OTA phaser questions

[MrStab]

Hi guys,

I've been collecting ideas for a phaser for a while and a coupla days ago i finally built what i have so far on the breadboard. it took about 2 days for me to realise there was an intermittent connection on my breadboard. so it was off to a good start.

So far, i have 2 fixed & 2 swept stages, a' la Causality IV, but i'm using the standard Ross LFO and plan to add more stages. There are a few other changes as well. I have this whole system where the signal to the phase stages is switched, status LED is activated and rate LED is dimmed/brightened by shorting one PNP to ground. i'm quite proud of that so i'm telling everyone. lol

I have a few random questions, that are mostly interchangeable with the Ross, Small Stone & Rick Holt's Causality IV:

• Mark Hammer suggests using parallel resistors to each OTA's Iabc pins when using more stages. Trouble is, i really like the mod where a pot is placed in series with the LFO output resistor. If i put the pot coming straight outta the LFO, then 3 resistors coming off of that (one to each LM13700), would that still fix the sweep problems i've heard associated with more stages? I've been looking up dimming in LED matrices, kinda similar, but can't find the info i'm looking for.
• This one's bugging me quite a lot: I don't have a scope, and I'm the kinda person who can't hang up a painting straight - it'll always need adjustment. So i have to ask: those of you who've tweaked the bias of the Ross LFO, particularly when using the LM13700, where did you find it most pleasing? I've tried turning regen up full and putting a pot in the LFO bias to change the sweep, and it does change, but it's so subjective that i keep thinking "maybe it could be better" and tweaking it. rinse & repeat. Even if no-one has a preferred position/resistance, any tips on finding the right sweep, either by ear or with a DMM or PC would be handy.
• In the Causality IV, the first fixed stage has a 4n7 cap, whilst the 2nd has a 47n cap. I noticed the Phase 100 and other phasers with fixed stages all had constant cap values, and to the best of my knowledge a cap ratio like this is used in Univibe mods, at least on swept stages. I made them both 4n7 and the phase effect was much weaker. Could someone explain this to me?

Cheers for any help!

[mth5044]

I cant really answer any of your questions, but be careful when changing the Iabc resistors. The input can only take up to 2ma. I imagine that using a single resistor to feed stages. Then adding more, would reduce the mA on each pin, making the phaser weaker sounding. If you have found a good value for each pin, you can add more stages with the same value without worrying about mA until your LFO can't give anymore.

[MrStab]

I've been reading a lot about Mark Hammer's experiments with Iabc tolerance, and i'm safely settled at 8k2 (well, a 47k and 10k in parallel). so it's something i'm concerned about already, but thanks for the reminder. i should be wary if the current limits particularly when trying new things out.

[MrStab]

i get that the current would be redistributed, but i just can't tell if having one common resistance (ie. the pot) between those parallel resistors and the LFO would make it a pointless exercise

[R.G.]

Mark Hammer suggests using parallel resistors to each OTA's Iabc pins when using more stages. Trouble is, i really like the mod where a pot is placed in series with the LFO output resistor. If i put the pot coming straight outta the LFO, then 3 resistors coming off of that (one to each LM13700), would that still fix the sweep problems i've heard associated with more stages? I've been looking up dimming in LED matrices, kinda similar, but can't find the info i'm looking for.

Use current mirrors. Put a 1-in/1-out NPN current mirror at ground after the LFO, then connect the output current to the input of a one-in/many-out PNP mirror at the power supply.

Separate resistors on the iabc inputs are an attempt to get equal currents to each OTA in the face of variations in the OTAs. It's much better than parallelling up non-identical OTAs. Current mirrors are the way to do this even more predictably. They will (reasonably) accurately cause equal currents to flow on the outputs, controlled by one input. The NPN/PNP mirrors I suggest are to get the current flowing into ground where you cannot get access to the insides of the IC.

As an aside, OTAs consist of one differential amplifier at the input, and several current mirrors.  

In the Causality IV, the first fixed stage has a 4n7 cap, whilst the 2nd has a 47n cap. I noticed the Phase 100 and other phasers with fixed stages all had constant cap values, and to the best of my knowledge a cap ratio like this is used in Univibe mods, at least on swept stages. I made them both 4n7 and the phase effect was much weaker. Could someone explain this to me?

Hold tight, Robin.

Each stage of a phaser consists of a single R-C filter arranged with an active device so that you get phase shift on the output signal without losing amplitude. You can get a maximum of 90 degrees of phase shift from a single R-C section. By ganging up many of these, we can accumulate much more than 90 degrees of phase shift.

But the results of a phaser's sound depend on not just how much total phase shift, but where in the audio frequency spectrum the 180-degree phase shift (and odd-numbered multiples of 180) points happen, because these are the frequencies where there will be a cancellation notch when we mix it back with the dry signal. So not just how many R-C sections matter, but also the value of the exact placement in the range of audio frequencies, from 20Hz to 20kHz.

Stay with me, we'll get there.  

The phase shift of a single R-C section is 45 degrees at the frequency where the impedance of the cap is equal to the resistor; or F = 1/(2*pi*R*C). But we're changing the resistance all to h@ll and gone; only the cap stays the same as it's swept. So the value of the cap sets the relative positions of the centers of the influence of each R-C section. They set the spacing of the eventual notches in the output.

That would imply to the casual reader than we ought to space those suckers out for best effect all over audio, right?

That's a problem. The human ear is not all that sensitive to notches in the bass range, and notches in treble get tiring, quickly. Unless you're shooting for some special effect, the cap values (and the values of the swept resistances) need to concentrate most of the notches in the midrange where we hear them better.

In the Univibe, they were doing two different things with the phaser - they were using it for both phasing (that is, making audible notches) and for vibrato, actual phase shifting. Vibrato needs to be spread out over a lot of the audio spectrum to sound better. They were compromising.

The issue of identical caps versus spread-out caps was the subject of a scholarly article in the JAES back in the early 70s, where the conclusion was that for best phasing, the caps should be identical. I don't know if that's right, but that's what HE thought. At the bottom of this is the question "how does it sound to YOU"?

[MrStab]

Use current mirrors. Put a 1-in/1-out NPN current mirror at ground after the LFO, then connect the output current to the input of a one-in/many-out PNP mirror at the power supply.

Thanks a lot for the input, RG. Hope you don't mind if i just address the first part of your reply for now, so i'm as thorough as i can be, then i'll get round to trying to understand the rest.

I have a lot of reading to do re. current mirrors, but it's about time i got round to it. Is this roughly what you mean?



I'm sure that's overly-simplistic and doomed to failure, but am i in the right ballpark? i'm definitely unsure whether i got the "one in-many out" half right.

cheers!

[MrStab]

[MrStab]

Really sorry to spam, i'll make this my last post until someone else replies (or until i go for the rest of RG's explanation), but i just wanna say that the above schematic seems to work! I'm only using one OTA atm, testing each of the current mirror's outputs seperately, so it's not entirely conclusive but it seems to be doing the job so far. Feel free to use it, anyone. Thanks RG!

[R.G.]

Really sorry to spam, i'll make this my last post until someone else replies (or until i go for the rest of RG's explanation), but i just wanna say that the above schematic seems to work! I'm only using one OTA atm, testing each of the current mirror's outputs seperately, so it's not entirely conclusive but it seems to be doing the job so far. Feel free to use it, anyone. Thanks RG!

Sorry, didn't get back to this til now. That's pretty much exactly what I meant. There are fancier, more-accurate current sources using more devices, but the simple version seems to work pretty darn well. Glad it's working out for you.