Please teach me about OTA all-pass/phase stages

[.Mike]

I've been working on an OTA-based phaser for a couple of weeks now. I'm kind of taking FrequencyCentral's Causality 4/6 design, extending it to more stages, and replacing the LFO with ElectricDruid's TAPLFO.

I'm having pretty good success. I have a fully functioning 8 stage phaser on my breadboard-- 2 fixed, 6 swept (3x NE5517). I ran out of breadboard room for the lfo circuit, so I hacked in the LFO from my tremolo, which uses the non-tap precursor to the PIC LFO.

I think I understand how the phase shift is controlled by current supplied by the LFO. I've had a lot of fun tweaking the range by adjusting the resistor that controls Iabc, and by controlling the range that the LFO covers. I'm using two inverting opamp stages to simultaneously filter the PWM, and take the LFO from 0-5v up to around 2.5 - 6.5v, and then feeding each OTA pair with 3.3k resistors.

Where my trouble lies is with the OTA phase stages. I can't for the life of me figure out what does what. It's only 4 resistors and a capacitor-- how hard can it be? I've been simulating the heck out of this circuit, and want to learn more.

Is there some sort of a "Technology of OTA Phasers" article out there that I have missed? Or, do any of you have a good topic or two here with information on how these work? I searched, but maybe I missed it.

Or barring that, could someone please provide a quick-and-dirty lesson on how to adjust OTA phase stages? Maybe even just explaining how to calculate the corner frequency of the filter?

I can't even find similar all-pass stage example circuits in the datasheets for the LM13600/LM13700 or NE5517.

Thanks in advance.

Mike

[R.G.]

An OTA is a voltage controlled current source. It produces at its output a current which is proportional to the difference between the voltages at its inputs. The capacitor lets through a current that is proportional to the input voltage and frequency. The cap current and output current are added at the output to make the phase shifted signal. The input signal through the resistors is to the inverting input while the cap current comes from the signal, so the phase difference ranges from 180 degrees (low frequencies, cap = open circuit) to nearly 0 degrees (high frequencies, cap appx = short). The output follower, well, follows. The feedback cleans up and "idealizes the addition at the output of the OTA stage.

This same thing happens in an opamp phaser with the cap current subtracted from the inverted signal current at the input of an opamp. But the OTA is unusual in doing this at the high impedance output. The resistance which the cap generates the phase shift with is the effective transconductance of the OTA, which um, varies.

[frequencycentral]

quick-and-dirty

The cap value is the easiest way to change the corner frequency. I used to use 33n for each stage, now I use 10n. Phasing can occur higher up the spectrum. Lowering the value shifts the corner frequency upwards.

[.Mike]

Thanks guys.

RG, I know that's going to help. I just have to read it a few more times, and read along while I'm looking at the schematic. That will have to wait until when it's not 2:30AM... heh! Tomorrow. I like building effects, but I like learning how they work instead of just doing circuit-by-numbers. Thank you

Rick, I've been trying values from 2.2n to 100n on both the OTA and the fixed opamp stages, since I've seen those values on other schematics. What I do is model the circuit to see how it changes the frequency response and sweep, then make my breadboard match to see what the modeling translates to in the real world. It's been a lot of fun, and I think I've learned quite a bit so far. With the flexibility of the LFO, I'm able to vary the amount of time spent on the low end of the LFO compared to the high end, and it has made for some really nice phasing. The square wave is pretty cool, too.

Mike

[.Mike]

OK, things are progressing well on my phaser experiments. I do have a question, though. I am trying to maximize the sweep and want to make sure I am correct in my thinking.

I understand that the Iabc is fed a current to make it sweep, and I understand how this is done by varying a voltage through a current-limiting resistor. I also understand that the absolute maximum Iabc on the NE5517 that I'm using is 2mA. I also understand that the Iabc pins sit two diodes above ground, which I measured as 1.3V.

My LFO is 0-5V PWM that I am filtering with a pair of NE5532 inverting opamp stages. I have rebiased the LFO, and it now runs between 1.4V (the lowest I could get an NE5532 to go) and 6.4v.

So if my peak is 6.4v, and the Iabc pin is at 1.4v, I should be able to calculate the proper current limiting resistor with Ohm's law, right?

The voltage would be 5 (6.4-1.4), and the current would be 2, since I want to feed each stage 1mA, and am going to do the standard one resistor for every 2 OTA stages. Sooo...

R=V/I = 5/2 = 2.5k resistor for each stage. If I wanted to feed the maximum 2mA/stage, I could decrease that resistor to, say, 1.3k to deliver just short of 2mA for each stage.

Am I on the right track? 

Mike

[R.G.]

I am trying to maximize the sweep ...
I understand that the Iabc is fed a current to make it sweep, and I understand how this is done by varying a voltage through a current-limiting resistor. I also understand that the absolute maximum Iabc on the NE5517 that I'm using is 2mA. I also understand that the Iabc pins sit two diodes above ground, which I measured as 1.3V.

This is correct. Keep in mind that OTAs have a bias current range where things work as expected of about 1000 to one; better ones run at 10k:1 or more.
Since the max current is 2ma, if your minimum current is down at 2uA or below, controlling current with a resistor gets tricky when...

My LFO is 0-5V PWM that I am filtering with a pair of NE5532 inverting opamp stages. I have rebiased the LFO, and it now runs between 1.4V (the lowest I could get an NE5532 to go) and 6.4v.

So if my peak is 6.4v, and the Iabc pin is at 1.4v, I should be able to calculate the proper current limiting resistor with Ohm's law, right?

... the lower end of the voltage is at "1.4V". It's not. It's at two diode drops. A diode drop is variable depending on the current through it. At a microamp or so, it may be down at 0.4V. That can (and has, for me) messed up tracking/control at the low end. It's something to watch for. But you're basically correct as long as you don't try to get the widest possible swing... which you are. 
Oscillators with 10,000 to 1 sweep ranges have been built with OTAs, but they generally involve active current sources feeding Iabc. So go ahead, but don't get discouraged.

By the way, you can offset the bottom end of that opamp output by putting a diode or two in series with a resistor to the minus supply. This knocks off a (kinda) constant voltage, lowering the output so the opamp output circuit can go lower.

The voltage would be 5 (6.4-1.4), and the current would be 2, since I want to feed each stage 1mA, and am going to do the standard one resistor for every 2 OTA stages. Sooo...

R=V/I = 5/2 = 2.5k resistor for each stage. If I wanted to feed the maximum 2mA/stage, I could decrease that resistor to, say, 1.3k to deliver just short of 2mA for each stage.Am I on the right track? 

Yes.

However, Mother Nature being the b... er, lady that She is, it's wise not to trust any two items to be identical. Using a separate resistor for each OTA can be a step in forcing a kinda-equal current to flow in them, at least at higher currents. The balance gets off more and more at low currents as the non-identicalness of the Iabc inputs becomes more prominent and the amount of drive voltage gets smaller. National does specify matching of both halves of the LM13700, so it's probably OK, but in general be suspicious of anything that tells you "It's OK, they come pre-matched." To the experienced EE, that's much like hearing "... and I'll still respect you tomorrow."

[.Mike]

Thanks RG. I'm not dead set on using the widest possible sweep possible when I finally build this thing-- especially if it gets increasingly difficult to get a nicely matched sweep. I am dead set on trying the widest sweep, just to see what it does, though!

I just spent a couple of hours playing around with this on my breadboard, going from 2 fixed stages + 2 swept, and then 2+4 swept, and 2+6 swept. I did notice that the turnaround at the bottom and top of the sweep were not perfectly matched. I will try using a separate resistor for each phase stage and see what it does, and then I can adjust how deep the sweep goes to see what that does, too.

Mike

[R.G.]

Thanks RG. I'm not dead set on using the widest possible sweep possible when I finally build this thing-- especially if it gets increasingly difficult to get a nicely matched sweep. I am dead set on trying the widest sweep, just to see what it does, though!

Good for you. Experimentation is always good. Remember, a monstrous mind is a toy forever.

[Mark Hammer]

Thanks RG. I'm not dead set on using the widest possible sweep possible when I finally build this thing-- especially if it gets increasingly difficult to get a nicely matched sweep. I am dead set on trying the widest sweep, just to see what it does, though!

I just spent a couple of hours playing around with this on my breadboard, going from 2 fixed stages + 2 swept, and then 2+4 swept, and 2+6 swept. I did notice that the turnaround at the bottom and top of the sweep were not perfectly matched. I will try using a separate resistor for each phase stage and see what it does, and then I can adjust how deep the sweep goes to see what that does, too.

Mike

When it comes to modulation effects, the "turnaround" is just about everything.  I applaud your efforts to get to a pleasing turnaround.

Note that the RC pair in the fixed stages determine where any additional phase shift contributed by those fixed stages kicks in.  Feel free to play with the cap values to move that "kick-in point" upwards or downwards to suit your tastes and the nature of the sweep range you already have.  

In many respects, I think the optimal application of such fixed stages is to make the upper 1/3 of the sweep range a little richer, since the typical response range of pickups (with the usual loading that happens) and guitar speakers can tend to work against the audibility of notches up there.  But that point is determined by the properties of your rig and preferences, not by some mathematical ideal, so dicker away!!

[.Mike]

Thanks Mark. That is interesting.

I am definitely hearing three different turnarounds at both ends of the sweep. I have it setup as 2 fixed, 6 swept right now, with as big of a sweep range as possible.

With what R.G. taught me yesterday, and you now teaching me that the capacitor can change the turnaround points, it makes sense that the sweep isn't quite even. I have seen values from 3.3n to 33n or higher for the capacitors. I aimed for 33n since that seems more common, but didn't have enough, so I... uhh.. kind of randomly picked values that were close. I also ran out of 27k resistors, so one pair of stages is using 24.9k resistors, the closest I have.

Today I plan on rebuilding the filter for the PWM. I was using a simple 2nd order opamp LPF @ 1kHz, but I was getting some bleed through. Setting the cutoff much lower eliminated the noise, but impacted the sweep. So today I am trying a 4th order multiple feedback filter to see how that works.

Then, I plan on playing with feedback a bit, pulling from different points, and injecting to different stages.

This is fun stuff.

Mike

[Mark Hammer]

Well don't run away with it just yet.  The cap value in the fixed stages doesn't change the turnaround (and especially its gracefulness) all on its own.  It sets what point in the overall sweep those stages contribute the additional  phase shift to produce an additional notch.

So, take an example of a standard fixed op-amp phase-shift/allpass stage where we have a 10k resistor going to the inverting input (from the previous stage), a 10k feedback resistor, a .01uf cap going to the inverting input, and a 22k resistor from the non-inverting input to ground.  The point where that stage contributes 90 degrees of phase shift begins around 1 / [2 * pi * .01uf * .022M] = 723hz.  Below that "knee" it contributes decreasing amounts of phase shift.  Drop the 22k resistor down to 10k and our max-contribution point moves upwards to just under 1.6khz.

It does NOT kick in suddenly, although actually, I'd like to see some sweep-generated scope shots of the notches created at different points in the sweep with and without additional fixed stages.  That picture would be invaluable for intuiting phasers and what fixed stages can do for you.

[Jazznoise]

Just gonna go ahead and ask, what exactly is the point of having Fixed Phase Stages in a Phaser?

[.Mike]

Just gonna go ahead and ask, what exactly is the point of having Fixed Phase Stages in a Phaser?

Why post here what can be easily covered in a 3 page topic... heh... !