JH. PolyKorg Clone
I have started to build a (slightly) updated version of the Korg PS-3200 synthesizer.
The PS-3200 was the last of three fully polyphonic, semi-modular analogue synthesizers offered by Korg in the late 70's. (See Ben Ward's excellent Korg PS site for detailed information, including user manuals.)
The concept of the PS-Synthesizers was different from other manufacturer's
early polyphonic instruments. Instead of using a small number of voices
and a clever keyboard assigning circuit, the "PolyKorgs" had a complete
synthesizer circuit, hard wired to each key. That makes a total of 48 VCFs,
48 VCAs and 48 voltage controlled ADSRs even for the smallest of the range,
the PS-3100. The largest of the range, PS-3300, even had 144 of these circuits.
The sheer number of synthesizer circuits called for an extremly economic circuit design, and it's a joy to look at Korg's design ideas which led to building blocks that almost did the same as in the better known "classic" synthesizers. And after many years of engineering and reverse-engineering electronic music circuits, I have learned to look at odd solutions not as "substandard", but as a source of creativity an individual character. Here's a list of some highlights:
|Single-Transistor Waveform Converter||creates triangle, saw, pulse and PWM from saw input, using one (!) transistor, one diode and two resistors per voice, plus two global control voltages||Pulse height also changes with pulse width|
|5-Transistor-VCF (Korg-35)||A Voltage controlled 2-pole (Sallen&Key) LPF built from 5 transistors||rather high CV feedthru|
|Single-Diode VC Resonance||The dynamic resistance of a simple diode is used to alter the feedback gain of the VCF||limited range of Q|
|"Expand" function instead of VCF Envelope modulation depth||Instead of scaling down the ADSR with a VCA, the a variable portion of the Envelope is just clipped with a single diode.||It's so remarkably close to ordinary VCA function that apparently nobody
takes notice. At least I have not read about it anywhere.
At slow Atack times, the Envelope appears delayed at the VCF (no effect until th eclipping point is reached). Usefull for Brass sounds, and not easy to emulate with conventional synthesizers.
|Minimum parts count Voltage Controlled ADSR||Three transistors, 1/2 of a LM324 and one CD4007 per voice. Plus some more involved control circuit, shared by several voices||Transistors must be selected in 13-tuples, not just in pairs.|
|ADSR detail (1):
One-opamp control logic
|1/4 LM324 is used as Flipflop, which is dynamically set by Gate-ON, dynamically reset by Gate-OFF, statically reset when the attack peak voltage is reached, and whose set/reset sensitivity is altered by a CV||Very odd "Hold" function, depending on the "Attack"-value. But very useful in practise.|
|ADSR detail (2):
Single-Transistor, exponential slope VC-Decay
|Using a single transistor per voice for VC Attack and Release is remarkable already, even though the A and R slopes are linerar. But the Decay slope is exponential, and this is achieved with a single transistor and two resistors per voice!||The Decay time range is rather limited. No ultra fast Decay, and no ultra slow Decay either.|
|Single-Transistor VCA||That's the "Korg standard" VCA, well known from other instruments like the MS-10.|
The PS Synthesizers contain a few obsolete parts, like the Frequency
Divider chip and the "Korg-35" filter IC.
Also, some of the circuits can be simplified with today's fast BiFET opamps, so tightly matched SK30A FETs can be avoided.
So a general redesign makes sense. I'm using 4000-series CMOS chips fo rthe dividers and for analog switches, I'm building the VCFs from discreete transistors, and I put it all into a standard 6U rack frame with double euro cards (160mm x 233mm). The "multiple" stuff will be on etched PCBs, and the single quantity stuff will be on Veroboard. I will provide circuit diagrams and PCB layouts (for the "multiple" stuff) on this site one by one, as I build the circuits. (This will be my redesigned version. Original circuits are available from Korg, and I will only put small excerpts on the web site for comparison and for educational use.)
The "Gate" circuits (VCF, VCA, ADSR), original Korg circuit
The following is a raw documentation of the circuits, as I built them.
Each Board is a 233mm x 160mm card with two DIN41612 connectors (32 Pins a+c), the upper connector is "I", the lower one is "II". These cards fit into a standard 19" 6U rack frame. What would normally be the front of the rackmount enclosure, will become the back of the synthesizer, so you can remove the cards from the rear - just like the original PS synthesizers. The connector side looks towards the front of the synthesizer, and from there it's wired to the front panel elements. The 6U frame is obviously not fixed into a 19" Rack - it's put inside a larger wooden enclosure.
Before I started, I made extensive Spice Simulations. Then, I did some breadboarding for the VCOs and the VCF/VCA/ADSR.
VCO Control Board (10)
VCO Control Board, Page 1: Supply Voltage Distribution
VCO Control Board, Page 2: Waveform Control for SG-1
VCO Control Board, Page 3: Waveform Control for SG-2
VCO Control Board, Page 4: PWM Oscillator
VCO Control Board, Page 5: Scale Control
VCO Control Board, Page 6: Vibrato VCA
VCO Control Board, Page 7: Frequency Control
MG-1 / S&H / VCF Control Board (11)
MG-1 / S&H / VCF Control Board, Page1: MG-1 Core, Waveforms, Noise
MG-1 / S&H / VCF Control Board, Page2: VC Waveform Select
MG-1 / S&H / VCF Control Board, Page3: Sample & Hold
MG-1 / S&H / VCF Control Board, Page4: Modulation Balance Mixer
MG-1 / S&H / VCF Control Board, Page5: Modulation VCA
MG-1 / S&H / VCF Control Board, Page6: VCF Control
Signal Generators Board (1) (2) (3) (4)
Photo 1 (component side)
Photo 2 (copper side)
SG PCB Layout
SG Components, 1st board (C down to G)
SG Components, 2nd board (F# down to C#)
SG Components (bottom; SMD)
"Gate" (VCF / VCA / VCADSR) Board (5) (6) (7) (8) (9)
Gate Schematics, Page1: common circuit for all voices of one board
Gate Schematics, Page2: individual circuit per voice
Gate PCB Layout (top)
Gate PCB Layout (bottom)
Gate Components (top)
Gate Components (bottom)
Check for a short between + and - supply next to LM358 on top left of PCB.
Make feedback resistors on RC4558's 100k (not 10k).
Components with solder connections on top side of PCB must be soldered in a reasonable order, i. e. before other components would get in the way of the soldering iron. For instance, solder in 4007 chips first, then 4M7 resistors and 1n5 capacitors next to the chip.
Forget about using transistors from the drawer and replacing those, which have too much tolerance, afterwards. Transistors must be selected. On each PCB, you need a group of 11 2SC945's for Decay (match for 2mV Vbe), another group of 11 2SC945's for Release (2mV Vbe), and a group of 11 2SA733's for attack (2mV Vbe). Also, you need a group of 49 2SC945's for the VCAs (matched for Beta, 20% tolerance). Even though there is a trimmer for each individual VCF (PS-3200 only has one trimmer per 12 VCFs) I recommend matching groups of 3 BC550Cs for the VCFs (2mV Vbe).
Signal Board (12)
Signal Board, Page1: Resonator CVs and Preset Volume
Signal Board, Page2: CVs for Balance, Resonator, AM and Ensemble
Signal Board, Page3: Audio Path
Signal Board, Page4: VCA 2 CV
Unlike the original PS-3200, my Clone has the Resonator section from the PS-3100 and PS-3300. Unlike these, the Resonator settings can be stored and recalled.
Most VC functions are Vactrol-based (as in the original), but I am using a very different circuit, modelling the two halves of a potentiometer rather than building two VCAs and mixing. The advantage is that the Vactrol's tolerances are less crucial here.
Patch Storage (16) (17) (18)
I could not get all of the schematics of the PS-3200's digital circuitry. I think one or two pages are missing. So I'm designing a new patch storage circuit from scratch, and it's a bit different from the PS-3200, too.
Changes: EEPROM instead of battery buffered RAM. Different ADC and DAC chips. Slightly different user interface for write operation. The 16 buttons are used to select 8 banks of 8 patches now, instead of only 16 patches.
(1) Programmer: Front Panel elements
and Bank / Patch select electronics
(Priority encoder and 8-ch.-MUX in a feeback configuration provide the "Radio Button" action. This is a veroboard located at the front panel.)
(2) DIG board: State Machine and Adress Counter
(New version! Previous drawing - jh_3200_digital_2.gif - did not work!)
(3) DIG Board: ADC, DAC, EEPROM
Circuits (2) and (3) are located on one veroboard: "DIG", or digital board)
(4) MUX Board schematics: CV MUX, DEMUX Sample&Holds, Preset / Edit switches
MUX Board Layout
MUX Board top side components
MUX Board copper side components
Note: Connect each pair of points with the same name. Only a part of th econnections are made by copper traces. There are many wires soldered on the component side! (This is a compromise between doing a full PCB layout and a veroboard construction. See picture for details.)
Photo of the Programmer
Photo of the DIG board (note the extra connector that goes directly to the MUX board)
Photo of the MUX board (note the mix of copper traces and extensive point-to-point wiring)
Photo of the three patch storage boards in front of the rack frame
Ensemble Board (14)
I didn't want to use the original Korg PS-Chorus / Ensemble circuit, which has two BBD delay lines.
Instead, I wanted a better "String Chorus" circuit, to get some typical string ensemble sounds. The classic Solina 3-BBD-circuit would have been an option. But I already have two such devices (Ensemble section of Polysix, and Dr. Böhm Phasing Rotor 78), so I opted for another 3-BBD-Classic: The ARP Omni.
For schematics of the original Omni "phaser" (that's how they called the 3-BBD chorus), see Peter Brown's excellent ARP Omni Page.
The 3 BBDs are modulated by 3 individual LFOs. There is a fast setting (normal operation), and a slow setting. Th elatter is used if you also run the Omni's synthesizer section thru the "phaser". This was the main reason for me to use this circuit: One setting optimised for strings, and one setting optimised for general synth use.
For the PolyKorg Clone, the User interface works like this: There is a single potentiometer for the Ensemble section. At center position (12 o'clock) the signal is passed thru without effect. If you turn the pot counterclockwise, the slow setting of the Omni "Phaser" is garually faded in. If you turn the pot clockwise, the fast setting of the Omni "Phaser" is garually faded in.
Thanks to Peter Brown and Mike Irvin for their info about the Omni Phaser!
Ensemble Circuit, Page 1: CV
Ensemble Circuit, Page 2: Audio
New: Amplifier and MIDI board (15)
This board contains a discrete balanced line driver and a headphone amplifier, complete with opto-electronic volume control for both.
Also on this board is a small MIDI-GATE interface module which I got cheap on the second hand market. Don't ask me about the brand of this interface (really, don't!) - any type that will put out inverted TTL level GATE signals for each individual key will do.
The concept for MIDIfying the PolyKorg is like this:
The Korg circuit needs +15V gates (15V powered CD4007). So I added a comparator (1/2 LM358) for each voice (included on the "GATE" boards. The comparator is inverting, and its input is biased to +5V, so a switch to GND will be enough to trigger a synth voice. Therefore a completely passive keyboard, with switches wired to a GND bar, can be used.
Also, more than one keyboard, or a keyboard and open collector outputs, can be connected to perform a wired-or function with inverted logic. This is important if you want to play the instrument from a keyboard and from a sequencer at the same time.
Now unfortunately the MIDI interface I aquired does not have open collector outputs. I could have used diodes to get the desired OR-function, but I chose a resistor solution instead:
With no MIDI interface connected, I'd had 100kOhm pullup on the GATE boards. With the MIDI interface, I have additional 18kOhm resistors going to the MIDI gates. A short to GND from the keyboard will override a +5V
from the MIDI interface (no harm, because there is the 18kOhm resistor), and without a key depressed, a 0V from
the MIDI interface via 18kOhm is strong enough to override the 100kOhm pullup on the GATE board.
And, last not least, the 18kOhm are a better protection for the CMOS components of the MIDI interface against ESD than a diode would be.
AMP / MIDI Board Schematics
Power Supply Board (19)
uA723-based Power supply
+/-15V electronics supply voltage
+10V reference voltage for potentiometers, ADC and DAC
+7.something (variable) voltage for ADSR peak
Front Panel - please click here for a high resulution pdf version (213kB)
===> PolyKorg Clone, Page 2
Please read the FAQ.
Back to JH. homepage