Polivoks repair: dead outputs, 120 volt conversion, scaling issues

This unit came to me almost completely dead: neither oscillator working, just a power light and a slight hiss.

Bad paper and oil filter capacitors were the ultimate cause of the dead oscillators, but once they were replaced there were serious issues with the keyboard tracking.

At this point it became tricky since the only schematics that seem to be available are in Russian!

I made some headway by replacing the op-amps on the keyboard circuit board. I used a reference table found here,

Russian Ref	Europa/USA Ref	Function
KR140UD8B	uA740 , uA741, LF351,...	Single standart AOP
KR140UD12	uA776, MC1776, LM4250,NTE888	Programmable low power AOP
KT315G	2N3904A	NPN transistor

which states that the russian op amps cross with uA741's, which are super standard monolithic op amps.

This fixed many issues, however, there was a gradual detuning of every note if the sustain was turned up.

With no circuit descriptions (that were in English), I was ultimately able to use the international language of schematics to find the sample and hold circuit on that keyboard circuit board. It consists of a fet, and an op amp and a few capacitors. Essentially, while the key is depressed the voltage flows through the fet and charges the .22 capacitors. Then the key is let up, the fet opens up and the capacitor remains charged to the same voltage, and continues to flow through the op amp (a7 on the schematic below). Since the op amp has super high impedance, the capacitor discharges so slowly that it holds at the same voltage for a long time, and the op amp output stays at the sampled voltage for quite a while. A good description is available here.

The sample and hold circuit thus demands a fet based op amp, so that it doesn't draw too much current from the capacitor and cause the sample voltage to decay too quickly. The 741 op amp worked perfectly everywhere else as a pin for pin substitution for the russian op amp, but not in this sample and hold circuit!  I replaced it with an original russian one, and everything worked fine. (I am sure I could have replaced it with any fet based op amp, but I had russian ones left over since I had replaced quite a few on this circuit board, having use a shotgun approach.)

Interestingly, here is a website which shows a sample and hold circuit using the 741... notice the addition of the 10k resistor, which probably helps the 741 work without drawing too much from the capacitor. 

Here is a picture of the circuit in question

Once all this was settled, the keyboard worked well. I installed the mods located here for external CV and Gate, and I also swapped out the transformer with one that can convert 120 volts down to the 34 volts, center tapped, that the polivoks needs for power (thus you can run it off from regular wall power). I installed an IEC connector and it is all set!

Rheem Mark VII Organ Repair

 

This gorgeous instrument would shut down unpredictably, or sometimes there would be excessive distortion on the outputs. My first suspicions were with broken solder joints, or perhaps failing electrolytic caps in the power section or on a circuit board. Sure enough, over time, I was able to detect low voltages from the power supply, as well as some AC on the DC rails. I replaced the filter capacitors, the rectifier, however the problem still persisted. This was an unregulated power supply,so there wasn't that much to replace or check.

I then disconnected,  board by board, each of the various circuit boards from the power supply, to see which area was loading it down, but there seemed to be no particular issue with any of the different rails.

This was a real mystery, until I decided, finally, to check the resistance across the terminals of the power switch. Although the switch was working mechanically...opening and closing as it should, there was a 30 - 70 ohm resistance across the contacts of the switch even when the unit was turned on.

I was able to disassemble the switch, and spray it with some cleaner and used my new fiberglass cleaning brush to clean the contacts, and all was fine. It has not usually been my experience that AC power switches can fail in that manner, but after 30 years or so of service, I suppose anything can fail.

OBERHEIM TVS-1 DIGITAL KEYBOARD AND SEM REPAIR

Although the schematics are easily found for the TVS-1, an actual circuit description of the digital keyboard circuitry is hard to locate. Having just fixed issues with this circuitry in two separate synthesizers, I decided to write my own.  In case anyone else is crazy enough to delve into this and sort it all out, this could be useful information, even if incomplete.


There are four purposes to the digital keyboard circuitry: one is to provide the correct  control voltage "cv", a voltage which will produce the correct frequency oscillation in the SEM module(pitch), the second produce is the "gate" voltage, a voltage which tells the SEM card that it is in use... enabling it to produce sound. The third purpose is to activate the sem cards in the right sequence (in unison, or right first and then left, etc). The fourth purpose is to astonish and stymie the repair technician with a brilliant, yet complex, array of logic gates.

The circuitry is designed to process two note polyphony, so after holding down one note, any second note played is sent to the other sem module.

This circuitry uses CMOS chips with a supply voltage of 9 volts... unlike many later logic chips (TTL) which only work with up to 5 volts.

1) the heartbeat of the circuit is A8, which is designed to produce a clock pulse which drives the other logic gates. It uses some RC components and a 4001 logic "or" gate. It's output, labeled "clock", should like like a clean square wave.

2)this "clock"  pulse is routed to A8, a "counter" device which sequentially routes pulses among 6 outputs, labelled "a1 -a6".

3) These signals (a1-a6) from A8 are routed through a1, then the keyboard, then a2, which essentially produces a type of serial output which contains information about which keys were pressed. This output is shown in figure 1... it essentially is a serial synchronous data signal, which produces voltage spikes which reflect which of the 37 keys were pressed. Ultimately, these spikes lines up with the "dac" signal, and depending where the spike falls along the dac signal, an analog voltage is produced for EACH key that is pressed, these voltages are higher if the higher keys are pressed. Note that the voltage spikes are recurring, synchronous with the DAC signal, and thus line up in the same spot each time with with the recurrent dac signal, as long as the key is held. See figure 2.

4)Now the beauty of the design can be seen... an analog voltage for EACH key can be produced, and by manipulating the "dac" signal, you can make that voltage be higher or lower, or if necessary you can curve the dac signal slightly making the voltage difference across the keyboard non-linear.


5)the "dac" signal is produced using the outputs "a1-a6", by Ic's a16 and a17, and a series of resistors making a voltage "ladder"... an old fashioned "dac". The signal,when all is right, should be a 3 volt sawtooth as seen in figure 1 and 2. Conveniently, the signal can be modulated up and down by 3 volts by using the transpose switches, and adjusted using the trimmers on the board. The length of one "DAC" wave corresponds exactly to the time period of the output on the "data" rail.

6)It is useful to note here that any issues in the circuitry described thus far will produce intonation problems for BOTH left and right voices.... issues with the "clock" signal, the "dac" signal, or with the circuitry used to produce the "data" signal.
Most of the next section addresses the circuitry that effects each voice seperately.



7) "LGate and Rgate are signals which are either high or low, low signalling the "gate is open", and the voice is in use.

8) "RTAKEN and LTAKEN are signals which go High when the right or left voice is taken.

----
When a key is pressed on the keyboard, a voltage spike is present on the "data" signal. Let us suppose that left voice first is selected. In this case, A11 pin 12 and 13 will be high, allowing  pin 11 to go low.

9) A12 pin 13 can now go High, which represents a clock pulse to latch A6.

10) The whole purpose of A6 and the network around it, is to "record" the position of the data pulse at that  time of the key press, and to repeat it each DAC signal. A6 takes that recording when a12 pin 13 goes high, which moves the clock on the latch. After that, at each  point when data lines a1-a6 are in that position, the network of logic gates a5,7,9, 8 and 11 will create a pulse which should line with the dac signal in the right position, so that the correct CV is produced by A18 pin 2. This voltage is stabilized by the small cap, and then run through op amp A6, which does not produce a voltage gain, but a current gain.

11) The gate is opened (goes low) on the first press of a key when a12 pin 13 goeshigh, a14 pin 4 goes low, then a15 pin 10 goes high. This, when synchronized with a data pulse at A14, causes a14 pin 12 to go low (and a14 pin 13 to go high. This happens at the first keypress, and initially opens the gate. After that, A15 pin 11 goes high with each successive DAC cycle,a15 pin 4 goes low, and a15pin 10 goes high.  and as long as there is a concurrent data pulse at 14 pin 9 (ie the key is held down), L gate stays open, and a sound is played from the left SEM. when the key is let up, the data pulse is no longer present at pin 9, and the gate is allowed to go high again. now a11 pin 13 is high, and the process can start again.

The process is essentially mirrored on the circuitry which corresponds to the Right Gate.


========================

Some other things:

Some issues with a weak resonance effect in the VCO's came down to a weak 3080, a transconductance op amp.

The customer was also concerned about driving the sequencer with an external clock,which did not produce the correct voltage for the TVS logic... for example a standard TTL square wave, from a 5 volt clock,  may not drive the TVS CMOS logic chips properly.These chips are designed to see logic highs above 7 volts (see http://www.allaboutcircuits.com/vol_4/chpt_3/10.html). Thus in the signal path between an external clock jack which I installed,   I added the small circuit below, which was effective in buffering the external clock input, and allowing it to accept a wide range of voltages. The second transistor is added to make the output square wave in phase with the input (as the first stage turns it upside down). The high value collector resistors mean that the circuit wont draw much current... you may notice it is not a wonderfully designed amplifier circuit, and more biasing could be done... however the transistors are either off or completely saturated, so I think all that extra biasing would be unnecessary.

Fatar SL-880 Repair - replacing crystal with oscillator

This keyboard would not turn on reliably. I did at first attempt a hard reset, pressing UP/DOWN/ENTER while turning on. This had some effect, but it still frequently failed. I also tried the instructions referenced here, which possibly helped but not completely.

I did find that when it was locked up there was no clock pulse around the crystal... however replacing the crystal had no effect. After downloading the datasheet for the micro controller: ST9036, I found that the crystal could be replaced with an oscillator, a device which creates a pulse on its own, eliminating the need for much external circuitry.

I had a 24mhz oscillator around, and connected it to ground and to the 5 volt regulator. I removed the old crystal and the small value caps that were around it (they are not needed with an oscillator) and made sure that the output of the oscillator went to the "osc in" pin on the microcontroller.

Somewhat surprisingly, this worked perfectly!


I glued the oscillator to the board, and the unit has worked fine ever since. It is not a pretty repair, aesthetically, but it does work!

I tested the idea with a breadboard first, connecting the osc output to the microcontroller with alligator clips

The oscillator is similar to the crystal, but needs to be powered with 5 volts, and doesn't need to have capacitors around it

The fixed board has an oscillator glued to it where a crystal once was.

ARP 2600 Repair: Blown Tantalum CAPACITOR, dead 15v rail!

This nice instrument came to me completely dead. One of the trimmers on the PS board was cracked... but even when replaced one of the rails (-15v) was dead.
The 2600 is actually fairly easy to troubleshoot...the issue boiled down to a shorted tantalum capacitor, visible in the second photo.
Tantalum caps are easily burnt out... I am not sure if the broken trimmer caused a spike which burnt out the capacitor, or if the shorted cap caused the trimmer to heat up and crack.

The burnt out cap is shown below.

Farfisa Mini Compact Distorted Divider

This Farfisa came to me with distortion on particular notes: the D's, except for the highest, were distorted.

The first step is dis-assembly. To get inside one of these, you remove the plastic cap on the top first (the red plastic thing). That is easy enough... just  a few screws.

Then you get underneath it (have it standing on its legs) and remove all the thumb screws (5) that are hanging down. You can then lift the whole chassis up and out of the wooden case(but you can't lift it very far, just high enough to sit i on top of the wooden case, as shown below.

Oscillator boards, which are accessible through the back, when chassis is lifted from the wooden case...

Looking at the distortion with a scope revealed that the 'divider' circuits were producing a modulated oscillation, they were not cleanly dividing the incoming wave... they were erratically triggering and including erratic harmonics which didn't belong.

This was a difficult one to troubleshoot, because the problem was somewhat intermittent.

I made some progress by replacing the tuned caps for the effected divider, C3a and C3b, especially when I used appropriate polystyrene capacitors (good and stable for oscillator circuits). But this just made it a bit less intermittent, but did not solve the problem.

I replaced the germanium transistors for that divider, thinking they might be noisy but this did nothing.I have seen them fail too, but when they do they cause silence, or they cause the divider to output its input tone and not divide.

Ditto for replacing the 1uF coupling caps which output the divided signal to the key contacts (circled in yellow below). I have seen these fail as well, but they either cause the divider output to go silent, or distorted in a different way: they don't seem to interfere with the frequency of the divider's output when they fail, and the distortion I was witnessing was a modulated frequency.

I did notice that putting a high ohm resistor across the caps (okay I will admit it was actually my thumb) that couple each divider to the next one downstream would cause the distortion to go away. Ditto for a high ohm resistor coupling one divider output to another! I realized at that point that the caps which couple one divider to the next were the culprits.

Those I have circled in red in the schematic below:

You will notice that the schematic calls for different values for these caps, between the different dividers (470pf, 1nf).

 I have on several occasions found these boards with ONLY 1nF caps installed as coupling caps between all the dividers, and they work fine. I am not sure if this is something Farfisa changed as time went on.

Anyway, adding a 1nf Polystyrene cap between the dividers fixed the problem.

Perhaps that original coupling cap, between the dividers, drifted over time and with heat... and did not feed the next divider downstream a good steady waveform to divide.

Anyway, I hope this helps anyone working on this issue!

Another blog by another guy who deals with this type of stuff and seems to have maintained his sanity is here.

Elka Synthex Repair

This Synthex came to me for two reasons... primarily for the replacement of some very bad potentiometers, which was pretty straightforward after I found some adequate replacement ones from Digikey, and secondarily due to some issues with the Square Wave output from the LFO.

This latter issue required a bit of diagnosis, an endeavor which was impeded by the strange (Italian) design of the Synthex. I came to find out that the Synthex is particularly hard to work on because 1) the schematic lacks any labels or circuit descriptions, and 2) the circuit boards are essentially piled on top of one another, so you cannot insert a test probe onto any circuit boards except for those on the very top.

As far as dealing with the schematic, I was able to find the LFO output on the 5850 circuit board, after searching for familiar patterns on the schematic. I found that the various LFO output shapes (sawtooth, sq wave, etc) are routed through a few 4066 IC's on that board, on their way to the oscillators or other point they can effect.

 As far as dealing with the inaccessible circuit boards,  I was able to do some testing by soldering tiny wires onto various points of the board, running the wires out through the tiny spaces between the boards, and then connecting my scope probe to those wires. This is an arduous process, but it does work.

Ultimately I found that the square wave LFO was getting into, but not out of, IC 2E, which is a 4066 switching IC. .I found that the reason was that IC 2A, a 40174, had a dead output, and was not activating the switch, and not connecting pins 8 and 9 of 2e as it should. Replacing IC 2A solved the problem.

I now understand why Synthexs' have a reputation of being difficult to fix!

The circuit boards, except for the top two, are impossible to access with a probe... so I had to pick some test points and solder small wires to them, and then reassemble and test.

The 5850 Circuit board, where the problem was... IC 2E has been removed here.

These Digikey Potentiometers were pretty close replacements... Linear 5k... however the shaft was a bit long and also the diameter of the busing slightly small.. but the shaft was easily cut, and the pots fit securely when tightened up.

The square wave was detected going into IC 2E right here... but not coming out the other side! The problem was a dead output on IC 2A.

The problem ultimately was not the with IC2E, but with the flip-flop IC 2A which was controlling it.