18V Electric Mistress with reworked LFO & VCO

[DrAlx]

A couple of years ago I built an 18V EM (Version 2) using the schematic from the Electric Mistress Mystery page.  Apart from the well known volume drop there was a surprising amount of clock noise.  I had experimented with various things to reduce the noise and managed to lower it a little but not to what I would consider an acceptable level.

I have also built the EM3207 clone of the 9V EM (which uses a LFO/VCO like the Deluxe EM).  That does not have the same noise issues.
I reckoned that the LFO/VCO of the original 18V EM was the problem, so I set myself the challenge of reworking the 18V EM with the following goals:

1) Get rid of heterodyne noise (due to RF picked up on the input line).
2) Replace the noisy LFO/VCO combination with something similar to the 9V EM.
3) Keep the simple SPDT switch to select filter matrix mode, instead of the more complicated DPDT of the 9V EM.
4) Have all the control pots work in the exact same way as the 18V EM. In particular ...
   The fastest sweep rate should match the original.
   The slowest sweep rate should match the original.
   The sweep (min delay to max delay) with the Range pot is at maximum should match the original.
   The sweep (min delay to max delay) with the Range pot is at minimum should match the original.
5) Fix the volume drop by adding a gain stage at the output.
6) Fit the whole thing in a 1590B on Veroboard. (I won't be running on batteries).

So I basically wanted the same sound and controls as the original 18V EM but without all the noise and the big box.  Point 4) was the trickiest to achieve but I got there within acceptable tolerances (a couple of percent).  Actually points 3) and 4) are related, and what allowed me to achieve both was a modification I made to the VCO that I'll explain below.

Here's a PCB layout with almost the same layout as the 25x21 hole Vero build shown in the photo:


Here's the schematic:


The Audio Path:

The audio path is the same as the original 18V EM apart from the following.

1) Added C2.  This forms a LPF with R2||R3 giving a 3dB point of about 11 kHz.  I took the idea of having an RF filter at the input from the EM3207.
2) Added C4.  This is taken from the 9V EM.  This cuts more RF in the pre-emphasis section, starting at about 23 kHz.
3) The balance trimmer RT1 can be used to minimise clock glitches at the BBD output.  The change in BBD gain is insignificant.
4) The boost stage based on Q1 fixes the volume drop.
5) The Electric Mistress does not actually use parallel multiplexing (as I explained in another post) so I used just one of the delay lines on the chip.
This allowed me to utilise a half-working SAD1024 that I have.


The LFO:

IC3 is the basic triangle/square oscillator.  I chose component values that would give me the same set of rates as the original 18V EM.  Why the 10k||33k for R22,R33 ?  Because that's what's in series with the rate pot in the original, and I wanted to match the measured waveforms from my reference circuit.


The VCO:

Apart from component value changes, IC4 is like the VCO in the 9V EM, but there is one subtle but important difference.  The extra resistor R29 means the discharge diode D2 is no longer directly connected to the comparator output. (i.e. R29 would be 0 ohm in the 9V EM).  I'll explain the purpose of this resistor.

The basic principle in this VCO is the same as in all versions of the EM.
The clock capacitor C16 is charged by a constant current source (Q2), so the voltage on C16 increases linearly in time.  When this voltage reaches the level of the CV (applied at pin 2 of the LM311), the LM311 output is pulled low,
causing the capacitor to be quickly discharged through D2.  The capacitor then starts charging again.

The VCO characteristic (i.e. mapping of CV to BBD delay) for this sort of VCO is a straight line, with CV on the horizontal axis and delay on the vertical.  The gradient of the line is inversely proportional to the rate at which the voltage on C16 increases while it is being charged.  Any one of the following changes will cause the capacitor voltage to increase more slowly, and so make the gradient of the line steeper.

1) Increasing the clock trimmer (RT4) resistance (so charge current is lowered).
2) Using a Q2 with lower hFE (so charge current is lowered). 
3) Increasing the value of C16 (so the charge current produces a smaller voltage increase on the capacitor).

The above are all equivalent and they all increase the gradient of the line.

Note: Having said they are all equivalent, there ***is*** a practical problem with using too low a value for C16.  Using a value in the 10s of pF can give a non-linear charactersitic at low CV levels due to the effect of diode reverse-current when D2 switches off.  I mentioned that on the EM3207 thread.  So I used a value of 470pF in my VCO like the original 18V EM.  This is 10 times larger than the 47pF used in the 9V EM.

Now for a fixed CV range, increasing the gradient of the line using RT4 will give a wider sweep range but a lower sweep ratio.
Here is an example with some made up values to highlight the point:
  RT4 set to min:  BBD delay is 1ms to 5 ms ==> Sweep range is 4 ms.  Sweep ratio is 5:1.
  RT4 set to max:  BBD delay is 4ms to 12 ms ==> Sweep range is 8 ms.  Sweep ratio is 3:1.

So it is not possible to set both a desired sweep range AND a desired sweep ratio with the single trimmer in the VCO.  The 9V EM circuit addresses this by shifting & scaling the output from the triangle wave generator so that is becomes a CV with a suitable range for the VCO.  That approach results in an extra op-amp, some scaling resistors, and the more complicated DPDT switch for filter matrix mode.

I took a much simpler approach.  I decided to shift the VCO characteristic instead of the CV, and that's what the extra resistor R29 is for.
As explained above, RT4 sets the gradient of the line in the VCO characteristic.  The extra resistor R29 allows you to set the vertical offset of the line.  Increasing R29 moves the whole line downwards, and providing R29 is kept small (under 500 ohms ) it has little effect on the gradient of the line.
This is how it works...

First consider the case when R29 is zero ohms.  Since C16 discharges through a diode, it never discharges all the way to 0V.  The lowest voltage it can discharge to is one diode drop.  Therefore the charging process actually begins with the capacitor voltage at about 0.6V.  So if the CV level is 2V, then the VCO period is roughly the time taken to increase the cap voltage by 1.4V.

Now consider the case where R29 is some non-zero value. R29 forms a voltage divider with R28.  When the comparator pulls its output low, the cathode of D2 no longer gets pulled to 0V but to some higher value instead (lets say 0.4V).  Accounting for the diode drop, C16 now discharges to a minimum voltage of 1V.  Therefore it starts its charge cycle with a voltage level of 1V instead of 0.6V.  Since the charge rate provided by Q2 is exactly the same as before, the CV level of 2V will now be reached in a shorter time.
So the VCO period decreases.  Therefore increasing R29 has the effect of shifting the line in the VCO characteristic downwards.

The effects of these two controls on the VCO characteristic (clock trimmer to vary gradient, and R29 to vary vertical offset) are only weakly coupled.  This means you can set the clock trimmer to get the desired sweep range, and then increase R29 from zero to get the desired sweep ratio.  That's basically how I arrived at a value of 220 ohms for R29.  Using this value and adjusting RT4 gave me the required flexibility to match the sweep range and sweep ratio of the original 18V EM,  with the added benefit of being able to drop the extra op-amp and DPDT switch used in the 9V EM.

If R29 is made too large (say over 1k) then C16 will discharge more slowly.
That's because D2 is effectively in series with R28||R29.  If the discharge period is significant, then the overall effect on the VCO period is less clear cut.
In theory it should still move the characteristic downwards because C16 discharges by less.  It's not a problem in this particular circuit though, and you can basically set R29 without worrying about it changing the gradient of the characteristic.

One other important component is R30. Together with C17, this prevents the VCO "thumping" when the range pot is at maximum and the sweep rate is high.  This "thumping" is not a clock tick but it can sound like one.
Note that if R30 was not there, then the series resistance going into C17 would be strongly dependent on the range pot setting.
In particular, with the range pot at maximum, the resistance would be zero and there would effectively be no filtering of the CV at all before hitting the VCO.  Unlike the 9V EM LFO, the filtering from R30 and C17 is light enough for the CV wave to still looks triangular.  In the 9V EM, the triangle wave is actually filtered twice before it reaches the VCO (once before the rate pot and once after).  So on that circuit the triangle wave looks almost sinusoidal at high sweep rates.  That is not the case with this circuit.

Given the high price and low availability of the SAD1024 I doubt anyone else will build this circuit, unless they have a vintage 18V EM and are keen to replace the guts of it with something without all the noise issues.  In any case, I was pleased with my VCO modification and thought it might be of interest.

[StephenGiles]

Many thanks, very interesting indeed!

[ElectricDruid]

That's a great piece of work, thanks for sharing. It's nice to see the analysis of the problem and then the development of the solution instead of just "here's the schematic!". I've been thinking of building a chorus next, and I think it could benefit from some of the ideas you've presented there.

Tom

[Scruffie]

Funny, i'd actually been working on a version with a 393 keeping the original LFO and using the 311 VCO today which being near rail to rail gave some rather large sweeps and a fair bit of dead sweep, i'll have to try your R30 approach.

[stringsthings]

That's really cool!  Any sound samples by chance?  The old SAD1024 flangers were really great.

[JC103]

I've actually got an old flanger kicking around with a SAD1024... and it's in a socket to boot! Might have to give this a whirl. Out of curiosity, could a TDA1022 chip be retrofitted in your adaptation?

[DrAlx]

TDA1022 is negative supply like the MN300x series. 
That isn't a problem, but if you want to make TDA1022 as close a fit to SAD1024 you would need to address the different BBD gain for the two chips.
Maybe with this sort of approach which I outlined for the MN3207:
            https://1drv.ms/i/s!AvrH61utWEtEiAbzPDPv5UK-geVK
You can control the BBD gain with R14, though not sure if its possible to get the same gain as the SAD1024. The TDA1022 appears to be lossy according to the datasheet.
The "Rds || 10k" comes to a value of about 1.1k to 1.2k based on my measurements on an actual SAD1024.
The point is that in in the original EM circuit, the impedance looking left from the 3n3 capacitor would be 10k in parallel with the drain-source resistance of the BBD output.  You need to have that same impedance for the frequency response to look the same.

As far as the VCO/LFO goes, the mod I mention above would not need tweaking assuming everything is still running off a regulated 12V.

[KHStudio]

Are the images still available somewhere?

Thx

[DrAlx]

Schematic:  https://1drv.ms/u/s!AvrH61utWEtEi1hXQjQDPCiV46Xc?e=8WBR4g
PCB layout: https://1drv.ms/u/s!AvrH61utWEtEi1oayThe2Upl7Z0X?e=4Wg774
Photo:         https://1drv.ms/u/s!AvrH61utWEtEi1mTTjjUznhTg6_7?e=uGbYjc

If you are on Windows or Linux Mint 18.3 you can download VeroRoute from
  https://sourceforge.net/projects/veroroute/.
The examples folder contains the PCB layout and a Vero layout.

[KHStudio]

DrAlx, Thanks for the links!

Btw, the schematic is hard to read & only blurs when zoomed in but I can somewhat make it out if that's all that's available.
I've been working on several different builds of the EM & was just interested in your work for additional options & learning.

[DrAlx]

Schematic in link looks ok on both my desktop and tablet. Maybe try download and view it in another app.

[bluebunny]

Schematic in link looks ok on both my desktop and tablet. Maybe try download and view it in another app.

Yep, I concur.  Link looks legible (and appears to be a preview).  "View original" is even better: 2048x1452 and crystal-clear.

[DrAlx]

Including C2 (2n7) was a mistake.  I didn't account for impedance of the input.
Having C2 is OK if the input comes from a low impedance source like another pedal (which is what I always have) but if you just have a guitar with passive pickup, then C2 sees (5k6 + pickup impedance) to its left and ends up killing high end.  Best to leave C2 out.

[Aurae]

Hi DrAlx and thank you for all this data. I was wondering if I could use the same values from the MN3207 schematic for an MN3007 ? Knowing that I would like to do a regulated 15V.

And what does the "Rds || 10k" actually stand for ? Just a resistor or something else  ?

[bluebunny]

Welcome to the forum!

And what does the "Rds || 10k" actually stand for ? Just a resistor or something else  ?

It's shorthand for a calculation: Rds "in parallel with" 10K.

[Aurae]

Thanks a lot and hope DrAlx can answer the other question

[DrAlx]

I will soon post a MN3207 based version of the above circuit.  I'm aiming to make it the best possible clone of the original V2 using a BBD that is still available. 
I'm waiting for the PCBs to arrive so I can do a final comparison against the SAD1024 circuit.

I tried to rework the LFO/VCO for 9V operation but couldn't manage to keep all the controls working the exact same way.
So in the end I decided not to rework any values and instead I just ran the LFO/VCO section off a 12V regulator, and ran the rest (audio chips, BBD and clock buffer chip) off a 9V regulator.
I have not tried a MN3007 but based on my measurements (posted recently on the EM3207 thread) and the datasheets, you will probably get a closer match to the SAD1024 using an MN3207 as opposed to the MN3007 (in the sense that BBD gain and hence wet/dry balance will be about 2dB closer match to the SAD1024 circuit).

Oh and ignore my previous posts mentioning circuit hacks to try and get an impedance of "Rds||10k".  It is not necessary if you just get rid of the emitter follower after the BBD.

[Aurae]

Thanks for these infos, I'm looking forward to the results for the 3207 !In any case for the 3007 I reach a result close to the original only by using it in 12v probably because of the LFO/VCO operation.

[Aurae]

How would you do both regulations here then? Powered in 18v DC then regulate in 12v and 9v with 78L12 and 78L09?

Do you already have a schematic?

[Ben N]

Following. With Madbean's Current Lover board unavailable, I have been looking for an EM project using obtainable parts. This looks exciting. Thank you DrAlx!