Ugly distortion from two jfet gain stages

[hans h]

Hi all, I recently posted my first topic here and got a lot of great answers, for which I am very grateful. I learned a lot since that post (biasing jfet by Rdrain with fixed Rsource) and now obviously ran into a new question!

I am trying to 'develop'  a two or three stage jfet overdrive. The type of tone I am aiming for is a nice little tube Amp, akin to seven caged tigers by stone temple pilots.

I started by converting the preamp of the supro 1606 and then drifted further and further from that design to get the kind of frequency response I wanted and to reduce noise. What I have now is the schematic I attached. This is my problem:

When I dig in hard I get a papery distortion on top of the main distortion. It sounds like a cheap telephone microphone which gets too much input, for lack of a better description.

On the first gain stage I solved it mostly by adding a 100k series resistor at the input, most likely because it brings the input signal down. However, I still get it at the second stage. Another possible culprit may be the cheap breadboard I have, but I do not have another one to try.

I also tried the back to back diodes seen in the rog supreaux deux, but that didn't help either. Have you guys and girls any idea what the cause is? At the moment the effect is only useful as sparkly clean boost but not as overdrive.

[antonis]

Better for our necks..



All I can see is an unbiased first stage Fet..

Move 1M resistor on FET Gate and try again..

[iainpunk]

Antonis is right! you need a pull down resistor to keep the gate at 0v when idle. and a diode parallel to that resistor (pointing towards ground) keeps the gate from forward bias, which can give some weird artefacting as well.

cheers

[jonny.reckless]

The first stage is letting too much bass through which will cause flubby distortion. Maybe shrink that 100uF bypass capacitor on the first JFET to something smaller. I'd recommend starting with 220nF and experimenting from there. Nearly every overdrive and distortion circuit has a bass cut or treble boost early on in the signal chain.

I'd also remove the 220pF Miller feedback capacitor which will be killing the high end. Also a gate stopper of around 100k in series with the gate of the second JFET will prevent any blocking distortion.

In summary:

• Remove the 3n3 coupling cap (short out) to allow gate to bias at 0V
• 100uF bypass on first stage -> 220nF
• Add 100k in series with gate of second JFET
• Remove 220pF Miller cap

You may then want to add a capacitor across the second 22k drain resistor to tame the harshness and reduce high frequency fizz. Perhaps start around 4n7 and experiment from there. I normally use 3 stages of JFET gain for a good distortion. You can get a nice light crunch from 2 though, if you fully bypass the second stage.

[antonis]

With all the respect Jonny, without Miller cap, I believe Q1 stage will squeal..

edit: I overlooked your suggestion about lowering Source bypass cap value..

[temol]

Here's the the original circuit in ltspice.


And here's my attempt to make it "sound" similar to Supro 1606 preamp. Blue trace it's tube preamp, green trace - jfet preamp. I have no idea how does it sound..
 

[hans h]

Wow, thanks for the in depth responses. I am going to try the suggestions given so far. First off, I put the input cap to the front of the circuit 😅. This did not really change anything with respect to the artefact I spoke about. But what I actually found out: I get the same type of sound when I bypass the pedal and dig in. Maybe one of the tubes in my vox is starting to go bad. When I have time I'll try out another Amp. Looks like I have some work to do!

[hans h]

I first tried the suggestions of Jonny. Your approach of removing noise (cap in parallel with Rdrain) is new to me. It kills les noise then the Miller cap but indeed I seem to retain a lot more high end which is nice. I ended up using 10nf.

The 220nf bypass was too shrill for me. I now use 680nf but might increase it some more. Indeed a lot better than the 100uf. For stage two I kept going back and forth between 100 and 10uf. I now have 10uf in that position.

Instead of 100k at the gate of q2 I used 47k because I figured that the gain pot also adds to the series resistance, except if on full.

I tried to get the coupling cap and output cap in the 50 to 40 hz range because I am now cutting bass (or amplifying it less) using the bypass cap.

By the way, I also filter power using 100r and a 100uf cap (not shown).

Rectification: I see that I noted 18v in, but this should be 9v!

[antonis]

Instead of 100k at the gate of q2 I used 47k because I figured that the gain pot also adds to the series resistance, except if on full.

Total Gate series resistance doesn't mind a lot so far as it's kept to a resonable value (1M or so..)

IMHO, "Gain" pot should be called Sustain-like 'cause, strictly speaking, it simply divides Q2 input signal amplitude..

[hans h]

@Temol: thanks a lot for the ltspice work. That is something I should pick up to in the future, but I am already sitting behind my computer too much as is. Three questions about the schematic: 1) why do you use such 'low' Rdrain values. Any specific reason? 2) what is the diode for. I have seen back to back diodes in the supreaux deux, but do not understand what they are used for. 3) any reason to use a lower gain/sustain/volume pot value? This will off course change frequency response, but that can be solved by changing the coupling cap

One question about the figures : what is on the axis of the figure with tube preamp VS your schematic? Is it frequency response? If so, my circuit looks a lot 'spikier'. At the moment it is either too spiky or too dull, so maybe your circuit will work better. It may also be the combination with the vox, which already has pushy upper mids.

Question @all: are there two types of distortion you can get from a jfet or only one? I could imagine that the gate can be clipped (channel fully open or fully closed). Also, the voltage headroom can be clipped. If there are indeed two types, is one of them desirable over the other? In that case lower input level would reduce gate clipping. Resistor in series with the bypass cap would reduce voltage headroom clipping (I think).

@antonis: I'll put the 100k back in if it does not make a difference.

[antonis]

Gate doesn't clip..
(neither Source or Drain..)

Signal is clipped, due to overdive..
(hitting supply rails..)

Lower Gate input signal level results into lower (actually higher due to phase reversal) Drain amplified signal..

Series resistor does nothing to signal level due to "infinite" Gate impedance..
(no current flow hence no voltage drop..)

Bypass cap is there for setting stage gain to maximum..
(it shorts Source to GND for AC..)

[iainpunk]

Question @all: are there two types of distortion you can get from a jfet or only one? I could imagine that the gate can be clipped (channel fully open or fully closed). Also, the voltage headroom can be clipped. If there are indeed two types, is one of them desirable over the other? In that case lower input level would reduce gate clipping. Resistor in series with the bypass cap would reduce voltage headroom clipping (I think).

yes and no.....
the top of the wave clips against the supply rail, and the bottom of the wave clips against the voltage on the gate, if there is almost no voltage over the gate to drain, the jfet is fully 'open' it then clips against either ground or the voltage on the input of the gate, you don't want the latter, you really don't want the weird effects that happen when you forward bias the base into conducting... its hard to describe, but its not a nice kind of distortion, since it sounds more like artifacts from bad tracking in a digital pedal, than some sort of nasty distortion. to prevent this, you can add a single clipping diode, so the diode drop + the source resistor voltage are never exceeded.

cheers

[antonis]

why do you use such 'low' Rdrain values. Any specific reason?

For higher Drain current, hence higher transconductance..

g_m = 2I_DSS/V_P  * (1- V_GS/V_P)

[hans h]

Higher current out for same voltage in gives higher tranconductance, but why do we want higher transconductance? For better driving the volume/ gain pots or for a different reason?

[anotherjim]

I'd put in a 10k or 22k resistor between the "gain" pot lug #1 and 0v because do you need it to adjust to silence?
I usually call a pot in that situation  "Drive".

[antonis]

but why do we want higher transconductance? For better driving the volume/ gain pots or for a different reason?

Both..
(High quiescent current assures higher load driving capability)

In case of Volume/Gain pots are considered appreciable load (of relatively low impedance compared to Drain resistor) and/or in case of we wish for a higher gain stage..
(the later been a must when dealing with JFET CS amps, compared to BJT ones, in distortion circuits..)

[hans h]

I finally got around to trying temol's suggested circuit (except with 100k series resistor at gate of q2). It is MUCH better than what I had so far. Seems like I have to start learning ltspice
Thanks a lot for that contribution. It is a nice basis for further tweaking. Nice warm sound, maybe a tad flubby bass, but a lot of character.