I know this forum hasn't gotten a lot of action these days (visit wise, that is), but then again I kinda like my solitude every now and then. Hopefully, however, this post doesn't become a soliloquy (if it does, then just refer to me as the old guy who talks to himself).

I'd like to mention a technical aspect regarding the difference between older and newer analog VCOs. Most people will agree that old oscillators sound thicker and warmer, and many assume that it has to do with drift. Well, drift does play a part, but there's another component that has, in my opinion, a greater impact on the shaping of the sound.

Take, for example, Moog's original VCO design, the 901 and 901A/B set. The core of those oscillators were built around a unijunction transistor (UJT). A sawtooth oscillator can be constructed with a UJT by simply adding a few resistors and a single capacitor. This was the heart of the 901 VCO (the rest of the circuitry consisted of wave shaping circuits and the linear-voltage-to-exponential-current-converter).

A UJT is essentially an electronic switch. This "switch" remains open while a capacitor connected to it's emitter is allowed to charge through a resistor. This slow charging of the capacitor is what produces the ramp side of the sawtooth. When the charge reaches a high enough value (the peak voltage) the UJT becomes forward biased (in other words, the switch closes) and quickly discharges the capacitor. After this action the circuit is reset and the capacitor is allowed to charge again.

UJTs were relatively stable for their time (60's) but they were not high precision devices. If you were to take a snap shot of a UJT VCO's sawtooth wave, that's exactly what you would see. But, if you were to take several snap shots of the wave at different times, then superimposed them together you would notice that they don't line up precisely. The reason is that there is a very minute variation in the charge/discharge time of the circuit. It's very small and hard to detect, but this ever-so slight variation causes a tiny phase shift in the wave form. Also, keep in mind that this shifting back and forth is taking place at audio frequency.

By the time the Minimoog had become available Bob Moog had redesigned the oscillators. He had done away with the UJT and instead began using a field effect transistor (FET) to discharge the capacitor in the circuit (a technique still used today in some designs). The FETs were much more precise than the UJTs and this resulted in a change of sound (there were still problems with drift, but that had more to do with the exponential converter than the oscillator core circuitry).

That was just one example, but what I was hoping to point out is that it was/is because of the variation in tolerances, component construction and circuit design that give old analog it's characteristic sound.

An interesting way to simulate this variation with newer, more stable VCOs is to use a sample and hold. Gate the sample rate with a pulse wave (preferably as narrow as possible) at audio frequency, then run the output of the S/H to the FM input of a VCO (patching the output back into the same VCO will also work). Sampling white noise will produce the most random results, but just about any high frequency source will do. The key is to use a very small amount of modulation.

Modulating a VCO directly with a noise source will also change the sound (of course), but the result will be different. By using the S/H the modulated VCO will track the sampled voltage with each cycle, thus simulating the instability of older oscillators. I've tried this and though the result is quite subtle it's also noticeable.

I'm sure there are many here who already know a myriad number of ways to warm up a thin sounding VCO. The technique I just described is just one way to add a little "life" to otherwise cold sounding wave forms.

If you notice any factual errors on my part please free to point them out. I'm always willing to learn something new.

Cheers!

Thank you. It's great to feel welcome.

The technical side of synthesizers isn't difficult to grasp if broken down into easily digestible chunks. I began DIY-ing audio circuits in the late 60s-early 70s even before I knew what a proper synthesizer was. When I discovered Moog synthesizers in 71 I began reading every book on electronics, synthesis and sound theory that I could get my hands on. By 1973 (at the ripe old age of 16) I'd designed and built my own synthesizer, which included: 2 VCOs, a "modulation VCO", what I called a cascade VCF w/ switchable attack, release or attack, decay EG, a VCA w/attack, release EG, and (wait for it) a programmable arpeggiator! I played it using a keyboard I'd salvaged from an old home organ I bought from someone for $30.00 USD. I dearly regret not having saved that little creation of mine, but being young and having a budget of $0.00 it was eventually cannibalized for other projects. *sigh*

Anyway...

A little more on ancient VCO technology.

IMHO, if there's any one component to blame for having caused so much design stress it's the charging capacitor. If, for example, you were to build a simple UJT oscillator like the one I described earlier (called a relaxation oscillator, BTW) and charged the capacitor with a voltage, the resulting wave would not be a perfect sawtooth. When a capacitor is charged with a voltage it charges exponentially, so instead of a nice straight ramp leading up to the peak you get a funny "humped" sort of shape. To give you an idea, imagine a sign wave beginning it's upward swing, then before it reaches the crest it's suddenly chopped. This type of wave form, though having an interesting mellow timbre, sounds nothing like a true sawtooth wave.

The solution to this is to charge the capacitor with a current, rather than a voltage. When a current is used the capacitor produces a nice straight ramp as it charges. The easiest way to convert a voltage into a current is to use a transistor. Applying a voltage to a properly biased transistor will produce a current at it's output (this circuit is known as a current source or current sink, depending on it's configuration).

There you go! A perfect sawtooth!

[A tidbit; the core of most if not all discrete analog oscillators are actually Current Controlled Oscillators. However, if one considers the voltage/current converter and wave shaping components as part of a single circuit, then "VCO" really isn't much of a misnomer.]

Now, however, the domino effect begins. A single transistor is far too unstable on it's own, so a second transistor can be added to produce a much more precise current. Connecting two matching NPN transistor into what is called a Darlington pair creates an excellent current source for our VCO example.

http://en.wikipedia.org/wiki/Darlington_transistor

So far, so good. In fact, an op-amp can be added to the circuit to increase precision even further.

The only problem is, at this stage the VCO is linear response. In other words, a change of one volt at the input will not produce a one octave change of output frequency. It's a shame too, because this is a very stable design.

What we need to get the 1V/Oct change is the dreaded linear-voltage-to-exponential-current converter...

...which I'll go into in my next post.

As always, feel free to point out any mistakes I may have made.

Cheers!

this is really interesting! Thanks for sharing your knowledge with us.

So, say a Yamaha or Korg synth- it's got a linear voltage scale, right? (Hz/V?) do the VCOs (which on the CS15 are ICs) just lack the linear-to-exponential converter in the circuit? or being IC oscillators is that a whole different beast all together?

That home-brew synth of yours sounds wicked! I met a guy who made one himself, all he had left from a massive flood was the remains of the keyboard.

You're welcome ctenophonic. I've always been fascinated by older synthesizer technology. The new stuff is great, but I'm definitely old school when it comes to circuitry design and technique. To me, the golden years of analog synths were from the late 60s to the early 80s. Just keep in mind that I'm working from memory here and may be just a bit rusty.



Good questions. The answers are a little complicated, but I'll try not to get carried away as I'm apt to do at times.

When it comes to discrete oscillators the core is by nature a linear response circuit (which consists of the charging capacitor, output buffer, and feedback loop to the discharging switch). However, linear response does not necessarily mean that it will have a specific change in frequency per every change in input voltage. One oscillator design may produce a frequency of x with a 1 volt increase, while another design will produce the same frequency change with only an 0.8 volt increase.

Back in the early days, there was no industry standard regarding linear response, so each synth manufacturer was free to give their design whatever specs they required. As an example, Moog's 901b and 921b VCOs were linear response (the 901a/921a "driver" was actually the exponential converter shared by these modules). If you were to apply the same voltage directly to a 921b linear input and a Korg linear VCO they would not have the same output frequency. Vary the voltage and the offset in tracking would be even more pronounced.

Even the so-called standard of 1V/Oct wasn't always strictly followed. The micromoog and multimoog were scaled at 1 octave per every 0.95 change in voltage. There was another old synth (I can't recall the manufacturer at the moment) that was scaled at 0.85V/Oct.

I'm sorry to say, but I'm not aware if a Hz/V standard has been established. I do have a young friend, however, who tried to use the Hz/V output from a Paia Fatman to run a linear VCO chip and discovered that it would not track properly.

There are ICs like the CEM3340 and SSM2038, but those are specialized VCO chips that I'll get into shortly.

I hope this was helpful.



I'm sorry to hear about your friend's losses. I experienced flooding two years in a row a couple years back and I'm still working on getting the place back together.

I built a large number of modules and general audio gadgets through the years only to sell them off or tear them apart to build something else. It never occurred to me that there might come a day when I would miss those old projects.

BTW, nice avatar. That's Mike Nelson being changed into Mikey by Tom and Crow's electrode-equipped glasses in episode 514 Teen-age Strangler. I'm a long time MST3K fan. If you haven't already, you should check out Rifftrax.com. Same guys doing the same thing to new movies.

Cheers!

This is some of the most valuable information on the internet regarding analog synths. Thank you!