Scoping and even some chip changes were done to discover that two of the CA3280’s were gone dealing with filter Polymod on Voices 2 and 3.

If not for the Voice indicator we built, then this would have been even more of a Headf**k than it was, and it was still quite one of those. The total chip count to sort this beast was 9, including two Curtis chips and two CA3280’s along with a handful of logic IC’s, as well as all of the power supply decoupling Tantalum bead capacitors being replaced with Electrolytics right at the very beginning.

Tantalum Bead Capacitors

There are long debates about the good and bad points about Tantalum capacitors, and there certainly are some good ones, but the bad point is that they can go dead short on power up of the synth, the exact reasons are unknown, some say that it is due to age and that Tants have a lifespan, but as it is solid state device (it has no liquid electrolyte in it like Electrolytics do, the clue is in the name), why should it suddenly expire. There is no denying that Tant’s do go short, we have replaced hundreds of measurably shorted ones in a huge variety of equipment.

The other side of the Tant is that it is used in the Audio path as well as decoupling the power rails on Prophets, and so far there has never been a problem with them in any machine when used in that mode, they have a limited amount of current and voltage fed through them due to the circuitry around them, and they seem to last forever. The sound quality of Tants in the audio path is definitely not Hi-Fi, tests that we did years ago with CD quality audio showed this, for Hi-Fi the Tants sounded awful, but probably their presence in a Prophet is part of the sound and if we ever found a dead one we would replace it with another Tant.

Our rule of thumb is that if one side of a Tant has a raw power rail from the power supply on one terminal and Ground at the other, then replace it with a low ESR electrolytic. Simples!
There are 19 of those in a Prophet 5 that we replace, 8 0n the CPU board, 7 in the power supply, 3 on the Voice card, and 1 on the left side of the front panel pcb, a mixture of 2u2 and 10 uFarad.

One other side of the Tant is that it has very low leakage, it doesn’t dribble its charge away with the gusto of most Electrolytics, there are a couple of Tants involved in the Battery backup system for the presets, they have never failed and have never been replaced, but if they were replaced by Electrolytics I would suspect that the lifespan of the new battery would be reduced considerably due to unnecessary capacitor leakage.

Many years ago we replaced the leaky Electrolytic capacitors in a Korg Polysix’s LFO and PWM sections with Tants because the calibration routines wouldn’t work with the original Caps, a few suitable Tants in there sorted that one right out, the issue being the slowest modulation frequencies, the leakage of old electrolytics just prevented the circuits from going down to 0.1 Hz or lower.

The Next Prophet

The next Prophet 5 back on the bench was another Rev 3 but with an entirely different set of problems, the keyboard was totally shot, and the Pots were very flakey, but that is not unusual for a 30+ year old Prophet that seems to have had minimum maintenance in that time apart from 17 bits of guitar string replacing the famous Pratt-Reed J-wires (thats a quarter of the keyboard).The J-wires themselves are getting scarcer to obtain and the price is going up accordingly, so I am experimenting a little with items obtainable fron the Jewellery world, Gold plated Copper wire, Solid Silver wire, obviously of the correct diameter and as it turns out, one of the standard jewellery wire sizes is the same as the J-wires in Pratt-Reed keyboards. I don’t know if they will be as good as the originals for a long period of time but the originals failed regularly and at least jewellery consumables will always be available, and relatively cheap compared to hunting down the originals, although there is a shipment of expensive original J-wires coming in (hopefully the last one).

The big fly in the ointment with this P5 is the battery backed up RAM, the chips are shot. If your create a few simple presets and store them (with a brand new battery installed) and then try to recall them a few minutes later, they will have changed noticably in a matter of minutes even with power on. If you switch the synth off and then back on again, the newly programmed presets will have changed some more. But this one got worse, Prophet chip death kicked in again, when DB6 ( a Data Bus Line) suddenly died across the entire synth, meaning that some buttons didn’t work, one in eight keys didn’t work, it wouldn’t tune and the ability to remember programmed presets was much worse than before. The logical thing to do first is to remove the relevent chips that are in sockets first, because if it is not one of those, then either tracks will have to be cut or IC’s cut out, neither of which are good things. It worked out that IC 307, one of the memory IC’s and responsible for DB6 in the RAM had failed, and after its removal the synth returned to a kind of normal where it would work fully in Edit mode, where all voices responded perfectly to all programmming functions, all keys worked and all switches did likewise. As two of the P5’s here seem to have bad 6508’s, we tried to design a slight upgrade to some later and more reliable family of RAM, the Hitachi 6116LP, they are 2k X 8 RAM chips and would in theory allow an extra bank of Presets, but we had to design a Daughterboard to be PiggyBacked on top of one of the EPROM’s, and while the result seemed good for about 20 minutes, there were soon some signs of memory corruption over time. The Battery was new, the new RAM chip was tested as good, there had to be an issue with the timing of the signals into the new RAM which while it is faster than the 6508’s. has different requirements to be reliable. A world wide Internet hunt for any help came to nothing, the result was, “we don’t know, good luck with that, when you sort it out tell us how you did it”.


We did design a PCB for this, but etching it at home proved to be very problematic, so for prototyping purposes we did it on some wirewrap style Veroboard, just to see if it worked, which it doesn’t. There will be no problem redesigning the PCB if and when we can make this work, the issue is the amount of time that will be taken up in testing to ensure absolute reliability

Actually two weeks later, after some research on the timing cycles and Truth tables of the three enabling pins on the 6116 (the old 6508 only has two) and trying to work out which of the three to not actively use, as the P5 inherently can only control two of them. First of all we went for the OE\ line, and this was not the correct choice, a detailed analysis of the Truth tables showed that the OE\ line was only effectual in 2 out of 8 logical possibillities for memory access, whereas the Chip Select function, CS\ was valid for at least 4 and was important to the timing of the memory write cycles also. Anyhow to get to the point, we rewired the daughterboard to use CS\ instead of OE\ and so far this has tested out perfectly, everything we program into the Prophet gets remembered correctly and multiple power cycles have not changed this. Strictly speaking, the Write Enable should be asserted at least 20nS after the Chip Select, and we half expected to have to insert a couple of 74LS00 gates in the write line to ensure this, but as the memory busses seem to run at 1MHz which is quite a slow speed even for old RAM, this thankfully seems unnecessary.

It seems that this synth is good to go but for one thing, the 17 Guitar strings in the keyboard, we have scrubbed and cleaned them endlessly but they are still unreliable, and we can only get 10 replacements from the US, and those are earmarked for another P5, the J-Wire shortage is proving to be a difficult problem, and as all of these three P5’s are all Rev 3.0’s with no MIDI, they still need their keyboards to work properly.

Despite the fact that this synth took a lot of work, the owner claims that this is a special sounding synth, so much better sounding than all the others he has played, and while testing it showed its calibration to be well out of spec, we haven’t touched any of the trimmers or tried recalibrating it at all, who wants a perfectly calibrated P5?

We could fully recalibrate the synth back to factory specs, but the client might not appreciate it.

One more interesting thing about this particular Prophet 5, until the RAM died, this synth was as it originally left the factory, no replacement Curtis chips, no new Op-Amps and no new Multiplexers. This is the only P5 we have ever seen in which this has been the case, every other one we have seen has had a long history of gradual failures sorted out accross the 30 years.

The Third Prophet

The third Prophet 5, which was the first of the three to arrive, will be the last to go, although a lot of work has gone into it over the last year or so, but it’s list of problems has been more interesting and difficult to diagnose than most. The original problem was that the presets in the first Bank were getting corrupted in a worsening way, every few switch power ups it would get worse, and then over time the whole contents of the RAM had turned to garbage. We put in a new battery, and replaced the 74C02 chip in the backup circuit with a new 74HC02 (which we believe to be equivalent in practise and far more reliable in time). Trying to program in a few Presets in the first bank didn’t work, the RAM didn’t remember anything at all. So it was logical that the RAM had failed, wasn’t it?

However before we got to debugging this, Sudden Prophet Death kicked in big time, the first failure was the 5MHz CPU clock module, a completely dead Synth. The only place reasonably convenient to get one of these from was Vintage Planet, who had inconveniently decided to shut up shop permanently. So we designed and built a replacement circuit on Veroboard with a 74LS00 chip on board to handle the Crystal and buffer it to feed to the LS74 that is the next stage. This worked for a short time, the synth tuned its oscillators, the A440 tone was spot on, and the synth mostly worked in non-preset manual mode, although the LFO was dead and the Filter on Voice 2 was just wrong. Replacing the CEM3340 in the LFO circuit cured that , but the weird Filter behaviour took a lot more to Debug, a replacement filter chip made no difference, as did testing all the surrounding components after lifting one end of all the resistors to measure them, and replacing the socket, and the Op-Amp associated with the Filter.

After a lot of time, we replaced the Filter Envelope chip and this fixed it, but we were fooled by the fact that the Envelope looked good on a Scope on an AC coupling setting, it was just when we measured the Envelope output from the chip with a meter that we found it was -8V and moving up and down correctly, whereas the others moved about 0V, so logically the chips output stage had failed. So far so good we had a basically working synth with dodgy RAM so we called a temporary halt while we researched a better replacement.

By the way, we had long ago replaced all 19 Tantalum Bead caps in the synth to try and stave off Sudden Prophet Death Syndrome (SPDS!), but this P5 still had a bit more suiciding to do, U319 failed, a 74LS138, and this meant total shutdown of everything, no keyboard, no switches, no tuning nada. Replacing this restored once again most of the synths basic functionality but another 74LS138 failed soon after. The P5 now seems stable for the last few months, as in nothing else has died since, but the Ram problem still remains. Some testing with Prophet number 2 on this blog and number 3 showed weird behaviour on the memory write lines on the no.3 and this points to a failure of the 74LS138 that is U318, evrything else about it is fine, but it looks like it enables writes to the NV RAM when it shouldn’t and whatever the contents of the databus at the time are inadvertently written to the RAM chips which explains the permanent gibberish in the NVRAM. We should have an answer to this shortly, when we can get back to it, and will report the results.

The results are in, we replaced U318 and the results are the same, so we will have to build a new memory board for this one also, everything else in the synth seems to work fine in every respect other than any semblance of memory retention.