Synth people: a ? about how to make a steady glide in sequencers

[Processaurus]

Hi, here's something I  wanted to pick your collective brains about, to see if someone could steer me in the right direction (circuits, patents, products, etc.).  Its something I've been thinking about casually since starting to play with analog sequencers, and that is how to get a sequence that has a linear glide to each step's voltage (set by a pot) from the last, like the second waveform in the picture.  A simple LP filter is used by most sequencer designs I've seen, but thats kinda low tech, and has drawbacks.  The output waveform is curved, the filter doesn't follow the stepping speed, and messes with the overall amplitude of the waveform.  I've done some reading on op amp integrators, but doesn't seem like the ticket either, because that gives a constant slope no matter how large the transition between two steps would be.


I thought about using a digital filtering chip, and clocking it with a higher speed clock, and dividing that down many times to clock the step sequencer.  That may end up looking like a simple LP filter though, just with the benefit of automatically following the step speed.

A different half idea of how to go about it is to have two multiplexer chips switching the pots wiper voltages, where one would be a single step ahead of the other, so then you could have a voltage that represented the difference in levels between the two steps, and use that to create the right slope somehow.

Theres definite potential for dead horsyness- Might be one for microprocessors, but I thought I'd ask, it would be great for effects that need a moving wave to shine, like phase shifters, traditional tremolo sounds, and flangers, where you don't want it sounding "steppy", but are terminally bored with triangular LFOs...

[Unbeliever]

Theres definite potential for dead horsyness- Might be one for microprocessors, but I thought I'd ask, it would be great for effects that need a moving wave to shine, like phase shifters, traditional tremolo sounds, and flangers, where you don't want it sounding "steppy", but are terminally bored with triangular LFOs...

The LP filter approach worked well for me in the infinitphase, exactly the application you're talking about (phaser with non-trianglular waveforms) - it sounds really good (to my ears). At high speeds (i.e. with the sequencer running at a high rate) I don't think you *want* the final output wave to vary completely from one voltage extreme to the other as it would sound too choppy - slow and deep / fast and shallow for phasers sounds good, at least IMHO. Also, I wanted to keep parts count down, but if you're doing a one-off you don't need worry about that, and the sky is the limit.

[Ashurbanipal]

I think this would work. Slew limiter aka lag time processor:

http://www.doepfer.de/home_e.htm

It's been a while since I had a synth with one, so not 100% sure. Analog Haven carries Doepfer and probably knows if that's the module your after:

http://www.analoguehaven.com/

Haven't ever tried to build one.

[Ashurbanipal]

This is the link to the module:
http://www.doepfer.de/a171.htm

[Unbeliever]

have a look at the 'Portamentiometer' here:
http://www.ori.org/~aaronc/synth/

Runs off +/-15VDC tho (all the best circuits do  )

[gez]

Take a look at the stutter trem thread.  Read the comments about trapezoid shaping.  Many ways to do this.

[puretube]

there once was a schemo in "Elektor", which did the hypothetical line in Proc`s middle graph.
(In german, it was called "Punkteverbinder" (points-connector).
Maybe I can find it one day...

[R.G.]

Adding on to what gez sez -

I think the slew rate limiter from EDN would give you your "integrated result". It slews to any intermediate value, then stops

[davebungo]

Provided you can somehow produce a voltage which represents the slope rate you need in a given period then I don't see any reason in principle why you can't control the slope of an op-amp integrator using a voltage controlled resistance e.g. using a FET etc.

[gez]

Provided you can somehow produce a voltage which represents the slope rate you need in a given period then I don't see any reason in principle why you can't control the slope of an op-amp integrator using a voltage controlled resistance e.g. using a FET etc.

See RG's comments about stopping (need to incorporate the integrator into a follower arrangement otherwise the slope keeps on going during the 'stops')

[davebungo]

Provided you can somehow produce a voltage which represents the slope rate you need in a given period then I don't see any reason in principle why you can't control the slope of an op-amp integrator using a voltage controlled resistance e.g. using a FET etc.

See RG's comments about stopping (need to incorporate the integrator into a follower arrangement otherwise the slope keeps on going during the 'stops')

I read the whole thread thankyou very much.  My comment still stands.  I'm not trying to design the damn thing just contribute some possibly helpful info.

[gez]

I read the whole thread thankyou very much.  My comment still stands.  I'm not trying to design the damn thing just contribute some possibly helpful info.

Apologies if I came across as rude, wasn't my intention.  Just trying to point out why what you suggest - unless I've misunderstood you - isn't a good option.  If you look (I think - can't see it while I type) at the third digram in the initial post the steps ramp up/down, then level out, then ramp up/down etc.  With an integrator it keeps on ramping up/down until it hits the rails, until 'direction' (and what if most of the steps occur above the trigger threshold?) is changed, you need someway of following the input signal to prevent this.

The circuit RG linked to in the stutter thread will do this, as will the OTA idea I outlined.

[davebungo]

Gez,
It's me who should apologise for my slightly flippant response. I took it the wrong way.
Dave.

[gez]

Gez,
It's me who should apologise for my slightly flippant response. I took it the wrong way.
Dave.

No need to apologise Dave, I have a tendency to just get straight to the point - I call it 'Spock Mode'.  Most of it is just to save time, but I appreciate that it can easily be misinterpreted as rude/arrogant.

All the best

[Paul Perry (Frostwave)]

Maybe a look-ahead linear interpolator is requied. Like this one: (see section 4.1.2)
http://homepage.ntlworld.com/henry01/waveform_synth/waveform_synth.htm

[R.G.]

A look ahead is definitely required. Unless you know what the next corner voltage to hit is, and how long you have to hit it, you'd have to violate causality to hit the right slope.

It greatly simplifies things if the steps are all the same size, as shown in the example. Making this work with random intervals as well as random voltages would be very tough.

So that means that you have to cheat just a little with your random steps. You have to have both the present step and the next step available to set your glide rate. Once you have that, you have the ability to sample the difference between this level and the next one, store it somehow, and when the step happens, use that stored difference as the control for the next integration slope.

As for how to do the integration, if you want to do it analog, you have to have a controllable integration resistor. There are not many processes that do that well, especially if you want a somewhat linear and predictable control of it. In non-switching things either an OTA as the variable resistor or an LED-dual-LDR module should work. JFETs are unpredictable, and a dual is what you really want; moreover, you need to supply the control voltage relative to the source, which makes control path tricky.

If you allow switching, a PWM'd resistor or switched capacitor integrator would do it.

So the block diagram would be:
- random step generator with both current and next step levels available
- sample and hold for difference between current and next
- controlled integration resistor for doing the integration

In digital, of course, you just do whatever you want.

I suspect that what you really want to do is exponentially shape the slopes once you get linear.

[Processaurus]

Thanks very much for the thoughts and direction, everyone.  RG, I hadn't thought of making it random, that definitely opens a whole new can of worms.  The logarithmic converter idea is interesting too, did you mean that that would be desirable for circuits where the sequencer's pots would seem to have the wrong taper, because everything would happen in a narrow range of the pots rotation? 
Paul, that link to the "lookahead interpolator" looks right on, when i have some spare time this summer I'll breadboard it and see how it works, who knows, maybe some else did the hard work already .  Gez, I'll re-read my chapter on integrators, i may have misunderstood them as functioning like slew limiters.

There where a couple goals of this possible project, one was to be able to emulate all kinds of odd LFO waveforms for modulation effects circuits, like trapezoidal, hypertriangular, the funky lopsided sinewave you get from old fashioned phase shift oscillators, etc.  Also it was to be able to use it as a analog synth accessory, as a fancy oscillator (like in Paul's link), to be driven by a higher register square wave from the synth.

[swt]

don't know if this might help, but take a look at ken stone's psycho lfo, i has a glide function...maybe you can take ideas from that one...

[Processaurus]

Here's one way this could happen I thought of this morning:  Start with a sawtooth LFO, derive a square wave from that to drive the clock for a Geofex type 40193/4051 based sequencer.  Have two 4051 sp8t analog switches, wired so the sequencer pots' voltage is being read one step behind the other.   Use the sawtooth LFO to drive some kind of voltage controlled PWM chip (I'll need to research these), this PWM signal in turn switches a 4053 spdt analog switch to switch between the two 4051 multiplexers, so it kind of linearly cross fades from the multiplexer one step behind to the one one step ahead, each step in the sequence.

Filter the jaggedies with a simple RC filter and then turn it out into the world.

You could feed the sawtooth LFO through a clipping amplifier to get faster (up to really fast, like there wasn't any) glide time, and more time hanging out on the step's actual voltage.



What's cool about this is you could have a bunch of switches (a set of dip switches even) to turn the glide on or off for each step, by using one of the other switches on the 4053 to turn the PWM to the original 4053 switch in the drawing on and off.

Think this would work out?

[Paul Perry (Frostwave)]

http://www.till.com/articles/scanner/index.html
Say no more! Nice one, Don.