EHX Microsynth schematic doubts

[DVB_master]

Hi dears,
I found this version of the EHX Microsynth Guitar schematic, which is dated 2002:
http://www.studio250.fr/docs/electro%20harm/ehmicrosynthesizer%20schemas.pdf

C8 and R13 in the Full Wave Rectifier section are respectively 0.047uF and 680 Ohm -> cutoff at about 5kHz
Another schematic shows different values for these components: 0.0018uF and 27kOhms -> cutoff at about 3.3kHz

Does anyone have a real unit to check the values of these components? Maybe they changed throughout various revisions of this effect??

What is the function of the squelch trim? If I understood correctly it adds a very very low offset, which is not very significant to me...
Thanks!!

[Rob Strand]

The 2002 version has been redrawn.
I have a 1996 version and the part locations do not agree with the 2002 version.

So while C8 on the 1996 schematic is 0.047uF it does not appear in the smoothing filter.
The smoothing filter is 1uF.
That didn't come out right.    In fact there's quite a few differences eg around IC3A.

Wait a sec, I'll convert my PDF or find the 1996 version on line.

[Rob Strand]

https://www.harpamps.com/schematics/gsynth1.pdf
https://www.harpamps.com/schematics/gsynth2.pdf
https://www.harpamps.com/schematics/gsynth3.pdf

[DVB_master]

Hi Rob,
I had both the 1996 and the 2002 schematic versions.

The 2002 version is not just a redraw of the 1996, rather it includes some fixes that were also observed by some users on the real unit. Such fixes are:
- 10k Ohm resistor in parallel with Square Wave pot
- 33k Ohm resistor instead of 3.3k Ohm in the 700 Hz input low pass filter
- 1k / 10k Ohm voltage divider at the output instead of just 1k resistor, when effect is bypassed
- 100k Ohm GAIN trimmer instead of 10k Ohm
- ... others
(I can make the full list if you want).

Here's the 1996 version with some of the above mentioned corrections: https://www.docdroid.net/u3cx2Xj/electro-harmonix-micro-synthesizer-revised-2017.pdf

-----

Now let's come back to the 2002 version.
What is strange to me are the R-C components near the squelch trimmer (screenshot is from the 2002 version of my first post):



Are these 0.047uf / 680 Ohm (as in the 2002 version) or 0.0018uF / 27kOhm (1996 version)?

Can anyone check this?

[Rob Strand]

The 2002 version is not just a redraw of the 1996, rather it includes some fix that were also observed by some users on the real unit.

I had to do something just when I posted and when I came back I found the old schematic and noticed more differences.   The 680R difference looked a bit weird.

What is the function of the squelch trim? If I understood correctly it adds a very very low offset, which is not very significant to me...

Even though the offset is small the signal level there might also be small.  I think it lets you tune when the signal gates off but it looks like it works in two polarities.

I started to analyse this in fine detail back in 2003 but I never finished it and never got back to it.

[Rob Strand]

I found this:  from swt (forum member), sept 2003

Corrections:
* R8=33k. R12=1k5. R21=22k. R68=12k. R89=47k.
* There's a 10k resistor between R105, and connection 1 on the pcb,
   so that's lifting the earth at the output of ic 1 by 11k.
*  R97 and pin 2 of ic 14 connects to ground.
*  From pin 1 to ground of ic 8, there's a 10k resistor, so that pot     becomes the result of the paralell 10k and the 100k pot.

Also...interesting,

* the fet is a bf245a, and it goes, source to V+, gate to pin 1 of ic 18, and drain to 12k resistor to V-.

*  All diodes 1n4148. .

* the preamp gain is a 100k pot, not a 10k as it's suppossed to be in that schem .

* The rest is correct.

* I've measured the original power supply and it's +8volts, -10volts
from ground. (as expected)


[Rob] Certain R12 is 7k5, as shown on the original.
      Pretty sure JFET is drain to V+ (not that it matters).
      R89 details not checked but new value looks OK.
------------------------
and this,

SWT 2003 fixes from original unit

Reference circuit:

eh_microsynth.pdf, drawn Fabian Hartery, 28 Sept, 1996 Rev 1 (three sheets)
   
Corrections:

R8  = 33k
R21 = 22k
R68 = 12k
R89 = 47k
R3  = 100k
Add 10k Resistor from A8 pin 1 to ground.
Add 10k Resistor between A2A pin 1 and R105 (non grounded side) ie. series with circled 1 line
A14 pin 2/R97 should be grounded.
JFET Q2 bf245a
Diodes 1n4148
Supplies:   +8V to +9V and -10V

[Rob 2003] All these corrections look OK to me, in fact I suspected 99% of them from the schematic.

[Rob Strand]

OK I just when over it.   Your new schematic covers most of these.

The exceptions are:
- Another 10k resistor in series with the "1" circles between the pages.
- The R12 1k5 vs 7k5 bug I suspected.  (because it's a standard rectifier circuit)
   [*** this fix is on the 2002 schem.]

It's so long since I've looked at this I would have to go over the whole thing.  It's a big circuit and there's several hours worth of detail to ponder over.

[DVB_master]

Hi Rob,
You are great! I also started to analyze this projects several years ago. I built my own PSU with SMD components and now I am trying to study the effect circuit.

Thanks for your help.

All of these corrections are present in the 2002 version and should be OK:
R8  = 33k
R21 = 22k
R68 = 12k
R89 = 47k
R3  = 100k
Add 10k Resistor from A8 pin 1 to ground -> resistor in parallel with Square Wave pot.
Add 10k Resistor between A2A pin 1 and R105 (non grounded side) ie. series with circled 1 line
A14 pin 2/R97 should be grounded.
JFET Q2 bf245a -> You are right, it does not matter whether drain or source are connected to V+. Data sheet states that Drain and Source are interchangeable.
Diodes 1n4148
Supplies:   +8V to +9V and -10V

About your post about exceptions:

- Another 10k resistor in series with the "1" circuits between the pages.

I think this is already in the above list. Right?

- The R12 1k5 vs 7k5 bug I suspected.  (because it's a standard rectifier circuit)

You are right.

------
I have tried to check all differences between 2002 and revised 1996. The 2002 revision has these exceptions:
- 0.047uF / 680 Ohm fiter in U3A feedback loop.
- Missing 22kOhm resistor between pin 6 of U6B and ground.
- 820 Ohm resistor (R97 in 2002 version) is connected between V- and D10,D11 cathode
- 0.1uF cap (C26 in 2002 version) is connected to Ground (not to V- as in the 1996 version)

My feeling is that the exceptions in the 2002 version are mistakes...

[Rob Strand]

While I have it go here, and click on the PCB pic.   I believe this is an original not a DIY.
https://smokingtip.wordpress.com/2013/08/23/ehx-micro-synth-guitar-synth/

At the top right, second row of resistors you can see the 3M3 resistor.  Next to that are the rectifier resistors  (15k and 7.5k)  Below that row is the R16 .    I can see  R16 = 27k.   I can't make out the cap.


Reading your post ATM ...

[Edit:  note the trimpot for the squelch?]

[Rob Strand]

I traced the underside of the PCB to resolve some of these,
http://farm3.static.flickr.com/2399/2165679711_84a204e13a_o.jpg

I think this is already in the above list. Right?

In the list but not on the schematics.

I have tried to check all differences between 2002 and revised 1996. The 2002 revision has these exceptions:
- 0.047uF / 680 Ohm fiter in U3A feedback loop.

Top side PCB pic implies 27k for the resistor.  At this point cap unknown.

- Missing 22kOhm resistor between pin 6 of U6B and ground.

I believe the 22k is required otherwise the circuit doesn't make sense. (It's a current source.)

- 820 Ohm resistor (R97 in 2002 version) is connected between V- and D10,D11 cathode

From the two PCB pics I'm pretty sure it goes across BE of the transistor like the 1996/2017 schem shows.

- 0.1uF cap (C26 in 2002 version) is connected to Ground (not to V- as in the 1996 version)

From the two PCB pics it looks like it connects to -V like the 1996/2017 schem shows.

I hope that is it.   Going through that stuff is making my head spin!

[Rob Strand]

Here's some incomplete notes I made about that effect back in 2003.  Perhaps before I got the fixes list.

As it stands, it is what it is.  There may be totally wrong things in there.
I kind of eye-balled the schematic and made notes. 
Some places I did some calculations.

If you can make use of it then all is well.  If you get frustrated with bugs or you disagree
with stuff then just bin it!
==========================================================
EH Microsynth Analysis
by R.Strand, V1.0, 9 Sept 2003

Overview
The Microsynth is a form of analogue synthesizer for guitar.  It consist of four signal generators whose frequency is derived from the guitar input signal:  Guitar, Octave-up, Sub-Octave and Square-wave (at the input frequency). These four signals are combined with a mixer, each signal having an individual level control.  The combined signal is fed through a VCF (basically an envelope follower) which is an upward sweeping third-order low-pass filter.  The output of the VCF is then fed to an amplitude control stage which modifies the attack of the output signal.  There is also an attack generating circuit, which has a trigger threshold, and performs three functions: control attack of the generated square wave,  control attack of the final output, control start of the VCF sweep. The attack circuits for the squarewave and VCF are really have separate function just that some parts of the circuit are common to both.

There are 11 controls on the unit which control various parameters of each of the stages:

    PreGain             Preamp Gain before all processing ckts
    Guitar                 Level of clean guitar signal before VCF
    Octave               Level of octave-up signal
    Squelch             Octave-up squelch
    Square               Level of amplitude modulated squarewave
    Trigger               Trigger level for note attack (VCF + Squarewave)
    Sub-Octave       Level of sub-octave signal
    Start Frequency    Start Frequency for VCF
    Stop Frequency    Stop Frequency for VCF
    Sweep Rate      Sweep rate for VCF
    Resonance        Resonance (Q) of 2nd order part of VCF

Note: There are two versions of the microsynth, a guitar version and a bass version.
          The only difference appears to be the range of the VCF sweep.

The rest of this document gives a detailed analysis of the EH Microsynth.
Power Supplies
The unit is powered from -10V and +8V rails, the two raill are derived from a single input supply.
The different + and - rails more than compensates for the differences in output swing on each rail of the opamps, as a result the opamp clipping slightly assymetrical.
Input Stage Preamp
The input stage preamp preceeds all other circuity.
The gain is adustable via the PREGAIN control.

Input impedance: 68k,  this will load the guitar pickup.
Variable gain:   x1 to x22
Guitar Signal
The guitar signal is derived from the output of the GUITAR control which is a simple pot connected to the input stage preamp.  The guitar signal feeds the pre-VCF mixer.
700Hz 3rd-Order Low-pass Filter
The 700Hz 3rd-order low pass filter preceeds all of the signal processing elements except the clean "Guitar" signal.  The filter's purpose is to extact the fundamentals of the guitar signal.
Parameters:
1st order:  750Hz
2nd order:  681Hz, Q=1.48
Full-wave Rectifier
The full-wave rectifier is generates the octave up signal and is used as a full-wave rectifier for the attack detection circuit.

The output of the full-wave rectifier is negative.
The input of the full-wave rectifier is high-pass filtered at 56.4Hz (10K//15K and 470nF).
The gain is 27k/15k = 1.8.
The output is filtered by a 3.27kHz low-pass filter.
The SQUELCH control  provides a variable DC offset at the output of the rectifier of -65.5mV to + 81.8mV.
Octave-up Signal
The octave-up signal is a (low Q 2nd-order) high-pass filtered version of the output of the full-wave rectifier.
The level of the octave-up signal is set by the OCTAVE control and the output connects to the pre-VCF mixer.

Loading causing the frequency of the high-pass filter to be affected by the OCTAVE control position.

With the OCTAVE control on minimum:
   R2 = 4.489k
   w0 =2.655E+03 (422.6Hz),  Q = 4.589E-01
  (or, re1= 277.6Hz, re2= 643.1Hz)

With the OCTAVE control on maximum: (nb: the VCF mixer looks like about 450ohms)
  R2=2.955k
  w0 = 3.272E+03 (520.8Hz),  Q=  4.397E-01
  (or, re1 =310.3Hz, re2=874.1Hz)

To simplify we use the geometric means of the real parts,
  re1_gm  = 293.5Hz,  re2_gm = 749.8Hz
  ie.  w0 =   2.948e3 (469.1Hz), Q = 0.4496

re1 only varies +/- 6% which is undetectable, re2 varies +/- 15% which is boarderline

Sub-Octave Generator
...

Attack Generator Level Detectors
Consists of: smoothing filter, low-level squelch ckt, log-amp, and trigger circuits.

The full-wave rectified signal feed a 4th order smoothing filter. The precise 4th order low-pass filter parameters have not been extracted, however, the following 2nd-order cascade approximation agrees within +/- 0.08dB. ( It was found by splitting the circuit into a 3rd order and 1st order circuit extracting the parameters from those then plugging the parameters into two cascade 2nd order in spice an tweaking frequency of the 2nd real stage.)

2nd Order Low-Q:   w0 = 353.8 (56.3Hz),     Q = 0.4572
2nd Order High-Q:  w0 = 268.0 (42.66Hz),   Q = 1.2618

The output of the filter feeds the low-level squelch ckt and the log amp.

The low-level squelch circuit goes high when the signal level is low, this corresponds to a filter voltage above -12.2mV.

The log-amp is pre-biased with a -455uA current.  Given the output of the full-wave rectifier can never exceed +81.8mV, the positive current can never exceed 81.8uA and hence the log amp is always bias with the same polarity and the diodes never get below 373uA of current.  If the rectifier is at maximum output then the current through the diodes is about 8.5mA.   If the diode is 0.64V @ 1mA then at 373uA Vd=591mV and the output of the log amp is at 2.366V.  At 8.5mA, Vd = 746mV and the output of the log amp is 2.98V.  The output of the log-amp therefore has a peak to peak swing of about 4 * 1.9 * 26mV * ln(8.5/0.373) = 0.62V. It's possible to write an equation for the log amp output as a function of the filter voltage,

Vlog = 2.404 + 0.1976 * ln(1-Vfilter/0.455)

The log-amp then feeds a 1st order high-pass filter with cut-off frequency 73.93Hz.  The high-pass filter essentially differentiates the log-amp signal to produce a positive pulse for start of the attack and a negative pulse for the end of note.

The high-pass filter feeds the TRIGGER control which is just a divider.  Two comparitors detect the start and stop pulses.  The attack detector is normally low and  goes high when the trigger signal is above 11.4mV. The stop detector is normally low and goes high when the trigger goes below -8.45mV.
Squarewave Attack Generator
The square-wave attack generator is basically an envelope follower which uses the DC output from the full-wave rectifier.

The rectified signal is filtered with an approximated first order filter.  The first order filter is a tap off the 4th order smoothing filter, the tapping means the response isn't exactly first order, it's actually more like a 1st order filter plus a shelving high-cut filter and even then that's only an approximation.  At high frequencies say above 100Hz it is largely a first order filter.  The cut-off is about 58.1Hz, with this cut-off the error is a maximum of 0.7dB in error below 100Hz due to the first-order approximation.

The filterered signal feeds an ideal diode peak detector and an attack circuit with a 1.03ms time-constant - both the input and output signal are negative.  The attack interracts with these two constants, however, the attack dominates the decay ckt by a large factor.  Note the idea diode clamps the voltage on the time constant cap, and this prevents the cap voltage going positive by a significant amount.

The decay time constant is complicated because there are two time constants, one from the low-signal squelch ckt and one from the stop detector.  The low-signal squelch has a time-constant of 32ms  and charges towards 6.3V.  The stop detector has a time-constant of 564ms and charges towards 6.3V. If both signals are active (which is unlikely)  the time constant is 30.2ms.
Squarewave Generator
The square-wave generator connects to the output of the 700Hz LPF.

The square-wave generator is an inverting gain of 10 amplifier preceeded by a 1st-order 58.95Hz high-pass filter.   It is likely the opamp at this stage clips.

The squarewave generator then feeds  into a low-Q band-pass filter.the square-wave amplitude modulation, which is fed by the square-wave attack generator.

....
Pre-VCF Mixer

VCF Attack Generator

VCF Sweep

VCF

Output Attack Generator
===========================
EDIT: Given we know SQUELCH is a trim pot there's obviously only 10 controls - which you can see on the front panel.

[DVB_master]

Hi Rob,
Thank you very much for the shared analysis and for your time! They are precious.

A 680 Ohm resistor in place of the 27kOhm would make the SQUELCH trimmer un-useful because the offset range regulation coming from the squelch trimmer would be too low, I think... Moreover, it seems that the squelch trimmer is removed from the XO version.
Since I do not have a 1.8nF capacitor for the RC filter on U3A, I will use a 2.2nF cap and a 22kOhm resistor. I think it would be fine as well (considering component tolerances). This solution will lower a little bit the gain of that stage, but cutoff frequency would be almost unchanged. What do you think?

A final question is about the CA3080 / CA3094. I see that some of the CA3094's can be replaced with CA3080 since the
Drive (emitter) pin of the CA3094 output darlington transistor is not always used.
- Is the the output darlington pair the only difference between a CA3094 and a CA3080? No difference in input / out impedance, etc.
- Can I just replace the 3094 with a 3080 when this output darlington is not used? (Did you try?)
- I also asked Renesas (new owner of the IC's) how the Not Connected pins are managed internally in the IC's, but they do not know... I asked this because I would like to use a single PCB layout to manage 3094 and 3080 (when possible) at the same time to check if there are significant differences...

[Rob Strand]

Thank you very much for the shared analysis and for your time! They are precious.

No problem.   Honestly 15 years have past and I'm happy to give them to someone that will use the info.
(BTW, stm swt did a lot of the leg work for that one.)

A 680 Ohm resistor in place of the 27kOhm would make the SQUELCH trimmer un-useful because the offset range regulation coming from the squelch trimmer would be too low, I think... Moreover, it seems that the squelch trimmer is removed from the XO version.

Maybe the 680 ohm is too low in general. I have a feeling some people have built it with 680 ohms and with 27k.  I wonder what the results were.

What do you think?

The preamp gain control has such a large range that the small change would have no observable effect.

final question is about the CA3080 / CA3094. I see that some of the CA3094's can be replaced with CA3080 since the
Drive (emitter) pin of the CA3094 output darlington transistor is not always used.
- Is the the output darlington pair the only difference between a CA3094 and a CA3080? No difference in input / out impedance, etc.

I've always treated them as being the same.   Most specs in the datasheet are similar provided you adjust for the different test conditions.

- Can I just replace the 3094 with a 3080 when this output darlington is not used? (Did you try?)

I haven't tried it but I cannot see why it wouldn't work.   By copying the internal circuit of the CA3094 you could probably add a Darlington to the 3080.  All the info about the Darlington is in the CA3094 datasheet.

- I also asked Renesas (new owner of the IC's) how the Not Connected pins are managed internally in the IC's, but they do not know... I asked this because I would like to use a single PCB layout to manage 3094 and 3080 (when possible) at the same time to check if there are significant differences...

It would be best to use jumpers or 0R resistors to select between the two.   I make it a rule never to go outside of known universe given in the datasheets.

[StephenGiles]

I'm on my iphone in my car with a sandwich in one hand at the moment, but I have some notes somewhere covering which 3094 configurations are troublesome to sub with a 3080 + darlington. I think the notes were given to me at EH UK around 1980 so may have faded over nearly 40 years!!!!

[Rob Strand]

I have some notes somewhere covering which 3094 configurations are troublesome to sub with a 3080 + darlington

I wonder what the problem is?

Here's an old app note.  (The original release of this was probably in the 70's or 80's.)
http://cms-content.bates.edu/prebuilt/ca3080.pdf

The opening paragraph pretty much says they added a buffer to the CA3080.

[DVB_master]

I have some notes somewhere covering which 3094 configurations are troublesome to sub with a 3080 + darlington

I wonder what the problem is?

Maybe some applications where the two output transistors must be matched (inside silicon this is basically for free, since the two BJT are probably close each other)

[StephenGiles]

Perhaps "troublesome" was the wrong word - "requires particular care" might be nearer.

[StephenGiles]

Can't locate those ca 3094 operating notes at the moment.

[Rob Strand]

Just to summarize the outcome:  (I don't know how I got through that yesterday I was pretty darn tired.)

The 2017 Revised Schematic is the most correct schematic.   (BTW: Thanks for putting that up.)
https://www.docdroid.net/u3cx2Xj/electro-harmonix-micro-synthesizer-revised-2017.pdf

The identified errors on this schematic are:
- There should be an additional 10k resistor wired between the circled (1)'s  on the two schematic pages.
   (based on swt's fixes to the 1996 schematic)
- The value of C5 has not been confirm on a PCB. (However swt's fixes to the 1996 schematic didn't say this needed fixing so it is assumed to be correct.)
- R12 should be 7k5 not 1k5.  (As it is a standard rectifier circuit.)
- R14, SQUELCH, is a trimpot.   I suspect it is set so the output of ICA3A pin 1 is zero volts. (However there may be a production setting for this.)

One thing I did note is the numbered wire connections on the actual PCB are *not* the circled numbers on the schematic.  The numbered wire connections are not shown on the schematic.

[Rob Strand]

Maybe some applications where the two output transistors must be matched (inside silicon this is basically for free, since the two BJT are probably close each other)

It should *Not* be critical or depend on matching.  IMHO if you copy the extra resistors shown on the CA3094 output buffer the behaviour should be very close.  If you plonked in a commercial Darlington transistor there you might be subtle differences.

Edited.