4046 PLL model scarcity.

Who needs a PLL? I'll now go silent ;-) ...Jim Thompson

d4-

in

rial

log

to

hing

That's OK, I'll try different values and measure on my own. I was thinking I'd like to keep the capacitance as big as possible, since then stray stuff is less important.

Keeping the impedance low also helps keep electrostatic 'stuff' from leaking in. Which is worrisome at low frequencies.

But if the chip likes a high impedance.. then I'll have to deal with it.

George H.

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Ok, :) Well, I'll say a bit more... I need one. First, I wanted the waveforms, same as those guys on that web page did, for an analog oscillator. Second, I wanted it for the clean square after my Schmitt trigger, for a freq-to-volt converter before I learned than the response that way was too slow. Then I realised I had a voltage already from the PLL, if the VCO was linear enough, and I could do divisions for frequency shifts, all sorts of fun stuff, just by staying with it, so I stayed. Given how cheap they are, it's a damn good idea to do that anyway. It's either that or learn to do it all digitally. Dieter Doepfer is far better at that than I'll ever be, and he hasn't solved it that way either.

I found the boom-box PDF via a bit of Googling, but I need something I can't do with op-amps and a flip-flop. They might get me a nice square wave in a deeper octave, but suppose I want a higher one? They don't get me a control voltage without frequency to voltage conversion, and that';s even slower in bass response, so I'd have to shift upwards, putting the flipflop in a PLL loop. And the PLL also gives me the CV ( I swear I said that before. >:) and it also gives me some nice raw analog waves like saw and triangle with a bit of coercion. In short, it's way too good an idea to ignore. There are whole synthesizers that cost hundreds that aren't this much fun.

Phil Hobbs wrote in news: snipped-for-privacy@electrooptical.net:

Indeed. :) Is one reason I like it. I was fairly chuffed when I realised my Schmitt trigger cleanup method let me use it instead of PC2 which most people (even those in the synthesiser business, earning money) would use. I started doing it that way too, and the results were as dodgy as the Korg MS10 and most other things I'd heard. With Schmitt and PC1 I could whistle faintly across a room, and the thing would be tracking pitch cleanly at amplitudes so low I could barely discern it in my own whistled tone. This is great because it means I can pick out a clean pitch long before I decide what any subsequent process will do with it prior to output. When Dieter Doepfer told me about his digital method, he surprised me by telling me that he didn't have it start trying to track pitch till a threshold on amplitude was reached. I still can't figure that out, I got best results once I had something that always tracks whenever there's anything to track. That way I don't have to wait for it to respond, it's likely locked on before there's enough amplitude to justify using as a gate threshold. I later learned that RA Penfold had also used a Schmitt trigger to clean up, and also to prevent lock to harmonic, but as I can use PC2, I get that anyway, because according to teh datasheet (and my tests) it won't stray. I'll soon test it with a bass guitar though, if it tracks that it out to track most things.

Anyway, caffiene or no caffiene, I've taken to rambling, so goodnight.

George Herold wrote in news:b908dd61-6c13-45d8-9023- snipped-for-privacy@m19g2000yqh.googlegroups.com:

Could be a trade-off between one source of nonlinearity and another. I know Winfield Hill likes to 'look the other way' but I suspect that trial and error might often be the only way to get best results from a 4046 in many specific contexts.

Stray capacitance might be ok anyway, if it won't change.

Actually that is something else I notice the latest version of Daqarta will do "Pitch to midi" I have no idea how well or if it works or not. Haven't bothered to download the newest version.

There was a realtime fundamental frequency display in earlier versions.

Martin Brown wrote in news:TQsqq.3100$ snipped-for-privacy@newsfe08.iad:

The program doesn't work well for me, chews all CPU I have, and won't let me select the channels I want for the sound output without using the main system setting for that. I like that they still support all of Win32 though, and a sound card was always a good idea for this, I made a small adapter to DC couple input and output in a Layla 20 or 24 bit interface which makes awesome lab (and laser scan control) gear considering eBay prices for those now. Results easily worth ten times the price. I might get a limited amount of realtime analysis from Goldwave, perhaps...

I don't know if my machine will cope with exotic realtime processes though, Daquarta chewed 97% of it just generating a sine wave! I'd have to record to Sound Forge and anaylse it afterwards. My next task will be simpler though, a MIDI to CV converter and a MIDI keyboard to play a few oscillators, 4046 VCO, LM311, etc... I think my ears will tell me soon enough what course to take next.

Take a look at the NXP 74HCT9046 data sheet.

The part is claimed not to have the dead zone problem, which is discussed in some detail on page 8 of the datasheet.

There's also a fair bit of detail on the relationship between the VCO frequency and the controlling voltage furthter down the data sheet.

-- Bill Sloman, Nijmegen

Lostgallifreyan wrote in news:Xns9F8CA5BDE8E8Ezoodlewurdle@216.196.109.145:

Right now, I'm having a hard time knowing what the hell the VCO frequency IS proportional to! With R1 at 47K, C1 at 10nF, I assumed that it was proportional to voltage like damn near any VCO that wasn't designed to follow the 1 volt per octave standard used for most musical synthesisers.

When I measure the voltage out of the MIDI to CV converter set to Hx per volt, the output voltage scales up correctly, doubling per octave, so clearly linear in proportion to frequency. The 4046 VCO isn't having any of it! Tones a keyboard semitone apart at the top end are far less separated in frequency than they are at the bottom end, and are actually MORE nonlinear than they are if I try to feed a voltage of 1V per octave, i.e. proportional to pitch. That obviously shouldn't work for a VCO that is supposed to be linear with frequency, but hearing it become even worse when fed by the correct voltage scaling makes no sense to me at all. The only thing that works as expected is that the actual frequency is always the same for the same voltage.

To prove that the voltage really is ok, I fed it to an LM311 configged as a voltage to frequency converter. The result was horrible as a musical waveform, horrible in speed of response, but it DID track perfectly, I could play fast scales on it that sounded like a rat-arsed little trumpet. :)

So as the 4046 VCO makes a clean square, and responds like a flea on crystal meth, I REALLY need to know why it can't track a voltage that IS correctly proportional to frequency. Again, this is no mere vague nonlinearity, it's grotesque, more nonlinear even than the difference in proportion between frequency and voltage.

Lostgallifreyan wrote in news:Xns9F8CEC6F73B0Bzoodlewurdle@216.196.109.145:

Meant: frequency and pitch. I also tried another 4046, and took out all extra gubbins on the pin deck that might be intering with it. At this point, hearing no change in this absurd result, I think I'll quit flogging a dead horse and hope that Dieter Doepfer finishes and markets his A195 pitch to MIDI device to solve this problem digitally. A PLL is no use to me if neither the output waveform OR the voltage can be meaningfully converted to a musically useful pitch. There may be ways round it (LM311 followed by a D type flip-flop, but this is pointless, the aim was to keep the system simple and that's clearly impossible now.

Only by rolling your own ;-)

(Though I suspect mis-measurement on your part... the 4046 should be essentially linear, frequency versus control voltage) ...Jim Thompson

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Not possible. An industry has not managed this over decades, bar a few amateur designs from whatever was cobbled together. Dieter Doepfer has struggled with this problem in analog and digital domains, and he's a master at this, one of the best electronic musioc hardware builders in history. His best efforts are still on the back burner because he's not satisfied they work well.

I tried two. I removed all but the VCO control voltage, and power supply, and output to a mixer, one capacitor and one resistor. No matter what combinations of resistor, capacitor, or 4046 (CMOS type, supposed to be the MOST linear), I got the same absurd results!

The closest possible method with any chance of success, is a VERY linear VCO (I already proved that the LM331 is at least that, if totally unsuitable on its own otherwise) in the loop, but then I might as well use an XOR gate (or apparently a 4 quadrant multiplier, as I read that's what's actually in a

Clearly this goes strihgt into the complex building of a PLL from discrete parts, something mentioned in AoE as difficult and unreliable in general.

This isn't a Spice problem, it isn't even a PLL problem anymore (the only remaining validity of the PLL is the lock indication as a way to determine that the input signal is a viable pitch, thus enabling output. Even that might be done better some other way.

All I'm left with is a Schmitt trigger cleanup as initial input process, which I discovered independently of RA Penfold who also uses that. The idea has been bombed back to the stone age and I think I'll leave it there, as far better engineers than I'll ever be have also turned aside. I had a go. I'll leave it at that.

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

As a last look I tried Googling for any hint of a different answer, and what I found surprised me, a LOT!

Check that file if you're still interested. If it's good enough for Bob Pease it's damn well good enough for me. What is even better, he uses the same idea I came up with, and as I'm no Bob Pease. That in itself means I must have been on the right track. 4046, but NOT its own VCO. (Presumably it's optimised for loop stability, and voltage linearity with frequency is unimportant otherwise. And the graphs I saw for HEF4046B pretty much prove this, though they didn't prepare me for the extremeity of error in practise). Anyway, Bob Pease used the LM331 and a divider in a better way than I proposed, but the basic idea is the same.

Even so if I ever do return to this, I'll hold out for a PLL whose VCO is adequately linear in the first place. If PLL's really are as useful as I (and apparently Bob Pease) thought they were for frequency to voltage conversion, then I hope some firm out there is making a single chip for the task. I'm ok both before and after that point.

Hacker is as hacker does.

A _much_ better arrangement for the PFD...

If you can't pencil the equations for yourself, you're a hacker ;-)

...Jim Thompson

The linearity of the oscillator with respect to the voltage-on-input-terminal is not as good as the linearity with respect to the current-through-monitor- FET, so you will benefit from using an op amp. Connect op amp output through resistor to "VCO in", (+) input comes from your control voltage source, (-) input comes from resistor from 'Demodulator out' to Vss (negative power supply). Usually, the resistor on the demodulator out would be the same value as the pullup on pin 11 ("R1" on my datasheet).