SergeModular

Hi list,
I'm trying to simulate the Negative Slew from this schematic (I'm assuming drawn by Ken - is he around these parts?):

http://www.cgs.synth.net/synth/serge/pic/serge_negative_slew_schem.gif

I'm trying to investigate how well the CV controls the negative slope, but I can't get the simulation to run properly, mainly because I can't see how on earth the main integrator (amp pins 8/9/13) resets: for this to happen base 6 will need to climb to about 2 diode drops above ground, but I simply see no route for any current to enter the 220k there - all the diode junctions in the two transistors would appear to prevent this. Anybody know if this schematic is accurate, or can explain how it does work (if indeed it does)? (I've checked that part of the circuit against the PCB layouts given in other pics there, and they look like they agree.)

I also think the (kind of) 'diode connected' transistor 9/10/11 would cause grief to the CA3086 if the 'diode' could in fact conduct: this would effectively put the transistor into saturation, which might be bad news for the chip, as I think it would turn on the parasitic PNP transistor (formed from base-collector-substrate), causing current to flow out of the substrate.

Has anyone any thoughts or ideas?

Thanks, Tim
__________________________________________________________
Tim Stinchcombe

Cheltenham, Glos, UK
www.timstinchcombe.co.uk

Tim,

Ken runs a Yahoo group (cgs_synth) for the discussion of the stuff on his site. You may want to post over there. He's typically very helpful and responsive. In fact please do as I'm very interested in the workings of these circuits...

-dustin

Hi Steve,

Many thanks for the response - that has more than answered my question!

>> Have you found a good spice model of the LM3900 / MC3401?


> I don't believe there *are* any manufacturers models for the
> 3900/3401 out there.

I suppose an email to National Semiconductors (2900/3900) or
Motorola (3301/3401) might be worth a try. Or try Bob Pease (c:

> Ack. It's unfortunate that the standard SPICE BJT model doesn't
> model such things as reverse junction breakdowns, as it does for
> diodes.

That's why it's called a model I suppose. Same reason that James
May's full size Airfix Spitfire wouldn't fly...

> Camenzind's excellent book for the stuff about parasitic devices,

It's on my reading list, which is longer than my remaining
life expectancy...

>> The other surprising thing is that switching on transistor
>> 6/7/8 doesn't cause excessive current to flow into pin 8
>> of the MC3401 and cause blowupedness


> Presumably the big discharge current merely circulates
> cap-33R-transistor, so the input pin isn't really involved.

Ideally.

>> Have you found the almost-exponential converter yet?


> I'm actually looking at improving the 'controlability' of the
> Plan B M10: it seems the core of that circuit is virtually a
> direct copy of the Serge envelope generator,

The designer / copier (delete as appropriate) is/was a member
of this list. Perhaps he'd like to contribute?

> maybe 15-20 percent of pot travel, and the relationship between
> rise/fall/timebase pots means there are large portions of travel
> which simply have no effect.

The same circuit crops up in several first generation Serges,
including the Neg, Pos and VCO. I tried strapping a decent
current source to the VCO and it was much improved.



Steve

Well, for anyone still paying attention, here is my take on how this Negative Slew circuit works.

First I want to say just how in awe I am of anyone who has such an intuitive feel for electronics that they can see and design such a circuit by apparently 'seeing the whole' so easily (or as the joke goes, Serge has probably _forgotten_ more electronics knowledge than I will ever know!). I can only dream of arriving at such a point.

So, first the easier stuff, and some notation. We have established that transistor 9/10/11 is actually used as a 7V zener, henceforth I shall simply refer to it as 'the zener'. The other transistors I shall call Qx, where 'x' is the pin of their base, so e.g. transistor 3/4/5 is Q4. Amplifier 10/11/12 is a simple inverting set-up, sending (approx) 0 to 4.2V at its input (the pin 9/330k), to 6V down to 0 at its output (the 4.2 is from the slight gain of 470/330, i.e. 6*33/47). Amplifier 4/3/2 is a simple adder/subtractor, providing a current into pin 13 of the main integrator (and due to the fact that there are diodes at pins 3 and 13, separated by only the 18k, I suspect this current has a very 'tanh' shape as the rate pot is swept, and the initial upswing of this curve is fairly close to being exponential): for the purposes of this description, we assume the current into pin 13 is kept constant. Q12 sets a quiescent current of about (6-0.7)/4k7 = 1.1mA being pulled from differential pair Q2/Q4. Thus the quiescent (0V at both Q2, Q4 bases) voltage at Q4 collector is 12-10k*1.1m/2 = 6.5V - in the quiescent state this is not enough to turn Q6 on.

Initially assume that both input and output ('3401 pin 10) are at 0V: because Q2 and Q4 bases are both at 0V, they share the 1.1mA equally, and so Q6 is not on; thus current into pin 13 keeps the 330n fully charged at 6V, and this is inverted to 0V at the output. Now apply a positive-going step-change at the input: this applies a differential voltage across the 18k between Q2 and Q4, turning Q2 on more, and Q4 more off; Q4 more off allows its collector to rise a little, which in turn means the zener can conduct, and this allows Q6 to turn on. Q6 on means the 330n can start to discharge, and since this is then inverted by the output amp, the output *rises*. However the output rising tends to turn Q4 on more, which in turn tends to make Q6 turn off - i.e. we have negative feedback, the end result of which is that the output will *follow* the input as it rises. If the input stays quite high, then the 330n is held at quite a low voltage.

Now apply a steep, negative-going step at the input that we expect to get slew-limited. The input drops to 0V quickly, but the output stays high: this puts a reverse differential voltage across the 18k between Q2 and Q4, which means Q4 tends to turn hard on (Q2 more off); this in turn will mean the zener is not conducting, and so Q6 is definitely off; thus the best the 330n can do is a slow charge according to the current input at pin 13 (and mirrored at pin 8), so the pin 9 output climbs linearly, and this is then inverted to give a linearly falling voltage at the output, pin 10. This is the negative slew action. If nothing changes, eventually pin 9 will max-out just below the 6V rail, and pin 10 will be just above ground.

If the negative step is not so steep as to cause the output to be slew-limited, we get the negative feedback action again (to me this seems the hardest case to explain...): at the start the voltage on the 330n is low, and the current into pin 13 (mirrored at pin 8) wants to make pin 9 climb, and hence pin 10 fall; however if pin 10 falls too much, this puts a differential across the 18k between Q2 and Q4 which tends to turn Q2 on and Q4 off; Q4 turning off allows its collector to rise, and lets the zener conduct, so Q6 is able to turn on; this effectively 'bleeds off' excess current through the 330n cap, and hence slows the rise of pin 9, in turn slowing the fall of pin 10, and again, the output (pin 10) will be forced to follow the input!

----------

Some months ago I came across the VCS, and since then I have fixed a Bananalogue one myself, and helped others fix theirs online - I have seen several comments to the effect that 'it is a simple enough looking circuit: slap your components in the PCB, tell us what is/isn't happening, and we'll sort the problems out, easy!'. The VCS also has a very contorted feedback loop (in fact it is probably harder to understand), which allows both the voltage to be followed, or become slew-limited when circumstances dictate. It may be that I am a bit slow on the uptake, but in my view, these circuits are emphatically NOT 'simple' circuits, but stand amongst some of the more complex synth circuits I've yet come across (I've barely looked at the positive slew, which looks equally complex!). As I intimated at the beginning - awesome!

Tim
__________________________________________________________
Tim Stinchcombe

Cheltenham, Glos, UK
www.timstinchcombe.co.uk

> > I don't believe there *are* any manufacturers models for the
> > 3900/3401 out there.
>
> I suppose an email to National Semiconductors (2900/3900) or
> Motorola (3301/3401) might be worth a try. Or try Bob Pease (c:

Actually contacting On Semi might be worth a shot (years ago - I think it was them - they were nice enough to provide an old app note, for the MC1495 multiplier, I think it was, which wasn't online at the time).

Nat Semi, in my experience, are a bunch of tossers! Again, some years ago I contacted them as there is/was a blatant error in their LM741 model - their support staff wouldn't even put my query through to their engineers, as I wasn't building a 100,000+ component/year's worth of business for them. Unfortunately in a fit of pique I deleted their condescending email - I really wish I had kept it now. Compare this to Phillips: when I spotted errors in one of their models, it was corrected at their website within about 2 days!

As for the Bob Pease suggestion, LMAO at the idea! :-)

> It's on my reading list, which is longer than my remaining
> life expectancy...

You have that problem too then?!

> The same circuit crops up in several first generation Serges,
> including the Neg, Pos and VCO. I tried strapping a decent
> current source to the VCO and it was much improved.

Maybe its poor control of the associated parameter contributed to its downfall as a useful 'circuit element', in favour of better methods?

Tim
__________________________________________________________
Tim Stinchcombe

Cheltenham, Glos, UK
www.timstinchcombe.co.uk

> As for the Bob Pease suggestion, LMAO at the idea! :-)

He's not keen on Spice simulations. You've seen that photo
in his book showing him throwing his Spice computer from
the top of the national Semiconductors building into the
car park?

>> It's on my reading list, which is longer than my remaining
>> life expectancy...

> You have that problem too then?!

+ kids!

>> The same circuit crops up in several first generation Serges,
>> including the Neg, Pos and VCO.


> Maybe its poor control of the associated parameter contributed
> to its downfall as a useful 'circuit element', in favour of better
> methods?

The VCO was affordable in 1972, but by 1976 is was a joke. I
believe it was the first module Serge replaced.



Steve


Tim
__________________________________________________________
Tim Stinchcombe

Cheltenham, Glos, UK
www.timstinchcombe.co.uk

> > As for the Bob Pease suggestion, LMAO at the idea! :-)
>
> He's not keen on Spice simulations.

Understatement of the year!

> You've seen that photo
> in his book showing him throwing his Spice computer from
> the top of the national Semiconductors building into the
> car park?

Indeed, I have that book.

Bob Pease seems to be a technological dinosaur, with a pathological hatred of computers, and SPICE in particular! Though to be fair, he obviously grew up with SPICE when it was in its infancy, and so would have suffered all its teething pains etc. He should try something like SIMetrix, which is what I have: it is not SPICE per se, but based on SPICE and heavily modified, so many of the model discontinuities and convergence problems are a thing of the past. But we can't all be born analog gurus, some of us have to work hard for an even meagre understanding, and as far as I'm concerned, SPICE is a tool like any other: use it well, and you'll reap the rewards; abuse it and watch out! As a tool to *aid* understanding, and not to *replace* understanding, is where Pease seems to have it wrong, in my view.

I've sent a polite enough request to On Semi asking about old SPICE models, but I shan't be holding my breath that they even acknowledge it (filling the 'required' field for 'company name' is always gonna be a killer!).

Tim
__________________________________________________________
Tim Stinchcombe

Cheltenham, Glos, UK
www.timstinchcombe.co.uk

> Well, for anyone still paying attention, here is my take on how this
> Negative Slew circuit works.

Sadly I'm not sure there really is much interrest here in the really
old Serges. Ken has an old 1973 and there are a couple of other
members with first generation stuff, but this list is predominantly
recent/STS. Maybe your post will help someone restore another
old beast and show a few others what they're missing.



Steve