Sunday, 4 October 2015

Doeper/ Serge VCS Expander Schematic

A description of the method I used to add Hold, Burst and EOR functions to the Doepfer A-171-2. An update to this post. The module is a licenced version of the Serge/ CGS DUSG/ VCS, so these mods will work on them too, although the pins may be different.

You can download the schematics and Fritzing file here. I wish I had done this at the time in spring as I've had to retrace my thoughts from incomplete notes. It's been fun but I might have some things wrong and I’m sure some aspects could be done better. So, please continue the discussion and post corrections and improvements to this circuit in this forum thread. I will update this post accordingly.


The idea behind the Hold circuit is simple: interrupt the integrator. If you wanted to go no further and keep this mod passive, all you would need to do is cut one trace and hook up a switch. On the A-171-2, I found a convenient place between pin 7 of the TL084 quad opamp and the 8k2 resistor (R39).

If we want to automate this, we need an analogue switch. Transistors can be fiddly and the common CD4066 won’t process all signals, so I used a DG201. This switch is ‘normally closed’, so with no gate on the command input, drain and source are connected and the integrator’s loop is closed. Pulsing it breaks the connection.


Referring to Tim Stinchcombe’s VCS analysis and comparing with the Doepfer layout, pin 4 of the LM3900 is high during the attack phase. The ‘not attack’ gate, which will become part of our EOR gate, can be found on pin 5 of the LM3900. If you observe this output with an oscilloscope you’ll notice we need to process it as the ‘not attack’ gate remains high until the next attack phase is initiated. This can have its uses, but it’s not what we’re after if we want an EOR/ variable length gate that can be used to ping filters etc.

This is where we have to get creative and ‘patch with ICs’: we use another switch on the DG201 to operate a logical AND function. The signal which chops our ‘not attack’ gate down to the right length is the EOC gate. We can tap this signal from the End Out jack on the A-171-2. As the EOC gate is ‘high’ at the wrong time, we want to flip its activity.

The need for a logic inverter conveniently also answers the question of how to buffer the inputs for this mod. I used an HCF4049UBC Hex Inverter. As the supply voltage also determines the logic threshold for this chip (lower voltage = lower threshold/ faster response), I chose to run it off 5V, supplied by a 78L05A regulator. I sent the new EOR signal from pin 15 of the switch to a jack and an LED.


Now that we have the means to process logic, the burst function is relatively simple. It’s a circuit adaptation of the classic Maths Trills patch where we used a logic gate to interrupt the loopback of the EOC signal to the trigger input. As with the EOR gate, we use a switch to function as a logical AND gate. The DG201 is ’normally closed’, so we need to keep the switch open when there is no burst command present. To do this, we use a spare inverter to flip the activity of the burst input. Looking at the schematic on Ken Stone’s site, the trigger and cycle inputs are OR-combined by diodes, so the VCS can be triggered and burst/ cycled at the same time.

The DG201 is powered from +/- 12V. Filter the supply as usual and add 0.1uF bypass caps for the ICs. The burst and hold inputs were conditioned by cutting negative voltages. The schematic says 4001, but I used 4148 diodes.  I added a manual gate by tapping 5V and sending it via an ON-ON switch to either the burst or hold jack’s switching contact.

It sounds more complicated than it is. Once I’d thought it through, I built it on stripboard on the fly, without the need for a detailed schematic. Lesson learned on that score! I hope these notes help and look forward to your comments and improvements.

Thanks to Dieter Doepfer and Chrisi & Erik at Koma Elektronik for their help on the subject of switches.

The usual DIY disclaimer: do this at your own risk, take care and have fun.

Saturday, 18 July 2015

Most Hazarai

Deriving a Beat clock and End of Loop pulse from the Electro-Harmonix ‘Stereo Memory Man with Hazarai’. An update to these previous posts.


I recently attended a talk with Charles Cohen, a musician who has improvised with a Buchla Music Easel for several decades. The Easel is a monophonic instrument but Cohen’s music is multi-layered. To achieve this, he uses long, looping delays to record in a ‘sound-on-sound’ fashion. In Berlin, he used an Electro-Harmonix ’16 Second Delay’, a vintage pedal that’s long out-of-production. A modern equivalent which is also capable of looping is the SMMH.

A stock SMMH will record loops and allow the user to over-dub. It’s fine for guitarists or other manual players, who can keep time with the previously recorded ‘template’ loop. But what if you want to record a clocked sequence from a modular? Laying down the first loop is simply a matter of starting and stopping the recording in time. The problem arises when you want to add a second layer. Although it may be at the same speed, it will probably be out of sync.

I didn’t realize this limitation until I tried Charles Cohen’s layering technique at home. He didn’t seem to have this problem, so what to do? I found a solution that requires a simple modification to the SMMH and - from watching Cohen improvise - a change in playing approach.

Here's what's possible: ehx-smmh-looper-mod.mp3


Unlike the ’16 Second Delay’, the SMMH does not have a clock output. However, its Beat and Loop LEDs do keep time, so this gives us the possibility to derive a clock and, importantly, a reset pulse from the logic signals that drive them.

If we unscrew the base of the SMMH and then compare the positions of these two LEDs with the PCB, we can see where their legs protrude (see above JPEG). If we locate the current limiting resistor R27 + R30 we see that each is linked to a ‘via’, or tiny hole in the PCB. This is where I chose to extract the signal by inserting and soldering a slim solid-core hook-up wire. You could use stranded, but solid gave me an easy, snug fit. A 1K output protection resistor could also be inserted here. I later added mine to the connection at the output jack.

As with the previous SMMH mod, I chose to drill holes in the enclosure with the PCB still mounted as it’s difficult to remove. I protected the PCB with paper and, fortunately, it survived.

In Use:

So, what do we have? The Beat clock is as we expect, but the Loop pulse goes high at the end of the loop. To use it as a reset signal for a sequencer we need to condition it. I use an A-162 Trigger Delay to stretch the pulse just long enough for it to overlap with one of the Beat pulses. This allows me to 'AND-combine' them in an A-166 Logic module. This gives a pulse on the first beat of the bar/ loop. I may find other workarounds or a fix, but for the moment this patch works well.

There’s one last mod I could envisage: a switch carrying 5V to hold the SMMH in perpetual record mode, like the 16 Second Delay. As I already have a trigger input, I just feed it a gate from one of the A-166’s inverters. New material can be added by opening a mixer’s Aux send.

Playing Technique:

Charles Cohen seemed to meditate before playing, taking a moment to consider his next performance. Having done this modification, I now wonder whether this artistic reflection might also have had a technical reason. On the SMMH, the Beat LED does not start pulsing until after the first, template, loop has been recorded. This gives us the opportunity to set the loop length and tap tempo before recording properly: just activate the loop with no signal and ‘record’ a blank template. Yes, that means we also gain a tap-tempo clock!

This modification is a ‘hack’: it’s quick and simple and opens up new possibilities. The great thing about the SMMH is that we can also record the effects in loop mode. Using the SMMH’s looper, I’ve been surprised by how little equipment I need to make interesting music - many of the quick ditties in the above demo were recorded with just one VCO, a set-up not too dissimilar to the Easel used by Cohen.

Friday, 10 April 2015

A-171-2 VCS Expander

Adding gated hold & burst functions and an End of Rise output to Doepfer’s Serge VCS. An update to this post.

A recent forum thread about the Serge 1973 envelope got me thinking about whether it would be possible to add its hold function to the VCS. Once I’d understood what was needed, a burst and EOR pulse were obvious additions.

Both Burst and Hold can be activated manually or by an external signal. Hold freezes the envelope in its tracks. It’s different to patching via a S&H as the envelope continues where it left off. Burst simply cycles the envelope at will.

Audio examples: Hold (cycling VCS FMs a VCO, button pressed to hold), Burst (first manual, then activated by 2nd row of sequencer).

The EOR is needed for quadrature functions with two envelopes (the other fires at the End of Cycle). Conditioning the pulse required some creative thinking, a case of patching with ICs! Its width can be varied, so it can be used to ping filters or delays.

These modifications are simple to implement with switches and logic inversion (schematic to follow). The additions borrow from my favourite envelopes and make the A-171-2 close to ideal. Yes, I could have just bought a Function but it wouldn’t have been as fun or educational!

Sunday, 22 March 2015

Serge VCS Modification

Modifying the Doepfer A-171-2 Serge VCS for more extreme non-linear curves and more manual control over rise and fall times.

The Serge Voltage Controlled Slope is a classic design and an integral part of many modular musicians’ systems. The Doepfer A-171-2 is a faithful recreation of the original circuit. It works just like the Bananalogue VCS that has been in my case since pre-Maths times. That’s good, and bad.

The VCS easily allows one to set different contours for its rise and fall phases. But its non-linear curves are not as extreme as those possible with Maths. Setting times on the VCS is harder, as the useful range is limited to about 20% of the potentiometer’s throw. It’s these differences between these two similar function generators that this modification tackles.

The feedback that generates non-linear shapes is pre-wired on both the VCS and Maths. Calibrating the VCS output voltage from 5V to 8V has only a marginal effect. But if one patches the VCS envelope output to its ‘Both CV’ input with VC Rise + Fall set to linear, the Serge yields the desired curves. This suggests the internal CV feedback loop is capped.

Indeed, if we look at the left of the schematic, at switches SWF & SWR we find 330K resistors limiting the amount of feedback to the CV mixer to about 30%. If we lower this resistance we’ll get more feedback. I desoldered and replaced the 330K resistors with 200K but you could also try soldering another value in parallel e.g. 150K (= ca. 100K, 100%) or 470K (=ca. 200K, 50%).

This part of the modification helped clear up an oddity about the VCS: namely, why making the curves more exponential actually increases the overall envelope time. Negative feedback should make it shorter, and vice versa.

If we study the same part of the schematic again, we can see -12V across a 1M resistor feeding the summing points. Given the gain ratio set by the 100K resistor, this offsets the rise and fall rates by +/-1.2 volts. I find this counter-intuitive, so I removed both 1M resistors.

Both my Doepfer and Bananalogue modules are fitted with logarithmic potentiometers to manually set the rise and fall rates. This means, when using the VCS as an envelope or slew limiter, changes in the first 50% of the pot’s throw are imperceptible. Typical envelope settings lie between about one and three o’clock. Tapering the A50K potentiometers on the A-171-2 with a 5.6K resistor between the CW/ ‘hot’ lug and the wiper solves this. The useful range now spans from nine to three o’clock.

To solder these in place, you’ll have to unscrew the jacks and remove the board from the faceplate. I tape Gaffa around the ends of my pliers to avoid scratches. While you’re there, you can measure the output between the A50K wipers and the subsequent 82K resistors to understand how the log pots choke the voltage. I did try an S-curve taper with two sets of resistors but the quasi-linearization suggested here by Daverj worked best.

So, what does it sound like? Here are two recordings:

feeback mod: exp-fall, unmodified, 0:08 modded, exp-rise, 0:16 unmodified, 0:23 modded.

pot taper mod: cycling, rise = zero, fall manually altered. Stock VCS then modded at 0:39

These simple changes have given my VCS more whip and made it easier to use. Thanks to Dieter Doepfer for helping me read his PCB layout, Ken Stone for publishing his schematic, Tim Stinchcombe and Dave Jones.

If you’d like to try this yourself, take the usual precautions to avoid damage to yourself or your module. I will not be held responsible. If in doubt, ask Doepfer or your technician to carry out the modifications for you.

Saturday, 21 March 2015

NAMM & Noodles

NAMM 2015 was a bit of a blur. It was my first visit and at times a little overwhelming. But it was lovely to meet friends old and new and I’m looking forward to next year. Thanks to Andreas Schneider and the gang for an unforgettable experience.

Thanks also to Dennis Vershoor, aka mono-poly, and his Noodlebar crew for a great time in Rotterdam. Dennis performed at the first Basic Electricity in Berlin so it was nice for me to play the return leg on his home turf.

This was the first time I’d gigged with my mini set-up, consisting of just a Clavia Micromodular, Faderfox LV2, delay and mixer - an exercise in Zen noodles!

NAVS (Berlin) - Noodlebar ::: February 7th 2015 from Streamline Media on Vimeo.

Friday, 5 December 2014

Mungo in Berlin

Recordings of some of the modules from the Mungo range.

John Pillans very kindly allowed me to spend some time with a case of his Mungo modules in October. I've finally got round to editing and uploading some of the recordings I made with these instruments:

The demos are comprehensive, but if you want more you can buy the album for the full, unedited recordings (17 tracks, 45+ mins).

The case contained the instruments lined up here as well as the p0 & r0 percussion and reverb modelling units. I was drawn to the w0 oscillator and the g0 granular sampler in particular, so there are more examples of those. For a primer on Mungo modules in use, read my post on the d0 here.

If the d0 took some time to fathom, what to make of a case full of Mungo? Well, I took them one at a time, using modules from my main system to provide a control or familiar ground. Even seemingly simple modules like the mixer had me scratching my head. In general, I think more visual indicators would be helpful to clearly show the status of outputs and modes. And, depending on the module in question, the Zoom function can be a source of confusion. The massive range and detail offered might make them more suited to the studio than the stage. Patching takes patience but the payoff is full flexibility and excellent audio quality.

A few words on the modules themselves: my interest in the g0 was real-time sampling and manipulation of blocks, rather than grains, of sound. In that sense, it's over-spec'd for me but the underlying technology means that the audio is smooth and clean. Of the modules in the case, this is the one I still pine for. I think John got a bit carried away with the hyperbole when describing the w0, but it really does combine a lot of features. It is digital, so factor in some filtering post FM. The f0 is nice if unspectacular but worth remembering that it too is digital and, as such, different. The m0 is the first EG/ VCA/ mixer combo I've had to use an oscilloscope to understand! The p0 percussion, v0 vocoder and r0 reverb are subtle instruments that require the detailed control offered by the Zoom function.

Watch the official videos to get more ideas and information, including on the n0 noise module. Or, if you're in Berlin, you can try some of the modules at Schneidersladen.

A belated thank you to those who came along to the Mungo workshop, Andreas Schneider and the Schneidersb├╝ro staff for hosting us and to John for taking the time to explain the technology and ideas behind his unique designs!

Wednesday, 3 September 2014

Patch Tips #27 - An FM Equivalent

An analogue take on digital FM featuring the Toppobrillo Triple Wavefolder.

Today's Patch Tip is inspired by MitchXI's description of a digital implementation of FM:

" ... in a yamaha style set-up, an oscillator is simply a ramp that goes from zero to one for every cycle - a phase accumulator - that gets fed into a lookup table that converts zero to one values into a sine wave. if you add a second oscillator's sine output to the ramp wave before the sine wave function, you get an fm equivalent ... to have an fm modulation relationship, you simply add the output of one oscillator into the other oscillator's sine wave function. the only digital computations that need to happen are addition, multiplication (for mod index/amount), and a table lookup."

If, like me, you feel digital is cheating, here's an analogue solution:

For the look-up table we can use the TWF's saw-to-sine function. Patch your saw (carrier) to the TWF via a mixer and trim the bias on a single channel to achieve a clean sine. Apply your modulator via a VCA or directly to the mixer and trim the amount of 'FM' to taste. In this example I used two VCOs/ two TWF channels and one modulator. I start with just the one voice which I pan left when I introduce the second voice:


At high modulator frequencies the result is pretty good. Because the saw-to-sine converter is analogue, it is sensitive to variations in amplitude and DC offsets. And, like Phase Modulation, this method has it's limits: If the modulator is too slow the 'FM' effect will be negligible.

This patch won't replace your thru-zero FM or phase modulation VCO - the maximum possible index seems similar to standard linear FM. But if your oscillators only have exponential inputs, the Triple Wave Folder offers another unexpected method of dynamically changing the colour of your sound.