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

LTspice simulation of EMS VCS3 VCO
MUFF WIGGLER Forum Index -> Music Tech DIY  
Author LTspice simulation of EMS VCS3 VCO
dot matrix madness
Before I was going to build the VCO 1 of the EMS VCS 3 according to Graham Hinton's suggested modifications, I set up an LTSpice simulation of the circuit (see attachment). Following transistors were used:
2N3969 as replacement of the two ME9002,
BC547C and BC557C as replacements for the remaining ones.
Since I can't get any functionality I was wondering if the error is in the wrong choice of transistor (models?) or somewhere else.
Thanks in advance for any suggestions.
Synthiq
Compared to Graham Hinton's schematic, Q11 should be a npn, not pnp.

After replacing Q11, I replaced the +12V dc supply with a pwl to ramp the positive supply to 12V in 1ms. I noticed an oscillation that stopped just before the supply reached 12V. Ramping to only 11V, the oscillation continues during the whole simulation. My guess is that some component value has to be tweaked to get the circuit to work at 12V.

[Update:]
It seems like transistor Q1 is used as a zener diode but the breakdown voltage between base and emitter is not modeled correctly. Because of that, Q2 never turns on so the current in Q6 is limited to 12V/68kohm so the majority of current in R10 goes through Q7 and cause the oscillator to oscillate at around 60kHz no matter what is done with the VC inputs. When I placed a BZX84C8V2L zener diode in parallel with Q1 I got the circuit to work as intended, even at 12V positive supply.

To get the saw-to-triangle circuit to work, I had to switch Q11 back to a pnp but with the emitter towards the positive supply and change R26 from 22kohm to 16kohm. The duty cycle is still not 50% and the dc level is all wrong so the sinewave looks like crap.
Graham Hinton
Synthiq wrote:

It seems like transistor Q1 is used as a zener diode but the breakdown voltage between base and emitter is not modeled correctly. Because of that, Q2 never turns on so the current in Q6 is limited to 12V/68kohm so the majority of current in R10 goes through Q7 and cause the oscillator to oscillate at around 60kHz no matter what is done with the VC inputs. When I placed a BZX84C8V2L zener diode in parallel with Q1 I got the circuit to work as intended, even at 12V positive supply.


Funny that you discovered this because it happens in real life too. I had a customer with a Euro Oscillator module who changed some other modules in his system recently which altered the way the power rails rose. If the +12V comes up first one oscillator went to 30kHz, but not all three. Then I changed a 2C746 that was poorly tracking in another oscillator for an LM194 and that oscillator did the same thing. The fix was to use a 10V zener. One was not used in the first place because David Cockerell built his prototypes from a bag of untested transistors that he bought cheap down Edgeware Road and was too mean to go out and buy a zener, then didn't clean up the design for production. And they are still being made like that nearly 50 years later.

It doesn't normally happen in a Synthi because it is a fixed system that always powers up the same way.

These circuits are not my designs, so caveat emptor.

Quote:

To get the saw-to-triangle circuit to work, I had to switch Q11 back to a pnp but with the emitter towards the positive supply and change R26 from 22kohm to 16kohm.


There have been early copies of that drawing circulated with that draughting mistake, it should be PNP. That's what happens when people don't respect copyright and IP.
dot matrix madness
we're not worthy Thank you very much Synthiq and Graham!
BTW, are there any other recommended switching transistors as replacements for the ME9002? The 2N3969 has become a bit expensive.
Best regards!
Synthiq
dot matrix madness wrote:
BTW, are there any other recommended switching transistors as replacements for the ME9002?

Besides a 400MHz bandwidth, no other parameters in the datasheet for ME9002 stand out as exceptional compared to a 2N2222 or 2N3904 so I would try one of them first. Their bandwidths are "only" 300MHz but I'm not convinced it will make a difference at audio frequencies.
dot matrix madness
Here is the working LTSpice simulation of the VCO thanks to Syntiq and Graham.
Compared to the original schematics following components were changed:
Q11 (the one with the unconnected collector; Q63 in Graham's modified schematic) is now a V10 zener diode.
2N3094 transistors instead of ME9002
All other transistors are either BC547C or BC557C, respectively.
Next step: doing a prototype in 3HE for euro rack.
The 10 turn frequency poti is already waiting in my drawer.
Synthiq
The way the 10V zener diode is connected now, it operates as a normal diode with 0.7V drop across it and it has to be reversed to work as a zener diode. The circuit works in both cases but with different frequencies and the temperature coefficient of the frequency is much higher when the diode is forward biased as it is now.

Transistor Q11 is a pnp transistor and they are normally biased so the emitter is the most positive pin but the opposite is true now. Again, the circuit works either way, but the triangle wave is more symmetric if the emitter and collector is reversed.

The sine output is still not correct so you may look into that before doing a prototype.

Since most npn transistors are actually BC547C, I tried to replace the 2 2N3904 with BC547C as well and I only saw a +1% change in frequency so they are not really critical so you might just stick to one type of npn transistor to simplify things.
dot matrix madness
Thanks Synthiq for putting so much effort into that. I'm still poking around to get things right and feel like this biologist trying to repair a radio.
I'm I at least right in the following points?
The very transistor is a PNP, has the emitter connected to +12V and its collector open.
A 10V zener diode in parallel between base and emitter with its cathode towards +12V.
In contrast to (most?) other VCOs the frequency goes down if a more positive CV is applied.
Synthiq
dot matrix madness wrote:
The very transistor is a PNP, has the emitter connected to +12V and its collector open.
A 10V zener diode in parallel between base and emitter with its cathode towards +12V.
In contrast to (most?) other VCOs the frequency goes down if a more positive CV is applied.

If you refer to your transistor Q1 in the original schematic, it was (and should be) a npn transistor with emitter to +12V and the base to the emitter of Q2. You could also use a pnp with the two pins reversed; what's important is that the base-emitter junction should be reverse-biased and you use the breakdown voltage of the base-emitter diode as a cheap zener diode. The breakdown voltage is never specified in the datasheet, it only says that the absolute maximum reverse voltage you should apply is 5V. So the only thing you can be sure of is that the breakdown voltage is higher than 5V but if you are unlucky it could be higher than 12V and your circuit will not work.

There is no reason to use both a transistor and a zener diode in parallel in a real circuit, the only reason I did it that way in my simulation was to minimize the changes to the simulation schematic. It is quite possible your original circuit would work in real life with a transistor there, but you will have more predictable results with a zener diode. An yes, the cathode of the zener diode should be connected to +12V.

In my simulations I also noted that the frequency drops with higher control voltage which makes it incompatible with everything else unless you put an inverter in front of the Q6/Q7 input stage. Or you may be able to swap the base connections of these two transistors to invert the function.
auxren
I wish we had more threads like this.
dot matrix madness
Thanks again Synthiq!
Synthiq wrote:

In my simulations I also noted that the frequency drops with higher control voltage which makes it incompatible with everything else unless you put an inverter in front of the Q6/Q7 input stage. Or you may be able to swap the base connections of these two transistors to invert the function.

For the use in the common 1V/Oct world Graham has corresponding modifications on his web site that consider these things.
Graham Hinton
dot matrix madness wrote:

For the use in the common 1V/Oct world Graham has corresponding modifications on his web site that consider these things.


The EMS oscillator has one big mistake regarding temperature compensation (in common with ARP VCO circuits): if you sum the input voltages with different value resistors directly on to the tempco resistor then each source has a different temperature coefficient! [Which shows how much these designs were ever checked and tested.]

The correct method is to use one resistor to obtain the one correct coefficient according to the tempco used, which then sets a V/oct sensitivity. Then put a scaling inverter in from of that to convert from 1V/oct (or whatever). I will be updating my EMS Modification site to explain this when I get a little less busy.
ricko
(Semi off topic, but relating to substitutions of transistors in simulators)
For simulations, don't forget that sometimes their SPICE models are pretty dodgy.

I just went through about 8 different SPICE models for a 2N3096 transistor, and they all differed significantly on key values. (I finally found that most models did not have anything for the avalanche values, which was not much good as I was looking at simulating an avalanche VCO.)

The reason is apparently because they use different curve fitting approximations for transistor/diodes, sometimes adjusted for the particular circuit simulator's target circuits. TINA-TI often does not want to oscillate, for example. If your circuit relies on some behaviour that is in the wrong part of the curve, the simulation may not be accurate on some simulators.
dot matrix madness
The caveat regarding transistor models and circuit simulations in general is certainly a good point. Anyhow I found simulations helpful to get some insight which signals can be expected at various points. Kind of circuit bending at the level of simulation, particularly if it is not a well established project for euro rack.
But I will first breadboard the circuit plus modifications before doing a prototype. Since I won't have time to realize that until X-mas there is no hurry.
Synthiq
ricko wrote:
I just went through about 8 different SPICE models for a 2N3096 transistor, and they all differed significantly on key values. (I finally found that most models did not have anything for the avalanche values, which was not much good as I was looking at simulating an avalanche VCO.)

Spice uses the Gummel-Poon model for bipolar transistors and as far as I can see this model doesn't have a parameter for breakdown voltage like the BV parameter available for diodes. If you need this you have to create a subcircuit with the transistor and an additional diode and then let your transistor symbol refer to that subcircuit instead of the BJT model.
MUFF WIGGLER Forum Index -> Music Tech DIY  
Page 1 of 1
Powered by phpBB © phpBB Group