Synth

These are my custom modules for a from-scratch Eurorack analog synthesizer.

Format

3.5" mono patch cables
3U tall, split into 84 HP horizontal units
10 or 16-pin ribbon power: +/-12V, 5V digital
Audio signals are typically a maximum of 10V peak-to-peak (i.e. between -5V and +5V)
Control voltages can either be unipolar or bipolar. Bipolar control voltages are typically 5V peak-to-peak (i.e. from -2.5V to +2.5V), unipolar voltages between 0V and 8V. The V/Octave scale is used for pitch information
Trigger, Gate or Clock signals are digital 0V-5V pulses typically used for timing and event signaling
minimum 100K input impedance
maximum 1K output impedance
connecting rail voltages to signal inputs should not fry any components (although it doesn't have to perform correctly)

Pinout:

Design goals

"fully-modular with defaults"
- semi-modular-style sensible default paths connected under the hood
- all defaults overridden with patch cables
- specifically targeting support for several modes:
  - 4 totally separate, independent synths, each controlled with constant VCO or sequencer
  - 1-voice, MIDI-,sequencer-, or constant-controlled synth with 4 oscillators, e.g. harmonizing intervals, mixing shapes/timbres
  - 4-voice MIDI-controlled synth
  - 2-voice MIDI-controlled synth, with a pair of oscillators for each voice
  - sub-combinations of the above, e.g.:
    - 1-Voice MIDI-controlled synth with a pair of oscillators, 1 constant-source independent synth, 1 sequencer-controlled independent synth
    - 1-Voice MIDI-controlled synth, 2-voice sequencer-controlled synths, with two oscillators for the first voice
LED feedback is useful
For multiple options (like VCO shape), use a CV that is binned to options
- Easiest solution is probably an ATTiny driving an analog switch
- uC seems like overkill, but it's got an ADC and lots of digital pins
- Lots of specialty components would be required for a native IC solution
- Firmware code can easily provide hysteresis and arbitrary numbers of divisions, unlike CDXXXX series CMOS chips
Outputs can use splitters, but inputs cannot

Tools

Based mostly on what I have available and have experience with.

TL07x
- Basic audio-quality op-amp
- single, dual, or quad
- note: not rail-to-rail, and asymmetrical clipping
- slew rate minimum 5 V/us, typical 13 V/us
- could be trivially substituted for any other general-purpose op-amp
LM13700
- dual operational transconductance amplifier
  - effectively an op-amp with control of the output current
  - could also think of it as a current-controlled resistor
  - makes voltage control easy
- see detailed behavior for equations
- sensitive inputs:
  - I_abc is limited to 2mA
  - differential input voltage is only linear at +/- 20mV
    - using linearizing diodes, linear region can be extended to +/- 60mV
  - differential input voltage is limited to 5V
- treat amp bias input as V- + 1.2V for the purpose of calculating I_abc
- 50 V/us slew rate
DMMT5401 (NPN) and DMMT5551 (PNP) matched pair transistors
- similar characteristics to 2N3904, 2N3906
- useful for current mirrors and exponential converters
- easier than trying to match transistors by hand (but here's a guide for matching)
- Unfortunately only come in SOT26 (3mm long)
DG403
- dual SPDT analog switch
- very generous on control ranges and input limits
- switching time in 100ns range
- on resistance in the 45 ohm range
- Unfortunately no DIP package - SOIC is largest size (10mm long)
- preferred over CD4066
  - CD4066 has max 20V supply differential, which makes it frustrating to use with +/-12V supply
  - when off, CD4066 connects the pins to the minus supply voltage, which makes it impractical in many situations
Tempco resistors
- resistance varies with temperature. Useful for canceling out temperature dependance
- not strictly required - classic synths would take some time to "heat up"

Otherwise, typically use 2N3904 / 2N3906 for BJTs and 2N2222 or similar for diodes.

Modules

Amplifier

LM13700
matched-pair PNP exponential converter
CV of 10V -> Gain=1
CV of <=0V -> Gain=~0, 6 decades below Gain=1
allow overdrive to Gain=2
- ideally this would begin to clip
mixdown
- mono - full mix -> left and right
- stereo split - A+C->left B+D->right
Direct stereo input jack
- mix arbitrary signal with mixdown
- e.g. for accepting signals from other synths

Oscillator

0-10V, 1V/octave, A0 to A10
Square/Triangle core via LM13700
Waveshapper for triangle -> sine
PWM control
- triangle wave applied to comparator with CV controlling reference. Triangle wave means CV is linear
abs(triangle) can be used for 2x square
- adds a lot of parts for little benefit
blending shapes isn't useful here, since we could use other oscillators for that
Option to sync with another osc?
- in theory, grounding osc at a regular interval would reset the osc over and over, effectively changing it's frequency
CV control for shape?
Room for more shapes
- white noise?
- PDS
  - https://en.wikipedia.org/wiki/Phase_distortion_synthesis
  - a bit of wild harmonics
  - multiplication techniques
    - log/antilog
    - https://en.wikipedia.org/wiki/Gilbert_cell
    - AD633 ($11-$13 /unit on DigiKey!?)
    - EL4450C ($4/unit on Digikey but out of stock). It's a little more awkward to use since it's max differerential input voltage for linearity is 2V
  - the hard part will be the "slightly higher frequency" bit
    - perhaps instead you have another osc (555?) running at a fixed frequency and clipping the sine
    - square PWM control could also adjust this clip frequency
  - In Max you do PDS by kink~-ing a phasor~ and using that to drive the phase of cycle~. Kink is somewhat straightforward to do - use a comparator to adjust the charge current after a threshold. But how you drive a sine wave phase with that I'm not sure
- down ramp wave with just a JFET
  - up ramp wave probably needs another JFET and an inverter
- sub octave
  - theory and example circuit: http://www.valvewizard.co.uk/uboat.html
- octave
  - full wave rectifier
- multiplying the triangle with a square wave with different phases and pulse-widths gives some neat sounds
- polynomial
  - The EL4450C datasheet describes using a multiplier ic to create a polynomial of the form (k1 v^2 + v) / k2
  - it sounds like crap on a triangle wave and does nothing for a square wave, but it adds a nice second harmonic to a sine wave
  - from experiments in Max, good values are k1=1.5-2.5, k2=2
  - but at that point, should you just use 2 oscillators?

MIDI-CV converter

digital MIDI input to OSC CV
DAC resolution:
- for 1 step = 1 cent on a 10V space with 1V/octave:
  - 10V / 2^x = (1V/12/100) => log2(10/(1/12/100)) = 13.55 bit minimum
- 16-bit DAC would be incredibly accurate, 0.18% resolution
DAC also needed for amplifier (velocity)
- Velocity is only 7 bits (0-127)
- builtin ATMega DAC (PWM) is 8bit, but very much in the audible frequency range (~450Hz)
- Even at a higher frequency, an aggressive enough low-pass filter would probably have a miserable response time
- instead we need a dedicated 7- or 8-bit DAC
  - TLV5620 is a cheap 8-bit quad.
  - Even though the data sheet seems to suggest that messages are 11 bits, this resource seems to use 2 bytes:
    - http://www.kerrywong.com/2013/10/12/interfacing-tlv5620-with-arduino/
- What's a reliable solution for reference voltages?
up to 4 voices
round-robin outputs on each key press
- use open output if available else replace oldest key pressed
use built-in 10-bit DAC for velocity control
Modes:
- off
- 1-voice: A
- 2-voice: A + C
- 3-voice: A + B + C
- 4-voice: A + B + C + D
- 4-voice-split: A + B, C + D
  - like standard 4-voice mode, but round robin is in 2 pairs of slots instead of 4 slots
  - allows for 2-voice bass and 2-voice melody
  - on mode selection, waits for a keypress to set the note to split at
portamento
- simpler envelope circuit could be used, but parameters need to be tuned for longer times over short voltage change
capable to expand to 8 voices?
Arpeggiator
- up, down, play order, random, other patterns?

Filter

4P LP(/HP?)
Cutoff frequency control has same range as oscillator
- 0-10V, 1V/octave, A0 to A10
filter we designed for class
- standard vactrols are available on Thonk
CV control of cutoff + resonance
Default to track cutoff with VCO-CV
individual filter knobs, plus a global knob for filter sweeps

A/D/S/R envelope

4 envelopes required for 4-voice polyphony
wired to VCA by default
save surface area with 1 set of ADSR controls for a pair of envelopes
fully modular means you could borrow one for VCF instead, or send to both VCA and VCF

2x is plenty I think?
OSC circuit, but adjusted to lower frequencies
bipolar (+/- 5V) and unipolar (0-10V) outputs
ideally multi-shape: sine,tri,squ,ramp up,ramp down
CV control for shape
Trigger input (e.g. MIDI gate or clock)?

Utilities

Buffer, Sum, Attenuate, Invert, Gate
Could be all-in one
- A input (default ground)
- B input (default ground)
- level knob attenuator
- gate digital CV in, default on
- (A+B)*level*gate output
- -(A+B)*level*gate output
- Constant CV out if A and B disconnected

Future ideas

Sequencer
- clock source for LFO? Or LFO input as clock?
- divide down clocks?
- likely digital
  - if MIDI output, can simplify duplicating VC outs for tuning, can leverage MIDI converter for this
  - however, key-based offsets of patterns wouldn't work
- 7-segment display of BPM for source clock?
- sweep control option
  - LFO or voltages to trigger and CV external LFO for 4-bar sweep
  - e.g. automatic filter sweeps
- random note drop feature
- Could be externally controlled, e.g. iPad
  - save space and iterate on features
  - allow it to be used as a drum machine too
Need some good percussive effects
- 808 would be a good reference
White noise generator?
Sample+Hold?
- combined with white noise creates RNG
- alternatively sequencer could do RNG digitally
- random is a lot more interesting if it can quantize
  - quantization is a lot cheaper in digital
- S+H has little use to me outside of random
Delay, overdrive, etc. are gimmicky and not useful, can be done via effects pedals

TODO

utility module
- Verify on breadboard
- design PCB
- order parts
- come up with a cute name
- design front panel
- assemble Rev. A PCB
  - bend the regulators first
  - plenty of 4066s
  - flip orientation of power connector
  - pot orientation - both channels:
    - right side (facing) to inside of board (from top)
    - left side (facing) to outside of board (from top)
- test with Neutron
  - constant output is only 9 to -7.8
  - high noise when used as a buffer
    - perhaps input signal is clipping?
    - white noise when the gain is turned down
- adjust design for Rev. B
- assemble Rev. B proto board
- fix prototype issues
  - sum pot is backwards
  - a bit of noise on the constant output near the center
  - verify output levels
  - hard click on gate
    - needs a LPF?
- assemble Rev. B PCB
- manufacture front panel
- assemble module
VCA
- Confirm amp on breadboard, finalize CV levels
- figure out where tuning trimmers are needed
- build prototype
- fix prototype issues
  - output is too quiet for headphones, need about 2x gain
  - CV range is too wide, adjust scaling for faster change in db/V
- finish PCB schematic for complete amplifier
- Create PCB
- design front panel
- manufacture front panel
- assemble module
VCO
- design core
  - core osc - square/triangle
  - wave shaper (sine)
  - pwm control
  - additional shape
  - shape selection
- Confirm amp on breadboard, finalize CV levels
- fix breadboard test issues
  - pitch is very sensitive to power supply noise
  - tuning is rather difficult
- Build prototype
- fix prototype issues
  - TBD
- design PCB
- Create PCB
- design front panel
- manufacture front panel
- assemble module
Envelope
- design in SPICE
- verify on breadboard
- build prototype
- fix prototype issues
  - TBD
- design PCB
- Create PCB
- design front panel
- manufacture front panel
- assemble module
MIDI
- select DAC
- write code to interface with DAC
- write code to interface with MIDI
- get opto-isolator
  - e.g. 6N138
- verify on breadboard
- build prototype
- fix prototype issues
  - TBD
- design PCB
- Create PCB
- design front panel
- manufacture front panel
- assemble module
VCF
- TBD
LFO
- TBD
build tools
- tiny Breakout pcbs for headphone jacks
- ~~transistor matcher circuit~~
  - bought matched pair BJTs instead
- quantized selector with hysteresis

License

Schematics, board layouts, documentation, etc. is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

If you're interested in commercial use, please ask first.

Any source code is licensed under MIT.

Many documents in reference are copies of resources from other parts of the internet (datasheets, schematics, etc) which I did not write - those retain their original copyrights.

Wong-Jhon / synthesizer