Electronic music - modular - interconnect notes
Eurorack
Power
Power connector
Every two pins side by side carry and connect the same thing.
The more standard 16-pin internal bus carries primarily:
- shared gate (not often used)
- shared CV (not often used)
- + 5V (few modules use it)
- +12V
- Gnd
- Gnd
- Gnd
- -12V
The 10-pin variant carries the last part, the 12V, Gnd, and -12V.
On 5V
Few modules use it (from the socket). Doepfer itself used it only in some old modules.
(So) not all power supplies provide it.
(So) modules that need 5V somewhere may choose to regulate it from 12V themselves.
On the CV and gate bus
This is basically one hidden wire shared by everything that cares about it.
In theory it's a nice ideay, avoiding a bunch of socket routing for things you usually wire up anyway. Say, the shared gate could e.g. be used to put the same sync signal on everything willing to listen, CV could be used for pitch.
However, only one module should ever drive it, which makes it somewhat error-prone and somewhat inflexible, so in practice a lot of modules avoid it, though various can be jumpered to use it in case you care.
Plugging backwards
Don't do that, because it
- feeds -12V into the gate bus (which may often be okay because it's not used much)
- shorts the power supply (+12V to Gnd and +5V to Gnd). If it's not protected against that, it may overheat a regulator, wire, or the supply.
Ideally, use sockets and plugs that have the plastic keying, so that you can't.
But modules may just have pin headers, so pay attention.
The ribbon cables, if they color one wire separately, should do so on pin 1 (the -12V end).
Interconnects
Plugs
Eurorack uses 3.5mm TS (mono) plug and sockets
- tip is signal
- sleeve is common mode signal reference (0V)
This carries either parameter or audio.
On ground
Sleeve is signal common.
Circuit-wise you can assume it's ground eventually, but consider you can invite more or less noise if you forget to think about various coupling.
Even if you do, this setup seems somewhat sensitive to inductive coupling, with little you can do about it (verify)
Panels may be conductive and grounded (to act as a shield), but don't make a design that doesn't work if it's not.
Voltage and function (Eurorack)
The precise meanings/ranges of voltages within eurorack isn't quite as standard as you'ld think, though most of it uses similar enough ranges that you don't often notice that.
Things should be designed that nothing breaks when given up to 12V [1],
and given the power supply, giving it more than that should be hard to impossible.
Nothing will be outside the -12..12V range, because the power supply isn't.
Gate, trigger, and clock are considered high above approximately 3V.
Gate
gate - pulses where you care more about the time it's on than the initial edge.
Usually gate output to gate input.
Parameter CV can also be used for gate inputs, because it often works out as a 'is high when above this voltage' threshold', but the exact threshold isn't so defined.
A few systems have/allow inverted logic on gates, i.e. you can switch them so that ~0V is considered on, and the higher voltage (whatever that synth system used) is considered off.
You may find older keyboards that do this, but almost all modern synths use only positive-polarity gates (low is off, high is on)
Trigger
trigger - signify a start
Things that interpret inputs as triggers will often care about the initial edge (rising edge on positive polarity systems).
the on state and falling edge are usually not relevant.
Trigger pulses may be on the order of 2 to 10ms.
Shorter is absolutely possible, but a little longer can mean that the receiving side is certain to notice it, even when they are cleaning up the input signal for stray signals.
Sync / clock
sync / clock amounts to a trigger at (usually very( regular intervals.
Like trigger signals, clock inputs may only react to (rising) edges, so require fast-rising signals.
In some uses, sync/clock and trigger are used in the same way - do something every pulse.
In some others, sync/clock is used more in a PPQN (like MIDI sync, DIN sync)
(Not to be confused with oscillator sync (which means one oscillator restarts another. If the controlling one is faster, that means the second one will follow the first one's frequency (often with some extra waveform edges))
Pitch CV
Pitch CV
Pitch CV is used to get all pitched sound producers to agree on what pitch they should be working at.
It's also used by filters, pitch quantizers, and such.
In eurorack is 1V/Oct (a.k.a. exponential), as once introduced by Bob Moog
There is another system, called Hz/V (a.k.a. linear), but is electrically less practical, so not seen much in modular.
While pitch CV input is exponential, you may occasionally see a linear pitch CV alongside,
to control things like tremolo from regular CV output.
Voltage-wise
1V/Oct literally means that every extra volt means an octave higher.
This means voltage to frequency is an exponential curve, hence the nickname.
There is no absolute mapping from voltage to pitch, in part because when you have tuning knob anyway, both to get things in tune with each other, and to set up detuned relations, you only need to get it roughly right.
That said, there are conventions to get things roughly on the same order. For example, 0V may well be a C (apparently stemming in a Moog convention?), and regularly it would be C2 or C1 or C0, etc.
CV voltage
(Because you care about this for DIY)
Eurorack is a little less standard than yould think.
Parameter CV often sits in -5..5V range (apparently some things extend to -8..8V(verify)), though some argue it should be 0..10V.
Actually a bunch of things take unipolar CV (e.g. in the e.g. 0..5V range) because they have no meaning for negative voltages - consider e.g. envelopes. This basically means negative parameter voltages won't do anything.
A lot of parameters will amount to linear control (e.g. twice the voltage is twice the thing),
mostly just because that makes a lot of things (that aren't pitch) sensibly controllable.
In some cases, it makes sense also, or instead, or switchably, to react with a different curve
The outside I've seen for Pitch CV is something like -3V..10V (-3V to 8V already the full 10+ octave MIDI range). But few things will take that much, and thirteen octaves is more than you will ever need from one module.
Consider that any single melodic use tends to stick to one to maybe four octaves (five octaves is already enough for, say, a bassy 50Hz to a beepy 1400Hz).
You can get away with taking maybe 0..5V, or less, more so if a module also has some way to shift up or down an octave or two. It's a not-unimportant footnote that getting stable tracking, and/or more precise use of DAC resolution, is easier on a smaller range.
Important to DIY:
- design things to not break on voltages up to -12..12V
- this often means a bunch of diodes (e.g. zener)
- e.g. even if it makes sense to never use negative voltages,
See also:
Audio
Audio is AC, in that it swings both ways around your Gnd
- amplitude-wise can be -5V..5V, but often more like -2.5V..2.5V for headroom
- There is also no true standard. Even modules from Doepfer, who gave us eurorack in the first place, and are German, vary a little on behaviour
- still, the values above are close enough that generally things just work, and only sometimes you need a but of amp/attenuate/shift. Which is regularly part of modules that most need it.
Physical
In practice, eurorack refers to a set of conventions followed by many companies.
The physical part is based on eurocard.
Row height
- is 3U rack height, which is 133.3mm (5.25")
- to DIYers:
- module front plates are often around 128.5mm high, because they assume you may be adding a rim between rows
- vertical screwholecenter-to-screwholecenter is 122.5mm (so 3mm from screwhole to plate's edge)
- Module boards may, on the inside, wish to be at most 109.5mm high to avoid not fittting due to variation in rails [2]
- meaning rails are expected to be under ~1cm
Module width and Row width
- units of HP, which us 5.08mm (0.2 inch) (in German: Teilungseinheit, TE))
- Rows are commonly 84HP (~42.7cm), 126HP (~64.0cm), or 168HP (~85.4cm), mostly for practical reasons such as
- 19" rack brackets are often 84HP, because that fits nicely in 19" rack's inner width, ~451mm
- modules have a practical minimum of 2 or 4HP, because you need 1cm to 2cm to fit on a knob or socket
- sizes are often multiples of 2HP or 4HP (for little reason, beyond that it's easier to panel up rows that aren't full)
Module depth
- no standard
- a lot of modules are 50mm to 80mm behind the panel. Among the deepest are 105mm.
- Cases tend be deeper just to not be incompatible with some modules, to allow power inside on the back, and to have that at enough distance to have not too much intereference
- when making a case, also consider it may be nice to have the case extend in front of the module plates, to protect the knobs during last-meters (or all) transport
Rails
- This is primarily about ease of placing modules
- The vertical distance is most important
- If the holes are fixed, the horizontal offset does too.
- Common pre-made rails include those from Vector Electronics, Schroff, GIE-TEC, and others. Your Mouser/Farnell/Conrad/similar will often carry one of them, or perfectly applicable variants rails, though do check their specs
- The differences between rails are looks, sturdiness, and ease of building. Row brackets can make that even easier.
- Expect this to cost you at least EUR18 per 84HP row.
- Many of the aluminium rails are meant to contain either square nuts (movable), or a pre-tapped inserted rail.
- The threading of these is typically either M2.5 (e.g. Vector Electronics, Schroff) or M3 (GIE-TEC, Tiptop Z), which only really matters to the box of screws you'll get (which should often be 6mm or 8mm long depending on whether you use washers).
- On the cheap
- you can get away with just wood as rails, and wood screws.
- Note that you want to make that bar replaceable, because moving things around a lot will slowly destroy it (moderately hard wood and pre-drilling can also help)
- On the slightly less cheap, you could screw the pre-tapped rails onto wood, possibly with a routed gutter. This should cost you maybe EUR5 per 84HP row.
- you can get away with just wood as rails, and wood screws.
- See also https://synthracks.com/eurorack-rails
Panel thickness is often 2mm, often anodized aluminium (Doepfer)
- with softer materials like wood or plastic you may want ~3mm for sturdiness.
- Or more, but that makes it harder to fasten pots, sockets, and such without some creative drilling.
Other considerations:
- consider that
- stood up and in general, a square area makes things easier to reach, and keeps interconnect cables shorter
- laid flat can be nicer on your arms, though you can only comfortably reach over ~4 rows of depth, so you may more easily go wider
- this also relates to where you add more rows later
- transport, both in terms of portability and potential damage. A little planning can go a long way.
- avoid making parts so large that they become hard to carry
- whether small or large, it needs a reason it won't fall over
- More depth is one way
- Some slotted pipe setup may be good too.
- consider also:
- solid suitcases, and e.g. make them hook open at ~90 degrees
- flight cases
- knobs not sticking out where a wall can hit them.
- or just raise the edges a bit
- a flight case / suitcase can make this easier
- a concave curve helps
- handles are nice. Can also be added later, but consider sturdy places for them
See also:
Note that you can get some correct sized,
CV DIY notes
CV means all things are communicated via voltage.
Which sometimes makes things easy electronics-wise, and sometimes makes controlling something hard - some things are are easier to control with voltage, but others are easier to control with current, resistance, or digitally, and each case means adding more parts.
Perhaps the most motivating / fun way is to get into circuit bending.
Get random sound producing toys from second hand stores, and mess with it. With more theory, you can mess with it better, but even without you can build some random nonsense.
On sockets and plugs
Normaled sockets are useful
When buying sockets, you may specifically want normaled mono sockets, not stereo - both have three pins but one of them is a lot more convenient in design: you can have it physically default to something when not plugged in.
They cost about the same.
"It's easier/cheaper for me to source stereo cables, can I use those?"'
The placement of connectors in TS / normed TS / TRS socket design means this will often work, but sometimes not.
You can't damage anything, but it will generally either
- do nothing at all,
- or seem be respond randomly (because you have disconnected it from the norming and effectively made it floating input, and are possibly using the cable as an antenna)
You could work around it in those cases. Up to you.
Consider what inputs do when not plugged
In particular to avoid floating input, many sockets will use a normaled socket so that that input is pulled to ground, or another known voltage.
But in many cases it can also make sense to have specific behaviour rather than nothing.
You'll need the kind of socket that has the tip connect to an extra lead to do this.
These are almost negligibly more expensive.
Often it's easiest to have sockets that mechanically go to one of their pins.
Which you can tie to ground, or in some cases another function.
- jacks may well use the feature where they can detect the presense of a plug and do something by default.
- Often by merit of being pulled to ground or connected to a basic bit of circuit
- Apparently this is called normalized, in reference to normalized synths(verify)
Input and output protection
Generating and accepting digital parameter CV
Parameter CV may be smooth in theory, but do consider that for e.g. effect control, on the order of a hundred steps is already plenty.
You may find this is true for a lot of parameter CV as well. (Unless you want to tweak something for a very specific phase effect or resonance or such)
As such, there are specific cases where you can cheat pretty hard.
Consider that while digipots are never high resolution, but they act like a resistor may make it easy to CV-ize what is currently a knob, or add circuit-bendy CV to random things (and they can cost as little as ~EUR2 for an 8-bit digipot).
If linearity isn't important, than a vactrol is even cheaper.
For more precision, you'll want to build a VCR, which is more involved.
On CV accuracy
Purely digital sound generation makes for sounds that are cleaner and more precisely repeated than you may want.
It's useful that this is possible, but it's boring that it always is.
Purely analog is the other extreme.
Old synths would be out of tune always, with other instruments and even itself, and change their pitch over time due to components warming,
particularly noticeable if it spanned multiple octaves.
It's not too hard to tame, and a little detune is nice, but it's something you need to manage.
Beyond that, most things aren't so critical either way.
Parameter CV used for modulation usually can be pretty inaccurate without trouble.
CV repeatability is sometimes useful, but consider that the repeatability of setting something yourself with a knob is pretty low, and yet enough for most any parameter.
If fluctuation is noticeable, they are often happy accidents (call it analog, why not) rather than problematic.
So if it comes from something digital, PWM with a lowpass, a cheap 8-bit DAC or digipot can all be fine (see more design notes below).
Pitch CV is different.
This means that anything that generates or accepts pitch CV (keyboards; VCO, applicable VCF; also consider things like quantizers) should be precise over their chosen range.
And note that pitch inputs on VCOs and VCFs regularly accept a wider range. e.g. -3..8V, and should be desiged to be somehow workable within that range, preferably accurate throughout.
Offset inaccuracies are generally much less important, particularly when there are tuning knobs (which are overall offsets),
but ratios/intervals need to be accurate to sound in tune.
So how inaccurate is bad?
tl;dr: aim for 2 cents, though you can easily get away with up to 5ish.
For more on cents, see Music_theory_-_glossary#On_cents
Some practicalities
- out of tune means out of tune with something else.
- If both are out of tune with some initial reference, and both are flat, it'll sound perfect. If one is flat and the other sharp, it's twice as bad.
- 1 or 2 cents out of tune can be intentional in some parts of electronic music - because the beats will make the sound feel fuller
- often done by ear, so you don't really need this accuracy between devices
- Old modulars and synths are notoriously hard to keep in tune, although this was sometimes more because of the oscillator's design than the pitch CV
- In general, anything that generates or accepts pitch CV might like a tuning knob (for offset).
- It lets you fix all offset issues by ear.
- For cases where error is incremental from a given point, then offset can help if you can stay within one octave and tune it for the best match in there.
- So this is mostly about stability (if analog) and resolution (if digital)
- If using a smaller range makes designs easier, consider doing that.
- Particularly if you can add a -2, -1, 0, +1, +2 octave switch - adding a stable voltage is much easier than having a larger range
- Consider that sure, an 88-key piano has seven octaves, each hand playing will typically stay within two octaves.
- Accuracy is also part of why V/Oct is the going standard, not V/Hz.
- While V/Hz is linear and sometimes easier to work with on paper, covering the human audible frequency range either needs a large voltage range, or using extremely precise components (well below 1mV in practice)
https://groupdiy.com/index.php?topic=40066.0
Digital pitch CV
ADCs and PWM/DACs are fairly easy for most parameter CV. Consider e.g. that in MIDI, things like velocity and CCs are 7-bit, and you never noticed.
Pitch is more picky.
Smooth pitch CV
While for various uses you can get away with ~5 cent accuracy, it is generally suggested that you if you to track arbitrarily you should aim for ~2 cents.
And for that to be stable you need a DAC - you probably can't get away with PWM (unless it's much faster, and filtered).
The accuracy also relates to the note range you want.
For example, for 5 octaves (=5V) you'll probably want 12 bits (6000 cents to ~2 cent accuracy means ≥3000 steps over the output range). You can do with fewer bits, if you accept using it on just 2 or 3 octaves, and add an octave switch (and tuning knob), but it may not be worth the trouble.
For reference:
- 8-bit = 256 steps, ~23 cents when the full range is 0..5V, bad
- 10-bit = 1024 steps, ~6 cents when the full range is 0..5V, not great
- 12-bit = 4096 steps, ~1.4 cents when the full range is 0..5V, enough
- 16-bit = 65536 steps, plenty for double the octaves
Pitch input - ADCs
For 1V/oct signals, a semitone is ~60mV, a cent is ~0.6mV, you ideally have accuracy down to a few cents, so want to aim for ~1..2mV (note that due to other components it's often not easy to do much better).
12-bit ADC seems minimum to get a precise and stable enough pitch input across a few octaves.
This partly because resolution should probably be precise to a few cents, and partly because fluctuation between two adjacent values should probably not be audible
E.g. on an Arduino's 10-bit ADC over its 5V range, that's ~4.9mV steps, so towards 10 cents. It functions, but flutter is audible in pitch
Sure you can reference most ADCs to a lower voltage to get a better mV accuracy,
e.g. 10 bits is enough for 3-cent accuracy on ~2.5 octaves.
But it'll just means you'll be ignoring a bunch of octaves on your input, and forcing the thing sending you pitch to play along.
It's maybe acceptable for DIY modules but less so in real designs.
Pitch output - DACs'
Pitch output can in theory be less precise, if you don't mind tweaking where that quantization goes.
If you only care about the keys on a keyboard, i.e. semitones, this is surprisingly doable. Say we want five octaves (6000 cents), and we have an 8-bit DAC (256 steps).
If we decide to use 240 of the 256 steps (makes it easier to spread the steps), then each step is 25 cents, which isn't enough for smooth pitch, but 2 more bits than we need to land semitones, because we only cared about the nearest 100 cent (a semitone).
Sure, that actually describes 0.94V/Oct, but that's one opamp and trimpot and calibration (which you wanted anyway) away from being 1.00V/Oct
This is what a bunch of MIDI-to-CV modules (e.g. Doepfer A-190, uMIDI2CV) are doing. MIDI semitones are already 7-bit things (0=C0 to 127=G10).
If you have another such DAC you could in theory add pitch bending, assuming you can control output reference, because spreading 8 bits on 1V (one octave) means ~4.7 cents.
You can do polyphonic if you want, as four-channel 8-bit DACs (e.g. TLC5620) are only a few bucks.
With a little trickery you can get a little more out.
In particularly supersampling (and/or lowpassing) means you can get an extra bit of resolution/value stability, maybe two, of effective resolution (but note that resolution only means accuracy when a few conditions are met, which is also why you can rarely get much out of supersampling).
Note:
- the ADC in most microcontrollers is a single-sided (rather than differential) ADC,
- meaning direct use can only do >0V(verify)
- You can shift and divide with an op amp, e.g. add 5V and divide by 2 to deal with -5 to 5 (verify)
You can get a
- 12-bit ADC for ~EUR2-3 (e.g. MCP320x series, multi-channel, MCP 4726, MCP4921)
- 16-bit ADC for ~EUR4, e.g MCP 3428)
"Microcontrollers often have PWM. Can we use that instead of a DAC?"
Dedicated PWM (and even modest speed, even bit-banged PWM) may be good enough for control CV.
Largely because you can filter it so hard that even if it's a millisecond or two late, you probably won't notice.
You will have trouble using it for pitch CV, probably even when it's more capable dedicated PWM.
- you probably still want a simple (fixed) RC lowpass filter on its output
- you might still wish to make it external to your controller IC, when that has any other computation to do.
On different formats, and mixing them
Synthesizers.com (5U, ~15V)
Physically
Q series modules adhere to the Moog Unit (MU), which is basically 5U.
- 8.75" high
- widths are multiple of 2.125", usually 1, 2, 4, or 8 times that (where 8 is a full 19" width)
Electronically
Waveform Voltages: +-5, i.e. 10V peak-peak
Gates: 0-5V positive ON, threshold 1.5V
DC Power: +15V, -15V, +5V
Internal power: Six-pin MTA .1 (+15, -15, +5, Gnd, and keyed against reversing it)
External DC Power interconnect: 6 pin DIN (carrying the same. Not sure about current rating)
https://synthesizers.com/technical.html
MOTM (5U, ~15V)
Height is 5U (8.75"), width units of 1.75"
+-15V and +5V
Patch interconnect: 1/4" (6.3mm) TS jacks
Audio 10V peak to peak
1V/oct pitch
Moog (5U)
On Moog V-trig and S-trig
Moog used voltage triggers (V-Trig) alongside switched triggers (S-Trig).
tl;dr: You don't need to bother with S-trig unless you have specific old Moog hardware that uses it.
What they are
In V-trig, zeroish volts is off, some higher voltage is on.
- This is basically the same as what later systems call gate
- This is easier to interconnect, interact with, and sometimes to abuse, than S-Trig
In S-trig, floating/hi-z connection means off, pulled to ground means on.
- This means you can use any passive momentary switch, without further circuitry.
- It also means you can safely connect multiple ones and they'll safely OR together.
- Yet using the result electronically is a little more annoying.
- Also it uses a different connector in modular moogs, which is a bit cumbersome.
Conversion:
- V-trig output to S-trig input amounts to a single resistor-and-transistor, which you can do within a cable because it needs very little current (and is powered from the voltage end).
- S-trig output to V-trig input you'll need power from somewhere else.
- In modular this is likely to be a (very) basic module that has a bunch of inputs and outputs, supporting the few specific devices you have that need it
Some digital controllers may support less-usual outputs (S-trig, or indeed Hz per volt) just because modern hardware makes it easier enough to do.
See also:
Buchla 200 (4U)
Serge (4U)
FracRack (3U)
Fractional Rack, a.k.a. FracRack or just Frac.
Used only by a few brands, like Blacet
Height is 3U, width units of 1.5 inch.
Patch interconnect: 3.5mm TS (mostly)
Voltage is +-15V
An American format
Note that while the height as Euro, the rails and screws are a little different.
There are some Euro modules that are fine in Frac's higher voltage, due to their design, but you basically need to know this per module.
See also:
Unsorted
1U is common in audio gear, to put more things in a 19" rack vertically.
Modular sees the occasional 1U tile modules, and the occasional 1U mults between rows
KOSMO Format - Look Mum No Computer's choice
- based on what he had
- first 15cm acrylic plastic (~3.3U, so a little larger than eurorack's 3U)
- then KOSMO 2.0, 20cm metal (~4.5U) and that he liked more room,
- in rack units that's and
- allows larger knobs and plugs, which seem to be his style
- electrically is eurorack (+12/-12V)
https://www.synthesizers.com/formfactors.html
https://www.muffwiggler.com/forum/viewtopic.php?t=122943&sid=b53b5a9a173d4af4ab449586a3888545