8 Stage Ross Phaser Switching

[moosapotamus]

Hi all. I haven't been here much lately due to life's myriad distractions. Eurorack synths has been one major distraction for me, among many others. But anyway, I am definitely still tinkering with stompboxes. So...

I happened to have a handful of LM13600 burning a hole in one of my parts drawers. So, I decided to make an 8-stage Ross phaser using PCBs from Tonepad (Ropez phaser).
http://www.tonepad.com/getFile.asp?id=99
http://www.tonepad.com/getFile.asp?id=100

I could use a DPDT toggle to switch between 4 and 8 stages. But instead, it would be nice to use a 4-position switch to select 2, 4, 6, 8 stages.

Will something simple, like this, work?



Or, is it necessary to also break the connection(s) between the unused stages? If so, then I'm stumped.

What is the best way to implement a 2, 4, 6, 8 stages switch in an OTA phaser like this? Thx!

[ElectricDruid]

I think it should work exactly as per your diagram. There's no need to break the connection to the unused stages.

HTH

[Mark Hammer]

Keep in mind the following:

1) More stages = more cumulative noise.  Many multi-stage phasers will stick a smallcap in the feedback loop of the last stage, to both keep hiss out of the overall phase-shift passband, as well as blocking it from going through the recirculation path and getting exaggerated.

2) More stages = more risk for oscillation with any reasonable amount of recirculation.  The schematics indicate that each allpass stage SHOULD be unity gain, but resistor tolerances can easily result in stage-gains a smidgen over 1x.  On its own that's nothing to be concerned about.  But several cascaded stages with a gain of 1.1 eventually gets you something that leans more towards 2x, rather than 1x, and recirculating that also reapplies that gain, leading to oscillation.  Francisco generously sent me one of the add-on boards for more stages, that it used to make an 8-stager, and I (too) easily got oscillation.  Judicious use of the Feedback control is recommended, as is trimming some of the bass off the feedback signal.  Francisco shows 1uf caps at the input the the Feedback pot, as well as its output.  You may want to drop those down as low as 100nf to keep the "honk" under control.

3) Selecting between the outputs of the 2nd, 4th, 6th, and 8th stages is not a problem.  One simply connects the 27k wet-mixing resistor on the 2nd op-amp (IC1b) to the relevant allpass stage.  HOWEVER, where do you take your feedback signal from?  Two-stagers never use feedback.  Would a 4-stager sound like one if it is taking its feedback from stage 8?  So it would seem that two switches are called for: one to tap the phase-shift at the desired stage, and one to select where the feedback is coming from, if at all.

4) I cannot recommend the "phase-filter" option highly enough.  I have never experimented with it beyond 2 allpass and 2 lowpass stages, so I have no idea what to expect with more than 4 stages.  But phase-filter without dry signal is just sexy as all getout.

5) Finally, the Ross is based on the 6-chip Small Stone, but changes a few things.  The Small Stone's "Color" switch changes between feedback and no feedback, and between triangular and hypertriangular wave forms.  It essentially switches between sweep shape and feedback that is optimized for long slow dramatic sweeps (wide, more feedback), and faster more bubbly ones (narrow, no feedback).  The Ross provides for continuously variable feedback, which is handy, but only uses the hypertriangular sweep.  Great for slow sweeps, but annoying when the speed goes up.  However, the LFO can be modded to provide either triangular OR hypertriangular, by copying the Small Stone circuit.  Here's the two of them, with obvious parallels.  I've installed a slide switch to go between the two waveshapes, and the triangle is MUCH better for faster bubbly sweeps.  The relevant 270k resistor is replaced with a series pair of 100k and 180k.  The waveshape switch connects either the 280k total to V+ or the junction between 100k/180k.

6) The Small stone use a clever trick, by sticking a 10uf cap to ground after the 10k output resistor that limits the current going to the OTAs.  It "filters" modulation above 1.5hz, almost like an automatic sweep width control that trims back higher speeds.  I've implemented that on Ropez builds and it works well.  I have also used slightly lower current-limiting resistor values than 10k.  One of my builds used 9k1, and thatthing sweeps ridiculously high in hypertriangular mode.  If you're familiar with the E.L.O. tune "Strange Magic", it sounds like that.

[moosapotamus]

Thank you, both! Very helpful.

I'm actually pretty far along with this build. Just waiting for enclosure to arrive. Unfortunately, the graphics for the num phases switch will only indicate 4 or 8 stages. So, it goes...

Mark - I've probably read most of your posts about this pedal and, after trying a number of the mods that you have shared and discussed over the years, I decided to implement the following...

• At the Regen pot, replaced the two 1uF caps with 220nF to reduce 'howling'.
• At IC4a pin1, replaced the 270K with a 180K and B500K pot for a sweep width control.
• Added switch to lift 27k clean mixing resistor for vibrato.
• Added switch for phase-filter.
• And finally, the switch to set the number of phase stages.

Regarding the concerns about hiss and also how/where to route the regen path, instead of starting with stages 1,2 for the 2 stages setting (as shown in my sketch), what if I wired the phase switch to start at stages 7,8 for the two phase setting then add 5,6 for 4 stages, 3,4 for 6 stages, and 1,2 for 8 stages (the reverse of what's in my sketch)? IOW, switch to the inputs of the filter stages instead of the outputs. I think this would do the following...

• Keep that small cap to reduce hiss (#1, above) always in the last stage.
• Allow the phase-filter switch to always be on the last two stages. In my sketch, phase-filter would only be effective in the 8-stage setting.
• Enable regen to always come from the last (or, second to last) stage no matter the setting of the phase num switch. In my sketch, the regen signal would always contain 8 stages of phasing no matter how many stages are selected. (Is it worth considering tapping the regen from stage 7 instead of stage 8?)

So, to switch the stages in reverse order like I am suggesting, instead of connecting the 27k wet mixing resistor to the output of the desired stage, the SP4T switch would connect pin1 of IC1a to the input of desired stage. Does that make sense? Am I missing anything? Thx!

[Rob Strand]

Regarding the concerns about hiss and also how/where to route the regen path, instead of starting with stages 1,2 for the 2 stages setting (as shown in my sketch), what if I wired the phase switch to start at stages 7,8 for the two phase setting then add 5,6 for 4 stages, 3,4 for 6 stages, and 1,2 for 8 stages (the reverse of what's in my sketch)? IOW, switch to the inputs of the filter stages instead of the outputs. I think this would do the following...

• Keep that small cap to reduce hiss (#1, above) always in the last stage.
• Allow the phase-filter switch to always be on the last two stages. In my sketch, phase-filter would only be effective in the 8-stage setting.
• Enable regen to always come from the last (or, second to last) stage no matter the setting of the phase num switch. In my sketch, the regen signal would always contain 8 stages of phasing no matter how many stages are selected. (Is it worth considering tapping the regen from stage 7 instead of stage 8?)

[/list]

I thought your switching was fine. 

I'd put the filter after the switch common so it always filters noise.

There's many possible ways to do the feedback.  You can even listen to the output of stages 1 and 2 but have the feedback coming from 8.   It makes more sense to use feedback from the same filter that you are using.   Not all combinations of feedback sound good.   The response produced for some combinations can be totally crap.   You can also invert the feedback.  I suggest running any options you have in mind through LTspice.

Also, for feedback, mixing using a dedicate mixer stage is preferred.  When you mix back to one of the phaser shift stages it screws up the phaser a bit (even though some pedals still do that because it doesn't require another opamp).

[moosapotamus]

Hi Rob. Always appreciate your thoughts. I thought I was just asking a simple question. Now so many things to consider.   

Mulling over all these ideas, I think I'll forget about the filter cap because it wasn't originally used in the Ross and it's one less thing. This is a nice sounding phaser that many have built, so I'm not going to sweat too many of the details. But I am still concerned about 1) how to switch the filter stages, 2) what to do with the regen path(s), and 3) making sure that the phase-filter works for all settings of the number of filter stages. So, this is what I am currently thinking...



I notice that the orig 4-stage Ross returns the regen to stage 2. Using a DP4T rotary, this arrangement returns the regen to the 2nd stage for all settings of the stages switch, except for the 2-stage setting.

This arrangement also keeps the phase filter on the last two stages for all settings of the stages switch. As an aside, a number of years ago I went knutz modding a Small Stone. One of the mods was the phase-filter on all four stages with four individual switches to independently set each stage to either all-pass or low-pass. I found the myriad combinations interesting but overkill. To my ears, low-pass on just the last two stages is all that's needed.

I'm going to wire this up and audition it before boxing it up. Thx!

[Mark Hammer]

The Ropez board provides for inserting the 4 additional stages between stages 2 and 7, OR bypassing them.  Theoretically, one could also leave in 5 and 6 and bypass 3+4 (or leave in 3+4 and bypass 5+6), to achieve 4, 6, or 8 stages, with feedback going from stage 8 back to stage 2.

As much as I like and endorse the phase-filter mod, I often wonder if there isn't a better way to do it.  First, switching where the cap end gets tied to (input or ground) results in an annoying pop, although one can simply make a point of only switching to that mode while in bypass.  Second, the cap values that are best suited to phasing may be suboptimal for lowpass filtering, although this is a matter of taste.  Certainly feel free to experiment.

Here's a quick demo of a tricked-out 4-stage Ropez (now sold).  I added the Mix level, and the W/N toggle selects between the stock hypertriangular LFO and a more triangular wave, which also yields faster speeds(as it does in the Small Stone).  The P/F toggle on the right should be self-explanatory.  The addition of vibrato mode and phase-filter allows the circuit to seriously transform from a plain vanilla phaser.

https://www.youtube.com/watch?v=v6Hrfc8Tscc&t=47s

[moosapotamus]

Nice sounding demo, Mark. I'm a big fan of that funky, slinky, swampy phase-filter sound, too!

After wiring it up outside of the enclosure and listening to the 2, 4, 6, 8 stage settings, I decided to just go with the basic two position, 4-8 stages, toggle switch instead of a 4-position rotary. I felt like I could get sounds similar to the 2 and 6 stage settings by just dialing back the Regen for the 4 and 8 stage settings. Happy to be building this. Nice sounding phaser!

[Mark Hammer]

Excellent!  It was after making use of the 4-stage add-on board that I understood the risks of oscillation with more stages.  In particular, is it your experience as well that 8 stages seems to start its sweep much lower down?  That apparent phenomenon was what led me to trim back on the low end of the feedback signal.

I'll also add that I find an "offset" control, that can move the starting point of the sweep upwards or downwards, can be an invaluable tool, though I'm not sure how to implement it in that circuit.

[moosapotamus]

Yes, I think that was a good call. I also like having the sweep width control. Even though it has a pretty significant impact on the speed, I think it helps cover the sound of 2 and 6 stage settings when only having settings for 4 and 8 stages. I'm still working on getting it boxed up. Psyched to give it a full workout after it's in the enclosure.

For the record, in addition to 4/8 stages, I decided on the following mods for the final build.

• Left the Regen returning to the 2nd stage for both 4 and 8 stage settings
• Replaced the two 1uF caps at the Regen pot with 220nF
• Replaced 270k from IC4a pin1 to V+ with 180k and B500k pot for sweep width control
• Switch to lift the 27k dry mixing resistor for vibrato
• Phase-filter on last two stages

I tried almost all the other mods I could find or come up with and those are the ones that floated my boat. In the past, I've tried Univibe cap values on a Small Stone and found it to make the effect slightly more subtle, kind of diminishing the 'character'. Curious to know if others agree/disagree. Wonder how equal cap values in a Univibe would sound.

Offset does seem like it would be useful. Maybe replace the LFO with a STOMPLFO?

[Mark Hammer]

Q: When the LFO is feeding more stages, and the same LFO current is split up among more stages, or alternatively feeding fewer stages, does the current-limiting resistor value need to be changed?

On the other hand, if one is sweeping all 8 stages all the time, but tapping the signal after some number, I would imagine a lower-resistance is needed, but no changes with different tap-points.

I know that OTAs can fry up real nice when too much current is coming into the I_abc pin.  So what sort of resistance range might be considered as usable?  I've dropped the current-limiting resistor from 10k down to 9k1 on one of my 4-stage Ropez builds and been impressed with the wider sweep.  But where does it begin to get dangerous?

[ElectricDruid]

I know that OTAs can fry up real nice when too much current is coming into the I_abc pin.  So what sort of resistance range might be considered as usable?  I've dropped the current-limiting resistor from 10k down to 9k1 on one of my 4-stage Ropez builds and been impressed with the wider sweep.  But where does it begin to get dangerous?

2mA max Iabc current for the 13700, IIRC. That said, a lot less is generally a good idea because of the variable offsets. The more current you use on the Iabc, the more CV feedthrough you get.

[Mark Hammer]

Finally got a Ropez populated board that had never worked, working finally.  Problem was that the wires from the speed pot to the board were tarnished enough that what looked like a proper solder joint was not.  So the Resonance seemed to do something but there was no sweep.

Once the problem was finally discovered and fixed, I went and did the mod to provide the two waveshapes of the Color switch on the Small Stone.  Then I swapped the 10k current-limiting resistor on the LFO output for 9k1 and it had the desired effect of widening the sweep (primarily extending it at the top end).  THEN, I incorporated the mod that the 6-chip Small Stone used for "automatic width adjustment", and strapped a 47uf cap from the OTA end of the resistor to ground.  That had the expected effect of reducing the sweep width once the speed got over .4hz.  Finally, wishing I had better control over sweep width I added a 5k variable resistor (pot) between the OTA end of the 9k1 resistor and the OTAs, providing about 14k of current-limiting resistance.  That additional control, with the cap to ground and the sweep-shape switch for the LFO, provided a lot of control over the feel of the sweep.  I can get nice calm bubbly phasing and ultra-wide slow dramatic sweeps.  I'm pleased.

Dare I add more?  We'll see.