A perennial question among modular synthesists is “do I really need a buffered multiple to connect to my oscillators?” The correct answer is “it depends” because there’s so many variables with the way different modules were designed. I figured it was time to flesh out those details so it didn’t seem like so much voodoo.
In simple terms, every control voltage source has an output impedance, and every control voltage input has an input impedance. When you have voltage and either resistance (DC) or impedance (AC) involved, there’s a chance the voltage is going to be reduced. The amount depends on the ratio between the output and input impedance. This reduction is usually too small to be an issue when opening a VCA or modulating a filter cutoff, but it can be audible when controlling the pitch of an oscillator, causing the pitch to go flat on progressively higher notes.
The One Output/One VCO Case
Let’s run through some examples. If the output impedance of your quantizer or sequencer or MIDI/CV converter is 1k or 1000 ohms (a common number encountered in vintage gear), and the input impedance of your oscillator is 100k (the defacto standard many follow in the Eurorack world today), the voltage is going to be reduced by 1/100. That works out to 0.01 volts getting lost for every volt of change.
In a typical modular synth that uses a 1v/octave system for pitch, an error of 0.01 x 100 cents per semitone x 12 semitones per octave = 12 cents of detuning per octave, which is quite a bit. The threshold of the brain perceiving detuning is usually stated as 4-5 cents, although if you are tuning two oscillators in unison at a high pitch, you may be able to detect even smaller changes than that.
Increasingly common among modern synthesizer module designs is an output impedance of 100 ohms or lower, which would yield 1/10 or less of that error – closer to 1 cent per octave, which should be below the threshold of perception in most cases – or at least a lot easier to live with. However, you can’t assume this; many new modules still follow the old “standard” of 1k output impedance.
Notice we’re not even talking about using a multiple yet – this is what happens when a single output is connected to a single input. Depending on the modules involved, you could get noticeable detuning even in this case; fortunately, most of the time you won’t.
The Multiple VCO Case
The opportunity for detuning increases as you connect more VCOs – particularly if you connect them using a passive multiple or simple cable splitter. In this case, the effective input impedance of connecting multiple oscillators is their average input impedance divided by the number of modules connected in parallel. Using the common value of 100k for input impedance, two oscillators would then have an input impedance of 50k; three oscillators would have an input impedance of 33k; etc.
For the same output impedance of our controller, that means the detuning could double when you add a second oscillator using a passive mult. If your modules have really low output impedance and really high input impedance, the error may still be so low that you don’t hear it (or at least, can live with it). If those two impedance numbers are closer together, then you’re probably going to have tracking problems.
How Does a Buffered Multiple Fix This?
An active, buffered multiple should have three features:
- a very high input impedance (say, 10 megohms) and very low output impedance (nearing 0 ohms when not shorted out; usually well under 100)
- no multiplication or division of the voltage going through it – it should multiply it by as close to 1.0000 as possible (although there is a worthwhile exception to that, which I’ll get to below)
- no offset in the voltage going through it – the multiple should not add or subtract any DC offset voltage
That first property of the buffered mult is what fixes our tuning issues, even in the single controller/VCO case. The last two properties are required to not introduce any new problems. Not all buffered mults meet all three of these requirements; the ones that don’t have been quietly discontinued or updated. I’ve had good success with the Intellijel Buff Mult and the Malekko Performance Buffered Mult; I also have a few of the brilliant AJH Synths V-Scale adjustable buffered mults arriving soon.
Note that if your controller is not outputting precisely 1.0000 volts per octave (1.2v if you’re using Buchla-standard equipment), or if your oscillators have their own tracking problems, a buffered mult will not fix their problems – you should spend some time re-calibrating them rather than trying to patch around the problem. The exception to this is the AJH Synth V-Scale, which has trim pots next to most of its outputs so you can tune out other problems including oscillators that do not have perfect tracking – it may be easier to tweak the V-Scale than to go through the tuning procedure for the oscillator.
A Case Study
The following video shows connecting a popular semi-modular synth – the Moog Mother-32 – to a pair of external oscillators (an Expert Sleepers Disting and a Mutable Instruments Braids), first using a passive multiple and then a buffered multiple (an Intellijel BM). You can clearly hear the difference in tuning accuracy – the Mother-32 stays in tune itself, as it has been calibrated taking its own internal circuitry into account, but some external modules fall flat in comparison at higher pitches.
The Disting and Braids have an industry-standard input impedance of 100k; using a passive mult, they look like a single 50k impedance to the Mother-32. The Mother follows the vintage synth standard of 1k output impedance for its “KB” output as well as for its own buffered multiple. Moog used a 1k resistor in a classic current limiting arrangement to make sure you didn’t damage the Mother’s outputs or cause disturbing dips in the signal when you inevitably shorted the connections while plugging cables in and out. Some other companies use a different circuit design that provides current limiting while keeping the output impedance much lower; that’s why you may not experience tuning problems when using their modules connected through a passive mult.
If you have a module that is experiencing detuning issues when driving external VCOs, then a buffered mult that matches our criteria above will cure it. In the video above you can clearly hear where patching an Intellijel Buff Mult between the Mother-32 and the oscillators tightens the tuning right up. Some Mother users have also reported that running its 1v/oct “KB” signal through a CV mixer also fixes their external tuning problems; that would work if the mixer had very high input impedance and low output impedance plus didn’t alter the voltage of the signal going through it when set to unity (a lot of “ifs”, but some units qualify). However, you cannot assume all buffered mults will fix all problems; if they have an output impedance of 1k ohms, they may still interact with the input impedance of some oscillators as do the mults in the Moog Mother-32.
To Review:
You might notice that not many synth module manufacturers state the input and output impedances of their connections in the specs for each module. It would certainly be nice if more of them did. For those who do, a 100k input impedance is considered normal (with higher being better – especially for buffered multiples); a 100 ohm or lower output impedance also seems to be typical. If you can find out those numbers, here’s what to look out for:
- To properly drive a VCO with an input impedance of 100k ohms without noticeable tracking errors, you would like your MIDI to CV converter, sequencer, buffered multiple, or other pitch voltage course to have an output impedance of ~100 ohms or under. (1k ohms on the output is fine for audio signals et cetera; it’s pitch voltages we’re focusing on here as being the exception.)
- If you have a module with an output impedance of 1k ohms for its pitch voltage, you really want to run that through a buffered multiple that has an input impedance of 1M or higher. If your buffered multiple has an input impedance of 100k, it’s not going to fix the tracking problems created by modules with an pitch voltage output impedance of 1k (although they will work fine if the pitch output impedance was 100 ohms or lower).
- If you are intending to drive multiple oscillators from a pitch voltage output, that output needs to have a very low impedance (100 or lower – the more VCOs you are driving, the lower this needs to be), or your VCOs need to have a very high input impedance (all of them – the lowest one sets the bar for their combined input impedance, and it only does down from there), or you need to use a buffered multiple to get good tracking.
- If you don’t know what the impedances of your modules are, use a known good buffered mult (like the Intellijel, Malekko, or AJH Synth) to avoid potential frustrations later with VCO tracking.
(For more on the types of multiples available and when to use which one, see my older article Multiples: Passive, Active, and Logic)
Measuring Tracking Errors
If you want to measure the effects of impedance resulting in voltage losses, try the following test that I learned from Graham Hinton on the Muff Wiggler forum. Get an accurate digital multimeter (preferrably one capable of 100uV resolution) and a set of alligator clip probes to clip onto the shaft and tip of a patchcord plug to help keep your hands free. Set the meter to DC Volts mode. An alternative is using a Mordax DATA in Voltage Monitor mode, which you can then just patch into the module under test.
Run a cable from the output of your sequencer, MIDI to CV converter, etc. directly to the meter, and measure the result at various octaves; there should be a 1.000 volt difference from octave to octave. Meters have very high input impedance, so they should not drag down the voltage on their own. An error of about 0.001 volts – 1mv – is usually considered acceptable, if you see something larger, you may need to recalibrate or replace your voltage source.
Now plug the output of your voltage source into a passive multiple. Run one or more outputs of the multiple to your oscillator(s), and spare output to your meter – this is the voltage that is going to your oscillators. Play the same notes again, and see if there is any change in the voltage you measure. Changes of about 2mv or larger should start to worry you. If you do measure a difference, try replacing the passive multiple with an active multiple and see if you get better results.
great video. very good for people learning synths. i’ve known these tricks for a while but seeing you make demo’s and explaining things clearly is still fun to revisit. keep up the amazing work.
This demystified buffered and non-buffered multiples. Alas, I have many non-buffered multiples and could use a buffered one. Well, back to buying modules, I suppose.
Hello,
I was about to acquire the intelijel buf mult, but then spotted the nifty polarity indicating 4ms unit. Do you happen to know if the 4ms works as well as you suggest the inteligel works?
I’ve not used the 4ms Buff Mult, so I can’t speak for it. In general I like 4ms gear…
There are two buffered mults that have me really excited these days:
– The AJHSynth V-Scale, for trimming the tracking of attached VCOs (it’s also a great buffered mult). I’m about to create a pair of videos on using them.
– The Frap Tools 333, as it contains 3 sets of 3 input precision adders going into 3 buffered outputs. I also just saw a video of the first production batch being tested, and they had colored LEDs, so it might also have the polarity indicator you liked on the 4ms. I hope to have one in a few weeks and will definitely report back.
Here’s a video on the Frap Tools 333:
https://vimeo.com/204007249
From the 333 manual:
Input impedance 100 KΩ ±10%. Output impedance <50Ω. Tolerance of <0.5%, with global offset within ±10 mV per section (or ±12% of semitone).
The <50Ω looks good but the other specs seem suboptimal.
The V-scale just says 'manual coming soon'. In the absence of a spec I'm tempted to just put one together myself.
I got the 4MS Buff Mult because I liked the idea of the polarity indicating LEDs, but the fact that it’s 3HP instead of 2HP like the Intellijel Buff Mult means that you’re left with an odd number of HP left which can be annoying to fill.
Then get another odd-hp module to keep it company! (wink)
Good read, thank you.
I just wanted to mention that the 1k output and 100k input impedance is the standard set by Electronotes in the 70ies. The reason for having this ration was not only the short protection of the involved electronics on the outputs, but also to have the option to do passive mixing.
Thanks, Matthias. Fortunately, today we have better solutions for short protection, mixing, and multiples; it’s unfortunate some are sticking with the old “standards” just because they were “standards” rather than looking at the side effects we learned about since then.
Hi Chris! Thanks for the article! It explains a lot to me! I have a question as I’m building modules to teach schoolchildren (and adults 😉 ) about synthesizers. I’ve seen/tried many schematics and most have the standard input and output impedances you describe and I must admit that I am using the same values, just because everyone else uses them as well. So, could you please point me in the direction of the better solutions for short protection, mixing, and multiples? So I can redesign the inputs and outputs that need it. My modules will for sure be wrongly connected and I expect a lot of outputs to outputs patches. Thanks!!
I’m afraid that hardware design was never my strong suit, and I haven’t done any in decades – so I’m not the person who can give yiou those solutions. However, there should be good examples available on the internet.
Just what I needed to know. 😎 Excellent explanation
Awesome article thank you! I think I can safely deduce that if I use buffered mults for OSC CVs, and passive ones for everything else, I should be good to go.
I think audio is less fussy.
Yes – you are correct. (And I’m glad you found the article useful!)