How analog oscillators work

[jupiter8]

The construction of analog oscillators came up in another thread and someone PM:ed me some question but instead of replying there i'd thought i'd post it here instead for everyone to read.

To begin with i'm not an expert on this subject. There are several on this board that knows far more than me about this subject. So the following might not be 100% accurate but should be pretty close.

So what was the question ?
How can you get several waveforms at once from a single oscillator ? In this case it was the Roland SH 101. To begin with the noise comes from a separate circuit.

Then you start with generating a sawtooth wave. How that is done is beyond this class for the moment but feel free to ask. You could also start with a triangle but as i understand it that's quite unusual and not what the SH 101 does.

So we have a sawtooth wave so that's one down,two to go.
You can,in paralell, compare the voltage of the sawtooth to another voltage with a comparator.
http://en.wikipedia.org/wiki/Comparator
If the saw is higher the output of the comparator will go high and if it isn't it will go low. So there's your square wave.

Note that by varying the voltage you compare to, the switch to low will happen at different times in the cycle resulting in different pulsewitdhs.


Note that the square and saw will have exactly the same frequency. They cannot differ when you do it this way.

The subocillator is simply a counter or something similar (i'm not completely sure what it's called.flip flop,counter,shift register).
Anyways the principle is that it has two states,hi and low (two different voltages cold be 5 and -5 volts for example) that it changes on a positive trigger. What that means is it changes state on every complete cycle of the main oscillator resulting in half the frequency IE one octave down. Again this will have exactly half the frequency. You cannot detune it slightly.



So why go thru all this trouble instead of simply building separate oscillators for every waveform ?
The answer is really simple: It's much cheaper.

Comparators and counters are really simple circuits so you'll add a lot of sonic power for very little cost.



And that's all i have to say about that for now.
Questions ? Corrections ?

[MadeInMachines]

Hi, I like this interesting post. I know practically nothing about this so was googling it in an effort to find out how electricity can be made produce different waveforms or indeed different types of waveforms.

From what your saying it sounds like the sawtooth is being manipulated using various circuits to result in the creation of other waves - so when i swithc from saw to square another circuit if kicking in on my synth that causes waveshaping?

How is the original saw created though?

[aciddose]

an integrator.

you keep adding to a voltage until it reaches a point. the output is compared with a preset level and if it is higher, a pulse is generated. that sync pulse shorts out the capacitor, returning it back to the beginning.

so an analog oscillator is a analog timing circuit. it's an integrator (filter) fed by the "frequency" input and an amplifier that generates a sync pulse to reset the filter.

compare this against a dco which is a digital sync pulse that resets a similar integrator. the sync pulse isn't generated from the ramp, it's generated digitally. so all timing is digital.

an important thing to note about comparators is that they're just amplifiers. you just subtract the two values and multiply the result by the gain, say 1000. then you limit that to the supply rails.

the consequence of this is that the mid point of the pulse wave will not be sharp. it'll be 1000 times sharper than the ramp, but the edge will generally be a lot smoother than the ramp's reset.

it's also possible to generate a triangle wave as the starting point and convert that into pulse and ramp. a ramp oscillator is easier to build though.

[ghettosynth]

The subocillator is simply a counter or something similar (i'm not completely sure what it's called.flip flop,counter,shift register).

To add a bit to this, one can use either a counter or a flip flop to achieve the same result. A flip flop is most often used. A counter is really just a collection of flip flops with each stage dividing the previous stage by two.

[jupiter8]

The context of this thread was some claimed the SH 101 had 3 oscillators. It hasn't. The important point was it has one and you derive the other waveforms from the saw so they're all in "sync". You can't detune the square from the saw for example and the suboscillator are exactly an octave lower (or 2 IIRC).

[aciddose]

A counter is really just a collection of flip flops with each stage dividing the previous stage by two

not always. you can also integrate the input pulses and reset at some point - it's the same circuit as the ramp oscillator core only rather than being fed by a constant frequency value it's fed by a highpass filtered (or differentiated) and scaled pulse.

you can then get fractionally divide the input. this is called an analog counter.

because you can use fractional divisions it means given a high enough frequency input you can divide to get all twelve notes of the scale, for example. that's how the analog dividers in organs work.

[ghettosynth]

A counter is really just a collection of flip flops with each stage dividing the previous stage by two

not always. you can also integrate the input pulses and reset at some point - it's the same circuit as the ramp oscillator core only rather than being fed by a constant frequency value it's fed by a highpass filtered (or differentiated) and scaled pulse.

you can then get fractionally divide the input. this is called an analog counter.

because you can use fractional divisions it means given a high enough frequency input you can divide to get all twelve notes of the scale, for example. that's how the analog dividers in organs work.

I know that there are alternatives, I was keeping it simple and clarifying the confusion regarding flip-flops and counters. The overwhelming majority of sub-bass circuits are done with flip-flops and the majority of divide by 2^n counters are also just chains of flip-flops (plus glue).

[ghettosynth]

The context of this thread was some claimed the SH 101 had 3 oscillators. It hasn't. The important point was it has one and you derive the other waveforms from the saw so they're all in "sync". You can't detune the square from the saw for example and the suboscillator are exactly an octave lower (or 2 IIRC).

That's right, it's one or two. While this may seem like it's a great intentional feature, it probably has more to do with the fact that flip-flops are often packaged in pairs on a chip. So, as long as you are dividing by 2 to get one octave down, why not get really low using the other flip flop to divide by 4.

So you get more flexibility for really only the cost of switch position.

Internally, the SH101 uses a CEM3340 which has the VCO and waveform processing on a single chip. Although unused, there is an available triangle waveform.

[jupiter8]

That's right, it's one or two. While this may seem like it's a great intentional feature, it probably has more to do with the fact that flip-flops are often packaged in pairs on a chip. So, as long as you are dividing by 2 to get one octave down, why not get really low using the other flip flop to divide by 4.

So you get more flexibility for really only the cost of switch position.

Sounds plausible,i need to check the schematics.

Internally, the SH101 uses a CEM3340 which has the VCO and waveform processing on a single chip. Although unused, there is an available triangle waveform.

Going a bit OT here but i find this interesting. The 3340 is a triangle core oscillator (or so i've been told).I always assumed it was sawtooth core for some reason. So it is kinda funny there's no triangle on it since it is what the oscillator produces (if my sources are correct).

From what your saying it sounds like the sawtooth is being manipulated using various circuits to result in the creation of other waves - so when i swithc from saw to square another circuit if kicking in on my synth that causes waveshaping?

The schematics i've seen (or at least remember seeing) the waveshapers is always "on" and you choose which wave you want. Slight and probably insignificant difference. What the SH101 does different from many other synths is that you can mix the output of the different waveshapers,most only offer one or the other or several at a fixed gain. I assume that is why people assume it has several oscillators.

How is the original saw created though?

Turns out in this particular case it isn't a saw wave but a triangle. You create a sawtooth by charging a capacitor and when it reaches a certain threshold (you compare it to a fixed voltage) you shortcircuit it so it instantly discharges. Comparators are cheap.

A triangle core oscillator i believe kinda reverses the process (IE you slowly discharge it in the same way you charged it but in reverse when you reach the threshold).

A gross common simplification (but still has some merit) in synthesizers are East Coast (Moog,ARP,SCI) vs West Coast (Buchla,Serge) and East Coast synths have sawtooth core oscillators and West Coast has triangle (there are other differences as well. East coast is all about the filters and West coast is all about the oscillators). That's why i always assumed the 3340 was sawcore (Japanese synths are West Coast).

[whyterabbyt]

Going a bit OT here but i find this interesting. The 3340 is a triangle core oscillator (or so i've been told).I always assumed it was sawtooth core for some reason. So it is kinda funny there's no triangle on it since it is what the oscillator produces (if my sources are correct).

According to the data sheet, it produces triangle, saw, square and pulse?

http://www.synthtech.com/cem/c3340pdf.pdf

which concurs with what I remember of my Digisound 80, and the Doepfer A-111.

[jupiter8]

Going a bit OT here but i find this interesting. The 3340 is a triangle core oscillator (or so i've been told).I always assumed it was sawtooth core for some reason. So it is kinda funny there's no triangle on it since it is what the oscillator produces (if my sources are correct).

According to the data sheet, it produces triangle, saw, square and pulse?

http://www.synthtech.com/cem/c3340pdf.pdf

which concurs with what I remember of my Digisound 80, and the Doepfer A-111.

The core doesn't have so much to do with what waveforms the oscillator produces. You can derive all the others from the core waveform (i believe). Sync works different on a triangle core than a saw core. I believe it's easier to derive a sine from a triangle than it is a sawtooth.

Aciddose knows this stuff much better than i do but i still think i have an edge in explaining it so "regular" people gets it,hence why i still try even though i may not get everything 100% right all the time.

[whyterabbyt]

The core doesn't have so much to do with what waveforms the oscillator produces.

Im slightly confused now. I know you can derive the other waveforms from the core oscillator, that wasnt what I was responding to.
It read to me that you are saying that the CEM3340 doesnt output a triangle waveform, despite having a triangle core, so I was pointing out that the CEM3340 datasheet says it does output a triangle waveform, and both of the CEM3340 based modular oscillators I have owned have had triangle outputs. Have I misunderstood what you were saying?

[jupiter8]

The core doesn't have so much to do with what waveforms the oscillator produces.

Im slightly confused now. I know you can derive the other waveforms from the core oscillator, that wasnt what I was responding to.
It read to me that you are saying that the CEM3340 doesnt output a triangle waveform, despite having a triangle core, so I was pointing out that the CEM3340 datasheet says it does output a triangle waveform, and both of the CEM3340 based modular oscillators I have owned have had triangle outputs. Have I misunderstood what you were saying?

My point was the SH101 doesn't have triangle wave which is kinda funny since it uses the 3340 which has and not only that,it is a triangle oscillator at it's core.

I in my turn thought you meant you can tell the 3340 is triangle core by looking at it's available waveforms,which isn't possible as far as i know.

Some wires got criss crossed there for sure.

[whyterabbyt]

Some wires got criss crossed there for sure.

erm, yeah. sorry bout that.

[aciddose]

triangle sync is only different in the case of soft-sync. for the triangle generally the sync will occur as if it were applied to a 'window' on the ramp.

normally for soft-sync you check if the ramp is below a certain level, if so the sync is allowed to happen. otherwise the ramp continues without syncing.

for the triangle, it's as if you check "if the ramp is above 30% AND below 50%", which produces slightly different results with respect to phase as well as creating variations on exactly which harmonics the sync happens more often. if i remember from my tests the windowed soft-sync will tend to lock on to harmonics of the master a lot more than the "less than" soft-sync would.