Saturday, April 4, 2009

Jitter, Revisited

Taking a break from the jiggy series to tell a "jittery" story...

A couple years before I met Kevin, Hans Zimmer hired him to build him a very large Moog modular.
(HZ just wrote Kevin to say it's still fine 16 years later, too!)
According to Kevin, they both discussed synths right and left then and what made them sound good.

Hans said that he preferred the sound of the Moog 901 oscillators.
At the time, Kevin didn't understand what he meant--An oscillator to him had pitch, waveform shape, volume and that was about it.
But Hans was certain this wasn't the case.

Two years later, right as Kevin was putting the finishing touches on Hans' big Moog, he found himself with many extra parts left over.
One part was a blank Moog 921B panel.
Kevin decided to use it to build a CEM3340-based oscillator, which is used in Jupiter 6's, Memorymoogs and many other synths we all know and love. (And Kevin's portable VCO!)

You'd think that using a VCO on a chip wouldn't be any different than what Moog used in a 921B... they're both oscillators, right?
When Kevin installed and made a patch with this new "CEM" 921B, he suddenly realized Hans was right.
There was some sort of sound the Moog's had that this new oscillator didn't.
In Kevin's words, the CEM3340-based oscillator "sounded razor sharp and sterile."

This keen observation wasn't thought of much for several years.
Then, a couple years ago, while restoring a Moog modular, Kevin made a video showing a Moog 901 oscillator plugged directly into an oscilloscope.
Here it is:
This video showed some sort of "jitter" in the waveform that wasn't present in the CEM3340 chip.

In an interesting and semi-amusing round of "shoot the messenger," Kevin was slammed when he showed this video to others.
They insisted that the video showed nothing that could actually be heard, and anyone who thought otherwise was probably hearing things.

Ken Elhardt took this position after looking for this jitter with his DAW (Digital Audio Workstation).
However, he didn't use an actual 901 oscillator. He used an mp3 of one. (That always struck me as odd, trying to analyze a digital sample of an analog waveform. But hey, I only change caps and rebush keyboards!)
Mr. Elhardt and others added that they prefer a very accurate and stable VCO and if they wanted any "jitter", they'd add it in with noise or whatever. (Personally, I'm not much for ultra-accurate, sterile-sounding oscillators.)

One person even said that Kevin had documented how Moog power supplies were actually responsible for this jitter.
This made me laugh, as the video clearly shows that he used a precision laboratory power supply, not a Moog power supply.
(More on that supply in future posts, as it's the heart of many things that will be featured in the "jiggy series".)

Finally, several people didn't understand what Kevin was showing at all and said whatever great Moog sound there might be is due to drift. (I hope these folks weren't the same ones who like their oscillators sterile!)

But one very well respected tech, Juergen Haible, also countered that Kevin was wrong.
This hurt Kevin's feelings because Kevin used to be friends with him.
He used to share parts, photos and music with him and hosted his site for free.

With all that in mind, I submit for your perusal:
It's a vco design made by Mr. Haible that he calls "living VCOs."

In reading his description, Mr. Haible describes things this way:
"...A good part of that special sound of early Moog and EMS oscillators is not because of any "randomness", "unstability", "instability" or "noisyness", as so often is said. A good deal of their behavior is because of that, but it is not the whole story. There are also some very deterministic factors in these old circuits which have been unpleasant side effects for the designers back then, but which are worth a closer analysis when we're designing a musical VCO today. "

Kevin found this amusing. (Once he reminded me of this whole jittery debacle, so did I.)
He pointed out that while Mr. Haible says what his VCO's aren't, he doesn't say what they are. No schematic or analysis was provided (at least yet.)
No mention of the word jitter either. ;-)

But Mr. Haible's board layout suggests a design very close to the Minimoog and Moog 921B oscillators. Many of the same exact parts are used as in Moog oscillators.
Mr. Haible's comments suggest that he learned something new here.
He designed his oscillators not to be super accurate, but to be musical instead.
He just built his "living VCOs" upon Moog's basic design.
He intentionally sacrificed perfection for musicality.
Isn't that what Hans originally observed and Kevin documented?

I think Mr. Haible changed his mind, and is skittish about using the word, "jitter."
No harm in a man having to change his position after being so insistent on it previously! (Though I suppose admitting that might be difficult.)

By the way, Kevin said he's has no hard feelings and would love one of his boards.
If we can find the extra cash, I wonder if Mr. Haible will sell us one? :)


jhaible said...

Thank you for the advertising, Mrs. Synthfool.

I have to disappoint you about some of your conclusions, though: I did not, in any way, "change my position" about VCOs since these debates you seem to recall so vividly.

The very VCO circuits that I'm offering as a PCB project today, have been part of my JH-5 synthesizers at the time when these discussions took place; what I've been talking about back then wasn't theory, but backed up with hardware I had designed and tested, and played every day, at the time.

I really wonder what makes you think I've changed my mind. Always learning, and never afraid of learning something new, but in that particular case ... I don't know what exactly you refer to.


Fountain of Filth said...

You don't expect me to read all that, do you? :)

jhaible said...

Never mind! :)

JH. (slightly amused)

Owl said...

More talk of what these vcos aren't, but I'd still like to know what they ARE.
I don't know what JH's POV was before, nor now, so it matters little to me if he changed it.
I just want to know what makes these vcos sound special vs other vcos.
How are they different?
I'm no expert but would love a simple explanation.
All I hear is what they aren't.
JH, can you explain (simply) why these are "living" ??

jhaible said...


I published some samples of the result (audio files) and some words about the theory behind it here:

(Recently updated)


Owl said...


I've read your description several times, but still cannot see any description of why your vcos are "living" or why they're different.
You state you like the sound of one vco over others, then fail to explain what the reasons are for that preference.
Nothing was actually explained.
You do seem to suggest drift.
If you mean drift, please say so.

I did look at Kevin's "jitter" video again twice and it's fairly obvious he was referencing one VCO against the scope's timebase, not another VCO.
The waveform does show some sort of movement in his video too, so I can't understand how drift or beating would be the explanation.

From all I can gather, you state you like the sound of a certain VCO, then made VCOs the same way.
But there's no explantion why you liked that VCO in the first place.

If faith and not technical explanations are in order, then perhaps your cause is more religious than you'd like to admit?

Thank you for your reply, but I must be dumb because I can't find what your explanation is explaining.

Also, "lose" is the word I think you meant to spell, not "loose."

sparenerd said...

Perhaps JH is citing "constant beat ratios" as linear oscillators can do?
That would make some sense, but only when paired with other such oscillators.

All good if so, but if CBR's then this really has nothing to do with single oscillator tonality.
It neither supports nor negates the jitter question.
It's just different.

JH please explain better.
I am interested too! :-)

jhaible said...

I'm a little bit puzzled here.

I think I have explained everything I possibly could about my VCO project, except giving the actual schematics.

I have a sneaking suspicion that all these "unanswered" questions are about the context in which the *Blogger* has put it - and that's beyond my capabilities to explain. I mean, I know about this ominous "jitter" experiment of Kevin, and that it was widely rejected by those who took a closer look, because of the trigger conditions of the scope that can show an excessive jitter where actually there's just a much smaller amount of jitter. I didn't take that experiment seriously, and still don't. So it's really hard to answer questions that obviously refer to that.


Kendall said...

And yet, Kevin apparently applied the same o-scope test to his CEM-based VCO and saw no jitter... so it couldn't be the o-scope, could it? Too bad we don't have a video of the CEM test.