Frequency Multiplier

[Uncle Krunkus]

Ah, very cool.
Even if we have the max multiplier at X8, and choose the cap based on that, can we still take taps off at lower multiples. Like can I incorporate the 5ths idea if my limit was X8, or should I do the X12 circuit?

[JovianPyx]

[JingleJoe]

Excellent ideas chaps I am thoroughly enjoying reading about this PLL multiplier thanks for sharing it!

I was also thiking about "reaction time" of the PLL I did some calculations and it seems like it needs atleast half a second to really respond to some of the frequencies I want it to maybe I got something wrong.... wait, yes I did
It only needs about 150 ms to respond to 55Hz (A two octaves below middle C I think) and I suppose 300ms to respond to the A below that.
These numbers are coming from the frequency division required in the feedback e.g. for frequency multiplication of 8: 55/8=6.875 so the time period is equal to 1/6.875=0.145, or 145ms.
This does not take into account the time it takes for the filter to respond.

[JovianPyx]

Yes, there is a delay. I don't know what it is exactly, a lot depends on the filter. I remember the discussion of this on SDIY email list forum when I started this. Lots of warnings - oh watch out for the delay. I might add that the warnings turned out to have come from people who had never used a PLL for a musical application.

From my empirical data of actually using it - it was a lot of nay saying for nothing. As I said - there is a delay. Is it enough to make it useless? Not by a long shot. And it does depend on the input frequency.

The delay manifests as a sort of short portamento, but from my perception, it was mainly contained within what I would call the attack portion of a note. It could be made longer intentionally by using a loop filter with lower Fc which turned it into a portamento effect. Unfortunately, I don't have a recording I can find and post (and the synth is down at the moment), but as I recall, it was such a short delay that it was musically insignificant.

The divide ratio inside the loop does exacerbate the delay, but again, I never found it to be a problem for making music.

In my opinion, this thing works well enough to try it. I would encourage anyone interested in using a PLL for music to breadboard it up and have fun.

[Uncle Krunkus]

I have never heard it, but I'm quite confident that a short portamento at the start would make the use I intend it for an even better waveshaper, and therefore very "musical". I don't find the point in going for absolute purity in oscillators. Purity isn't that interesting. Everything down the track from the oscillators will effect the purity of the wave, and I'm glad of it.

[Uncle Krunkus]

I was just about to put a 4046 on the breadboard, and I suddenly realised that what I want is not X8. It's X2, 8 times. In other words, 8 octaves up! Or X256.
What are the chances that is gonna get up and running?!?

Well, it's going on the breadboard this morning, so we'll soon see what we can do.

[JovianPyx]

hmm. not sure why my last reply didn't make it...

Anyway, as far as 256x frequency multiplication - the 4046 IC will do it - whether it is useful will be entirely decided by the "perpetrator"


My advice - break out the solderless breadboard and try it. All you will waste (since you own the breadboard and the ICs) is a bit of time that you can call an autodidactic education

[JovianPyx]

Here is what I tried to post before which I hope answers some other questions.... EDIT ADD: Oh - and this is most likely a local network problem and NOT wonderful electro-music.com

------------------------------------------------------------------------------------

Yes, there is a delay. I don't know what it is exactly, a lot depends on the filter. I remember the discussion of this on SDIY email list forum when I started this. Lots of warnings - oh watch out for the delay. I might add that the warnings turned out to have come from people who had never used a PLL for a musical application.

From my empirical data of actually using it - it was a lot of nay saying for nothing. As I said - there is a delay. Is it enough to make it useless? Not by a long shot. And it does depend on the input frequency.

The delay manifests as a sort of short portamento, but from my perception, it was mainly contained within what I would call the attack portion of a note. It could be made longer intentionally by using a loop filter with lower Fc which turned it into a portamento effect. Unfortunately, I don't have a recording I can find and post (and the synth is down at the moment), but as I recall, it was such a short delay that it was musically insignificant.

The divide ratio inside the loop does exacerbate the delay, but again, I never found it to be a problem for making music.

In my opinion, this thing works well enough to try it. I would encourage anyone interested in using a PLL for music to breadboard it up and have fun.

---

Bottom line is that I would use as small a multiplication factor as is needed since it does "slow" the system down.

[Uncle Krunkus]

Well,
X128 would still yield interesting results, even X64 would be worth trying.
I'll give it a go, and keep you posted on the results.

[J3RK]

Thanks for all of the information you posted Scott! I've been struggling with PLL concepts and basic use of the 4046 for this sort of thing for some time, and this just about clears it all up for me!

I'm at work, so couldn't go through this with the fine-toothed comb that I will later, but if I just want to make one (external to the 4046) VCO track another, (within reason) where (which pin?) do I take the error voltage from to control/adjust the second VCO?

I'm sure with a combination of your info and the many datasheets I have collected I'll eventually figure it out, but thought I'd ask anyway.

I think to keep things simple initially, I'd take the square/pulse output straight from the comparator after the VCO core, and put it into PC1 on the 4046. This way, it's got something consistent to work with. I'm thinking of this in the context of a dual-VCO with a simple "lock" function. (maybe add a pot into the filter to sweep the range a bit)

Nice thread!

[JovianPyx]

[J3RK]

Thanks!

So, if the primary (the VCO being tracked,) is exponentially controlled, could one then take the error voltage, and either send it through the linear FM of the secondary VCO, or inject it past the exp circuitry? Or would this still not have the desired effect? It will take a minute for this topology to sort itself out in my head.

[JovianPyx]

[J3RK]

Very helpful, thanks!

I actually do want to implement some form of frequency offset (probably in 3rds, 5ths, and octaves,) much the way you do with your circuit. To start though, I want to build something on a simple level, that I can see function, and then expand from there. I'm basically thinking of different way of combining two VCOs in a dual configuration that add a bit of variety. Right now, I'm using a tri-core (basically interpolated between a 258 and VCO-1 core.) I've built a crossfader that fades between the triangle output and pulse, and then it also has PWM. So you can fade the varying pulse with the triangle, and get some interesting animation and in-between waveforms.

Combining two of these, with a PLL, and some internal cross modulation routings, should make a dual version fairly versatile. It may not actually be the best way to achieve it, but its something that was already somewhat intriguing to me, so this is a great start. Having them differ in frequency while locked in phase would be fairly useful I think.

I agree, that an identical CV signal is the best way to make two VCOs track together. And possibly just tune, and/or have a voltage offset switch could be another simple way of getting multiples.

Lots of good information to absorb though, thanks!

[droffset]

So I've been sitting around for the past week thinking about how there must be a way to get an octave or two up on a vocalsynth idea I'm playing with.

And then I found your 4046 multiplier article.

And then I found this thread, where it's all being discussed and explained.

Pretty cool.

[The Real MC]

I know I'm late to the party, but the old Eventide H9xx pitch shifters are based around a PLL multiplier core.

[Smada]

Hi guys, sorry to drag up an old thread but I have been trawling through google for information on audio range frequency doublers and stumbled upon this thread - which seems to offer exactly what I want. However, I have one question - is the PLL frequency doubling circuit only suitable for square wave inputs or will it work with a sine wave input too? The iniital thread post seemed to be focused on square waves so I wanted to be sure before I build the circuit. Thanks for any help you can offer!

[JovianPyx]

[Smada]

Thanks for the quick reply mate, I will hit up the lab and let you know how it goes!

[Grumble]

I'm thinking about making a triangle waveform into a saw wave, this is because the parts I'm using for wave forms are only capable in making sine, triangle and squares.
First I invert the signal, and make two top detectors which make a small puls at the top of the triangle and a puls at the bottom of the triangle.
This will be fed into a set-reset flipflop, which in its turn opens /closes a cmos switch to conduct the slopes from the original signal and than from the inverted signal.
So effectively you end up with a sawtooth signal of twice the frequency of the original triangle wave.

[thermionicjunky]

For those looking to go beyond square waves with the 4046, just reference the waveshapers in Thomas Henry's X-4046 VCO. He did all the work already.

http://birthofasynth.com/Thomas_Henry/Pages/X-4046.html

[Grumble]

For me diy means that I like to "invent" my own wheel.
The fun for me is thinking about what I want and how to do it, not copying someone else's work.

[elmegil]

On the other hand, looking through and understanding someone else's work (and I'd particularly cite Thomas' for this sort of thing) helps understand the principles that then let you invent your own wheel....

[JovianPyx]

My only point was how to make a 4046 behave and properly track an external input. Certain signal conditioning is required, such as the rail to rail input requirement (to get a clean output signal). Of course - you can do whatever you want as long as it gets the results you want.

I too have played with making other waveforms from the 4046 output, it's not really that hard and had actually been solved well before TH presented his circuit (not to take anything away from TH BTW - he does great stuff). The thing to understand from what TH (and others before him) had done is that the capacitor is in a sort of floating "flying capacitor" circuit (neither side is grounded) and that the way to get a signal from the cap (without screwing up the 4046 operation) is to use very high input impedance amplifiers (FET input opamps) - which is apparent in the TH circuit.

Other things to be aware of with respect to frequency multiplication (to put the thread back on it's title) are the loop filter and not getting carried away with too much multiplication. The loop filter needs to account for the fact that a larger divider in the PLL loop will slow the response of the PLL because it takes more input cycles before an edge change inside the PLL occurs due to the divider. The circuit cannot be expected to track as quickly with a divider than it does without. One easy thing to do is to make the loop filter adjustable. It's a simple RC filter (or can be) so it's very easy to adjust. I had very good success using a divide by 12 circuit. "Good success" meaning there was no discernible portamento or delay in the PLL catching up to the input signal.

[Grumble]