dmlandrum

Modular synth design: V/Hz vs. V/Octave
« on: July 16, 2010, 06:27:32 PM »
A long time ago, I had wanted to build a modular synth. Indeed, that's what got me started in this whole DIY thing to begin with. I had given up, to be honest, but I never shook the idea, and now I finally say, to hell with it, I'm gonna do it.

In the "make some fundamental decisions about the design" department, my question is about V/Hz vs. V/Oct. The vast majority of analog synths and just about all modulars are V/Oct. It turns out it's more complicated to design modules to respond to V/Oct, but I guess it greatly simplifies other aspects of the design...? Actually, that's one of my questions. Just what is the advantage of V/Oct that it's worth it to take simple module designs and make them complicated?

The follow-on question is, is there some overriding reason not to design a modular synth around V/Hz, aside from compatibility with commercial modules? If I'm not mistaken, Buchla modules are all V/Hz, and those happen to be my favorite modular systems around.

Thank you all for the help. I like to think I'm not a beginner, but hours of searching turned up no useful information on these questions.
Darren Landrum

Be comforted that in the face of all aridity and disillusionment
And despite the changing fortunes of time,
There is always a big future in computer maintenance.


Curtis

Re: Modular synth design: V/Hz vs. V/Octave
« Reply #1 on: July 16, 2010, 07:49:07 PM »
Off the top of my head the major advantage of V/Oct is that the range of the control voltage isn't too large for a synth requiring multiple octaves of pitch generation - starting at, say C1 = 1V, then the next octave up will be C2 = 2V, C3 = 3V, C4 = 4V etc.  To get a large pitch range out of the synth doesn't require massive amounts of CV.  The downside is that in order to translate the V/oct CV into musical pitches you need to convert the linear change in CV to an exponential change in pitch, which requires (not surprisingly) a linear-exponential converter.   The common versions convert the linear CV into an exponential current, which then controls your VCO by changing charging/discharging times in a capacitor.  The exponential converter is easy to build but is difficult to control, being very temperature sensitive.

V/Hz require no expo converter, as the CV itself will be scaled exponentially, so if the CV for C1 is 1V, then C2 = 2V, C3 = 4V, C4 = 8V etc.  The obvious downside is that you're going to have a synth with reduced octave range because after several doubles of the CV you're going to run out of headroom.  Generating sub-C1 CV's (eg LFO's) will be difficult too because as you halve the CV further you're taking it closer to the noise floor.  Finding a way to generate the exponential CV in the first place can be tricky too - resistive voltage divider ladders are one way, but finding appropriately-sized resistors will either dictate the use of custom laser-trimmed units, or living with some degree of note "wonkyness".

The Moog Taurus bass pedals have an exponentially-scaled CV using a special resistor voltage divider network (the need for a wide-ranging synth in this application obviously isn't so great - only 13 pedals to press!)

The Minimoog has a linear CV and exponential converter in the VCO (multi-octave with LFO's in a synth with 10V rails dictates the need for a way to expand the usefulness of the limited CV range)

« Last Edit: July 16, 2010, 07:56:10 PM by Curtis »
--------------------
http://www.thethirdending.com
--------------------

dmlandrum

Re: Modular synth design: V/Hz vs. V/Octave
« Reply #2 on: July 17, 2010, 10:57:38 AM »
Yeah, I kinda thought that would be the primary advantage. It's just too bad that putting exponential converters in every module increases their complexity by large amounts, at least if you want good temp stability. I guess I'll have to search and play around with some ideas. Thanks for the help.
Darren Landrum

Be comforted that in the face of all aridity and disillusionment
And despite the changing fortunes of time,
There is always a big future in computer maintenance.

PRR

Re: Modular synth design: V/Hz vs. V/Octave
« Reply #3 on: July 18, 2010, 02:17:35 AM »
20Hz-20KHz is 1,000:1 range.

We pick a "note". What accuracy do we need? What will be "out of tune"?

The difference between (european equi-tempered) notes is about 1.06. This is split into 100 Cents. The "tolerable" pitch error varies, but say 5 Cents. It works out to about 1 part in 333,000 apparent accuracy(?).

Take 10V as your top voltage. You need 0.03 milliVolt precision. Common older opamps have 2mV errors and 1mV drift. This is 100 times worse than we need. Consider: 20KHz=10V, then 20Hz is 0.5mV and the next note at 21.2Hz is 0.53mV. With just 0.5V short-term error drift, "20Hz" could come out zero Hz or 40Hz. We don't come close to "note" accuracy.

Yeah, we don't play 20Hz much. But even at 80 Hz thud we are liable to be 30 Cents off.

We could say that we don't play 20KHz. Few traditional instruments are played much over 2KHz nominal. They have higher harmonics. In subtractive synthesis we start from a low buzz-tone and filter the overtones; but some synth work is additive where overtones are inserted to taste.

If we work at 1V/8ve then a semitone is 60mV and a Cent is 0.6mV. At least on paper we can get few-Cent accuracy with exponential, not with linear.

The actual tolerable pitch error depends on freq. Bass can be some sloppy. Fiddle has to be pretty good (or mask the error with vibrato.... and here we get into the history of performance practice complicated by recordings....). In fact what we want is to keep any off-pitch beat rate too slow to notice on one note. (This makes linear more tolerable than you might think.)

A particular advantage of V/8ve is that if you want two oscillators to sound in octaves or thirds (or arbitrary interval), or track a filter a musical-interval up from a buzz-tone, you just add a Volt to the CV of one. You can do the same with linear by using an amplifier: 2X gives octaves and variable gain gives arbitrary intervals. However in a non-trivial pitch-patch it may not be easy to apply the right coefficients in the right places.

Some special cases: the keyboard for V/8ve is a string of identical resistors. For linear it is a mess of values; like a switched attenuator except we tolerate 5%-2% errors of gain, we need like 0.2%-0.1% error of pitch. Maybe today you can score 40-some 0.1% resistors in 12th-root-of-2 intervals from Mouser, but not when Bob was toying with the idea.

PAiA had a well-considered linear synth. Oscillators had limited range without switching; but then so do all musical instruments (bass, cello, viola. violin) except piano and organ. The oscillator was a unijunction and a few other parts. Stability was not a lot worse than an early ARP at 1/100th the cost. The real breakthrough was a 4-bit(!) CPU to do some management (and this is about where I got off the synth bus).

If you want a good synth, there's computer software which will do Switched-On Bach or Age Of Electronicus better than the originals (mechanically).

If you wanna mess with synths, do whatever seems best for your level of interest.

I've told this story. One of the best "synth" works I ever heard, the musician went into the Synth Lab and found that the synth was dead. But the tape deck worked, there was a microphone, and a large rubber band. The result was very musical, moreso than most of the drek I heard. It's not the fiddle, it's the fiddler.
« Last Edit: July 18, 2010, 02:28:59 AM by PRR »

StephenGiles

Re: Modular synth design: V/Hz vs. V/Octave
« Reply #4 on: July 18, 2010, 10:41:56 AM »
 ;D
The best place to find a helping hand is at the end of your arm!

dmlandrum

Re: Modular synth design: V/Hz vs. V/Octave
« Reply #5 on: July 18, 2010, 12:37:53 PM »
So... Just build some sh*t and use it.

Okay, I can do that.  ;D
Darren Landrum

Be comforted that in the face of all aridity and disillusionment
And despite the changing fortunes of time,
There is always a big future in computer maintenance.