neil's webbly world

me@njohnson.co.uk
research
learn Russian
  

Jen SX1000 Project


Join the Jen SX1000 group on Yahoo!


On 23rd April 2009 I gave a talk at Anglia Ruskin University about my modifications to the Jen SX1000. You can now download the slides (PDF, 1.7MB).


This page documents my restoration/repair project on a Jen SX1000, henceforth called "Jenny". I acquired Jenny from Tony Allgood at 2002's UK Synth-DIY meeting, with the proviso that it needed a lot of work, not least being a replacement M110 chip.

Undeterred, I brought Jenny back to Cambridge and started looking around for a replacement M110. I found plenty of hens teeth, rocking horse manure and cockerel eggs, but no M110's (they seem to be pretty rare these days). It dawned pretty quickly that a simple plug-and-go repair was going to be out of the question, so that lead me to start thinking about a more complete restoration and modification project.

The result is described on this page, Jenny being completed at the end of July 2003. The results are shown in the following photos:

[Click] Jenny enjoying the sun in the garden.
[Click] The enhanced rear panel, with new mains connector, headphone socket, signal interface, analogue control interface, and MIDI socket. The labels were laser printed onto paper, laminated in our office laminator, and then the labels carefully cut out with a scalpel and glued onto the rear panel.
[Click] The M110 replacement oscillator module installed inside Jenny.
[Click] Inside shot showing the additional circuits and wiring to the front panel PCBs.

User Manual

Thanks to the very helpful Bill Fox I now have a scanned copy of the SX1000 user manual in PDF format. It is rather basic, but it does come in Italian, English, French and Spanish!

Jen SX1000 User Manual (PDF, 2.4MB)

Audio Samples

Now you can here Jenny in action. The following are a few attempts at capturing some of the fun you can have with a Jen.

JenDemo.mp3 (1.1MB) A few notes with varying amounts of PWM, LFO and filter twiddling. The only processing done was a bit of compression to even out the levels a bit.
JenDemo2.mp3 (604kB) Something a bit more together. Two tracks of a soft brass-like sound, plus some drums and reverb.


Contents


1. Introduction

Back to Top

Jenny was in a bit of a poor state when I got her. The photos below show the basic layout. Pretty obvious are the missing coloured knob caps, which all Jen's suffered from due to weak glue.

[Click] Jenny's front--you can clearly see the three rows of knobs, some of which are missing their coloured key caps.
[Click] This shows the back end of this little monosynth, rather devoid of any connections other than a single mono output jack.
[Click] The left-hand side controls the single oscillator, noise source, LFO and, just at the bottom right-hand corner, the output volume.
[Click] The right-hand side controls everything else---the VCF and the VCA.
[Click] The upper PCB is the noise source (switch selects off-white-pink and the pot sets the level mix) and Glide pot. The larger PCB is the oscillator waveform shaping and the LFO. At the top-right is the volume pot on the VCA board (see below).
[Click] The upper PCB is the VCA, which is mostly the ADSR for the VCA. The lower PCB is the filter, which is a 4-pole low-pass filter based on two LM13600 dual-OTAs. The resonance control almost takes the filter into oscillation. The top four pots control the filter's ADSR. The bottom four mix the various control sources (manual, LFO, envelope) and the resonance.
[Click] This is about as simple as a mains PSU as you get: transformer, rectifier, a couple of caps and three voltage regulators (+5V, +12V, -12V).
[Click] And here it is folks---the very rare M110 from SGS-Ates, very rare, and in this case, half dead.


1.1 Knob Cap Colours

Back to Top

The knob caps are often missing on Jen's due to glue that over time seems to weaken and let go. The table below lays out the cap colours on the original machine.

red
tune
red
octave
yellow
vibrato
yellow
speed
white
frequency
white
resonance
yellow
L.F.O.
blue
env.level
red
waveform
red
pulse width
red
P.W.M.
red
level
blue
attack
blue
decay
blue
sustain
blue
release
red
glide
gold
noise
gold
level
deep red
output volume
blue
attack
blue
decay
blue
sustain
blue
release

The colour scheme is as follows:

  • red for VCO,
  • yellow for LFO,
  • white for VCF,
  • blue for ADSR,
  • gold for noise,
  • deep red for master volume.

The Jen originally came with a set of large patch sheets that went over the entire front panel, with red and blue marks to set the controls for a particular 'patch'. Someone scanned it and posted it to Vintage Synth Explorer. Here is a blank one (800kB) you can print out and use for your own patches.


1.2 Original Schematics

Back to Top

Here are scans of the original (pre-modded) schematics (copied from Jezz's site):

  • Page 1 (833kB) - Master oscillator, voice generator, PSU
  • Page 2 (793kB) - Waveform generator, LFO, glide osc.
  • Page 3 (529kB) - Filter, envelope generators
  • Page 4 (659kB) - Filter (continued), noise generator, VCA

A single-page version is available.


2. Condition and Problems

Back to Top

Ok, so you know what she looks like, now for the technical stuff. I catalogue here what I see as problems that need rectifying to get Jenny into a decent state of operation.

Item Problem
Pots To begin with, all the pots are way too noisy and scratchy to be useful, so they will need relacing.
Mains I hate permanent mains leads, for two reasons:
  • They are usually too short most of the time (Sod's Law)
  • Makes carrying equipment around more hassle, as you end up tripping over the lead, or treading on the plug, getting it caught it doors, etc.
So, plan to fit a standard IEC mains socket on the rear.
Oscillator The main issue is the half-broken M110 master oscillator. I say half-broken, because the keyboard scanning and trigger/gate works, but its the oscillator outputs that seem dead (this might have something to do with the range switch being a make-before-break, so shorting together two outputs momentarily...something else to replace).
Since the oscillator also resets the sawtooth at each cycle the sawtooth output does not work.
Lack of "grunt" Many people have commented out the lack of "grunt" or "punch" that the Jen has, often citing the digital oscillator as the cause. However, the EDP Gnat also has a digital oscillator, and that seems to pack a powerful punch, so I think there is hope for Jenny.
Knob caps The missing knob caps look ugly, so replacements must be found somewhere, or a completely new set of knobs.


3. Modifications

Back to Top

This section catalogues the various modifications I have made to Jenny. They are not in any chronological order, so you may find some of the time-related references in the wrong order!

Please note that these modifications were made to my Jen for my interest and amusement only. If you try any of these mods yourself and damage or break your Jen (or someone else's Jen) then I cannot be held responsible. The details are presented purely for discussion.


3.1 VCO

Back to Top

Sub Oscillator

Previous sub-oscillator designs have constructed dividers from JK flipflops, limiting any practical size to 2 octaves (divide-by-4). I chose to use the 4024 counter, giving up to 7 octaves below the output of the main oscillator.

The circuit is quite simple, taking input from the main oscillator square-wave output, and selecting one of the outputs of the divider stage. Which one you choose is more a matter of personal taste, and you have lots to choose from! I left myself enough free wire to reconfigure the output if I get bored with the current setting.

Note: I could have used a rotary switch to select the divisor, but there isn't any room on the front panel.

Sawtooth Multiplier

Now that the level knob for the noise/sub-oscillator is going spare, I decided to add a sawtooth frequency multiplier (Schematic). It consists of two parts:

  • Variable Threshold Comparator --- when the sawtooth level reaches a threshold set by the front panel control then the sawtooth waveform is reset;
  • Variable Gain Block --- resetting the sawtooth early will obviously reduce its amplitude, so this gain block keeps the output at a constant amplitude.

The current circuit is configured to double the frequency when the control knob is at halfway, then all the way up to about 10 times the input frequency.

Update: A small mod to allow the LFO the gently modulate the saw trigger point. It uses the PWM LFO pot, marked "P.W.M.", to inject a small amount of the LFO signal into the sawtooth multiplier bias circuit, modulating the position of the the sawtooth reset point. It is subtle, but gives the sawtooth a little extra movement to the sound, thickening it up.


3.2 VCF

Back to Top

Filter Pedal Input

The same mod as for the swell pedal input (see below) but this time increases the filter cutoff frequency.

Normalised Mod Wheel Input

If you fit the MIDI board then you might also want to make the normalised VCF control input be the Modulation Wheel control voltage. In that case use a jack socket with normalising contacts and connect it up to the appropriate output from the MIDI board.

Inverted Envelope

Nothing original here--its an inverting amplifier wired across the envelope amount pot, as described with schematic and board layout on Bluebear's site (link below).

Remove Input Filter Capacitor

I had read on Håkan's page that the VCF suffers from input high-pass filtering by its input capacitor. Closer inspection of the input stages shows the 1uF input capacitor forms a high-pass filter with the 220-ohm input resistor, whose cutoff frequency (-3dB point) is around 720Hz.

The solution is remove the VCF input capacitor (replace it with a wire link) and to fit both the VCO output and the noise output with their own DC-blocking capacitor and output resistor.

Another BIG change to make is to replace the filter integrator capacitors. The original ones are cheap'n'nasty disc ceramics---fine for decoupling, but bad for audio. These have been replaced with 270pF polystyrene capacitors (the closest I had at the time) and sound much nicer.

I have also tweaked the offset and resonance trimmers to improve response of the filter. Unfortunately the design of the filter is such that the Q seems to be frequency-dependent (greater Q at higher frequencies). I may look into the maths of this filter and see if there is a way to correct this (perhaps turn it into a state-variable design).

I also note that on the schematic, the output of the VCO level pot goes direct to the VCF's input capacitor. In practice this is not the case---there is a 47k resistor in series with the wiper.

The circuit for the VCF input stage is a two-input mixer---one input from the VCO and one from the noise generator (see elsewhere). Operation is quite simple: the dual linear pot mixes the two input sources (VCO and Noise). The 33k resistors bend the linear response to almost-log, and is a cheap way of making a log/anti-log pot. The 47k resistor and 4u7/10n capacitors couple the two DC-biased signals into the VCF input.

Filter Response Analysis

Online (especially eBay auctions :-) there is a common misbelief that the SX1000's filter has a 12dB/octave response - the usual kind of response for a 2-pole filter. However, the schematics cleary show four filter sections, so the obvious conclusion is that this ia a 4-pole filter, with a 24dB/octave response. In theory.

Until now...

I have run an experiment. Using the white noise source to provide a flat spectral input to the filter I have measured its response. The result confirms that the filter exhibits a response far steeper than -12dB/octave, confirming the suspicion of a 4-pole design. The analysis was done using Visual Analyser.

Filter Response shows -24dB/oct

The graph shows a slope of about -20dB/octave. This is slightly shy of the -24dB/octave you would get from an ideal filter, but I believe this discrepancy can be explained by component tolerances between the four filter stages - you would only get the full 24dB/octave slope if all four filter sections were tuned to exactly the same frequency. In the SX1000 the resistors around the filter are all 5%, and while the capacitors in my test SX1000 have been upgraded to polystyrene (from the original ceramic) the tolerances of those capacitors is also 5%.

So, to reiterate - the Jen SX1000 has a 4-pole -24dB/octave filter. Yay!!!


3.3 VCA

Back to Top

Mains Switch

The first mod to make is replacing the fixed mains lead and rear rocker switch with an IEC socket, and to replace the 10k LIN output volume pot with a 10k LOG switched pot, thereby (a) replacing the nasty mains lead with a nice IEC socket, (b) bringing the power switch to the front panel, and (c) fitting a proper "audio" curve volume control.

I have also added a blue LED to provide visual indication of when it is turned on, now that old neon rocker switch has gone.

[The blue LED in action!]

Fortunately, the 30mm x 22mm hole occupied by the old mains rocker switch is just the right size for a standard IEC socket. Well, I'll qualify that---the one I bought from my local Maplin shop was 1mm too big in both axes, so a careful bit of trimming with a Stanley knife (0.5mm off each edge) and I have a mains socket that is a perfect fit.

This leaves a round hole to the right of the rear panel, which is used for the headphone socket (see below).

HAZARD: This modification involves Mains Wiring. ONLY attempt it if you are competent with mains voltages, or get an electrician to check your work before you connect it to the mains!!

[View the new switched pot on the VCA board]

Swell Pedal Input

Adding a swell pedal (increases volume on top of what is set on the front panel) is relatively simple---we just need to feed in some more current into the emitter of the CA3080's current converter transistor.

[View the VCA board mod]

Headphone Amplifier

Again, another simple addition, but quite useful for solo playing when you don't have an amplifier to hand. We tap off the audio signal from the wiper of the master volume pot and feed this to a stereo low-power amplifier based on a TL072 (Ok, not HIFI but sufficient for this application).

In case you're wondering why I chose to build a stereo amplifier, rather than just connect left and right phones to a mono-amp, its to allow future expansion to stereo (perhaps a panner or chorus unit?)

The additional take-off lead from the VCA board can be seen in the mains switch picture (link above).


3.4 LFO & Noise

Back to Top

LFO Wave Switch

The standard LFO generates one waveform, a triangle wave. Not happy with this, I have modified the LFO to both operate over a wider frequency range, and also do rising and falling ramps.

To select which waveform is generated I had to fit a switch to the front panel.

[This is what it looks like]

The circuit for the modified LFO shows the components that have been added or changed (the oroginal components are left unmarked). The switch determines the output waveform:

  • A - rising ramp
  • B - triangle
  • C - falling ramp

The estimated frequency range of this LFO is 0.2Hz to 212Hz, enough to have plenty of possible uses.

For the basic modification (excluding the wave shape switch) just remove the switch and any corresponding components. Then the modification to the original Jen LFO (see schematic page 2) is made with the following five steps:

  1. Remove 22k resistor from pot track (live end) to ground;
  2. Replace 150k resistor with 4k7 resistor;
  3. Add 100R resistor from pot trace cold side (CW end) to ground;
  4. Insert 47k resistor into pot wiper track;
  5. Replace pot with 100k log pot.

External Audio Input

A simple modification allowing external signals to be fed into the VCF requires the addition of a mono switched jack socket. The circuit is simple, adding the socket to the sub-oscillator feed to the noise selector switch.

Analogue Noise Generator

The original noise source in the Jen is an MM3857N digital noise generator. It sounds horrible, more like a steam engine chuffing along!! So, this is soon replaced with a true analogue noise generator, based on a reverse-biased transistor junction.


3.5 Keyboard

Back to Top

Here is the circuit (corrected, revision B) of the first version of the M110 replacement module. I've squashed it onto one page, so it fits within the limitations of the free version of Eagle.

Note: Missing from the schematic is the information that IC6 is a ULN2803A octal driver.

Pictures of the M110 replacement module:

  • top, showing all of the passive components fitted, a few chips, and waiting for the micro and DACs,
  • bottom of this compact PCB, showing the 40-pin DIP header ready to sit in place of the dead M110. The additional connector is for additional power supplies, the MIDI input, and the CV/gate/modulation outputs.

Schematic and PCB

Eagle Schematic and PCB files available to download:

Whichever one you choose, they both expand into a subdirectory called Jen1KProc

If you want to make your own board, I suggest you download one of the above files and then extract the .brd file and send it off to PCB-Pool who will turn it into a PCB for you. I used them for my board without any problems. You might wish to add a solder mask and/or silk-screen position print. I kept my board simple and cheap :-)

Software

Software is available in the following files (depending on your choice of decompressor:

(Note: the cformats.tpl is a template file for Programmers File Editor, which I use for source code editing.)

Both archives expand into a sub-directory called OS. I have stripped out the intermediate files, so there are just the source files, the RCS directory if you want to see how the development progressed, and the final jen1k.hex binary file for programming into an ATmega8L.

Note: this archive does NOT include the AvrX free open-source RTOS which is needed to compile the code. I suggest you install it in the directory c:\usr\lib\avrx.

You will also need a copy of AVR-GCC to actually compile the code; I use WinAVR version 3.3. You might need to frig around with directories in the makefile for the Jen software (especially where the compiler can find the AvrX libraries and headers).

Alternatively, just use the jen1k.hex file to program an Atmel ATmega8L. The AVRDUDE program included with WinAVR is pretty good. I use an STK200-style programmer that was kindly donated by Insight Memec. Thanks guys!!

Note: the ATmega8 fuses need to be programmed with the following bytes:

  • Low Byte Fuse: lfuse = 0x1F
  • HIgh Byte Fuse: hfuse = 0xD9 (no change from factory default)

If you find any bugs in the code, please do let me know so that I can fix them.

Construction

Building this board will be slightly fun due to some areas being a little squeezed together. If you use a through-hole-plated PCB then assembly will be easier. The only trick to be wary of is fitting all components before fitting the 40-pin turned-pin sockets, which otherwise get in the way of soldering.

Depending on the size of your programming header you might have to attack it with a knife to shave off the top millimetre of plastic, which would otherwise foul the bottom of the case. Just don't cut your fingers off!!

Installing the Processor Board

Three changes to the main oscillator board before fitting this board into the M110 socket. These changes are:

  1. Remove the resistor-capacitor pair that drive the base of the PN2222A transistor;
  2. Remove the PN2222A transistor and link the B-C pads;
  3. Remove the 560R load resistor connected to the collector.

These changes remove the level converter on the output of the master oscillator. I found this gave so much distortion of the 2MHz clock signal that it was better to run the signal at 5V direct from the 74LS221.

Several changes are necessary for the keyboard contacts PCB. They are:

  1. Remove the 5k6 pull-up resistors;
  2. Remove the redundant caps and trimmer;
  3. Replace the old M110 socket with a new turned-pin type.

You need to use 40-pin IC sockets to space the processor board away from the keyboard PCB. It is best to use turned-pin IC sockets. Fit turned-pin SIL headers to the PCB, and then carefully push a 40-pin socket onto the header. You can see in the above photo the header strip and the IC socket.

One final modification to make is a hard link from the Modulation CV output to the normalised input of the VCF pedal input. This allows you to control the filter cutoff from the Mod Wheel of your MIDI keyboard.

Calibration

The only calibration on the board is the CV buffer gain trim. This sets the scale of the CV output, which should be +1V per octave. If you have a calibrated +1V/oct VCO then use that, and adjust the trimmer until the Jen is in tune.

An alternative method requires a digital voltmeter, the more digits the better. With the meter measuring the CV output, play the lowest note on the Jen keyboard and note the voltage reading. Now play the highest note, which is three octaves up. The voltage should jump up by exactly +3 volts. If not, adjust the trimmer and repeat. Repeat the low note-high note cycle until your meter shows a jump of +3 volts.

The MIDI channel is determined by the settings on the MIDI channel header You can either add an external selector switch (perhaps mounted on the rear panel) or hardwire it on the board. If you leave all connections free this selects MIDI channel 1. The channel number is one greater then the binary value set on the header (all links in place = 15, so channel = 16).

Finally, choose whether you want your Gate signal to be +5V high or +12V high, and fit a link accordingly.


3.6 Interfaces

Back to Top

Rear Panel

After some concerted panel-bashing I now have enough new jack sockets to support all the additional connections into the system.

[This is how I fitted them all in]


4. Suppliers

Back to Top

Here are details of suppliers for various parts used during Jenny's restoration.

Rapid Electronics
A good supplier, used for many of the other parts, including the replacement mains transformer (part no. 88-0300)


5. Links

Back to Top
Bluebear's Page
Has circuit diagrams, datasheets and simple modifications (sub-osc and inverted envelope, which I have borrowed :-)
Sound-On-Sound Review
Gordon Reid's review of the Jen in the respected press!
Håkan Eriksson's Page
Has modified his Jen with a mod to the filter input and mods to sort out the differing noise and oscillator waveform amplitudes.
Till Kopper
Till has a comprehensive range of synths, including a Jen SX1000.


6. Servicing

Back to Top

The SX1000 is a fairly reliable machine. However, as well as the M110 dying there are a few other common faults that are quite easy to repair.

Broken glide oscillator
Symptoms: You play a key, and the note sounds. Then you play a different key but it sounds the same pitch. This fault even occurs with the MIDI retrofit.
Cause: Both the M110 and the MIDI retrofit need the glide oscillator to move from one note to the next. If the glide oscillator has stopped working then the pitch will not change.
Fix #1: The glide pot needs cleaning. First try turning the glide control full circle a few times (first left, then right, then left, etc) to see if that helps. Otherwise open the synth and squirt some switch cleaner into the pot. If that fails, replace the pot.
Fix #2: The glide oscillator is a simple circuit based around a single op-amp. It is located on the small PCB at the lower-left of the front panel, together with the noise generator (MM3857N). The original was a LF351. A good replacement is the TL071 or TL081. Both are readily available from online suppliers or your local electronics store (e.g., Maplin in the UK). Just remove the old chip and fit the new one, making sure it is the same way round.

Copyright © 2001-2014 Neil Johnson