Adding stages to OTA-based phasers anything else to change?

[Mark Hammer]

Okay, so last night I whipped up a little 2-stage add-on for the Causality4MkII phaser that I made from Rick Holt's clever design and nice board layout.  The only change was use of 13600 instead of 13700.  It "works", but the addition of the two stages via the "patch-in" points provided result in a change in the location of the notches, not just an increase in their number.

Specifically, the notches seem to sweep so low that there are unpleasant resonances produced.  The same thing was also true when I added 4 more stages to a Ropez.

Now, there is no fault with the basic designs or the layouts.  Everything works.  I just can't help but feel that there is some sort of change required in the manner in which these stages are being driven when the number of stages goes up.  I know there are reasons to be concerned about driving the stages with too much current and potentially frying the OTAs.  I don't know how to calculate or identify the boundary between safe and risky, though, so I don't kow what changes ought to be made.

For instance, should the current source (LFO) feed those added stages via a lower-value series resistor to permit more current shared across the stages?  Should the value of the resistor remain the same but a second one of equal value be used in parallel to feed the additional stages?  Shuld there be some other change made to accommodate that same LFO driving more stages?  Presumably whatever is the right course of action is also applicable to virtually any OTA-based phaser.

I don't know what to do, and I imagine there are others in the same boat.  Any ideas or suggestions?

[StephenGiles]

I don't like having to turn the light on at 6:15pm

Are there any schematics out there for multistage phasers using 13600/13700? I'm sure I've seen one but can't put my finger on it.

[StephenGiles]

Yes, this one uses 8 stages.

http://www.musicfromouterspace.com/analogsynth/PHASESHIFTER2007/PHASESHIFTER2007.php

[Tonemonger]

 R.A.PENFOLD ?
He likes 136+700s

[Mark Hammer]

Yes, this one uses 8 stages.

http://www.musicfromouterspace.com/analogsynth/PHASESHIFTER2007/PHASESHIFTER2007.php

You will note that Ray Wilson uses eight 20k resistors in parallel to "distribute" the control current to the OTA stages.  The Small Stone, Ross/Ropez, Nobels, and Causality4 each use a single series resistor to directly feed the Iabc pin on all OTAs.  The DOD FX-20 feeds the LFO to one fixed resistor which then splits into two parallel 4k7 resistors, each feeding 2 OTAs.

I could explore further, but there appears to be no real consistent pattern to what normally gets done.

[StephenGiles]

Maestro MP1 uses 1k resistors to all 13700 stages!
Behringer vintage phaser uses a single 47R then common to all 13700 stages.

[Mark Hammer]

Now you can see why I am so confused!

[StephenGiles]

It goes on - Moog 12 Stage phaser - 30K  resistors to all CA3094s!

But look here -

http://www.jhaible.de/varislope_filter_phaser/varislope_filter_phaser_sch1.pdf

Now that's something else!!

[mrslunk]

I think it depends on the circuit.
For the lm13700
Iabc of each OTA needs to be <2ma
How you feed that current in is up to you...

Iabc feeds into a current mirror. So from what i understand, the voltage at Iabc will be about a diode drop or 2 above V-.
My transistor theory is not particularly good though, this is mostly assumption based on observation of schematic and what my multimeter tells me.

So, say you've got 4 ota's (2 LM13700, for eg)
You could
a) feed a current source through a current limiting resistor then straigh into all your OTA's Iabc
b) feed a current source through a 4 current limiting resistors then into each ota.
c) feed a voltage source through an appropriately chosen resistor, say <2k for 9v(max) into 4 ota's (<1ma for each ota)
d) feed a voltage source through 4 appropriately chosen resistors (say, 10k seems to be pretty common for 9v stuff) so that you have Iabc <1ma for each ota.

d) Is best in my opinion.
a and c both require you to have all your OTA's wired up at all times, otherwise you may feed too much current in and fry your oldmates.

I think the ross phaser uses (c), each OTA should be recieving around 0.1mA.. i think. (LFO around 4.5v +/- 2.5ish)

i've got a 12 stage ross prototyped, I found that reducing the 10k resistor from the base of the LFO buffer to Iabc's, to about 4.7k(ish) or lower worked well
however, in working that out, i think i fried about 4 or 5 ota's by not having some of them socketed whilst testing stuff. Yumm.

I'm current jimmying around with and OSCar filter clone. 10k resistors to each OTA from a voltage source. Haven't fried any yet. fingers crossed.

Hope that helps.
Pete

[StephenGiles]

Which brings me back to this:



This old chestnut    I used a 30k resistor last time I had this on a breadboard - what do you think Pete?

[mrslunk]

The first thing that comes to mind for me is to jimmy around with the gain of the non-inverting amp feeding into the 20k resistor->OTA pin 1.
2M pots aren't particularly nice, there are pently of other ways to get gains from 1->200
you could even drop down the gain on the opamp and decrease the 20k resistor to feed more current in.
By how much etc, etc you'd need to know what voltages you're running at and the like and i'm afraid my theory isn't good enough to work it out without and breadboard and multimeter.

[StephenGiles]

The first thing that comes to mind for me is to jimmy around with the gain of the non-inverting amp feeding into the 20k resistor->OTA pin 1.
2M pots aren't particularly nice, there are pently of other ways to get gains from 1->200
you could even drop down the gain on the opamp and decrease the 20k resistor to feed more current in.
By how much etc, etc you'd need to know what voltages you're running at and the like and i'm afraid my theory isn't good enough to work it out without and breadboard and multimeter.

Yes you're right, that non inverting op amp is simply dropping a voltage on the 20k which is roughly proportional to the input signal, and then as a current on the other side of the 20k, energises the OTA. The trick is that the peak of the voltage at the OTA output is also proportional to the input signal, so when applied to a VCF the sweep will be dynamic, as adjusted by the start and stop frequency pots.

[PRR]

If the control signal is a voltage, use a separate resistor for each OTA.

Two OTAs on one chip, the two diodes are well-matched and you can get away with one resistor. If you could buy bulk factory-fresh OTAs, then there is some good chance four chips from the same tube will match close-enouh to share a resistor.

In today's odd-lot world, don't bet on matching. Use separate resistors.

As Peter says, te maximum Iabc is 2mA, but error increases at 1mA. And withthe Darlington buffer, there's no real need for high current. Since the chip will more than cover the audio range, stay with 1mA max.

At +/-15V (total 30V) then 30V/1mA= 30K per Iabc pin is fine. We often run +/-12V, and there is a volt or so at the Iabc pin, so 23K is fine; we see 22K and 27K used.

> this one uses 8 stages.

It feeds the Iabc pins from a current-source referenced from ground toward -12V. So max resistor voltage is 11V. This designer allowed a separate 20K resistor for each pin. Max Iabc is 0.55mA; except R14 will only pass ~~1mA total for all 8 stages, so 0.125mA max each.

Figure what the original Iabc range was. If it did not have separate equalizing for each Iabc pin, fix that. Then ponder how to get the same Iabc range with your added stages.

If the wobble source is a voltage from an opamp, you can probably just add more of the same resistors to the added stages. Past a dozen, you should wonder if the opamp can pass the added current.

For that particular 8-stage: R14 sets the total currrent to split to 8 stages. For 16 stages you probably want to cut R14 in half.

[frequencycentral]

Any progress Mark? Have you experimented with different cap values for the phase stages?

[Mark Hammer]

So last night I applied what was discussed here, and re-arranged the way the two additional swept stages in one of my C4MK2 were driven.  Instead of feeding all 4 stages from the same 4k7 resistor, I simply added a second 4k7 resistor in parallel and fed the additional two stages via their own 4k7 resistor.

Improved things dramatically.  Very nice, now.  The bubbly fast Leslie sound is much better.  Nails the "No Quarter" tone.  Now I have to see if I can improve the performance of the 8 stage Ropez I have.

[frequencycentral]

That's great Mark. I'm all excited to build my next one now. I wonder if I should layout a new board for a Causality 6? I'm also wondering if there would be any mileage in having a second Range control for the extra stages. I think that might sound really cool - two sets of notches, one 'bright', the other 'dark'?

[Mark Hammer]

Now I have to see if I can improve the performance of the 8 stage Ropez I have.

So I attacked my 8-stage Ropez and fed each set of 4 stages with its own 9k1 resistor.  The sweep is noticeably better, but it has this quirk that it sounds kind of "doubled" at the top end of the sweep when the speed is more than about 1.2hz or so.  What I'm wondering is whether this is a result of each set of 4 being controlled by a discrepant current, due to mismatched resistors.

The other thing is that the sweep is not nearly as wide as it is on a 4-stager I made that used a 9k1 current-control resistor, rather than 10k.  That one sweeps like it was an A/DA Flanger: waaaayyyyyy up.  Just keeps on going and going.  Not this one, though.  Should I assume that the current output capability is a function of the number of stages being fed?

I'm going to see if I can tinker with the resistor values and improve the sweep.

[Mark Hammer]

Ha!!  Solved.

The 1st issue Small Stone used an LFO almost exactly like the Ross.  One of the features it had was the inclusing of a medium-value electrolytic cap to ground from the OTA side of the current-limiting resistor coming out of the LFO.  This had the effect of acting like a lowpass filter, and taking out some of the sharpness in the "turnaround" of the top part of the LFO sweep.  This had the effect of softening the sweep at faster speeds, but leaving the hypertriangular sweep in place for slower speeds.

Well, I liked it, so I copied it, and stuck a 10uf cap to ground from the OTA side of the 10k (now 8k2) resistor on the LFO output.  That was fine as long as I was only driving 4 stages.  Once I added 4 more stages, and found out they would work better with a separate LFO output path, I neglected to add the softening cap on that control current path.  

The upshot was that 4 stages had the peak-softening on the LFO, and 4 stages didn't.  So when the 4 stages without the cap would continue to sweep somewhat higher, the other 4 stages with the cap would "hang back" a bit, as a more triangular wave.  Although all 8 stages are in series, the net effect was that their summed phase shift had a sort of "jerk" built in at the top of the sweep.  I just experimented by lifting that 10uf cap, and bingo the sweep was smoother.

And, of course, the reason why the jerkiness only appeared at faster frequencies was because the added cap only exerts its peak-softening effect at faster sweep speeds.

I am now using two paralleled 8k2 resistors for driving the two sets of 4 stages.  Not a superwide sweep, but noticeably wider than before.

I also found that I could not max the resonance control when using 8 stages, without risking ugly oscillation at an unpleasantly low frequency.  I swapped all the 1uf caps around that pot for 100nf, and the woof monster has been tamed.