mini filter sample + hold schematic

[nocentelli]

i breadboarded a cut down version of the ggg schematic of the maestro filter sample hold (no envelope mode) and made a few small changes, and i've drawn up a schematic ahead of attempting a vero. I wonder if some kind person could cast a second pair of eyes over it, perhaps point out any glaring errors, or even suggest improvements...

i've never got the FSH to fire up properly before now, so i'm very pleased at the moment, but the changes are pretty simple:

- Swap out the FET buffer between the OTA stages and replace with unused darlington buffer
- Swap collector and base on noise transistor
- Limit maximum voltage from the 100k bias trimmer with a 47k resistor

The noise transistor pin swap seems to make it much easier to dial in the random voltage blips (thanks to aquataur who suggested this http://aquataur.hilpold.net/aquataur/musicstuff/fsh-1.html). The 100k bias pot now acts more like a purposeful frequency range control: 12 o'clock gives a full range of frequency blips, turning it down progressively lowers the maximum frequency of blips (which can produce a wonderful burbling squelchy bass sound with a sub-octaver), and turning it up pushes the voltage blips to the upper frequencies. The extra resistor just allows a more useful range from the pot.

[Kevin Mitchell]

I'm surprised to see no feedback here. I gave up on my FSH1 after noticing some bad capacitor and resistor substitutes I have made when I built it. Though I never gave up trying to recreate the filter, sample/hold effect.

The schematic shared is a great idea and should definitely be applied for a more up-to-date FSH-1 workalike. I did try to breadboard the circuit above and noticed that at the 330n and 470k of the S/H U1B have the connections flipped around. So the 330n should go to -9v and the 47k to ground.

I do believe I had an error on the breadboard. The guitar signal dropped dramatically at the 1st - input of the lm13700 (after the 100k). This could be an error on my part and not a mistake in the schematic. I'll share my experiences when I'm back at it. I then referred back to the original schematic and used a Jfet as a buffer once again. After some more tinkering I've got it running! The noise transistor is very crucial. About half of the 2n3904s I've used had no effect as a noise source. I found one decent one and used pins E and C as the original schematic suggests.

I'm in the middle of a few projects but putting together a *better* FSH1 is for sure on my list. Thank you, nocentelli for sharing your prototype. I have a few ideas of where I'd like to expand on this schematic.

[nocentelli]

Thanks for the feedback. I went ahead and drew up a vero (with the correct arrangement of resistor and cap in the LFO)and built it. However, it is plagued with a weird problem: On powering up, it works fine for a few minutes and then the effect slowly degrades over the next few minutes. By "degrades", I mean the clarity of the filter blips becomes slowly less pronounced, and the guitar signal under the filter seems to become grain-y and lo-fi. Unplugging the power and shorting the power rails "resets" the effect, but soon after powering it back on the same thing happens. I haven't had the energy or inclination to debug (I'm not even sure what I would be looking for, to be honest) but if anyone can see a problem with the schematic - apart from the 47k/330n / ground / -9v mix up that kevin points out - that migjt cause this issue, I'd be very grateful. I don't recall the circuit displaying this behaviour on the breadboard, so it may just be an error in my vero layout.

[frequencycentral]

Have you tested the voltage at the charge pump when you get the grainy?

[nocentelli]

No, but I feel inspired to dig it out and have look now.

[Kevin Mitchell]

I recall having a similar issue when looking for my errors in the noise section on breadboard. Though it wasn't a recurring problem as I sifted through and swapped jfets and transistors. I've read that the operational requirements for the 2nd jfet (getting +9 and -9) is a bit crucial. Likely depending on the circuit in it's entirety. I recall swapping that one out a few times before being satisfied.

Put an audio probe to the noise transistor and follow the signal until you find the error. You should hear the hold effect at the jfets and the pnp filters out some of the noise I believe. So start with the noise transistor or follow CV to the transistor. It's likely a short or a need for a tryout for what works best in the two positions I have mentioned (noise transistor and Q6).

I'll fix up the circuit as per your drawing and share some measurements if it would be of any use to you.

[ElectricDruid]

I do believe I had an error on the breadboard. The guitar signal dropped dramatically at the 1st - input of the lm13700 (after the 100k). This could be an error on my part and not a mistake in the schematic.

An OTA like the 13700 can only cope with very small signals (a few mV) so it's typical to see hefty potential dividers on the inputs, as here (100K/3K3). I'd say what you saw was probably correct.

Tom

[Kevin Mitchell]

An OTA like the 13700 can only cope with very small signals (a few mV) so it's typical to see hefty potential dividers on the inputs, as here (100K/3K3). I'd say what you saw was probably correct.

Last night I went ahead and omitted the jfet buffer, used pin 7 and 8 of the lm13700 and changed the 12k to 22k - as per the schematic in the original post. It works with really no sound difference to my ears. So it must have been a breadboard issue. I think I had a problem with a capacitor to ground during my first attempt though I'm not positive.

Anyways it's now set up exactly like the schematic above. I'm working on the envelope section now.

There are a couple of things that I'm experiencing that I've been trying to improve. The resonance pot works so subtle I'm not even sure it's doing much at all.
I don't like to mess with the depth or the bias pot since they in a way set the frequency and consistency of the sample/hold effect. If I mess with them I find myself undoing my adjustments and trying to keep them at the sweet spot. I think the depth pot in a way also sets the center frequency of the effect. The sound of the guitar at certain frequencies isn't pleasing to me.
EDIT: It seems that after adding in the envelope trigger section the guitar sounds a bit better. Also the resonance pot seems to act a bit better. I'm sure the conclusion may be obvious to others but not myself. Maybe this has to do with current load of the circuit all together though I didn't think this would effect the guitar sound. Perhaps the load of the original circuit (as a whole) helps set the *neutral* level of CV (or rather - pin 1 and 16 of the lm13700)thus causing the unpleasant balance I've mentioned when the envelope sections is omitted as per the schematic in the OP

Here's a sample of the circuit in the first post. Only difference is using pin E and C of of the npn and also I've added a LED -> 1k5 -> gnd from the output of IC1B. I covered speed, depth and bias in that order.

[ElectricDruid]

Are you sure there shouldn't be a resistor to ground from the positive input of U2A? That's fairly unusual to not have it. The second OTA stage has the 3K3 from both inputs, notice. But the first one doesn't. The output from the filter could be reasonably hot.

Dunno if this is a problem. It just struck me as a bit unusual is all.

HTH,
Tom