hardware details

On Youtube I found a Stylophone 350S teardown and repair video by Marco Reps, that shows yet another chipset version (revision "6510 B"?). His (already replaced?) amp IC SN76018 had a SIL pinout and he has reverse-engineered some of the ciruitry (showing schematics). He didn't answer e-mails, but in his video he told that a shorted capacitor killed his amp IC, so he rewired it to a TDA2030. Also his different rocker switch type (with wiping contact) does not seem to fall apart when removed from the PCB.

Stylophone 350s Vintage Synthesizer - Teardown, Repair, Reverse Engineering
https://www.youtube.com/watch?v=vFdiA97dq3k

My octave divider is the CMOS 7-stage ripple counter "Motorola MC14024BCP, M 9 38" (14 pin DIL) as the octave divider. It is basically a flipflop chain that halves the VCO frequency in each step. My power amp IC is "National Semiconductor LM380N, /925" (14 pin DIL). Despite the 2 huge 9V battery compartments this model can run very well on 10V (tested with halfway empty 9V transistor batteries). Both batteries are wired in series (no intermediate tap) for supposed 18V, and there is also a 3.5mm power supply jack (labelled "BATTERY ELIMINATOR", positive center). I strictly recommend to feed it with not more than 10V to 14V DC ("6V" or "7.5V" setting of unregulated PSU) to avoid damage.

The foam rubber on battery lids were very crumbly and fell apart by any small touch, thus I had to vacuum the residues out. Also 2 Piher trimmers had strange white blooming (like mould fungus) on them that could be easily wiped off; may be the plastic sweated out some chemicals. This may be by battery leak vapours, but another Piher pot looked ok. The volume pot is made by ITT with white plastic package. caution: The rocker switches are held by each 2 screws at the PCB back and fall apart when removed. Thus to avoid mess, only remove and clean one switch at a time and hold it with the other hand during removal. These switches contain up to 3 metal contacts, each with ends sticking in 2 holes of a L shaped sheet copper leaf spring. Look carefully how they are placed (if in doubt, carefully unscrew another switch to compare). Clean the contacts with cotton swabs and isopropanol. The opposit contact side are solder nails on the PCB. These may fall out or misalign if you try to resolder them without switch in correct "on" position holding these nails down. (I accidentally broke a leaf spring when I tried to bend it higher; fortunately it could be soldered back together.)

Strange is that (seen in revision "6510" and "6510 C") at one side of the "reiteration" switch the trace shorts all 4 contacts (making that contact useless) in a strange curly shape that looks like when originally the lower contact was designed to be separate and changed later to fix a bug.

important: reduce operation voltage

I conclude that the 18V operation was only done because the at that time only available zinc carbon cells had strong voltage drop when running empty. Suitable voltage regulators did not exist in 1970th or were expensive discrete circuitry, so the maximum possible initial voltage was used to extend battery life. But perhaps the manufacturer also wanted to squeeze out as much output power as possible through that tiny amp (installing a heatsink on MFC6070 may be a good idea if it runs hot). Reducing voltage may cause distortion at maximum volume setting, but unlike 1970th nowadays it is anyway uncommon to play instruments loud on stage through a that small internal speaker or amplifier (and who wants to use an external 35 Ohm speaker?!). It also may be that semiconductors were simply too new technology, thus manufacturers did not know that they will decompose from operation near their maximum possible voltage despite they work well in short laboratory tests. In modern IC datasheets are often sentences like "Product is rated for operating voltage of 15V. Absolute maximum voltage of 18V may not be exceeded. Running near this voltage will reduce reliability and may cause premature failure..." Thus I conclude that by modern terms the MFC6070 would be only suitable for 15V and rated wrongly.

But the big warning on the official Stylophone 350S sales website (and the many websites copying from it) IMO sounds very much like propaganda to raise panic for selling expensive snake oil (like the "Stylophone ACM" module upgrade voodoo) for the sheer sake of moneymaking. Even owners of a 1st generation Stylophone 350S can certainly easily buy a cheap regulated 12V power supply to operate it without any risk of damage. Also installing a chain of about 5 1A silicon diodes (e.g. 1N4007, each dropping 0.7V) in series with the battery is a cheap and simple measure to reduce the operating voltage by >3V, which should be enough to keep the MFC6070 in its safety margin. I even doubt that the instrument is that ultra-rare; may be that Dübreq only counts their nowadays customer list as the total number of owners.

Unlike modern digital instruments, there are no unreplaceable special parts (ASIC, SoC, chips with firmware, displays etc.) inside, thus so long it looks mechanically reasonable, any broken 350S can be fixed by adapting standard parts.

theory of operation

The key contacts respond on connection to GND (that is what the normal stylus does), which triggers the decay envelope (if enabled) and plays a tone with pitch depending on the resistance of that connection (much like Gakken SX-150; due to their wiring as a resistor chain, pulling multiple "keys" to GND makes the higher note succeed). With "reiteration" (chopper) enabled, the reiteration stylus apparently pulls to GND through a pulldown transistor modulated by a square LFO. Because the other stylus stays always GND, using both together makes the tone only toggle between notes when the normal stylus is on a lower note (a multivibrator with 2 real toggling stylusses would have prevented this). To prevent waiting on a note, the chopper LFO starts only when its stylus touches the keyboard.

AFAIK no other Stylophone used this kind of dual-stylus mode. But the method how they are connected through a square LFO can be likely retrofitted to upgrade any cheap standard model.

For professional circuitbending the 350S has high potential. Because everything is analogue, potentiometers may be installed as drawbars or to change the LFO and decay speeds. Connecting a capacitor to GND at the decay switch makes slow attack. The LFO signals may be also routed elsewhere. Strange is that vibrato (intentionally?) fades in very slowly when the switch is pressed.