Barberpole Through-Zero Flanging

[StephenGiles]

I refer you to the H&SR Dual Gate linked above for capacitor discharge and recharge circuitry!

[Keppy]

Thinking about Keppy's suggestion.  Not quite sure I understand how the capacitor starts recharging after it has been discharged.  As long as the discharge diode is switched on the cap can't be recharged.  So the discharge diode needs to stay conducting just long enough to empty the cap and no longer.  Problem is that the comparator that switches on the discharge diode leaves it switched on for 25% of the overall cycle (unless I've understood things wrong). Also the cap voltage doesn't charge linearly as pointed out by Keppy.

That's why the comparator is hooked to the CLOCK pin. That pin is triggered by a sharply rising voltage that causes the state of the DATA pin to be transferred to Q. I had been thinking the SET pin would then retake priority after the pulse and begin charging the cap. Upon checking the datasheet, I think I got that wrong, not because the SET pin wouldn't take over after the clock pulse, but because the SET pin would cause the clock pulse to be ignored in the first place and the cap would never discharge. So yeah, I don't think my idea will work. Thanks for checking me.

You seem to be thinking along the same lines as me, only with more clarity. Edge-triggered one-shots with the ramp voltages taken from the timing caps seem like a workable, non-microcontroller way of generating offset ramp voltages from a single LFO.

[DrAlx]

An advantage of generating the ramps indirectly (i.e. by having a square wave set the overall timing, rather than having the "ramp caps" set the timing) is that if you generate the initial square wave using the common triangle/square wave LFO, then the triangle wave gives you the mix control signal for the VCA.

See waveforms in pic here:

http://1drv.ms/1HEqEEf

I've assumed the one-shots as being triggered on the downward slopes of the square waves and this gives the notches that control the discharge diodes for the ramp caps.
The peaks of the triangle wave correspond to the middle of the ramps.

[DrAlx]

Something else just occurred to me (and should have been obvious from the start).
For a barberpole flanger based on 2 BBDs as described above, you only need a single VCA and single triangle wave control signal for the mixing.

Assuming the VCA gives a gain factor "G" that varies from 0 to 1, then you can mix things by taking a simple 2-input  op-amp adder circuit with equal mix resistors.

Input 1:  FlangerA
Input 2:  ( FlangerB - FlangerA ) * G

So when G is zero we get all FlangerA  at the output and when G is 1 we get all Flanger B.

A bit of thought shows that you don't need to pre-create the flanger sounds before mixing them.
Breaking each flanger sound down into its component Clean and Delayed parts...

     FlangerA = Clean + DelayA
     FlangerB = Clean + DelayB

and we can write

   (FlangerB - FlangerA) * G = (DelayB - DelayA) * G

Therefore instead of having a 2-input op-amp adder, we can use a 3-input adder (with equal mix resistors) as follows

Input 1: Clean
Input 2: DelayA
Input 3: (DelayB - DelayA) * G


So the single VCA just needs to operate on the difference between the two delayed signals.
As before, we get the FlangerA sound when G is zero and the FlangerB sound when G is 1.

[DrAlx]

I LTspiced some stuff.

I think you can get both ramp voltages as well as a control voltage for the VCA with a single quad op-amp (LM324), a dual multi-vibrator (CD4528), and a dual analogue switch.

http://1drv.ms/1DqIpTk

As I pointed out above, a single triangle wave is all that is needed for the mix control, and a triangle wave with a trough in the middle of ramp can be made to work just as well as one with a peak in the middle of each ramp.

I took a regular square/triangle LFO using an LM324.
I couldn't find a model for the CD4528 one-shot multivibrator so substituted something else that gives a short duration pulse on every rising edge of the square wave from the LFO.  (The other half of a CD4528 would be wired to give short pulses on each falling edge of the square wave).
A linear ramp voltage can be produced using a simple integrator, and it should be clear that when we use two of these to give two ramps we can match the ramp gradients by trimming the R values.

I haven't used an analogue switch before, but I've assumed that the one-shot can be used to close an analogue switch to reset the cap.

EDIT: In case anyone is wondering why I have modelled the op-amps with a negative supply set to Vee = 0.65V, it is because I found that a real world LM324 with 0V at the negative supply pin actually swings no lower than 0.65V.  I found the spice modelling works more accurately if you pretend the negative supply is 0.65V rather than 0V.

[DrAlx]

A better demo video clip has been posted on the original link.

http://www.mrblackpedals.com/products/shepards-end

The first 10 seconds shows the barberpole effect quite well (the rest of the video not so much).
It sounds to me like the sweep goes back down a bit after the zero point has been reached.
I am wondering if each component flanger (I reckon there are just 2 of them) is only faded out of the mix after its zero point has been passed?
That would make sense to me since if you fade it out AT the zero point then you won't hear the zero.
If I'm right it means not all the notches in the frequency response are moving upwards for the whole cycle.
I can't see any other way to get both barberpole and through-zero though.

Still sounds pretty good.

[armdnrdy]

You can hear it at the end as well. The rate is just faster. You hear the flanging continuously "going up" but...you don't hear a downward sweep.

But...I agree...the beginning of the sound sample is a much clearer example.

[StephenGiles]

Yes, the first 10 seconds is just the demo we need - but the rest is rubbish, a complete mess

[Mark Hammer]

A better demo video clip has been posted on the original link.

http://www.mrblackpedals.com/products/shepards-end

The first 10 seconds shows the barberpole effect quite well (the rest of the video not so much).
It sounds to me like the sweep goes back down a bit after the zero point has been reached.
I am wondering if each component flanger (I reckon there are just 2 of them) is only faded out of the mix after its zero point has been passed?
That would make sense to me since if you fade it out AT the zero point then you won't hear the zero.
If I'm right it means not all the notches in the frequency response are moving upwards for the whole cycle.
I can't see any other way to get both barberpole and through-zero though.

Still sounds pretty good.

I would think that perceptible through-zero requires that there be somewhere to pass through TO.  IN other words, some time  has to be spent on the other side of zero, in order for the zero-point to be detectable.  Like yourself, I find it hard to imagine barberpole and TZF co-existing, at least in any way that can be clearly heard.

[Keppy]

It sounds to me like the sweep goes back down a bit after the zero point has been reached.
I am wondering if each component flanger (I reckon there are just 2 of them) is only faded out of the mix after its zero point has been passed?
That would make sense to me since if you fade it out AT the zero point then you won't hear the zero.
If I'm right it means not all the notches in the frequency response are moving upwards for the whole cycle.
I can't see any other way to get both barberpole and through-zero though.

I would think that perceptible through-zero requires that there be somewhere to pass through TO.  IN other words, some time  has to be spent on the other side of zero, in order for the zero-point to be detectable.  Like yourself, I find it hard to imagine barberpole and TZF co-existing, at least in any way that can be clearly heard.

Check out the documentation: http://www.mrblack.jackdeville.com/manuals/shepards-end-manual.pdf
The WAVE control goes from Upward Cycle (CCW) to Thru-Zero (center) to Downward Cycle (CW).
That makes it seem like it doesn't do Barberpole and TZF at the same time.

[DrAlx]

I'm guessing the WAVE control doesn't change the gradients of the two parts of a triangle wave, but rather changes the amount of upward time vs amount of downward time. e.g. there will be a setting near the end of that control that gives a long sweep (where delay goes from 7ms to 0ms) followed by a very short sweep in the opposite direction (e.g. where delay goes from 0ms to 0.2ms) before the whole cycle repeats.  That would explain why I'm hearing the zero point near the end of the cycle but not at the very end.

[DrAlx]

Actually come to think of it, a much more plausible implementation is that the two ramped delays actually only ever go in one direction (e.g. always ramping upwards to give increasing delay) and for the WAVE pot to actually control a FIXED delay for the "clean" signal.

In other words, each of the 2 flanger sounds are not made by mixing a clean signal with a (ramped) delay,
but rather by mixing a fixed delay with a ramped delay.

See the three slides here...

http://1drv.ms/1g4HqDD

[StephenGiles]

Actually come to think of it, a much more plausible implementation is that the two ramped delays actually only ever go in one direction (e.g. always ramping upwards to give increasing delay) and for the WAVE pot to actually control a FIXED delay for the "clean" signal.

In other words, each of the 2 flanger sounds are not made by mixing a clean signal with a (ramped) delay,
but rather by mixing a fixed delay with a ramped delay.

See the three slides here...

http://1drv.ms/1g4HqDD

That makes sense.

[StephenGiles]

This fascinating thread seems to have died a death when it was getting very interesting.

[armdnrdy]

This fascinating thread seems to have died a death when it was getting very interesting.

I have a feeling that DrAlx hasn't given up on it. I would imagine that he's working diligently on a build. 

[StephenGiles]

This fascinating thread seems to have died a death when it was getting very interesting.

I have a feeling that DrAlx hasn't given up on it. I would imagine that he's working diligently on a build. 

Aha, I'll keep quiet then - back to the manual for my new Tascam DR-05!

[puretube]

U.S. pat. # 2846913



(65 years ago...)

[StephenGiles]

It will be DSP Fv-1 based.

It is!

[Ice-9]

It will be DSP Fv-1 based.

It is!

Yes it is  Jacks' Fv-1 stuff is quite amazing .

[StephenGiles]

Any developments on all this ?