MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

(No longer) stumped by VCS3 filter problem
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3  Next [all]
Author (No longer) stumped by VCS3 filter problem
Scot Solida
very frustrating very frustrating very frustrating

My VCS3 developed an issue recently and we (my techie son) and I can't seem to get to the bottom of it.

The trouble is that the outgoing signal is massively amplified - well beyond what it should be. For example, and incoming signal measured at 1.8 Vpp or so, passes through the filter, but is spat out the other side at a whopping 14.4 Vpp (when resonance is down and cutoff wide open)

Oddly enough, as far as we can tell, the filter seems to be operating normally in every other way. And in fact, the filter's own self-oscillation comes through the output at the expected 1.8 Vpp. It's only the incoming signal that is massively amplified.

We've checked all the obvious (and some not so obvious), such as:

-PR6 calibrated to produce 261Hz when Response (resonance) is full and Frequency (cutoff) at half
-the diodes themselves (in fact, they've been replaced)
-the caps (also replaced the electrolytics).
-trim pots, the panel pots (replaced the former as a precaution).
-we searched for broken or bridged traces
-checked continuity throught the circuit
-Q26 (2N5163) has been replaced twice. Each attempt used JFETs selected with pinch off voltages within the 2.5-3.5 V range specified for that transistor.
-R84 and R85 have been replaced. The voltage after this divider is 300 mVpp, assuming an input of 1.56 Vpp (measured before R84).
-R93X, Q29, and Q30 have also been replaced.
-All of the voltages on the test points on the schematic match up fairly well with what we're seeing.

And yet... The base of Q31 shows 14.8 Vpp.

Any help or suggestion would be most welcome. We are absolutely flummoxed, and I desperately want to get my VCS3 back into action. A screen grab of the filter schematic:

Synthiq
Without the feedback network R88 and R87 in parallel with C37 and Q26, the gain is even higher than you see now so my guess is the problem is related to these components. If the dc level at the emitter of Q31 looks reasonable (=close to 0V), I think R87 and R88 are fine, too. One test you can do is to remove C37 from the circuit to see if the gain returns close to unity again. If this is the case, the problem is most likely limited to the area around C37 and Q26 and the biasing of Q26.
Scot Solida
Thanks for the suggestion. The dc level at the emitter of Q31 is pretty much 0V, yeah. Omitting C37 produces self-oscillation at a level of 1.8 Vpp, which is what the self-oscillation has always been. The self-oscillation has always been pumping out at the expected level - only the signal coming in is being grossly amplified by the time it hits the output.

confused
guest
R85 may have come loose (the input isnt being attenuated properly).
Scot Solida
Thanks. R85 has been checked, replaced as a precaution, rechecked, but to no avail.
guest
what is the AC voltage at the base of Q22 with normal signal applied? the signal is either coming in too hot, or its getting amplified too much. if the feedback oscillation is working ok, then its probably not being amplified too much. does the resonance knob have a reasonable response for its travel?
Scot Solida
We get roughly 280 mVpp at the base of Q22, with an input of 1.33V measured from the input at R84.

With PR6 calibrated to produce a self-oscillating tone of 261Hz with the Frequency knob at ‘5’ (as per the service manual) the range of self-oscillation when sweeping the Frequency knob goes from 15Hz to 24kHz.

The Max level of self-oscillation signal is 1.8 Vpp at the output, just as we’d expect.
guest
ok, that sounds right (at least according to the schematic). what are the AC levels at all of the emitters of Q27-31 with a signal applied? usually, the input to a differential amplifier is 20mV to 50mV, and above that gives a lot of distortion, so the feedback loop must be compensating in some way. unless of course there is something further up the chain that is missing, like a high impedance on the output stage of the thing thats supposed to drive it.
Scot Solida
The emitter of Q27=525 mVpp
The emitter of Q28=242 mVpp
The emitter of Q29=968 mVpp
The emitter of Q30=500 mVpp

The emitter of Q31=13.8 Vpp
The base of Q31=18.1 Vpp

This is assuming an input (sine) of 1.31 Vpp at a frequency of 263Hz, Response (resonance) all the way down, cutoff cranked full.
Synthiq
I have done some LTSpice simulations of a somewhat simplified circuit and was able to set it to oscillate at 261Hz with an amplitude of 1.5Vpp if C37 was removed.

With C37 in place and a 200ohm resistor in series to represent the typical JFET on-resistance when Vgs=0, I also saw a very distorted 14Vpp output signal when the input was a 1.8Vpp sine wave.

With C37 in place and a variable resistor in series to represent the JFET on-resistance, the maximum gain was 30 times when the resistor was 1ohm but the gain dropped to 17 when the resistor was set to 200ohm. The gain can be reduced further by increasing the resistance but was still 2-3 when the resistance has reached 1.5-2kohm and the self-oscillation started.

So it looks to me as the circuit is working as expected and I have no answer to why it didn't behave like this earlier but maybe something was preventing the JFET from turning on fully. If this is a problem, you can try to place a few hundred ohm in series with the JFET to limit the maximum gain or limit the voltage to the gate to some value less than 0V.
Scot Solida
Thanks for the reply. cool I am fairly certain that the circuit is not working as it is supposed to be - unless you mean Q26 or the feedback circuit specifically, which might very well be doing what it is supposed to do. The filter output, however is far, far louder than it ever has been, and far louder than any other output level on the machine. Oddly, the filter output is one of the few that isn’t specified in the manuals. However, the ring mod and reverb sections, for instance are expected to put out a max level of 6 Vpp and 5 Vpp, respectively.
guest
i was scratching my head about that JFET. since there is no DC current path, its being used as a variable resistor (as pointed out by synthiq), but why run the audio signal back through it then? without peak detection that will make for asymmetrical gain. im sure that sounds good, but is that the point? i figured it would be some sort of AGC. but, lower resistance of the JFET gives higher gain, and a higher input signal gives lower resistance, so there isnt negative feedback. or maybe i have this backwards because of the phase shift around the filter. for DC, a lower resistance is higher gain, but for AC a lower resistance is lower gain.
Scot Solida
Are you you referring to the gate of Q26? We've been wondering about that as well... Our thinking was that this is being done to linearize it in a way similar to what is described in this application note: https://www.vishay.com/docs/70598/70598.pdf

Maybe this isn't the case though.
Synthiq
Scot Solida wrote:
Are you you referring to the gate of Q26? We've been wondering about that as well... Our thinking was that this is being done to linearize it in a way similar to what is described in this application note: https://www.vishay.com/docs/70598/70598.pdf

Maybe this isn't the case though.

That seems reasonable to me too, but the implementation is different as it doesn't use the drain voltage but the output voltage. So the ratio between the drain signal and feedback signal will depend on the JFET on-resistance and will not be constant, which I thought it should be.

quest wrote:
i was scratching my head about that JFET. since there is no DC current path, its being used as a variable resistor (as pointed out by synthiq), but why run the audio signal back through it then? without peak detection that will make for asymmetrical gain.

The gate voltage for a NJFET is measured between the gate and the most negative side of the transistor, not the source. So with a positive drain voltage the gate voltage will be fixed but for a negative drain voltage the effective gate voltage will increase as much as the drain voltage. By adding 50% of the drain voltage the effective gate voltage will at least increase the same for positive and negative amplitudes.
guest
Synthiq wrote:
Scot Solida wrote:
Are you you referring to the gate of Q26? We've been wondering about that as well... Our thinking was that this is being done to linearize it in a way similar to what is described in this application note: https://www.vishay.com/docs/70598/70598.pdf

Maybe this isn't the case though.

That seems reasonable to me too, but the implementation is different as it doesn't use the drain voltage but the output voltage. So the ratio between the drain signal and feedback signal will depend on the JFET on-resistance and will not be constant, which I thought it should be.

quest wrote:
i was scratching my head about that JFET. since there is no DC current path, its being used as a variable resistor (as pointed out by synthiq), but why run the audio signal back through it then? without peak detection that will make for asymmetrical gain.

The gate voltage for a NJFET is measured between the gate and the most negative side of the transistor, not the source. So with a positive drain voltage the gate voltage will be fixed but for a negative drain voltage the effective gate voltage will increase as much as the drain voltage. By adding 50% of the drain voltage the effective gate voltage will at least increase the same for positive and negative amplitudes.


thanks for the explanations, this makes sense to me. as its an old synth, perhaps they didnt have the linearization scheme fully worked out yet, and this was the best they could do. there are a lot of variables in there, both for the linearization, and setting the JFET to the right resistance based on its parameters, and no trimmers. seems like it could easily be at the wrong operating point. perhaps the transition between full resonance/oscillation to no resonance is super sharp, and the majority of travel is in the full damped postion (JFET essentially shorted), which would give a really high gain.

is there a position of the resonance knob, just below oscillation, where it works correctly?
Scot Solida
Thank you both... this discourse is helping us get our heads around the issue, I think.

Quote:
is there a position of the resonance knob, just below oscillation, where it works correctly?


Actually, you may be on to something here. There is indeed a point... right around self-oscillation (which does in fact seem to pop in quite quickly) where the incoming signal is actually a bit below unity. If I crank the cutoff full, the self-oscillation goes higher than the audible range, and I get something like what I'd expect to hear. I think maybe this way the answer may be found...
Scot Solida
Unfortunately, we (my techie son and I) have gotten no further with this. We tried experimenting with Q26, replacing it 2n5163s that had higher and lower pinch-off voltages. Both of the alternatives were within (but at the oppoiste extremes of) the ranges specified on the schematic. The only significant difference was a shift in the position of the Response (resonance) pot where self-oscillation occurs. hmmm.....
guest
ive been thinking about this a fair bit, and i think im with synthiq at this point, and am wondering if this circuit can function right. the input signal is really high, such that it needs negative feedback to have a reasonable output level. the only way to get this, is to have a lot of feedback. this is fine, until you get to higher frequencies, and phase reverses, and now its positive feedback. then the large feedback causes resonance. so, to not have that, you need small amounts of feedback, which give large gain. is it possible there are other components in the cabinent that we arent seeing on the schematic?

if you like, like try clipping a 470ohm resistor across R85 and see what it sounds like.
Scot Solida
Quote:
is it possible there are other components in the cabinent that we arent seeing on the schematic?


Alas, no... this is it.

Quote:
if you like, like try clipping a 470ohm resistor across R85 and see what it sounds like.


I think my son may have already tried that, but I'll ask him. If not, we'll give it a go...

I am wondering if we've been focusing on the wrong place... We've been suspicious of Q22.

I hope we get to the bottom of it soon... I miss my little space traveller. waah
guest
have you traced the signal through the whole synth yet? input a sinewave, and look at the amplitude at all of the transistor emitters? if there is a problem with Q22, it will show up as a reduced or distorted signal on Q27. the signals should also look identical, but of opposite polairty, on all of the transistor pairs. it possible that something is not right in the diffamp, and the gain is too high. you can check how much it amplifies at each stage. and maybe check the signal on either side of C38, just to make sure its at the right level, and making it across the capacitor.
Scot Solida
The base of Q27 (around 250 mVpp) is higher than Q28 (around 500 mVpp). This is making us suspect Q22 and Q25... At lower input levels, Q27 is still higher than Q28 as well. Nothing appears to be distorted and both appear to be nearly identical other than level and phase.
guest
Scot Solida wrote:
The base of Q27 (around 250 mVpp) is higher than Q28 (around 500 mVpp). This is making us suspect Q22 and Q25... At lower input levels, Q27 is still higher than Q28 as well. Nothing appears to be distorted and both appear to be nearly identical other than level and phase.


do you have the values backwards here? 250mV > 500mV? either of those seem really high for not having any distortion. the fact that they are different makes me suspect Q27 or Q28. if you turn the cutoff frequency all the way up, do the levels become more equal?

are all the DC levels correct? is there 0.6V drop across each diode?
Synthiq
guest wrote:
Scot Solida wrote:
The base of Q27 (around 250 mVpp) is higher than Q28 (around 500 mVpp). This is making us suspect Q22 and Q25... At lower input levels, Q27 is still higher than Q28 as well. Nothing appears to be distorted and both appear to be nearly identical other than level and phase.


do you have the values backwards here? 250mV > 500mV? either of those seem really high for not having any distortion. the fact that they are different makes me suspect Q27 or Q28. if you turn the cutoff frequency all the way up, do the levels become more equal?

are all the DC levels correct? is there 0.6V drop across each diode?

I have noted something similar in my simulations but wasn't sure if it was caused by the simplifications I did to the circuit. I see a larger amplitude if the dc level is higher as guest also suggested. My interpretation of this is that there is an inherent offset in the circuit and the feedback corrects this by unbalancing the diff-stage so the dc current is different on the two sides causing the dynamic load impedance to be different as well. So lower dc current -> higher dc voltage and higher load impedance -> higher signal swing.

I use a 100mVp input signal and get clean sinewaves.
MikeDB
I would guess that R97 is faulty, or failing that R93. Check both with a multimeter.
Scot Solida
MikeDB wrote:
I would guess that R97 is faulty, or failing that R93. Check both with a multimeter.


Thanks for stepping in to our mystery MikeDB. Which R93? (there are two, one is labeled R93X and that one I've already tested. My son may have done the other... I will check with him and we'll do it if he hasn't.)
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3  Next [all]
Page 1 of 3
Powered by phpBB © phpBB Group