The Secret Life of LFO's

[george]

or everything you wanted to know about LFO's but were afraid to ask ...

anyway now I have your attention ... can somebody point me to some resources or explain how LFO's work in, say, the Zombie or other choruses?

Thanks
George

[mattv]

There is a ton of info on LFO's at www.geofex.com

[acromarty]

An LFO is just a low frequency oscillator. Looking at the zombie chorus circuit, the two opamps and the RC feedback components comprose a phase shift oscillator, where the oscillation frequency is determined by the feedback components and the rate pot.

As far as I understand the operation of the circuit, the output from the LFO feeds into the VCO input of the 4046 phase locked loop, which modulates the clock frequency of the MN3007 analogue delay chip. So the delay time varies smoothly (or not) as the LFO output changes, and when this delayed signal is mixed with the original input signal it produces a comb filtering effect. The variation in the delay time changes the notch frequencies in the comb filter, and this gives the characteristic swoosh/sweep of the effect. The depth pot on the LFO output varies the size of the signal to the 4046 and determines by how much the delay clock is changed and how wide is the sweep of the comb filter frequencies.

Does this make any sense?
All you experts please enlighten us if I have got it wrong.

Cheers

[Mark Hammer]

EVERY lfo is intended to work akin to keeping your hand on some knob and tweaking them in a cyclical manner.  Different sorts of sounds and feels can be created depending on how that tweaking occurs.  In more technospeak, this is partly a product of the LFO waveform.  So,  for instance, many tremoloes these days have a triangle/square switch which selects between two waveforms the lfo is capable of producing (and just about any and every 2 op-amp LFO will do this).  A triangle waveform creates the sense of the parameter being steadily increased then decreased, while a square wave makes it seem like it is turned on then off repetitively.

Othe waveforms can be created though, some that don't immediately strike you as a waveform or else are produced in ways you don't think of.  For example, you can make an LFO waveform with a counter/sequencer.  This is used in the Z-Vex Seek-wah, Ooh-Wah, and there is a circuit for something like this at geofex.  Here, a counter that is capable of providing some maximum voltage steps through a sequence and provides that voltage at each of some number of separate outputs.  Stick a trimpot at each output, and mix the trimpot outputs together and you can produce a cyclical change that varies according to the trimpot settings.  So, set 4 trimpots in an 8-event counter/sequencer to max, set the next 4 to min, and as the sequencer keeps repeating its cycle you produce a square wave.  Adjust all those trimpots so that #1 is maxed and each successive trimpot is set a bit lower, and the sequencer produces a descending "ramp" or sawtooth waveform.  As you can imagine, all sorts of other possibilities are available (just one of the things that makes pedals like the Seekwah and Oohwah so nifty).

Changing the shape of waveforms is not only achievable by "building" waveforms as in the manner described above.  Any of the techniques that synthesists have used to shape waveforms for oscillators to change the timbre can also be used to shape low frequency waves as well.  Think of a sweep up/down as an "event" whose onset and offset can be used productively and altered and you'll get the idea.  

Tim Escobedo has a "simple square-wave shaper" at his circuitsnippets site which he borrowed from an old Polyphony article.  That article, in turn, simply piked it from a standard synth circuit floating around.  An attack/decay envelope generator in a synth simply provides a voltage that ascends at some rate, and then descends at some rate.  Essentially, that envelope starts out as a square-type wave, but one control limits how fast it rises, whie another limits how fast it decays.  Tinker with those and the output can be all sorts of waveforms.  Tim, and his source material, adapted that principle to take an *audio* square-wave and re-shape it so that it can be tweaked to rise or fall (or both) not quite as sharply as it started out.  In between a simple event generator and an audio waveshaper, we have LFO circuits that also incorporate this variable rise/fall time control to produce other waveforms besides a triangle or square.

Some sorts of waveshapers play not so much with the ascending and descending "straightaways" but wiuth the "turns".  One LFO waveform which is felt to be desirable for sweeping some types of effects at some speeds is referred to as a "hypertriangular" waveform, which is a typical triangle at one end and more like a sine-wave at the other.

LFOs can be combined to produce more lively and richer modulations.  An article I read in the Journal of the Audio Engineering Society about 20 years ago or more addressed the elusive "fat" sound from synths.  In the early days of synthesis, before digital sampling of existing "fat" acoustic sounds became popular, people would complain that keyboard synths sounded wimpy and uninspiring.  The authors of this paper tried to find out how much modulation you needed to make a sound from an standard poscillator sound animated, lively, rich, and attention grabbing.  Peak interesting qualities seemed to emerge around 3-4 sources of modulation and after about 5-6 you were way past the point of diminishing returns.  So, to illustrate, perhaps one LFO would vary the pitch a bit, another would vary the pulse-width or some other aspect of the waveform, while another would vary the amplitude.  They weren't all at the same speed and were certainly not all modulating by the same amount, but this would have produced a "fatter" sound.  The CV inputs that some effects pedals have (and this can include inputs for "expression pedals") permit the user to combine multiple LFOs and feed their mixed value into that input to provide more animation.

One term you will hear occasionally is "periodicity".  When an effect is swept in a "periodic" manner, that means you know exactly when it is going to sweep up and down.  Very predictable, very repetitive, very boring, and sometimes very intrusive.  Indeed, it can be the very obvious repetitive nature of a modulated effect that *distracts* away from the notes and focusses too much attention on the up and down stuff.  Here, there are several solutions.  One solution commonly used is to avoid using an LFO and use a noise source as a way of generating random voltages and random modulation as a result.  That has its moments and its pros and cons.  Another way is to use multiple LFOs that are not synchronized and may even output different waveforms.  Mix 'em together and you can get modulation that, if not strictly random, is noticeably less periodic, and so less intrusive or attention-rivetting.  When you have a chance, stick a pair of phasers or pair of chorusses or flangers in series and set their sweep off from each other a bit.  You still hear the effect but you don't get quite as seasick.

Many of the synth DIY sites have tons of schems for LFOs.  I have a couple of older articles on these things at my site too.  Check out the hypertriangular clock, and the SEWAR project (http://hammer.ampage.org or go directly to hammer.ampage.org/files) for some ideas.

[george]

Thanks all for your replies, I guess my immediate concern is to understand how the LFO on the Zombie works.

This is because
- It works but the LFO sounds uneven ie the ramp-down is much shorter than the ramp-up
- I've tried what I understand is Mark's detick mod (U2 pins 2 and 5 connected up to a different Vref) but the LFO stops for a considerable part of it's cycle.

Please note I didn't build on the RG PCB, I used IC experimenter's board.  I'm fairly confident the circuit is right (otherwise I don't think it would work at all)

I've triple checked components coming in and out of U2a and U2b (or rather their equivalent on the quad TL064 I'm using) and they are all correct.

Any help with this would be much appreciated!  I can measure voltages, unfortunately I don't have a CRO and I think my DMM is a bit slow in picking up varying voltages even with the rate turned right down ...

Regards
George

[Rob Strand]

It works but the LFO sounds uneven ie the ramp-down is much shorter than the ramp-up

The LFO and and VCO (4046) are interellated, you should make sure you are attacking the right problem.

The symmetry of the LFO is determined by the BIAS voltage on the - input of the first opamp.  That opamp is a schmitt trigger and it's output saturates agains the +ve and -ve rails.  To get equal on and off times (up and down times at the ramp) you would make the bias point dead centre between the two voltages at the output of this opamp (you can measure this at low speeds).

Keep in mind this this symmetry in *time*, ie. an equal amount of time rising and an equal amount of time falling - such an LFO is a good place to start. By fixing the up/down times the VCO operating frequencies are stuck. You can either play with the resistors on pin 11 and 12 to put the delay times where you want them, or add a manual control like say on the Boss BF2 - this effectively shifts the DC level at the VCO input.  Changing the opamp bias changes the delay setting *and* and up/down times, you should get the up/down time how you like it first *then* play with the delay time.

How the LFO sweep *during* and ramp-up is a different matter.   You can have equal up and down times but the *amount* of time spent say sweeping through the longer delay region can still be short or long (usually it's short).  To change this you have to change the way the LFO drives the VCO, that involves a number of design changes.  I can only suggest you read through the hypersine stuff on Marks site.

[george]

The LFO and and VCO (4046) are interellated, you should make sure you are attacking the right problem.

hmm I've checked the 4046 too, all the connections and components look fine ... the only change I've made is add another cap that can be switched in in series between pin 6 and 7 to get the Hammer Badge Sound. R11 and R12 are the correct values.

The wierd (and annoying) thing is that Mark's detick mod doesn't work.

[Rob Strand]

George, can you describe your problem exactly because I'm not quite sure if the actual LFO up and down ramps are unequal, or the time spent in the long delay region is too short for your liking.  I don't mean to be condescending, I just want to make sure you know the difference between the two problems - both sound bad on the chorus.

Anyway,  supposing the up and down ramps are off what you can do is temporarily replace the 10n cap on the LFO with a larger one like 100nF or 1uF.  Next connect a 10k resistor (large as possible so you can just see the LED) in series with an LED.  Connect this between the the output of the first  LFO opamp and ground and time the on and off times with stop watch.  Just to be sure, do the same with the LED between the same LFO opamp and +9V (don't forget to reverse the LED connection).  In the first case LED ON = ramp down, and in the second case LED ON = ramp up.  This is just way of quantifying the ramp up and down times.

In one of your previous posts you said,

(or rather their equivalent on the quad TL064 I'm using)

What opamp did you use?  If you have used a bipolar-input opamp instead of a JFET input the input bias current of the opamp will cause the ramp up/down  times to get out of whack.  On the second LFO opamp, you should try dropping the 4.7M  to 470k and increasing the 10n cap to 100n - this will eliminate (hide) any bias current issues.

[george]

Thanks Rob, no your not being condescending - you've made me think about what *exactly* is going on.

I'll think about it, do your LED tests and get back to you.

RE the op amp, it's a TL064 ie so bipolar input current isn't the issue ...

Thanks
George

[george]

I'm ba-ack ...

To Rob:

The problem I'm having is the brief out-of-tune sounding pitch increase as the LFO sweeps up to the shorter delay range (? I think that's right - I only know if you decrease the delay on a delay pedal while you're playing a note the pitch goes up ...?)

I tried replacing the 10nF cap in the LFO and it reduces that effect considerably - it sounds much more like my CE-2 (I checked the CE_2 circuit, *very* similar LFO)

The only downside is that the maximum rate is much slower - I'm going to check the Keeley website for his CE-2 mod that speeds up the maximum rate.

I might also reduce the LFO cap a bit to see if the rate speeds up enough ..

Thanks
George

[Rob Strand]

Hi George,

That pitch blurr effect is caused by *both* the maximum delay and how quickly you sweep.  By making the sweep slower you are just hiding the blurr effect.   Speeding-up the sweep up again will simply cause the "problem" to return.  (Changing the cap is how to change the sweep speed).

What you want to do is to decrease the range of the sweep. To a large extent you should be able to do that with the depth control.  You could also try slightly decreasing the two 1nF caps on the 4046.

[george]

Thanks Rob, I'll try that tonight

BTW I notice my zombie ticks *more* when it is bypassed; I have tried using shielded cable on the input and output jacks to no effect.  Any ideas how I could fix this?

Regards
George

[Rob Strand]

George, I pushed some buttons on my calaculator and I think the Zombie VCO values (on the CD4046) are *way* off (a lot more pitch blur) for the CE-2 sound. It could even be a bug, in fact I can't see how you would get a CE-1 sound (as is claimed) even with the CLONE switch open.  From what I can see if you drop the 68k on pin 11 to somewhere around 18k to 22k, and set the CLONE switch to open, it should be closer to a CE-2 sound - I don't have a unit to test it on but it looks that way.  It would be good if you could try this because it might explain why people say this pedal is a little noisy.

BTW I notice my zombie ticks *more* when it is bypassed; I have tried using shielded cable on the input and output jacks to no effect. Any ideas how I could fix this?

That's pretty weird, especially if you are using true bypass.  It's possible that your wiring passes the ground signal though the PCB somehow in bypass mode.  Some for of  fault would also cause this.

[george]

thanks for that Rob, the 68K resistor change would be easy.

BTW the I think CE-1 sound is supposed to be with the switch CLOSED, to short circuit the second 1uF cap to increase the delay time?  With the switch open the 2 1uF caps are in series giving 0.5uF total capacitance, reducing the delay time and supposedly giving more of a Small Clone sound.  That's my understanding of it anyway.

As for the noise, there seems to be 2 issues, firstly the ticking: this seems to be reduced by putting the LFO on Vcc/2; secondly it is fairly "swooshy" when your not playing through it but this doesn't particularly worry me (yet, I've only played it at home so far).

As for the ticking when bypassed, could it be the switch?  I'm using the one that you get from Jaycar, some people on this forum say they're a bit dodgy ...

Another factor could be the rate and depth pot leads passing pretty close to the input and output jacks. Maybe I need to shield them with aluminium foil or something.

I'm also using a crappy plastic box, but I don't think it would make any difference if I used a metal one because all the electrical noise is being generated inside the box?


Thanks again
George from Sydney

[Rob Strand]

BTW the I think CE-1 sound is supposed to be with the switch CLOSED

I believe the CE-1 mode is when the switch is open.  This doubles the clock frequency (as you described) and consequently halves the delay time, the aim being that the short delay time emulates the lower number of delay stages on the CE-1.   From what I can see the VCO values don't give to small clone, CE-2 or, CE-1 regardless of the switches.  The  68k mod I described will put you in the Small-clone/CE-2 range.

this seems to be reduced by putting the LFO on Vcc/2

Are you using true bypass?, ie. is your switch disconnecting the guitar input from the effect input on the PCB?

If so, what I suggest is an experiment.  In bypassed mode short the audio input at the effect input PCB and see if this removes the problem.

could it be the switch?

The switch could exagerate the problem but I don't think it is the root cause.

What could be happening is the combination of that switch and the 5Mohm input impedance is cause a kind if oscillation.

If the shorting experiment above removes the problem I suggest changing the input cap from 1n to 10n and reducing the input resistor *on the opamp* to 1MEG.

Another factor could be the rate and depth pot leads passing pretty close to the input and output jacks.

That type of thing is really asking for these type of problems.  That and a plastic box can make things worse.  So yeah, it would be a could idea to shield these, the shield has to be grounded though.  Another option is to tightly twist the wires to each pot.  Given you are using a plastic box it's probably a good idea to solder a wire on the metal part of the pot and connect that to ground.

When you connect the LFO to it's own VCC/2 (which *is* a good idea) you should also try putting an electrolytic cap across the new vcc/2 - like the other one has.

[Rob Strand]

I believe the CE-1 mode is when the switch is open.

George, I'm *sure* this is what was intended but at this point the unit is fairly close to the Small Clone when the switch is *open* (!), when it is closed it doesn't do a CE-1 (it doesn't emulat any known chorus).  Like I mentioned before I think there is a bug with the delay timing on this one.

[george]

Hi Rob

how's grand final fever? (you live in Melbourne yeah?)

the only reason I say switch closed is CE-1 mode is because I have a CE-1 and have built a Tonepad Small Clone (I've also got a CE-2, fanatical or what?).  Switch closed sounds more like a CE-1 or CE-2 and switch open sounds more like the Small Clone.  At least I think the switch is closed, I'll check ...

Anyway I've tried your 68K mod and it sounds good, there seems to be a more even sweep up and down and the pitch blur is less pronounced.   I'll be playing with people tonight, I'll have a better idea of how it sounds (and ticks) then ..

l8r