Oberheim Xpander – Episode 1

Oberheim Xpander

Overview In 2020 I researched the original OB-Xpander analog synth and posted my story about the prototype here. I concluded that whilst it might be possible to recreate it from OB-8 PCB’s and modified OS software, the effort was not worthwhile for a 4-voice synth. However it might be worth recreating the production Oberheim Xpander with its 6-voices but the time needed to design the five PCB’s meant it was never going to be viable for me, although people have recreated the Jupiter 8 and Memory Moog from scratch.

Then it all changed! In Spring 2021 I noticed that Syntaur had NOS Oberheim Xpander PCB’s for very reasonable prices; both the Voice and Processor but no Pot PCB. They also had the Power and Display boards but these are not needed as I will use a modern PSU and new VFD replacement displays. Before buying the PCB’s I carefully researched the availability of components, especially rare DAC’s and old ROM, RAM, and concluded everything was easily available except the 14-bit DAC. Even the numbered buttons could be bought from Mouser (if 16mm square), and the CEM3374 and 3372 are available at just £20 each.

I priced up a full BOM, which came to a total of around £2000. This is an expensive project but cheaper than a second hand one at £4000, and I would get a brand new PSU and set of displays, so my new build would be future proofed. The downside is the resale value is probably low but then I don’t plan to sell it!

Xpander Panel Mockup

The Elephant This all sounds promising but the “elephant in the room” is the panel. Whilst I may be able to get a metal case manufactured, and I can buy new wood sides, I sure can’t make either a new Mylar film or a silk screened panel. The completion of the project is reliant on someone else doing this, as replacements for existing Xpander panels.

As at Spring 2021 there are two possible contender projects which seem active, although it could be a long wait for these projects to be in production! However it will probably take me 6 months to build the Processor and Voice boards, so I am hoping the timing works out! I will build a plywood base to mount the Voice and PSU and work out a temporary panel to hold the Processor and Pot Boards.

A smaller “elephant in the room” is that I don’t have a Pot PCB, and I will have to make one. This is tricky to get right in terms of dimensions before I have a proper and accurate front panel, so I will probably use a set of smaller prototype PCB to prove the new Xpander works and then get a full sized one made when I have a panel. A one off PCB this size is around £300, so its a major expense and something I have to get right.

Progress in April The Processor and Voice card PCB’s have been ordered from Syntaur and the CEM3372 and CEM3374’s acquired. I have started to identify the type of capacitors needed in the Voice board using detailed photos of the Xpander and Matrix 12 PCB’s.  During May I hope to acquire the new VFD displays and some of the hard to find components (DAC’s, RAM, ROM’s). In terms of critical path I can build the Processor Board first but really need a Pot board working to see if the CPU boots up and loads the software. I also need a power supply!

Xpander Mockup Panel was created by Sunshine Jones and its his copyright – More in Episode 2!

Obie-Wan – Episode 2

Modulation Panel

Introduction In January 2021 I set out to recreate the Oberheim OB-1 in a Waldorf KB-37 keyboard and Eurorack case. The OB-1 was one of the first analog mono synths with presets and luckily the schematics are available, unlike the PPG 1003 Sonic Carrier. It also has a full set of hardware controls, unlike the later Moog Source with its membrane control pad. The OB-1 also sounds fantastic with a SEM VCO, a Roland VCF and a set of CEM based ADSR’s.

The Obie-Wan project is part of an AMSynths plan to diversify from analog modules, to complete mono analog synths, ending up with replicating analog poly synths (like a rack munt SCI 600).

Progress Update By late March I had the analog PCB’s (VCO, VCF, VCA & ADSR’s) at prototype testing stage, and set them aside whilst I focused on getting the rest of the synthesizer boards designed and manufactured. I also got the left hand MODULATION panel manufactured, as a test of grey and white printed on black anodised aluminum. The quality is very high but the grey is a bit too light compared with the white lettering. I may go for a darker grey.

CPU and TOUCH Test

First CPU Test I populated the Tauntek CPU PCB, which replaces a lot of the Oberheim processor board with a PIC18F2525 processor and a smaller number of CMOS chips. This solution was my project breakthrough, as it reduced the number of chips and the size of the CPU board.

Before getting the large PANEL PCB manufactured, which holds all the pots and switches, I wanted to check that my substitution of momentary switches for the original capacitance switches was going to work.  I designed a small TOUCH PCB that contains the 10 momentary switches with indicator LED’s, and the CMOS logic that deals with the pressing of switches and the lighting of LED’s. The switches I am using are the same as used in the Prophet 5 (E-Switch E5501), whilst wider versions were used on the Oberheim OBX.

Preset Switches

I made a wiring loom to connect the TOUCH PCB to the CPU board, and after checking the voltages were ok, I powered the two boards on. Major Success! The new CPU sequenced the LED’s at power on and then rests at Preset 1. I was able to change to any of the Presets, and enter and exit WRITE mode successfully, after correcting three errors I had made in the schematics. This proved the TOUCH circuit and the CPU board, and it also avoided my mistakes being made on a costly PANEL PCB.

The next set of PCB’s Originally I planned to have a single large PANEL PCB that would contain all the scanning and the residual analog circuits that don’t fit on the individua PCB’s. However this was too much circuitry for a double sided PCB and I didn’t want the expense of a 4-layer board.

I have separated out the scanning, modulation and analog circuits with a LFO PCB set behind a 16HP MODUATION panel to the left, and an ANALOG PCB, that sits behind the main PANEL PCB, and contains the additional analog circuits. My PANEL PCB hold the pots and switches, the 4051 multiplexers and the TOUCH circuit. I have to be careful not to have more than 2 PCBs in depth behind the right hand side of the design to avoid interference with the KB37 power supply.

The MODULATION section provides connectivity to the Aftertouch, Velocity, Bender and Wheel CV’s from the KB37 using short patch cables. The PITCH and GATE CV’s are derived from the Doepfer bus with the KB37, so only 4 cables are needed. The main OBIE-WAN panel is 84HP wide, which leaves 7 HP spare at the right (I may add a BBD Chorus).

I also designed a SYNTH PCB which holds the DAC and Sample & Holds, and a CPU PCB that replicates the Tauntek processor board. Both these PCB’s are 90mm in height and can fit on the base of the KB37 with ribbon cables going to the PANEL and ANALOG boards, and power coming across from the Doepfer 16-pin bus.

I checked the PCB’s for accuracy and placed an order on 02 April and ordered a set of E5501 switches from Mouser. Work continues on designing the 4 remaining PCB’s.

Power Rails The original Oberheim OB-1 has rather a large number of power rails which I need to replicate:

• +18V for the VCO cores
• +/-15V for the Op Amps and analog circuits
• +10V for CMOS logic
• +/-7V for CMOS logic
• Additional +5V for the PIC

There are also two reference voltages; AREF for the PIC microprocessor at +4.5V and VREF for the DAC at +10.69V.

For one of the 2019 Behringer projects I tried out the Oberheim SEM VCO with +/-12V analog power and +15V for the VCO JFETs. Whilst I didn’t proceeded with the  Behringer project, I did get the VCO to work successfully. This is going to be my approach on this project, as the KB37 has a 15V switched power supply that is then regulated down to +/-12V and +5V for the Doepfer power bus. The OB-1 has a central +10V power supply but I am devolving this and the +/-7V rails to local linear regulators on each PCB that needs it.

Next Steps The plan for April is to bench test the VCO and VCF, complete the design of the remaining PCB’s, test my CPU and SYNTH PCB’s with the Tauntek PIC microprocessor. I am aiming to complete the Obie-Wan project by the end of May! More in Episode 3!

Obie-Wan – Episode 1

Introduction The Oberheim OB-1 was the first analog mono synth with presets, closely followed by the PPG 1003 Sonic Carrier. Announced in January 1977 at Winter NAMM and released back in November 1977 the OB-1 is a rare synthesizer, even though hundreds were made, it is even more rare in the UK where Oberheim had no distributor and the UK price was over £1200.

The OB-1 was innovative in 1977 but its market started to be eroded in 1978 as analog poly synths with patch memory entered the market. Whilst the Prophet 5 and  OB-X were considerably more expensive, the Jupiter 4 (1978) and 2-voice Pro Mars were closer in price to the OB-1. With only eight presets the OB-1 looked dated by 1979 as did its pre-microprocessor design. Moog did attack the preset mono synth market in 1981 with the Source but ditched real time control.

The two oscillators have a SEM heritage, and the VCA uses a CA3080 OTA. but the legacy ends there. The filter is a world away from a SVF with a Roland-like 2 or 4 pole OTA filter. The envelope generators were initially OTA based but replaced with CEM3310’s in the Mark 2 OB-1.

The modulation options are fantastic, with noise as one of the sources. The OB-1 is a precursor to the OB-X and OB-8, which harness the full set of CEM chips to great effect.

Why an Obie Wan?  I had the idea of building a replica of the OB-1, initially as a wide 84HP euro rack module.  So why is the OB-1 an attractive synth to replicate:

• Variable waveform VCO’s.
• Sub octaves on both VCO’s.
• OTA 2/4-pole filter (not a SEM SVF).
• Fast and snappy envelopes (CEM3310).
• Fantastic bass and lead sounds.
• 8 user presets, upgradeable to 64!

How hard could this be?

Inside an OB-1

Analog Boards The first and easy task was to design a set of analog PCB’s, one for each major component (LFO, VCO, VCF, VCA+ENVS). The schematics are easily located and I have a high resolution copy of an original service manual. Each PCB is 70 x 105mm and this is familiar analog electronics which is easy to get right. A set of PCB’s were ordered in late February 2021. The LFO board will need to be revised to include the performance controls – see below.

Digital Boards The OB-1 uses a ton of CMOS chips to scan the switches and pots, store presets, strobe the memory and S&H’s, and light the switch LED’s – as well as controlling the analog synth. Although the Z80 was launched in July 1976, and I remember the impact very well as I was studying electronics in the UK – it took another year for the Z80 to be accessible to synthesizer manufacturers, as software development was needed – which was more expensive to set up than using CMOS digital logic. You needed a computer to create the software!

The Z80 powered the first set of poly synths in 1978, starting with the Prophet 5 and the Jupiter 8. Prior to this PPG, SCI, ARP and Oberheim were using CMOS logic in their synths. and making use of high cost RAM memory back in 1977, which seriously constrained the size of storage. The OB-1 is frugal on memory and uses just 1k bit of CMOS RAM. This means that serial processing and 6-bit values for potentiometers is required. The Oberheim OB-1 case is large and for a reason, there are a lot of circuits in there!

Game Change Replicating all the CMOS logic in the OB-1 processor board was daunting and I knew it would take up  too much PCB space to make the project viable. Thanks to Tauntek the CMOS processor has been replaced by a modern PIC18F2525 processor on a smaller PCB. Whilst this PCB is still too big for my project, it can be redesigned to fit. The new processor PCB effectively upgrades the OB-1 into the 21st century with 64 preset patches, no need for a battery, although we still have the 6-bit data values with 64 values and only half of the MIDI standard of 128.

Waldorf KB37

The KB37 Host The Waldorf KB37 is a fantastic piece of design and engineering, very well made and with space for 107HP of Euro Rack. I initially used it with my SH05 modules, but it seemed wasted on a mono synth with no integration of velocity and aftertouch. Patch cables were needed everywhere and it was just a bit messy. I nearly sold it, until I released what it true role was – to host my OB-1!

Fitting the OB-1 into the KB37 was not as easy as I first thought. There is only 25mm of depth in the right hand side of the KB37 to clear the power supply. This restricts the PCB depth to use one layer, two with care. There is a Waldorf PCB at the bottom of the case facing the rear with the 5V and 12V regulators, Doepfer style power headers and the digital circuits for keyboard scanning and MIDI.

I decided  to tailor my design to the KB37 and it is not transferable to a rack mount case. I think the advantages of having the keyboard and performance controls outweighs the limitation of not being an 84HP super wide module. The KB37 brings MIDI in and out and instantly makes the OB-1 MIDI capable, it includes an on board arpeggiator with glide and the aftertouch and velocity from MIDI or the keyboard can control the OB-1 filter cutoff and the volume by using an updated performance control module.

More in Episode 2!

Jupiter One Synthesizer

Overview The Jupiter One project is an original Jupiter 4 Voice Card turned into a complete single voice synthesizer, with the addition of a replica Controller PCB and the addition of a new Pot PCB that holds the switches, LED’s and potentiometers. The project started in 2018, but was paused for a year whilst I worked on Behringer projects, and restarted over the Xmas holidays of 2020. This post describes the work to get to a completed and full working synth.

I decided to replicate the Roland Jupiter 4 Controller functionality, which translates the 0 to +5V  potentiometer voltages into the CV’s required for the voice circuit even though I am not planning to add in the preset memory capability at this stage of the development. It also contains the LFO, Noise and half the ADSR’s, so I had to build at least 50% of it anyway! The Voice Card tends to use +15V on/off signals whilst the panel and DAC of the Roland design uses +5V. There are various level shifters on the main PCB which I have simply omitted and used +15V on the Panel PCB where needed.

A future project could be a replica Pro Mars but the Roland design is incredibly complex, very inefficient in terms of components and space and requires a lot of trimmers. The OS is locked up in a 8048 chip, so it not easy to modify the code to add more presets. A replacement microprocessor was available as the IO product but that appears to be history. I am replicating the Oberheim OB-1  preset monosynth instead, using a new microprocessor (already programmed) to reduce the complexity.

Controller PCB This new PCB holds the LFO and Noise circuits, plus the CV control stages for the VCO, HPF, VCF and VCA and six clock generators for the two ADSR’s. The LFO of the Jupiter 4 and Jupiter 8 are very similar and I have used the later design as it uses a CMOS analog switch to select the waveform. Initial tests proved the LFO worked well with a minor resistor value correction to get close to the original frequency range, but with a higher top frequency of 200Hz (not 70Hz) which I have retained.

The Noise circuit worked first time and uses the same 2SC828R transistor from the legendary TR808, which delivers a wide noise spectrum. In fact the noise level was too high and I had to trim in back to 2V p-t-p, as it would bleed through the JFET on/off switch at higher output levels.

Oscillation The VCO based on the ua726 chip worked first time, with the correct sawtooth and square waveforms, and frequency range selection, but the pulse width circuit needed some work as its uses a DP4T switch which I have not be able to source. The PW on/off switch was stuck on LFO modulation, and needed +15V as the reference voltage not +5V (level translation issue). Pulse width selection now works, with 50%, 40%, 20& and 10% widths, although the 10% is very narrow more like 5%!

The LFO bleeds slightly into the VCO frequency even with the modulation turned down, so this needs some attention.

Filters The HPF and LPF control voltages were initially too low and too high, although the filter did resonate correctly. The HPF control circuit works well in my 8104 filter module, so I correctly suspected a mistake by Roland at R111, which should be 47K not 100K. The LPF control voltage is being driven too high by the non existent Foot Pedal input which I have now removed. Both filters now work well and trimming is good.

The original filter S&H circuit uses the digital arpeggiator clock as a sample source, this is not available as it needs the 8048 microcontroller. I have used the LFO as the clock and S&H circuit from the Jupiter 8, with the front panel slider controlling the level of S&H into the VCF using a BA6110 VCA, so this could be preset in a future version with patch memory. My first attempt used an expo transistor pair which made the slider only responsive at the very top, I changed this to a single transistor linear response (as used in the JP-8).

Envelope Generators The original Roland design is unique, and a very complex way to get to a voltage controlled ADSR. It uses CMOS clocks and switches to get the voltage control and was replaced by a single IR3R01 chip in the Jupiter 8. The voice card contains half the CMOS circuit, so I have added the other half to my Controller PCB. I have proven the design by building the AM8007 Env Generator module, so I am confident I can get this working.

The filter ADSR worked immediately, whilst the amplifier ADSR was dead. I replaced the buffer Op Amp at IC21 which is a 4558 chip.

VCA and Output The voice card contains a BA662 VCA with a volume control on the front panel, and its own dedicated ADSR. This was easy to get working properly. This is also a final analog volume control before the output signal goes to the rear mounted Output jack socket.

LFO Modulation The original Jupiter 4 has a rather nice modulation panel to the left of the keyboard. I have replicated this using a thumb joystick that provides vertical pitch bend and horizontal LFO modulation depth to the VCO, VCF and VCA using the same three selection switches, and two rotary knobs.

Inputs and Outputs My Jupiter 01 has four 3.5mm jack socket inputs; traditional pitch and gate, and then X and Y inputs which over ride the Thumb Joystick control voltages and enable velocity or aftertouch to be patched into the synth. There is a single 6.35mm jack socket mounted on the rear PCB which also houses a Meanwell power supply.

Pot PCB My initial design used DP4T slide switches for the PWM, LFO waveform and Key Follow selection. But the switches could not be wired as dual four way, so I need to rethink.

Chorus Options The Jupiter 4 used a BBD chorus circuit to add some much needed depth to the single VCO sound, which I suspect I could do with as well! I could recreate the MN3004 fixed rate chorus as an internal PCB or go with a BBD-320.

Outcomes More work is needed on my PCB design. This set of tests proved many of the circuits but also showed up some issues – the 4053 LFO switcher needs +/-8V rails and the mounting of the Roland voice card needs to be extended as the gap between the PCB’s is too tight. So this project goes onto the back burner for a few months, before I get a final set of PCB’s designed.

Prologue Fast forward to Spring 2021 – and a redesign. The 4-way switches will be replaced with pots and the LFO waveform with a rotary switch. This means some mods on the Controller and Voice PCB, and some panel layout changes. The big change is to add a second VCO to give more depth to the sound, just like the Pro Mars. There is space next to the original voice card where I could squeeze the VCO in, and its all digitally controlled in terms of waveform and range. Just need to add a detune knob.

Ultravox Custom Mini-Moog

Ultravox Mini-Moog

Overview I came across a picture of the Mini-Moog that was customized by Roy Gwinn for Ultravox and was intrigued by what the changes were for and how they were implemented. The actual Mini-Moog has long since been scrapped, as it had a very hard life touring and then being painted grey. It was in use from 1980 to 1983, until the later songs used Billy Curries bigger collection of synths and Chris increasingly focused on playing bass guitar.

The modifications were aimed at two limitations of using the Mini-Moog live in 1980; reproducing tight bass lines locked to the drums, and the tuning instability. The  modifications were quickly obsoleted by the introduction of MIDI in 1983 and digital synths.

The custom Mini-Moog was a key part of the revised Ultravox band and sound from June 1980 and the Vienna album. It featured in the bass lines of the album and was played by Chris Cross. I cannot be 100% sure about the modifications but I have collated as much information from the Internet as possible. Big thanks to Roy Gwinn for creating these modifications in 1980!

There are three visible modifications to the Minimoog;

• Left Hand Side – Toggle switches, red push button, input jack and LCD display.
• Lower Front – Extra potentiometer and a 7 segment display.
• Env Generators – Toggle switches.

VCO3 Voltage

Early photographs show only two toggle switches on the left hand panel and no jack socket or red button. The displays are Digital Multi Meters that measure voltages and not music values such as tempo (bpm) or pitch (note values).

Trigger Sequencer This was a modification to the Mini-Moog and used to pump out a steady stream of eighth-notes, which could be transposed up the keyboard by Chris. By keying different notes, a bass line was produced with the unwavering perfect tempo of the machine.  This rock solid tempo had a hypnotic effect to it, which was an important part of the new Ultravox sound, which also used early Roland drum machines like the CR-78.

It looks like VCO3 was used as a LFO to drive the trigger sequencer using the external gate input, then switched into the filter and loudness envelope generators by the toggle switches in the Modifier section. The LFO/sequencer was not in sync with the drum machines and Warren simply played along with acoustic drums. Later on the drum machines were sync’d up as the clock source and the control panel further modified. The CR-78 does have a trigger out for this.

The first incarnation has only two toggle switches on the left panel, possibly external CV and Gate on/off. The second incarnation (1983?) has an extra push button, switch and input jack socket, which I presume is for an external clock from a drum machine, with the extra switch and push button controlling the clock in. This setup would have been easier to use on stage with Warren switching on the drum machine and Chris bringing in the clock on the Mini-Moog to create the bass sequence.

VCO1 & 2 Voltage

DDM Readouts There are two DDM displays; the LCD on the left shows the voltage going into VCO3 and therefore the trigger sequencer tempo, and four 7-segment LED displays show the voltage going into VCO1 and VCO2 – with an extra potentiometer (green cap) marked OSCS Find Tune, to the left of the display.

When VCO3 is switched to LFO mode for sequencing, it is disconnected from both master tune, pitch wheel and the keyboard. It is therefore totally independent and not affected by other pitch controls. The extra fine tune control is connected into the Tune CV output which goes to VCO1 and VCO2.

This voltage is monitored by the DDM before the keyboard and pitch wheel CV’s are added. You can see a display of 52, which might be +5.2V! This would be correct as the Tune control varies from 0 to +10V and its usually set mid way.

I wouldn’t be surprised if the 7-segment display predated the LCD display, as tuning would have been a big issue and LCD’s came out later. Early Mini-Moog’s were renowned for temperature instability and even later ones would have struggled in the hot conditions of a live stage. You can see Chris adjusting the tuning and checking the readout at the start of songs, like All Stood Still.

Env Generator Switches These are unmarked toggle switches, which I think switch the gate input on/off for the filter and loudness envelopes, and therefore whether they are driven by the trigger sequencer.

Replicating Today With the advent of MIDI sequences the Ultravox Mini-Moog trigger sequencer is redundant and its really easy to recreate the 1/8th note sequences on a DAW or MIDI sequencer. If you are into old school CV and Gate and have a Behringer Model D, you can patch the on board LFO to the Gate inputs of the VCF and VCA, to recreate the same effect. Unfortunately this is not possible on the Behringer Poly D, as even though there is a 4th VCO which can be set as a LFO it does not have an external output.

A better approach is to use a 8-step trigger sequencer with toggle switches (e.g. Nosie Engineering – Bin Seq), and drive this from a MIDI clock or analog clock. This delivers the driving 8th notes but with skips and rests, which is probably the way I will go.

Pitch instability of VCO’s is still with us on analog synths, but generally they are more stable than the early Mini-Moog’s. So there is not much call for an onboard fine tune pot and a DDM these days, although your DAW can use tuner plugins to get to the same result, or there are euro rack modules that do the same.