Help with JFET tremolo idea

[MrStab]

Hi guys,

I've been trying various things to try and "design" a straightforward, useful, minimally-distorted tremolo that uses JFETs to cut the signal in and out. why JFETs? well, LDRs are apparently banned (in new products) by RoHS, and using an OTA resulted in a right-royal thump, which i couldn't get rid of. plus i like the choppiness of a good ol' transistor. i have it more or less working with a JFET, but i have a few questions.

so i have this basic setup, a JFET pulling the signal to ground in an AC-coupled part of the signal path, between two opamp buffers (one will become a gain stage):


the Gate of the JFET is hooked up to an LFO i took from this Stompboxology article:


www.moosapotamus.net/files/stompboxology-mo-tremlo.pdf


(B=vref, buffered from a spare opamp, as suggested in the article)

so i have a few questions.

-the square wave is strong and works well, but doesn't cut the signal completely. also, the triangle wave is too weak. i've read that the tri being weaker is a common thing, so i'm not too worried as i could attenuate the square later on (which i assume R1 on the LFO schematic does?), but how could i increase the intensity of the effect? R2 on MY schematic is currently 4k7 iirc, but lowering it seems to have minor benefits, and it doesn't work at all with no resistance. R1 on my schem is a 10k pot for "Depth", but adding more resistance lessens the intensity, as you'd expect. attenuation of the signal just before the JFET, as seen in the input of the Schaller Trem (which uses a NPN), doesn't help much.

-am i right in thinking, based on a poor understanding of other posts i've read, that the vref in the LFO controls symmetry/duty cycle? i had issues with symmetry, but a trimpot & LM324 in place of a TL074 seems to help a lot. i think i read that FET opamps aren't ideal for LFOs. could've just made that up, though.

-in relation to the above: say i do have to tweak the LFO bias in each unit - if i'm using another opamp to invert the LFO for a stereo trem, would it be okay to use the signal path's vref? i'm guessing using the uneven bias would be bad for that.

-do you guys think there's any significant advantage to having both sine(ish) and triangle options? i have both working, but both are subject to the same problem vs. the square wave. personally, i wouldn't mind either sine or triangle as in this application, i don't sense much practical difference.

-any other comments/suggestions?

a large part of this problem is that i'm not being calculating enough with my use of JFETs. i just can't figure out how to transplant the system from circuits such as the EA Tremolo or Kay Tremolo into this one. a NPN worked great, for the record, but it distorted in an ugly way. so any help in increasing my understanding of JFETs, via. this scenario, would be really appreciated.

cheers!

[amptramp]

Univox used bipolar transistors in this circuit:



Look at the lower left of the schematic and you will see a phase-shift oscillator coupled to a bipolar transistor that varies the loading on the input.  A JFET is a depletion-mode device meaning that the gate has to be biased below ground for anything to work.  A MOS enhancement-mode device like a 2N7000 would be a better bet since the bias can be within the rails.  The jack in the diagram is for a switch that cuts off the tremolo.  Note that the level control above also varies the bias on the transistor so the input cut gets deeper as the tremolo depth is increased.

[MrStab]

thanks for the schematic, Ron. any chance you have a link to a bigger image? i can see the outlines upon zooming in, but i can't figure out the values & how they relate to one-another.

i tried a 2N7000 at one point in my rudimentary setup, but no doubt it could benefit from the added circuitry. the problem i had was that it was "gated" almost, ie. no smooth transition as the LFO wave ramped up.

what's the other BJT for? the one furthest bottom-left.

thanks!

[MrStab]

FWIW, i tried a 2N7000 with the signal going from Drain to Source (instead of to ground) at one point, and it worked quite well, but still had traces of the problems i'd encountered the usual way (and the turn-on threshold). as most non-opto tremolos vary resistance to ground, i figure that's the "right" way to do it.

[midwayfair]

See if you can find CultureJam's Stutter Tremolo. There's a way to build it with "shunt mode" that completely cuts out the signal.

[MrStab]

the only vague lead i could find on the Culturejam Stutter was on a site that requested payment to view the whole document, unfortunately. everything else i can find relates to Shoot the Moon. thanks, though, i'll keep an eye out for it. does the Stutter use a LDR? i was so annoyed when i found out i couldn't use em for sale, because they worked really well, even in square mode (trems should be in the FX loop and nowhere else!). i even emailed a company a coupla weeks ago to see if their RoHS exemption was still in place - no dice, it expired at the end of last year.

i built the EA Trem a long time ago, and i can't actually remember if it was full on-off, so it's possible i have been successful in using the JFET in a similar fashion to that circuit, and my expectations are off. one thing i've noticed is that there's a cap between the EA's LFO output and the FET, but in my circuit, higher cap values made the effect weaker whilst lower ones killed it altogether.

i've just realised i erroneously said the Kay Tremolo uses a FET - it doesn't. interestingly, it uses both a PNP and NPN to shunt the signal. i've heard the Kay distorts too much, though, which is not o-Kay for what i want to achieve. ha. also just realised the other BJT in Ron's link was probably the LFO itself. just had to clear up my misinformation.

i'm all out of ideas. you wouldn't think that, from the wall of text i've just posted. lol

[midwayfair]

the only vague lead i could find on the Culturejam Stutter was on a site that requested payment to view the whole document, unfortunately. everything else i can find relates to Shoot the Moon. thanks, though, i'll keep an eye out for it. does the Stutter use a LDR?

Stutter's all transistors.

Just PM CultureJam, I'm sure he's got the document somewhere.

[samhay]

Q1 is one half of a voltage divider. Where's the other half? Add some (more) series resistance between C1 and Q1 D.

[merlinb]

LDRs are not banned from new products. ROHS exists primarily to stop dangerous material entering landfill, but effects pedals are the sort of thing which never goes into landfill. An enforcement body would have an uphill struggle to convince a court that one or two LDRs in an effects pedal could constitute an environmental risk. They're still used in many non-disposable audio products, and probably always will be.

[R.G.]

I've been trying various things to try and "design" a straightforward, useful, minimally-distorted tremolo

Part of your difficulties stem from a misconception that non-distorting tremolos are straightforward. They're not. Smooth, clean volume modulation done quickly is hard to do well.

that uses JFETs to cut the signal in and out. why JFETs? well, LDRs are apparently banned (in new products) by RoHS,

LDRs are banned in the EU by EU's RoHS, which made many manufacturers stop making them. Some still do, but they're not sold in the EU.
This is only an issue if you are (a) in the EU and (b) want enough to make pedals to sell as a commercial enterprise. If you want to make one pedal for yourself, there are ways around this.

and using an OTA resulted in a right-royal thump, which i couldn't get rid of.

As I said, good, fast amplitude modulation is hard. OTAs have problems with noise and signal levels. OTAs which show thump are modulating the DC level as well as the AC signal. It's a design problem to be worked on.

plus i like the choppiness of a good ol' transistor. i have it more or less working with a JFET, but i have a few questions.
so i have this basic setup, a JFET pulling the signal to ground in an AC-coupled part of the signal path, between two opamp buffers (one will become a gain stage):
the Gate of the JFET is hooked up to an LFO i took from this Stompboxology article:
so i have a few questions.

-the square wave is strong and works well, but doesn't cut the signal completely.

That is a problem because you have committed the common mistake of not including the source impedance feeding your circuit and load impedance fed by your circuit as parts of the circuit. They're almost never shown on schematics, but they're part of the circuit nevertheless.

In your case, you can make the square wave cut the signal more completely by including some resistance in series with the input. The way a shunt impedance (i.e. your JFET) works is by forming a voltage divider with the source impedance of the signal that drives it. If your JFET turns on to a resistance of, say, 200 ohms, which is common for JFETs, then what you get for signal reduction might be:
- almost nothing if the input is fed out of a low impedance opamp output directly
- only 1/5th if the input is fed from a 1K source impedance
- 1/100th (-40db) if the input is fed from a 20K source impedance.

The moral of the story is to put a resistor in series with the input to get good shutoff.

also, the triangle wave is too weak. i've read that the tri being weaker is a common thing, so i'm not too worried as i could attenuate the square later on

Triangle is "weak" because even if the tremolo was working perfectly, the amount of time spent by the signal being full loudness with a rectangle LFO is larger than the time spent being loud with a triangle. Your ear integrates this time spent loud versus quiet and your brain decides that on average, the triangle wave is quieter. The answer is simple - you can't get equal loudnesses with square and triangle wave modulation with only an LFO. You have to amplify the triangle ouput up or attenuate the rectangle output down until they're equal apparent loudnesses. Note that since you're trying to get the maximum loud-to-quiet ratio from your LFO, you also can't do this with your LFO. It's best done cleanly by changing the signal gain or attenuation with a second switch section when you switch from rectangle to triangle (or sine) waveforms in the LFO. Remember - fast, good, cheap: pick any *two*.

(which i assume R1 on the LFO schematic does?), but how could i increase the intensity of the effect?

As above, diddle with the series resistance on the INPUT and change the after-trem gain with a gain switch on the after-trem amplifier.

R2 on MY schematic is currently 4k7 iirc, but lowering it seems to have minor benefits, and it doesn't work at all with no resistance. R1 on my schem is a 10k pot for "Depth", but adding more resistance lessens the intensity, as you'd expect. attenuation of the signal just before the JFET, as seen in the input of the Schaller Trem (which uses a NPN), doesn't help much.

-am i right in thinking, based on a poor understanding of other posts i've read, that the vref in the LFO controls symmetry/duty cycle?

No. Vref in most integrator-schmitt trigger setups (this is one) controls the middle/DC level of the waveform. Symmetry is controlled by the peak voltages of the output of the schmitt trigger with relationship to the Vref. In practice, if you're doing the simple thing of letting the schmitt trigger section bang against its power supply limits, then yes, moving vref unbalances the symmetry by effectively unbalancing the center point versus the power supply limits of the schmitt, but you can achieve a balance with a variable vref by limiting the highest and lowest output levels of the schmitt trigger with diodes or zeners to something smaller than the power supply limits. The vref will then move the DC level and the peak to peak will remain the same.

i had issues with symmetry, but a trimpot & LM324 in place of a TL074 seems to help a lot. i think i read that FET opamps aren't ideal for LFOs. could've just made that up, though.

Neither the TL074 nor the LM324 are great for the schmitt trigger in this. If you want full control, you have to not rely on the power supply limits of the opamp doing the schmitt trigger, and insert limits smaller than the power supply after the schmitt trigger, adjusting the integration RC time constant to get your frequency back down.

-in relation to the above: say i do have to tweak the LFO bias in each unit - if i'm using another opamp to invert the LFO for a stereo trem, would it be okay to use the signal path's vref? i'm guessing using the uneven bias would be bad for that.

For simple, cheap designs yes. Design for manufacturing in the face of component variation is a whole field of study in itself. If you're designing for making and selling multiple units, yes, you'll have to tweak every one (and deal with the boy-genius-customers who immediately open up the box and twiddle the pots     )  or do a more sophisticated design that is in sensitive to component variation. 

-do you guys think there's any significant advantage to having both sine(ish) and triangle options? i have both working, but both are subject to the same problem vs. the square wave. personally, i wouldn't mind either sine or triangle as in this application, i don't sense much practical difference.

Both have the same issue vs square wave, and the same solution, as noted above.

a large part of this problem is that i'm not being calculating enough with my use of JFETs. i just can't figure out how to transplant the system from circuits such as the EA Tremolo or Kay Tremolo into this one. a NPN worked great, for the record, but it distorted in an ugly way. so any help in increasing my understanding of JFETs, via. this scenario, would be really appreciated.

Everyhing is linear (i.e. nondistorting) if the signal is small enough. There are tremolos that use NPNs just fine - they just first attenuate the signal down to less than about 25mV peak, do the modulations, then amplify it back up to full level. Bipolars have their problems as modulators. So do JFETs, just different problems. JFETs have issues with the minimum resistance they can get to, and distortion for signals more than Vth, which is very close to Vgsoff from the datasheet. JFETs can be moderately "cured" of this issue, for signals up to about 3V peak by using drain-gate feedback resistors and about doubling the LFO signal size. But again: fast, good, cheap: pick any *two*.

[R.G.]

LDRs are not banned from new products. ROHS exists primarily to stop dangerous material entering landfill, but effects pedals are the sort of thing which never goes into landfill. An enforcement body would have an uphill struggle to convince a court that one or two LDRs in an effects pedal could constitute an environmental risk. They're still used in many non-disposable audio products, and probably always will be.

Your country of residence matters. I don't know how it is in the UK, but environmental bodies in the US have enforcement authority. They can, for instance, direcly impose fines without court action and without asking any other body for enforcement. Worse yet, some of them have armed "enforcement" authority. There was recently a case where agents of the USA Environmental Protection Agency (EPA) showed up at a mining area in Alaska with full SWAT gear - military arms, bulletproof vests, helmets, etc. - to enforce a cease and desist notice. It's up to the person being enforced upon to go hire an attorney and get the action reversed, with the need to retain the attorney for (in general) about $10k up front, and going to court costing $100K up.

So check with your country's laws about whether the enforcement agency even has to go to court to enforce fines on you. The situation is of course very different if it's your single pedal that no one else ever gets to mess with.

[amptramp]

Found this larger schematic:



I had mine apart and knew the value of the resistor going from the collector of the tremolo oscillator to the jack (which goes to a shorting switch), but I don't have it now.  The values do not appear to be critical.

[kaycee]

See if you can find CultureJam's Stutter Tremolo. There's a way to build it with "shunt mode" that completely cuts out the signal.

I think this one was done by one of the other guys over on BYOC, CJ may have supplied some boards or contributed to it. Do a search for stutter Trem over at BYOC. I built one, it had 4 controls, rate, depth, volume and cycle which I think changed the wave shape. It was a very fast very deep square wave type Trem, sounded a bit like the Repeat Percussion. My build is up in the show off section, I auctioned it off earlier in the year as I use the RP and one based on Liquitones univibe type Trem.

[MrStab]

thanks for the v. informative replies, guys. there's a lot to reply to, so quotes are out of the question but i'll try to see if i've understood everything...

so a court case could act as a... wait for it... buffer! against prosecution. hmm, i'll need to look into that. i don't expect anything bad to happen, and i love breaking the law, but i wanna be sure i don't run into unnecessary problems. i'd probably make about 50, after a lengthy period of trying to figure it out & test it. or just give up completely, but all the different parts of the circuit are working to some degree, so i'm confident for now.

voltage divider is something that's been mentioned twice. reminds me of a post on another thread by RG, in reply to a question about the Schaller Trem input impedance, where you said it was integral to the function. so in basic terms, what i should be expecting from a transistor is a pot. i'd considered this at some point, but wound up using an actual volume pot (to ground) when experimenting, which wouldn't be the same effect (but beneficial in attenuating?). so a 100k or so pot as a variable resistor in series with the signal path, pre-JFET/NPN would help, i'm assuming.

another thing i've neglected to conceptualise is how the different waveforms would be perceived - a strangeness i've encountered when putting recovery gain stages after other tremolos only to find that getting the "right" level was finicky. i think i understand that.

i also think i understand the diode rail-limiting thing. similar to how they're seen in the audio path sometimes? in the case of an LFO and the non-rail-to-rail nature of most opamps, how far from the rails do you think i should go re. diode choice?

thanks for the bigger schem, Ron. i'm starting to see more parallels between different circuits.

at the end of the day, i'll probably need access to a scope to ensure this works right (or as right as it can be given the components/requirements), but it's gotta sound audibly decent first. thanks again for all the help, sorry if i forgot anything. i'm not around my gear at the moment but i will be in a coupla days.

[MrStab]

further thoughts:

besides the thump, an OTA seemed to work quite well - i just couldn't find any other precedents outside similarly-themed datasheet suggestions, descriptions of failed attempts or successful attempts with little documentation. so after a point, i figured it'd be too much of an uphill struggle when there's a wider precedent of using optocouplers or transistors.

so, if i understand the "second switch" part of your post, RG, attenuating or amplifying the signal itself would be a better approach than trying to alter the LFO's output between square and triangle modes. hopefully i could pull that off without too much back-forth routing proximity on a DPDT or something.

admittedly, i've only used about 4 tremolos in my time. to what extent do you guys think triangle waves should cut out entirely? i mean, i think it definitely needs more intensity, but it's just about usable at faster times.

could anyone point me to a schematic of the peak-limiting diodes in action? just so i have an idea of placement. also, how should i determine the ideal value for the minimum resistance to the Gate (R2 in my first schem)? i'm guessing this'd be dependent on the JFET somewhat, and the impedance at the Drain, as alluded to (i think).

i notice that some of the triangle > psuedo-sine circuits use what looks like hard diode clipping in order to create something "hyper-triangular". not wanting to use one of the more complex circuits i found (for uncertainty of added benefit), i tried that, but i put the diodes in the feedback loop to see if it worked based on my newbish understanding of soft-clipping. would that be at all beneficial over the harder clip? or is this paragraph complete nonsense?

now i've absorbed some of these points, i'm optimistic!

[bool]

Hmm ... but why wouldn't you use something like a Kay tremolo ... as a "source of inspiration", tweak it to suit your requirements and call it done? If I understand correctly what you're after, this would be quite "there"?

[Transmogrifox]

I agree with bool -- it's hard to get much more simple than something like a Kay tremolo without adding complexity.

As for your first post, maybe there is something you can learn from the VGA used in the Orange compressor.  It uses a JFET shunt to ground in a manner similar to what you are attempting.  Perhaps there is something in compressors worth considering here.

For example, what if you mildly modulate an OTA gain stage and mix it with the original input signal, where the OTA output is out of phase with the input.  Even though the amplitude modulation is not very deep on the OTA, you might be able to cancel with the input to an acceptable depth, then boost the gain to the desired output level.  Summing out of phase signals usually gets you closer to perfect cancellation than brute-force variable gain elements.

[MrStab]

if i implied a need for simplicity, i only meant on the user's behalf - i'm willing to do what it takes! well, within reason. i'm still kinda dumb.

Transmogrifox, that is a really clever idea. i like it! would the VCA thump not be audible if used to cancel? even as an artifact in the resulting waveform? just to entertain the idea: i wonder if maybe it would work using a FET as i'm currently doing, but inverting and mixing with the clean signal via. opamps.

as far as i know, Bool, the Kay tremolo only has one wave shape on offer, and both my own arbitrary attempts with a NPN and opinions i read of the Kay made me feel that it was easier to prevent distortion in a FET. though it was mentioned earlier in the thread that a BJT could be made to work.

[Transmogrifox]

The idea with the OTA summing with out-of-phase signal is to optimize the OTA gain change region to minimize thump.  For example, when you try to bring an OTA near to its shut-off (-infinity dB gain) region, this brings BJTs close to the knee, turn-on/off area.  This results in a much larger common mode voltage change. 

If the OTA starts at a reasonable amount of gain where it is already well-on, then the increase in bias current (transconductance gain change) doesn't result in nearly as much voltage change on the output.  By avoiding this low gain area you can minimize the thump voltage.

You compensate for the lack of zero-volume depth by cancelling the lowest gain signal to nearly nothing, then let the gain go up from there.  Hopefully that makes sense.

You can test the amount of DC voltage change on the output of the OTA by putting a pot on the gain control input to increase current there, and hold a DMM on the output to see what region of gain change causes the most DC voltage shift.  You will probably find a "sweet spot" in there somewhere that you don't see the output change too much...but then maybe I'm wrong and the DC offset dominates. 

It's an idea, anyway.

[Blitz Krieg]

I use the RP and one based on Liquitones univibe type Trem.

Liquitone's is the Forum Vibe?  what is the "RP" trem?