How do the 18V Electric Mistress LFO & VCO work?

[Scruffie]

I like the simplicity of the design of the 18V mistress lfo/vco/clock set up and it's use of the second half of the 4013 for the LFO rather than wasting it and i've had it on the breadboard for 3 days now tinkering, the original gave a frequency sweep of 20 - 200kHz (confirmed against 2 original pedals) now, i'm using a MN3207 with a clock buffer and was hoping it'd be a simple case of halving the 470pF clock cap off the current source to double that to 40 - 400kHz to accommodate having twice the stages however this was not to be the case... it seems at most it'll clock up to 300kHz.

Now ignoring that for a moment, I can't even get it to give the full 20 - 200kHz that it should, I split up the resistor bias network to make diddling the values easier to try and get that full sweep but I just can't, at least not without messing up the sweep.

So how does it all work, I can see the separate blocks but how each functions is beyond me.

I especially don't get pins 8, 9 & 14 (let's call it comparator 4) and also how the 4013 is being used in the LFO. I set up pins 1, 6 & 7 (comparator 1) like the 311 of the 9V mistress and dropped comparator 4 and was able to get the higher clock frequencies but then lost the low ones, although it still needed work as it didn't quite behave correctly.

Anyone have any insight in to it all?

18V Mistress Schematic


and the 9V for reference
http://www.metzgerralf.de/elekt/stomp/mistress/images/1981-electric-mistress-v6-schematic.gif

[StephenGiles]

In a word.....no! Probably easier to speculate what substances they may have been (mega-subjunctive ) on at the time!!!

[DrAlx]

Hi Scruffie.

Just had a quick look.  I can't give you an answer yet, but can tell you the questions I ask myself.

1) I think the charge path of the clock cap seems to be the PNP (like the other EM circuit).
Lowering the emitter resistor on the PNP will increase the charge rate.  How close to 400kHz can you get going that route instead of decreasing the clock cap?
I don't know how close that PNP is to its maximum allowed current, so maybe it is already maxed out.

2) What is the the discharge path for the clock cap?
For the 9V EM, it is clearly through a diode that is switched on/off creating a discharge path to ground.  That is not the case here.
I can see IC4d doesn't have a pull-up resistor on it's output pin (unlike all the other comparators) so
understanding that might be key.  Is it discharging directly into that output pin 14?  
The only other option is via the input of IC4c, but aren't comparators not supposed to draw any input current?  I am guessing it discharges into pin 14.
Actually come to think of it, can you confirm what the cap waveform looks like.  Is it clearly saw tooth because of a charge-discharge process?

3) What sets the min and max capacitor voltages (the threshold voltages) during the charge-discharge process?
It seems to be a combination of the voltage at pin 8 of IC4d and also pin 9 (which is set indirectly by the range pot and IC4c).
You might be able to confirm that with a scope.

BTW, are you trying to test all this in filter-matrix mode?

[Scruffie]

1) Lowering the emitter resistor (first thing I tried as this is the 'clock trim' in the other mistresses) doesn't get it much closer.

2) I don't have a scope, using the frequency counter on my DMM

3) I can't confirm it with a scope but having played with the 680/150R combo I can confirm that it does have an effect of the maximum frequency, still it'll only take it up to about 350kHz and then the bottom of the sweep is gone (think it was something like 80khZ at the bottom at best, maybe 100).

I'm testing this all in swept mode, full range, slowest rate, exactly as I had the real units tested, matrix doesn't make much difference though.

[DrAlx]

The sweep ratio (i.e. ratio on max/min clock rate) is set by the LFO rather than the VCO.
I posted something on this on the EM3207 thread recently.
I built a couple of EM3207 and both had different sweep ratios.
Changing the VCO rates does not improve the sweep ratio.  
The difference was down to the LFO waveform being wider on one than the other.
(Note that the range pot also doesn't have a strong effect on sweep ratio).

To prove the point I showed how you can mod the resistors in the LFO stage to get big sweep ratio.
I could then get massive sweep ratio but it stops it sounding like EM.
The EM sound needs the sweep ratio to be in the area of around 4:1.

I'm surprised that it is 10:1 in the original 18V version.  The 9V version is definetely lower.

[Scruffie]

The sweep ratio (i.e. ratio on max/min clock rate) is set by the LFO rather than the VCO.
I posted something on this on the EM3207 thread recently.
I built a couple of EM3207 and both had different sweep ratios.
Changing the VCO rates does not improve the sweep ratio.  
The difference was down to the LFO waveform being wider on one than the other.
(Note that the range pot also doesn't have a strong effect on sweep ratio).

To prove the point I showed how you can mod the resistors in the LFO stage to get big sweep ratio.
I could then get massive sweep ratio but it stops it sounding like EM.
The EM sound needs the sweep ratio to be in the area of around 4:1.

I'm surprised that it is 10:1 in the original 18V version.  The 9V version is definetely lower.

I realise that which is why it was a joint question, increasing the 680k divider with the 1M does improve the ratio a bit but not drastically.

I was surprised too, it certainly sounds 'nice' with the lower ratio but it certainly seems to be missing something, I think the closest I got was about 6:1, the lowest of the original units measured had an 8:1 ratio which i'd settle for.

Oh btw, here's some scope shots of an original unit for the cap discharge http://www.metzgerralf.de/elekt/stomp/mistress/repair-volt-v2.shtml

Edit: Forgot to mention the reason the 9V is so low is the supply voltage, the deluxe electric mistress has a much larger ratio but much the same LFO.

[DrAlx]

Thanks Scruffie.  I new the Deluxe had the same sort of LFO but didn't figure that the sweep ratio was that much wider.  I never took into account that it was running at 18V.

That link was useful, but it's a pity there are no time-axes on those graphs.
If decreasing the charge time (via smaller PNP resistor or smaller clock cap) didn't help then maybe it is something limiting the discharge process?
I'm sure now that Pin 14 usually floats (at the cap voltage) but gets periodically pulled low to discharge the cap whenever waveform E goes below 0.09V (the voltage at pin 8 on Ralf Metzger's schematic).
There's a delay between the cap reaching the threshold set by pin 7 (which causes waveforms D and E to start to go low) and waveform E reaching that 0.09 V value.
So there is a response time between the capacitor threshold being met and the discharge being implemented.
I can't tell if that is the limiting factor (no time axes on the graphs) but it would be my best guess.
It's also not clear if that 0.09V figure reported on the schematic varies appreciably during the sweep, 'cos there is a link to the LFO control voltage via the range pot.  If it goes lower then 0.09V then it will serve to increase the response time even more.  Have your tried using a slightly bigger value than 0.09V and see if that helps you reach the higher clock rates?

I don't about the LFO but may have a go at simulating it 'cos I would like to know how it works too.

[Scruffie]

I changed the clock cap to 220pF, the 150R resistor to 820R and the 680R to 10R and it sweeps up to 300kHz now, the problem seems to be that if you raise the frequency you loose a lot of the bottom of the sweep as it seems to struggle charging as opposed to discharging.

I have got it up to 400kHz and more but with too much alteration it messes up the sweep waveform.

[DrAlx]

I know this is an old thread, but I've just put together this LFO/VCO and thought I'd post my findings.
I made a compact perfboard build, and the the only change from the above schematic is I've used a 78L12 voltage regulator instead of the Zener/741/Transistor combo.

Ralf posted some measured voltages on his Electric Mistress Mystery page, and I don't get the same voltages.  I measured around 12.45V supply (regulated) on my cheap scope while Ralf got 13V.  The biggest difference I saw was that my control voltage got nowhere near as high as the 10.6V reported by Ralf.  (I could only get those higher voltages by lowering the 1M/680k ratio at pin 4 of the 4013).  I was using 1% resistors.  I am guessing the tolerances on the original EM were lower.

Here are my measurements in sweep mode

Max Range, Slowest Sweep
CV Min:   0.34 V         fCLOCK: ~222 kHz       BBD delay: 1.15 ms
CV Max:  8.70 V         fCLOCK: ~30.76 kHz    BBD delay: 8.32ms

Min Range, Slowest sweep
CV Min:   0.068 V       fCLOCK: ~256 kHz       BBD delay: 1 ms
CV Max:  0.519 V       fCLOCK: ~212 kHz       BBD delay: 1.2 ms

The original EM has no "rounding off" of the triangular CV waveform at high sweep rates as seen in the 9V EM or Deluxe EM.
But I did see a shift in the range of CV values.  At the fastest sweep rate I found I got slightly small ranges of CV values.

Max Range, Fastest Sweep
CV Min:   0.47 V         
CV Max:  8.58 V     

Min Range, Fastest Sweep
CV Min:   0.093 V     
CV Max:  0.527 V 


So I had no problem getting to over 200kHz but couldn't get the low 20kHz minimum fCLOCK reported by Ralf,
Even if I made the CV go to over 10.6V by tweaking the 1M/640k ratio, I still got just under 30 kHz at best.

So I started to investigate the VCO.
I measured the clock cap voltage during the charge/discharge process and found that at slow clock speeds the cap does not fully discharge.
I can't see how I can make it discharge any quicker into pin 14 of the LM339.  I could try a lower cap and lower charging current but it will be a major PITA to tweak things on this perfboard build


Regarding the original aim of trying to get this LFO/VCO working at double fCLOCK (so you can use 3207 BBD) I would suggest the following approach...

Ditch the VCO section that uses 2 parts of the LM339 and use just one (based on the VCO of the 9V EM or deluxe EM). The reason I say this is that the diode in the above picture provides too much of a delay. If you measure the voltage at pin 9 of the LM339 you will see a high voltage with a periodic "notch" that appears during the discharge process.  That notch looks exactly the same, regardless of fCLOCK, and is always 2.1 microseconds wide (and 2.1 microseconds is 476 kHz).  I think its too close to the target max fCLOCK of 400 kHz.

One thing you have to realise when you use the VCO of the 9V EM is that the cap discharges through a diode so it will never fully discharge.
It can only go down to 0.6V because of the diode voltage drop.  So the CV must be kept above 0.6V (I'd say > 1V).  The CVs I posted above fall below 0.6 V so the LFO section would need to be reworked to give a suitable range of values for the VCO.

So the procedure would be to start with a 9V EM VCO (but using the LM339).  Find the range of CVs that give fCLOCK from 40 kHz to 400 kHz.   Then tweak resistor values in LFO of above picture to give that CV range.

[DrAlx]

... you loose a lot of the bottom of the sweep as it seems to struggle charging as opposed to discharging.

I've looked at this in more detail and its definitely the discharging (or rather lack of discharging) of the clock cap that causes the bottom of the sweep to go.

The problem with trying to increase the threshold voltage at pin 8 (in order to make the lower comparator react sooner to the diode switching off) is that the comparator will also react sooner to the diode switching back on.  So the low end of the sweep goes because the clock cap will not properly discharge.

If you look at the waveform at pin 1 of the LM339 you will see a voltage level of around 8 or 9 volts with a short periodic "notch" where the voltage drops to near ground.
The left wall of the notch is steep showing that the diode switches off very quickly (and its the diode switching off that triggers the clock cap to start discharging)
The right wall is less steep and has a bit of a curve to it (looks like the graph of a cap charged through a resistor).  So the diode switches on a little less quickly than it switches off.

I placed 100pF between pin1 of the LM339 and the supply rail (ground worked just as well). This causes the left wall of the notch to be practically the same as before (i.e. diode switches off just as quickly) but the right wall is much shallower.  So the diode effectively takes longer to switch on, which in turn means the clock cap gets more time to discharge properly.

I found 100pF at pin1 had little effect on the high clock rates (212 kHz dropped only a bit to 204 kHz) but I could get the low end down to around 24 kHz if I made the CV high.
Without the 100pF at pin 1, the best I could do was just under 30 kHz.  (I'm still not sure how the measured 20 kHz was achieved on the original EMs.  )

I don't know if you persevered with your attempts to double the clock rate Scruffie, but thought I'd mention this cap at pin 1 trick. It might stop the bottom of the sweep from disappearing when using diddled resistor values.