303 clone and Q27 burn out (remedy?)

[Martin Czech]

> 
> I used to believe this also, but the 'character' of the 303 oscillator
> without any filtering is very noticable, particularly with the square wave -
> possibly due to the 'rounding' of the wave from the saw to square convertor.
> i noticed the same thing with the oscillator in the SH09 - I replaced the
> filter and vca in a 09 with a completely different design, but the character
> of the synth is still the same - which IMO must be down to the non-ideal
> waveforms from the vco.
> I suspect a few Curtis-3340-haters might agree here too ;-)
> 
> I would say the 303 oscillator contributes 40% of the sound, the filter 50%
> and the distortion in the VCA and elsewhere another 30%.
> (120% total due to synergy ;-)
> 
> 

I have now the schemo right in front of me, yes the "sqare" converter is only 
an overdriven pnp, this is really no ideal sqa.

And now to the scr discharger:

As far as I can see the operation is as follows:
Q24 is a saturated npn (diode), if a little leakage current flows
thru Q27, Q24 will establish a temp compensated base potential for
Q25. This potential is about teh emmitter potential of Q24 + 0.6V. This
is the reference potential for the whole SCR switch. The temperature
compensation will prevent thermal run away of Q25. If the timing cap
C33 is discharged by the expo converter such that the emitter junction
of Q25 gets forward biased Q24 turns on and subsequently Q27, which in
turn raises the base potential of Q25, this is the SCR snapback. The
snapback point would depend on the thermal behaviour of Q25 if there was
not the temp compensation with Q24. If the SCR has triggered the base
potential of Q25 rises to 12V (supply) minus the diode voltage of D25,
ie. about 11.4V (this is possible because of R101, which disables the
clamping action of Q24). I think the diode D25 is necessary to enshure
that the SCR will turn off quite fast, because the emmitter potential of
Q25 has only to rise to 12V - 2x0.6V =10.8V to turn Q25 off.  SCRs tend
to not turn off, they are still on at very little currents, so I think
the additional diode makes the shut off faster and safer.

I think the whole arrangement does only reach to the 12V supply because
of the shut off latency of Q25 and Q27, these transistors are heavily
saturated, it takes a lot of time to clear the base zone from all
carriers. This also causes the flat portion of the saw wave and the
time of this flatness should be independend of osc. frequency.

Could some 303 owner check this ?

I said that Q27 and Q25 are saturated when the SCR is on, ie. transistor
beta will be very low, maybe 2 or 1.5 or so. This means that a large
ammount of the transistor current flows through the base, especially in
the case of Q27 where there is no base resistance nor diode. If Q27 is
not suited for such an application, it will shure burn out. Most good
audio transistors with high Ft and high beta have shallow base junctions
that are very sensitive. So the problem of Q27 burnout could be addressed
with using (perhaps) lower beta, switch type transistor with very high
specified max base current. The same applies to Q25 and  Q24 (remember
tempco compensation, Q25 and Q24 should be the same type)


As I now understand the VCO operation, I'd say that it would be possible
to take a normal saw vco, add a monoflop thing for the flat time, adjust
level and offset in order to feed the "sqa" converter (copy this part of
the circuit) and it should sound identical. No need to use the SCR. This
is of course only valid if other (temperature) instabilities won't affect
for example the tuning. But since there is only one vco that should be
no problem at all.

The vca seems to be an ota, replacing it with an 3080 should make no difference,
if the resitor network gives the same amount of distortion. 

So there is only the filter left...  A funny circuit. A dual pot for
more output at resonance...  And the foot of the current source Q11
has no direct path to ground, as I would have expected.  It seems to be
fed by the current source (and AD output) Q39 via VR5.  This is really
strange... also VR3 seems to interact there...

Do you know the reason for this ?

m.c.