Understanding the EHX Y-Triggered Filter

[Mark Hammer]

Bit by bit, I'm repairing and bringing back to functionality a bunch of pedals that have been sitting unused for too long.  One of them is my old EHX Y-Triggered Filter.

It's an interesting beast, with a somewhat idiosyncratic sound.  I think it's a swept bandpass, but you know I'm not all that sure anymore.  It doesn't ever really seem to get thin, the way a bandpass would when swept high.  The circuit, even with the much more readable redraw from Scruffie (thanks, mate!) doesn't look like anything else out there, or at least isn't immediately recognizable as such.

The "Range" switch is essentially two different resonance settings: one high and one low.  Naturally, I just HAD to mod it, by installing a 3-way toggle to change the value of C13, for a higher and lower sweep range.  More amenable to bass now, when C13 = 6800pf.

It has a single bidirectional control that does direction and sensitivity in one knob.  Turn it clockwise and it sweeps down, counterclockwise sweeps up, and no real sweep in the middle.  But the correspondance between sweep and picking is not the 1:1 one finds on many envelope-controlled filters.  It just doesn't feel like a sensitivity control.  Moreover, as you can see below, the sidechain starts out with a clipping circuit, using a back-to-back diode pair that restricts dynamic range.  Now why would you want to do that if you were trying to extract an input signal envelope?  It also triggers weird, occasionally not accepting further plucks.

All of which leads me to realize that the name of the pedal does not lie.  It is NOT an envelope-controlled filter, it is an envelope-generator-controlled filter.  That is, somewhere in all of that (and I honestly can't figure out where), there is a preset envelope/transient generator that is initiated by the input signal reaching some threshold, which triggers a fixed sweep.  There may be something adaptive about it that might yield a changing decay time, but I can't see it.

Actually, I probably lied.  I gather it initiates two equal-but-opposite fixed sweeps, that cancel each other out when mixed 50/50.  Hence the other part of the name: Y-triggered.  Two parallel sweep generators converge on one point, and the knob adjusts their balance.

I cannot for the life of me figure out what sort of a filter topology this is.

So, if you have any ideas about how this beast works, I'd love to hear them.  A search for Youtube demos finds only two.

[robthequiet]

A couple of initial questions -- I take it R28 is your up/none/down sweep control? And unclear about the nomenclature on the switch, "on <-> up"... (Oh, "on" means "down")

Just from the description it vaguely reminds me of a multiplier circuit but is it really a kind of ring mod? What would happen if we feed a sine wave or square wave through it? Deep waters for my limited swimming skills but intriguing nonetheless.

Update -- went and looked at the you-tube vids and found the ehx ad from way back when. So it's an envelope filter,  and the upper part of the circuit is the "rms" and the OTA section is the drive, hence R28 adusts baseline current on OTA1 pin 5 to be modulated by the diff at pins 2 and 3. But the RC networks around U2 threw me for a loop.

My best guess without totally circuit-solving. Maybe the "Y-trigger" has to do with the signal taking a fork at the switch to either the cold or hot side of R27.

The claims that it would make your guitar into a synth were a bit exaggerated, imo.

http://www.ehx.com/blog/ehx-flashback-1976-y-triggered-filter

[Mark Hammer]

I take it you mean R38, rather than R28

"Up" in the drawing refers to the position of the slide switch, when installed.  With the switch in that position, the resonance is higher.

In 1976, ANYTHING that resulted in a filter sweep with each strum or string-pluck was perceived as being synth-like.  Keep in mind that synthesis was not all that prevalent at the time, other than some very big and very expensive installations at university or commercial studios, or in the hands of wealthy musicians.  Guitar synthesis was just an emerging idea.  The very idea of a pedal that could provide filter sweeps for under a hundred bucks was pretty spectacular.

[duck_arse]

that's funny, I've just been playing with up/blank/down. I'd hazard a guesss, under duress, that C5 and C9 might be timing sweeps, being tants and all.

[Mark Hammer]

Worth considering.  I'll try tacking on more capacitance and seeing if that makes a difference.

[robthequiet]

So R38 acts as a mixer/pan control between a low sweep and a high sweep?

[StephenGiles]

It would be interesting to substitute the rectification circuits with the very clever envelope generator in the Space Drum which should match current levels required by the CA 3094.

[jatalahd]

Mr. Duck might be on right tracks, to me it seems Q1 and Q2 are wired as simple switches. When off, the caps across them charge up relatively slowly  and when on, the caps discharge fast through the open switch. If this is the case, the waveform at the ends of R38 could be a mirrored sawtooth (ramp) wave, which, as a control signal to the OTA, could cause a sweep. But still, for me there are quite a lot of things in the circuit that I can't follow so well, especially around the OTA.

If we start to analyse block-by-block from the top left, there is a simple clipping amplifier U1B and a full-wave rectifier U1 (funny to do this for an assumed square wave input). Then I can't really say I could identify the two remaining op-amp configurations, although the one on the far right is again a basic non-inverting amp, with a few added caps around it.

Just sharing my thinking, but usually I take the wrong assumptions on the first go. It is an interesting circuit and I am also keen to know all details of it.

By the way, in the shown schematic the collector of Q2 should not be connected to the collector of Q4, it just does not make any sense.

[robthequiet]

I'm having a hard time interpreting the signal path from the schematic, specifically the input jack.

[Mark Hammer]

So R38 acts as a mixer/pan control between a low sweep and a high sweep?

Well, between two equal-but opposite envelope voltages, yes. 

In something like a Mu-Tron, or any other similar pedal, changing sweep direction is a matter of either adding the rectified signal voltage to some basic standing level (usually zero), or subtracting it from some arbitrary maxmum value.  The rectified signal, however, could vary, depending on picking.   

If my suspicions are correct, and this design uses a fixed envelope voltage, then the middle position of R38 is able to cancel out the two opposite-but-equal envelopes....something a comparable control couldn't do in the usual envelope-controlled filter.

[duck_arse]

+1 on Q2 shouldn't have CE shorted. and, does U1A appear to be ...

adding the rectified signal voltage to some basic standing level

... as mark so eloquently puts it?

[robthequiet]

Sorry if I'm getting noisy on this thread -- at any rate, I'm with Mark on the idea of the fixed envelope. I suspect that it is U2B that releases the stored charge when voltage at the inverting input goes over (or under) V/2. Looking at the size of R9 and all those diodes, it resembles a ramp wave generator. So the wave gets delivered to the base of Q1 and flows to the differential amp formed by Q3 and Q4, which are biased (correction: the current is directed) by R38 (pot, not bias resistor, on second guess) to favor either up or down output.

So the OTA gets the Iabc from R38 wiper and puts input-modulated current out on pin 8 to the R23 and R24 junction into what I presume is the filtering part. I would love to have a live circuit and a scope to help figure this out. Could C17 be the primary filter cap?

What is throwing me now is that the input seems to enter at the left side of C21 which also seems to be connected directly to the output jack. Can someone help to de-confuse me on this?

[jatalahd]

What is throwing me now is that the input seems to enter at the left side of C21 which also seems to be connected directly to the output jack. Can someone help to de-confuse me on this?

The Y-filter stompbox has one pot, one slide switch and the actual stomp-on, stomp-off  foot switch. This foot switch is directly connected to the output jack in the schematic. To me it seems that in the schematic the foot switch has been drawn to the state where the effect is in by-pass mode. Toggle it on and you can find the signal path when the effect is activated.

[Mark Hammer]

Sorry if I'm getting noisy on this thread -- at any rate, I'm with Mark on the idea of the fixed envelope. I suspect that it is U2B that releases the stored charge when voltage at the inverting input goes over (or under) V/2. Looking at the size of R9 and all those diodes, it resembles a ramp wave generator. So the wave gets delivered to the base of Q1 and flows to the differential amp formed by Q3 and Q4, which are biased by R38 to favor either up or down output.

So the OTA gets the Iabc from R38 wiper and puts input-modulated current out on pin 8 to the R23 and R24 junction into what I presume is the filtering part. I would love to have a live circuit and a scope to help figure this out. Could C17 be the primary filter cap?

What is throwing me now is that the input seems to enter at the left side of C21 which also seems to be connected directly to the output jack. Can someone help to de-confuse me on this?

Another vote for a fixed envelope, that should have dawned on me earlier: NO ENVELOPE RIPPLE.  Even the very best envelope-controlled filters, that use full-wave rectifiers, and optoisolators (both of which reduce ripple) have audible ripple.  If there is no ripple, that HAS to mean the envelope is not coming from an unstable source, like the guitar itself.

Which, if you think about it, makes the ad copy claims about sounding like a synth more legitimate.  It also make wanting to understand the circuit and know more about what is setting the envelope parameters.  Being able to vary the attack/decay in the unit would be VERY cool.

[Scruffie]

I fixed one recently and have voltages if that'll help you get yours back to life Mark

Also, that schematic, while I redrew it one to one to the schematics I had (and there were 2, they differed around the Q1 section if I recall) did have some mistakes compared to the actual unit (for a start the V.Ref is shorted out) so take it with a pinch of salt.

Also without any studying so this may possibly be a stupid statement, the clipping may just be an attempt at poor mans compression to reduce ripple.

The one I fixed had the same non-response to picking as you mention too for what it's worth.

[jatalahd]

Mark, regarding attack and decay, did you try to swap the tantalum caps C5 and C9 as pointed out by Duck_? This is just a hunch, but I would say attack times for both sweep-up, sweep-down control signals are equal (about 15 milliseconds), determined by (R13+R19)*C5 and (R12+R14)*C9. For corresponding decay times my guess is R19*C10 and R11*C8.

Waiting for someone to challenge me on this

[StephenGiles]

Mark, regarding attack and decay, did you try to swap the tantalum caps C5 and C9 as pointed out by Duck_? This is just a hunch, but I would say attack times for both sweep-up, sweep-down control signals are equal (about 15 milliseconds), determined by (R13+R19)*C5 and (R12+R14)*C9. For corresponding decay times my guess is R19*C10 and R11*C8.

Waiting for someone to challenge me on this

............possibly!!!!!!

I currently have no chance of breadboading this because my wife hopefully returns home from 2 1/2 weeks in hospital with trench foot. No that's wrong - should be pseudomonas infection and cellulitis (look it up, interesting pictures!!) so nursing duties when at home!!

So I'm viewing this from the hospital via Teamviewer to my home PC on the recently upgraded "free" wifi!

Where does the trigger circuitry end, I can't get my head around this - is it pin 1 of U2, or is U2 and associated components just a conditioning filter for the trigger pulse which energises the offending "envelope generator", which, unless I am persuaded otherwise, should be replaced with something more useful with adjustable attack and decay times.

[jatalahd]

Where does the trigger circuitry end, I can't get my head around this - is it pin 1 of U2, or is U2 and associated components just a conditioning filter for the trigger pulse which energises the offending "envelope generator", which, unless I am persuaded otherwise, should be replaced with something more useful with adjustable attack and decay times.

My current understanding is that, yes, all the 4 op amps in the top row are used for generating the trigger signal and pin 1 of U2 delivers short trigger pulses to the base of Q1.

Downward sweep: In steady state, the base of switching transistor Q1 is at V_B, hence the switch is closed and the cap C5 is fully charged (it charges inversely, from V_B towards 0 volts), collector of Q1 is close to 0 V. When the trigger pulse arrives (assuming it is a downward pulse), Q1 conducts and discharges C5, then collector of Q1 rises to V_B. This would be an instant voltage rise, but it needs to charge C10 through R19, which slows things down. C10 pushes base of Q4 towards V_B and the emitter follows the voltage rise of the base (VBE diode is free to move because of R20). After the trigger pulse has ended, Q1 is closed and C5 charges towards 0 V through R19 and R13. This reflects to Q4 as just explained and this causes the prominent sweep (V_B towards 0 volts)

Upward sweep: at the same time when the trigger pulse affects base of Q1, it is seen inverted at the collector of Q1. In steady state, the base of Q2 is at ground and the switch is closed, C9 and C8 are fully charged to V_B. The inverted trigger pulse is now upwards pulse and it triggers base of Q2 through C10. C9 discharges immediately through Q2 and this puts base of Q3 to ground, Q3 is open and lets C8 to discharge through R11. After trigger pulse ends, Q2 is closed, potential rises so that Q3 is closed and C9 charges from 0 to V_B through R14 and R12.

So C5 charges from V_B towards 0 V, and C9 from 0 V towards V_B, simultaneously. The RC paths have the same RC time constants. These are the two opposite control signals, which cancel each other at the middle of R38 (the sum of voltages stays constant in the middle of R38).

Sounds good, but I am not sure    ... Actually just realized that C8 would charge up faster than C9, which makes a small hole to my theory Back to thinking mode.

[StephenGiles]

Thank you, something to think about!

[robthequiet]

If I had one of these puppies I would test the difference between a series of long soft input signals vs. a short large signals, to test the charging/triggering rate. In other words, can the triggering be manipulated by the input or does it automatically draw a charge from the power supply, or both? The Youtube videos seem to indicated that it has a very smooth release, perhaps accounting for the overcomplicated (hat tip to StephenGiles) circuitry up top.

I have a question about the schematic -- the sleeve of the input jack goes to ground, but also seems to go through R30/470K to a ground point near the output jack switch. Kinda the way Q2 appears to have its collector and emitter tied to ground. Or am I misreading the schematic symbols?