Some musings about the Ross and Small Stone phasers

[Mark Hammer]

Thanks to Jeorge Tripps  posting the schematic way back, and Francisco Pena generously producing a PCB layout over at Tonepad, a great many people have become acquainted with the LM13600 OTA-based Ross Phaser (as the "Ropez").  It's a nice little phaser.

The Small Stone, like the Big Muff, has lived through several generations.  Most people are familiar with the 5-chip issue, though it was preceded by a 1975 6-chip issue.  The 6-chip issue uses a pair of CA3094s to form the LFO.

Last night, I was looking at my binder of phaser schems, and realized that the Ross actually resembles the 6-chip Small Stone a lot more than I had realized at first.  Indeed, the LFO is VERY similar in design, in more ways than both of them using two OTAs.  There are some interesting differences, though.

• For speed and speed range, the Ropez uses a 3.3uf cap between V+ and the OTA/buffer junction (see pin 12).  The 6-chip Small Stone uses a 1uf cap.  The Ropez uses a 500k pot in series with a 4k7 fixed resistor between the output of the LFO (pin 9) and the Iabc input pin (pin 16) to set speed.  The Small Stone uses a 1meg pot and 10k resistor.  Not having used a 6-chip Small Stone I couldn't say which had the more usable or pleasing range of speeds, but this suggest that it may be worth tinkering with part changes in those components to season the speed to your tastes and needs.
• The outputs of the LFO or both the Ross and Small Stone feed the Iabc pins of the phase shift stages via a 10k resistor.  However, where the Small Stone tacks a 10uf cap to ground from the "free" end of the 10k resistor, the Ross does not.  That cap would seem to be something that may have some impact on the waveshape of the sweep, and altering it may get you a sweep form more to your liking.  I know I've dickered with the value of that 10k resistance on mine so as to get a much much wider sweep, but the sweep shape at those widths doesn't quite cut it unless the sweep is set VERY slow.  I'm gonna try out adding on a smallercap (say, 2u2 or 4u7) and see if that helps out.
• In the Ropez/Ross, there is a 270k resistor between V+ and the Iabc pin (pin 1) of the first OTA in the LFO.  In the 6-chip Small Stone, the Iabc pin is tied to V+ through a 180k and 100k resistor in series (280k).  In one of the color switch positions, the 100k resistor is shunted so that Iabc now connects to V+ through 180k.  The Color switch alters both the amount of regeneration of the phase-shifted signal, but it also introduces more sweep width in the "color on" position.  This suggests that the 270k resistor on the Ropez can be tinkered with.  I've played with the 10k output resistance of the LFO to get different sweep widths, by varying it between 2k2 and 50k or so, but it may well be that I ought to play with the 270k resistor to get changes in sweep width accompanied by waveshape changes that suit the width.

[Arn C.]

Excellent Mark!  I plan on building this soon, so this will give me some more things to try. 

I do have one question on the pots for the ROPEZ.  The two 500KC, are these reverse audio?

Thanks!
Arn C.

[Mark Hammer]

The Regen pot is a standard volume-type 500k log pot.  The taper is actually produced by the 27k resistor in parallel with it.  I don't even know what taper I used for the speed control.  Whatever the case, it never seems to be the right taper for me!

I also strongly encourage you to experiment with replacing the 10k series resistor on the LFO output that limits the sweep current.  I stuck a 2k2 resistor in series with a 25k pot, but I think that a 50k pot might have been a better choice.  Certainly DO NOT go below a minimum 2k2 resistance or you can damage the 13600 chips.

As the resistance decreases, the sweep goes much higher, though no lower.  As is, the pedal is set for a sweep width that is a compromise between what works best for slow sweeps and what works best for fast sweeps.  If you want, you could actually use a 3-position toggle to select between 3 different resistance values and widths, since the continuous variability may be of little use and even create difficulty in replicating sounds.  Depends on yur needs and preferences.

If you go that route, I would think that a switch selecting between 2k2, 10k and 47k is probably a suitable arrangement, with 10k being a little under 5x the widest sweep resistance (2k2) and 47k being a little under 5x that.  The 47k setting would be best suited for the fastest speeds for more "bubbly" fast Leslie type sounds.  To do it, you'd install a 47k resistor instead of the 10k value, and use the toggle to switch in parallel resistances of 12k (for 9.6k equivalent value) and 2.4k (to achieve 2.3k).

[puretube]

That cap would seem to be something that may have some impact...

forms an (ultra-) lowpass filter with the 10k...

[Mark Hammer]

So does that end up "rounding off" or triangulating the LFO waveform more?

[SeanCostello]

So does that end up "rounding off" or triangulating the LFO waveform more?

If it works like lowpass filters in other phase shifters / modulation effects, it would round off the triangle waveform, but the amount of rounding depends on the rate of the LFO. At a very low LFO rate, the waveform will still be fairly triangular, while at higher frequencies, the waveform will be more sinusoidal. Think of it as a filter that can attentuate the upper harmonics of a triangle waveform, leaving you with a sine wave.

A more important effect of the lowpass filter is to reduce the LFO amplitude at higher frequencies. If the lowpass filter cutoff is set to a VERY low frequency, like 1 Hz or so, the fundamental of the triangle wave will be noticably attenuated as the LFO speed goes up. This results in much less modulation depth as you increase the LFO to vibrato rates. Which is very very nice if you have a phase shifter with a single knob, like the Small Stone.

The same trick works in digital, BTW, although it is trickier because you can't just run a lowpass filter on the LFO and use that to directly control the phase shift coefficients, due to the warping of the bilinear transform. 

Sean Costello

[Mark Hammer]

Okay, much clearer, and very nicely explained.

Now here is the interesting part, I suppose.  Should one attempt to use a variable series resistance feeding the Iabc pins, in the manner I was suggesting, the effect of the added cap to ground would depend on: a) the LFO speed, and b) the setting of the sweep width pot.

With a 10k resistance and 10uf cap, that corner frequency is set around 1.6hz, something that will affect "throb rate" and "bubble rate" speeds, but not slow speeds which will generally be in the <0.5hz range.  If the output resistance is changed to 2k2, that corner frequency is now 7.2hz, which is close to the upper range of the most commonly used LFO speeds, meaning that the "starting" LFO shape is essentially unchanged across most of the normal speeds.  If the output resistor is increased to 47k to produce least sweep width, that corner frequency drops to about .34hz (one complete cycle every 3 seconds), which will affect the waveform of all but the slowest sweep speeds.

Here is the tricky part...  The "native" waveform of both the Ross and SS is apparently a hypertriangular, or parabolic, waveform, which is close to triangular as it sweeps to the higher frequencies, and more sinusoidal as it sweeps to the lower regions.  How perfectly so, I have no idea, but Sam Hoshuyama assures me it is very close to the ideal waveform for such purposes.  While it is only my personal hunch, rather than anything widely supported empirically, the usefullness of hypertriangular sweeps is very probably confined to slower sweep speeds, where the changes in notch location occur at the sort of rate where decellerating during the more "interesting" parts of the sweep pays off for the listener: a bit like slowing down to drive through a village and speeding up once you exit the village and find yourself back on the highway.  Were one flying in an airplane, "slowing down" over a city provides no real increment to detail observed, and by analogy, changes from rounded to triangular extremes of the sweep cycle probably matter very little when the LFO rate is 3hz and higher.

So, in principal anyways, the addition of the cap to ground may help in terms of the adjustment of sweep width with increasing speed, but it works counter to the desired direction with respect to waveform vs speed.  On the other hand, as I suggested above, the listener probably has a hard time telling the difference between purely sinusoidal, purely triangular and mixed waveform sweep cycles once the speed gets above a certain frequency, so if the hypertriangular waveform is "corrupted" at faster speeds, it won't hurt anything at those slowest speeds where it matters most.

All of this brings me to the final destination, which is that people can experiment for themselves with sweep waveform in both the Ross/Ropez and 6-chip Small Stone, by playing with the cap value.  Having both a variable series resistance, and the choice of a couple of "filter" caps on that control current line, would allow for both crude waveform control as well as sweep width.  Obviously, one would not be able to achieve pure parametric control such that width is completely independent of speed or waveform, but one could generate different "feels" to the sweep without having to radically overhaul the LFO.

Incidentally, I suppose on of the advantages of DSP and digital control is that a single speed control could conceivably alter sweep waveform with rate such that the waveform is always "best" for the given sweep speed.  The Small Stone certainly aimed for that sort of "one-knob wonder".  It didn't fail in that mission totally, but it strikes me that a purely DSP-based unit could more authentically and fully realize that one-knob miracle, producing a phaser that not only adjusts waveform and sweep-width with speed, but regeneration too, such that slow speeds got you wide hypertriangular sweep with lots of regen, and faster speeds got you more symmetrical shallower sweeps with less regen.

[puretube]

So does that end up "rounding off" or triangulating the LFO waveform more?

`t works like Sean stated,
in the case of the hyper-wave, mainly the height of the peaks (which in fact resemble "harmonics" of the sine) get attenuated, leaving the lower part of the modulation-wave ("sine-ish") intact, which overall results in a rounder wave with lower amplitude, at higher rates.

[edit]:just got off the scope, where you had me going...

[Mark Hammer]

Incidentally, the "chop" switch in the old E-H Pulsar does something similar from what I gather. selecting between a more or less unfiltered squarish LFO output, and a more triangular filtered one.  Note that this is a very different LFO than the current production Pulsar, which provides more direct continuous parametric control over the LFO waveform.

[SeanCostello]

Here is the tricky part...  The "native" waveform of both the Ross and SS is apparently a hypertriangular, or parabolic, waveform, which is close to triangular as it sweeps to the higher frequencies, and more sinusoidal as it sweeps to the lower regions.  How perfectly so, I have no idea, but Sam Hoshuyama assures me it is very close to the ideal waveform for such purposes.  While it is only my personal hunch, rather than anything widely supported empirically, the usefullness of hypertriangular sweeps is very probably confined to slower sweep speeds, where the changes in notch location occur at the sort of rate where decellerating during the more "interesting" parts of the sweep pays off for the listener: a bit like slowing down to drive through a village and speeding up once you exit the village and find yourself back on the highway.  Were one flying in an airplane, "slowing down" over a city provides no real increment to detail observed, and by analogy, changes from rounded to triangular extremes of the sweep cycle probably matter very little when the LFO rate is 3hz and higher.

I always thought that the idea was for the sweep to sound consistent in ALL regions, rather than spending too much time in one region. Having a rounded bottom and a triangular peak means that the LFO will spend more time sweeping through the lower frequencies, in terms of time taken to sweep through a given amount of Hertz. However, we don't hear in linear units (Hertz), but rather in log units (pitch, critical band spacings, choose yer poison).  A sweep that is triangular in linear space will sound like it is slowing down in the high frequencies, and zipping through the lower frequencies. To make a smoother sounding sweep, you need to either warp the linear triangle wave into log space (the expo convertor in synth filters and some phase shifters does this) or you need to generate a waveform that more closely resembles a linear waveform warped into log space.

Incidentally, I suppose on of the advantages of DSP and digital control is that a single speed control could conceivably alter sweep waveform with rate such that the waveform is always "best" for the given sweep speed.  The Small Stone certainly aimed for that sort of "one-knob wonder".  It didn't fail in that mission totally, but it strikes me that a purely DSP-based unit could more authentically and fully realize that one-knob miracle, producing a phaser that not only adjusts waveform and sweep-width with speed, but regeneration too, such that slow speeds got you wide hypertriangular sweep with lots of regen, and faster speeds got you more symmetrical shallower sweeps with less regen.

It is a lot easier to gange (sp?) these controls together in digital than in analog. In analog, you may need to resort to voltage control techniques, where in digital you just need to avoid using constants, and have access to the various variables of the algorithm. The tricky thing is that most digital algorithms have tended to err on the side of exposing as many parameters as possible, in order to give the user more power. This is good, but the fact is that you can't fit 40 knobs on a stomp box, and most of the settings of those knobs will sound horrible. So, you do have to make decisions on what knobs to expose, and what they control under the hood.

Sean Costello

[Mark Hammer]

However, we don't hear in linear units (Hertz), but rather in log units (pitch, critical band spacings, choose yer poison).  A sweep that is triangular in linear space will sound like it is slowing down in the high frequencies, and zipping through the lower frequencies. To make a smoother sounding sweep, you need to either warp the linear triangle wave into log space (the expo convertor in synth filters and some phase shifters does this) or you need to generate a waveform that more closely resembles a linear waveform warped into log space.

Actually, it's sort of inverted from that.  The rule of thumb in psychophysics is that as some dimension increments, it takes more physical units of change to elicit a perceptual change.  In other words, it doesn't take much change in notch location at the low end for me to notice the notch location movement, but once you get higher, it takes hreater and greater changes in notch location forme to notice it.

[Processaurus]

So does that end up "rounding off" or triangulating the LFO waveform more?

`t works like Sean stated,
in the case of the hyper-wave, mainly the height of the peaks (which in fact resemble "harmonics" of the sine) get attenuated, leaving the lower part of the modulation-wave ("sine-ish") intact, which overall results in a rounder wave with lower amplitude, at higher rates.

[edit]:just got off the scope, where you had me going...

Thats a nice design feature, because usually people want to turn the depth on their modulation effects down as they turn the rate up, to get the same percieved amount of effect. 

On a side note about LFOs in phasers, I always liked the Small Stone's hypertriangular sweep on slower settings, but never for faster ones.  On the Phase 100, which has a regular triangle LFO, I never liked the slow sweep, but as it goes faster it sounds better than the Small Stone going fast.

[Mark Hammer]

Interesting observation, but in some respects not really a fair comparison.  Since the Small Stone is OTA-based, the LFO waveform has a more direct relationship to the sweep feel.  In the case of the Phase 100, it's LDR-based (well, mostly; there is the trifling matter of those 4 unswept stages ), so the relationship between LFO properties and LDR behaviour is not absolutely direct.  Since they both use sweep width presets, there is still the matter of what they've set the width to internally.  Easy to imagine that the one aimed for a compromise favoring slow sweeps and the other aimed for a compromise favouring fast sweeps.

FWIW, after I came home this evening, I tacked on a 2.2uf cap in my Ropez where the 10uf one was in the Small Stone.  I found the fastest sweep a lot more pleasant to my ears, and a much closer fast Leslie approximation than before.  Note that this also includes a 50k pot in series with a 2.2k fixed resistor instead of the stock 10k current-limiting resistor.

I was working on a 9-stage Ross/Ropez tonight with phase-filter option for the last two stages.  Nine stages?  Yeah.  I liked the idea Ross had with their orange-box 5-stager that added a fixed stage in the feedback loop so it could mimic the regen of a 6-stager going 5 stages back, and figured if 5 is good, 9 has to be 4 better!!  So, nonfeedback goes for 8 stages (last two switchable from allpass to lowpass), and the regen loop passes back to the beginning through a 9th fixed stage to achieve the needed inversion.  I'm going  to stick a 4.7uf cap where the 10uf one goes on the Small Stone and see if that improves sweep feel.  The perf build is actually going well so far, thanks to the ease of layout the 13600s afford and the bonded wire from Small Bear, so I'm looking forward to liftoff sometime this week.  Just hope someone in town has more 27k resistors tomorrow.  Certainly 8-stage vibrato is going to be a little nuts.