Understanding the EHX Y-Triggered Filter

[mookyj]

I'm finally done with the PCB for this. I need some help - is this ground plane done well or not? I don't have ground loops, but I wander if it is OK regarding current flow. Please help!



Single layer, with two SMD capacitors (for PS filtering right under ICs).

in reviewing what you have here:

You have it grouped by circuit function rather well. I would go back and review your placements within each grouping as they can be improved. That is always the case in 2nd and third looks, easy to say after it is done. I have marked up a few that you may find helpful. I would strongly consider moving the power input section the the upper left of center. Think about current flow as you you supply each circuit. Not a fan of single layer for reliability (non plated holes) and for lack of best impedance paths for signal flow. shortest return path is underneath forward path. Shortest distance is not always shortest impedance return, split planes is an example of that.  In the layout it does not have a short return path in bypass mode as return is forced up and above the switch and led circuit. For single layer you do provide much better return path that typically seen in some designs. If staying with single layer you can improve it some, that said with the observation you have applied plenty of effort to create more gnd pour in the layout.  Personally, I would default to a bottom layer ground plane would allow you to improve the V+ rails  and still allow a coplanar top ground pour also for much of your circuit.

[rankot]

Thanks a lot, mookyj! I tried to keep power section close to input jack, because I use it's middle contact to disable power if pedal is unplugged. Is it really necessary to move power into top left corner? I use single side PCB because it is complicated to me to order dual layer from abroad, and I can get single layered locally. You're right, I really shouldn't be grouping input and output signal capacitors near each other. Is this better:

[mookyj]

Thanks a lot, mookyj! I tried to keep power section close to input jack, because I use it's middle contact to disable power if pedal is unplugged. Is it really necessary to move power into top left corner? I use single side PCB because it is complicated to me to order dual layer from abroad, and I can get single layered locally. You're right, I really shouldn't be grouping input and output signal capacitors near each other. Is this better:

It is incrementally better for sure.  As to disconnecting when unplugged, if not using a battery wouldn't matter to switch it.  I get why on a single layer, but I think I would try both and you can determine for yourself.   I would prefer not to return the power supply current back along the output. People do it, but I would be concerned with picking up noise, disturbance in any kind of gained pedal. With a battery or a very well filtered pedal board supply, I would say no worries. Anything coming from the wall wart could end up in your audio even with the filtering you have on board being so close to the output.  You have done a lot with a single layer.

[rankot]

I would prefer not to return the power supply current back along the output.

Now I really need some help

[mookyj]

I would prefer not to return the power supply current back along the output.

Now I really need some help

it is easier than you might surmise, what would normally be switched by the jack would go to ground, move the series diode to the top as I mocked it up and that will straddle the trace going to the 1K Resistor, The lower circled section would need to move down some and you have plenty of space to do it. By rearranging the placements of the filter section after the power jack you will minimize the space needed. you could turn the jack upward it you want.

 

[rankot]

What about this solution - I left jack switching and moved jacks and 3PDT sw. upwards, so there's enough place for 9V battery now. Connected ground using one jumper, because it follows V+ current. OK?

[mookyj]

What about this solution - I left jack switching and moved jacks and 3PDT sw. upwards, so there's enough place for 9V battery now. Connected ground using one jumper, because it follows V+ current. OK?

My answer may frustrate you, but I would ditch the battery.  THe long switching routes essenstial distribut any noise from the wall wart along your anoalg sections. Kudos on clean work, I do like it.

Mike

[rankot]

Well, this could be the final version. What do you think, shall I remove the jumper and connect the ground plane here (red mark), or not?

[mookyj]

Well, this could be the final version. What do you think, shall I remove the jumper and connect the ground plane here (red mark), or not?

I would move the jumper closer together and connect the right side ground. If you find any issue with it, it is easily cut/scored and peeled.

Mike

[Mark Hammer]

I'm about 2/3 of the way stuffing the board I etched the other day.
I dug my own unit out of the basement, and dickering with it just now, several things are clear to me:

1) It really needs a sensitivity control.  The unit is NOT an envelope-controlled filter that corresponds directly to picking strength, but rather employs a triggered envelope.  Triggering is not entirely independent of input level.  So something needs to be adjustable around U1.1 to allow it to retrigger reliably with both higher input levels, and lower ones.  You will note that R3/C1 roll off the bass sensitivity of that stage around 1500hz.  Changing the value of R3 to adjust the gain monkeys around with that rolloff.  So, in the absence of any strong rationale for using it, I'll suggest replacing R1 with something like a 250k variable resistor in series with 100k fixed resistor, to vary the gain of the stage between 22.3x and 75.5x (stock is 58.4x)

2) A filter range switch is useful.  On mine, I installed a 3-position toggle.  I replaced the stock 2200pf cap (equivalent to C22 in the redraw) with 2200pf and 6800pf in series.  The switch bridges the one or the other or maintains them in series, yielding three ranges.  Thinking it over I should have probably used 3300pf and 6800pf, since the highest range when using 2200 and 6800 is useless.  Putting 3300 and 6800 in series gets you 2200pf, which returns a stock sound.  The unit sounds really nice on lower ranges.

3) We know a transient is generated, and that two versions are produced: a rising and a falling version, equal in amplitude/sweep, and cancelling out when in 50/50 mix.  What we don't know - or at least I can't for the life of me figure out from just looking at the drawing - is where that transient is produced.  It would be nice to know a little more about how the time constants are determined.  Being able to vary the rise or fall time would be nice.  I'll haul out my scope later tonight and see if I can spot where appropriate intervention points might be, as well as learning a little more about the transient that is normally generated.

4) The Range switch is also a quirky one.  It sounds like it selects between two resonance settings.  It could really use a 3rd setting in between the two provided.  The stock arrangement provides a high-resonance setting, and another that is so bland it deters use.

I don't want to overcomplicate it, but this could be a much friendlier and more maleable filter pedal with just a couple of tweaks.  If it had variable sensitivity, the stock up/down control, a 3-position filter range switch, a better selection of resonance settings, and a 2 or 3-position toggle for fast/medium/slow sweep, it could be a really nice pedal.  Let's get this done.  Meanwhile, I'll continue stuffing my board and see if it fires up.

[Mark Hammer]

At the wiper of the Up/Down pot. the DC voltage  is high then momentarily swings low and back up again when sweeping up.  In the down-sweep setting, the DC voltage at the wiper is low, jumps high in response to triggering then returns low again.  NO movement when the control is centred.  NO great surprise there.

But the big surprise is that the schematic shows 4 transistors, and yet my board has only three.  And my personal big surprise is that the redraw actually includes a Speed switch.  Duh!  On mine, the default electro setting the time constant is 5uf.  Strapping in 10uf in parallel yields a modest, but still noticeable slowing of sweep that works well in both up and down directions.  I tried 33uf, and it worked fine in upward sweep but lousy in downward, for reasons I don't quite get.

[rankot]

But the big surprise is that the schematic shows 4 transistors, and yet my board has only three.  And my personal big surprise is that the redraw actually includes a Speed switch.  Duh!  On mine, the default electro setting the time constant is 5uf.  Strapping in 10uf in parallel yields a modest, but still noticeable slowing of sweep that works well in both up and down directions.  I tried 33uf, and it worked fine in upward sweep but lousy in downward, for reasons I don't quite get.

Mark, I've traced this using your photos, and my schematic use only three transistors. There are two extra ones used to provide transistor output for LM13700 which is missing compared to CA3094. It could be also done with one transistor (using LM13700's buffer), but Scruffie said that he gets better results with two transistors, so I left the final schematic that way.

However, there are some photos available showing some kind of mod with the fourth transistor, but I didn't have time to trace that too.

Could you, please, confirm resistor values asked before (see schematic at https://www.diystompboxes.com/smfforum/index.php?topic=113930.msg1167164#msg1167164), resistors  R2, R4, R5, R6 value. They seem as 1k to me (if photograph colors are right), not as 10k as shown on original schematic.



And this is the schematic with Mark's mods, if I understood him well:



Both filter/speed switches shall be on-off-on type.

[Mark Hammer]

1) The photo of the copper side with the additional transistor is not from my particular pedal.  But there are several components soldered to the copper side on mine, so it is not unreasonable to think that EHX  modified the design and tacked on the additional transistor.  We need to remember that the pedal was made when there was NO layout software.  You would work it out on paper and then use rub-on tape to make the pattern on a transparency, photograph it or otherwise turn it into an image, and use that to make boards via a photosensitive method.  So there were "disincentives" to changing the layout, with add-on components being the preferred choice for smaller producers.  EHX was certainly "big" at the time, but nowhere as big as they are now.

2) The suggested mods to the circuit are generally faithfully reproduced in your drawing, but with a few changes needed:

• I suggested a 100k+250k pot for Sensitivity, based on reading the schematic as having a 270k fixed resistor in the feedback loop.  When I looked at mine, it was a 390k resistor, which would recommend use of a 180-220k fixed resistor, in conjunction with the 250k pot, to get a little more and a little less sensitivity.
• You show C27 (4u7) and C13 (10u) as additional caps to place in parallel with default value C12 (4u7).  The difference between 4u7 and 4U7+10u is just big enough to be perceptible as a slower response.  The difference between C12 alone and C12+C27 would be negligible.  If you wanted to make C27 an even larger value, like 22u, it might be justifiable, but I can't see the addition of a parallel 4u7 cap worth the effort.  It might be best to just stick with a 2-position slow/fast switch.

I tried looking for the possible shape of the transient produced last night.  But my scope skills, and the capabilities of the scope itself seemed to be insufficient.  I could see when and where it went high or low, but not the entire transient from start to finish.  So I can't tell you much about it, unfortunately.

[MonkeyKult]

What a coincidence, I just came across an old Y-Triggered Filter that was in pretty bad shape, and I just finished restoring it just yesterday. It was tricky, cause there were a lot of discrepancies between the circuit on my pedal, and what I’ve read in this thread and elsewhere. I tried my best to piece together and here is what I found:

1. First, the similarities; I do have an additional 5 uF cap on the back side in parallel with the 10uF (not a tantalum, despite the schematic) to give 15 uF total across the resistor. This seems to work well.

2. Mine only has 3 transistors; 2 PNPs and 1NPN. This was a cause of a lot of confusion, as it was hard to reconcile these. Not sure which is correct.

3. Most notably, the circuit across the 4558 around R1 is totally different on my version. Instead, my R1 is 47K and there are no diodes in parallel. At first I thought this was a mistake, but I must have one of the first versions cause the resistor values are printed on the CB. So this is how mine is configured.

Lastly, mine needs to have a battery and a 3.5 mm DC plugged in to work properly.

Otherwise, I have to admit this sounds pretty good. I changed all electrolytic capacitors and the 1 tantalum. I’m pretty pleased, though it was a challenging rebuild.

[rankot]

If I can see well, those mods with a transistor from the image I posted before, should look like this, although I'm not so sure about resistor values (and transistor is 2N5087 actually):



Tried this mod with ltspice, but doesn't seem to do much.

[Mark Hammer]

Congrats!  It's a fun pedal.  It tends to be confused with the DR. Q, that came out around the same time, but behaves quite differently.

Mine has the 3.5mm phone jack as well. I replaced it yesterday with a more standard 2.1mm barrel jack.  Wasn't easy machining the hole to make it bigger.  I think over the years I've replaced a few of the non-polarized capacitors as well.  I also replaced the original stompswitch with a DPDT, but now that I look at it, I think I'll stick a 3PDT in there, with a status LED.  If it was a rare beast, I'd be concerned about its "vintage" properties.  But I'm confident that it was produced in sufficient numbers that mine doesn't need to be preserved intact as a museum piece.

[rankot]

Hmmm, shorting E&B of this additional transistor DOES make some change to the sound. I don't have this on a breadboard, but ltspice shows there's something going on. When they're not shorted, this mod seem to do nothing.

[Mark Hammer]

So, I have Rankot's board etched and stuffed, and modded (naturally) to have slow and fast time constant and three different filter ranges (two additional below stock).  I used 2700 and 5600pf in parallel with the stock 2200pf to get a medium and low range.  Sounds good.

Once I figured out that BC327 have the opposite pinout of a 2N5087, everything works as it should, except that the sound is very distorted and very dull.  I also need to turn my guitar volume down to about half or less to get reliable triggering.  It is possible that I may have mistaken values - the curse of using 1/8w metal film resistors. But I also wonder if the components around the LM13700 are appropriate.  The odds are that it is MY mistake, rather than anybody else's.  BUt I just thought I'd put it out there.  It would be nice to finally have this off the bench, so I can get to other things.  In the meantime, I'll go over the component values.  I'm usaing the layout posted here:

Both sides separated:

[rankot]

I'm still waiting for my PCB to arrive, our post is at strike for two weeks

However, you may check resistor values which I was not sure about, as I wrote few posts above:

Could you, please, confirm resistor values asked before (see schematic at https://www.diystompboxes.com/smfforum/index.php?topic=113930.msg1167164#msg1167164), resistors R2, R4, R5 and R6. They seem as 1k to me (if photograph colors are right), not as 10k as shown on original schematic.

I did use 1k on my layout, maybe they should be 10k?

Also, what you used is not the last version of my layout, there could be some mistakes there as well. I did finish a layout with some "bells and whistles", however, didn't want to publish it until I build it and make sure it's working. However, here it is, for reference, together with appropriate schematic:

[Scruffie]

10k.