Another New Design.. i present you the "DUI distortion"

[DougH]

There is another thread about JFETs being widely variable. How much will that impact this circuit?
-Joe Hart

Different JFET p/n's will typically have different amounts of gain but the biasing will be consistent.

[alanlan]

- It's not a 'moot' point - A JFET may be modeled as a current source, like every other transistor, but we don't care about Id - we care about Is as that is the terminal feeding the lower JFET.

In JFETs, id = is, as long as gate source is reverse biased.

- How can a constant current of zero be a current source? Electrically speaking, it's nothing. And if that is the ideal situation, how would the output voltage not also be zero? In fact, your ideal situation means that there is no battery/power source either because even leakage currents can't get through - and they always do.

I'm talking about AC signal variations in id about a DC operating point.  Of course there is a DC bias current flowing continuously through both JFETs, but to signals, the upper JFET acts as a stiff very small AC current source i.e. a very high AC resistance.  If you think about it, id = gm.vgs.  So, if vgs is very small, then so is id.  If you measure and/or simulate the mu amp, you will see.  It is because the upper JFET is a very high AC resistance, that the gain of the circuit is so high.  The lower JFET will modulate id in response to the input signal and because the load on the drain is so high, the output voltage is also very high (it will quite easily hit the rails).  This is also why the output resistance of a mu amp is large and why, in general, it must be buffered by a high input impedance follower if you wish to preserve the amount of output signal.

- The output impedence at the source of a JFET is low, which according to Ohm's Law, and not a model, means that it is closer to a voltage source - the opposite of a current source.

The impedance of a JFET is variable depending on the voltages across the terminals.  You can't just say the impedance is this or that without defining the conditions.

are you saying then that the output impedence at the source of a source follower is very large as well?

No.  A source follower has a fairly low output resistance.

The upper JFET is a source follower that is also being driven by the signal on it's source - which, as R.G. said in the quote, doesn't contribute any (voltage) gain, but DOES help to move the load.
The top JFET is not a current source. At all. It provides current to the bottom JFET, but that doesn't make it a current source. The best way to think of it is that it 'bootstraps' the gm of the lower JFET. 

Well, the upper FET has no choice but to be a source follower since the cap shorts the gate and source together as far as signals are concerned.  The upper JFET is configured as a very low value AC current source and therefore a very high AC resistance and in turn produces a high voltage gain for the circuit.

If you are really interested, I can post the maths, but I don't want to bore anyone to tears.

[dschwartz]

hi everyone,
last night i experimented with this mods:

- larger opamp feedback to ground cap: originally 104, i changed to 474. the result is a fatter, thicker sound, very good for leads, but not as good for chords, since bottom end starts to become flubby as it distorts. The sustain and attack is nicer, but i felt muddiness on the lower strings..i reminded me to mesa-style leads..

- no  clipping leds: i let the opamp distort by itself, this gives a big muff type of sound, but softer. Better than a big muff, but not my style of sound..

- just 1 clipping led: very interesting, the sound became over compressed and fuzz-face-ish, nice octave effect, not as noticeable as an octavia, but is there.

about the FET discussion..
I´m still not clear about the u-amp output impedance, is it high or low?, the tone control seems to lower the volume, as expected, but do we need a buffer?

[alanlan]

I´m still not clear about the u-amp output impedance, is it high or low?, the tone control seems to lower the volume, as expected, but do we need a buffer?

The output resistance of the mu amp is very high, so your tone circuit will lower the output quite a bit I would have thought (although it will be frequency dependent because the tone circuit is so by its very nature).  I can't honestly say whether it is necessary, but if you want more output after the 2nd mu amp, a buffer is worth a try.  Don't forget also that you have 2 mu amps in series so that's a lot of gain in the circuit anyway.

[dschwartz]

yes, without the tone control, the output level is pretty high so maybe there´s no harm on that.

[JDoyle]

Alanalan,

It took me a while to figure out what you were saying, and I get it, and you are are completely right. However, in the future if you could in some way differentiate when you are meaning DC vs. AC variables, or point out that your lowercase lettering means just that, it would help a great deal. I realize that it isn't always the easiest to do. I also realize that it is my fault, and that my confusion is based upon the fact that, as I'm not an EE, I don't immediately see 'vgs' as being a different quantity from V_GS - at least when written on the internet. (for instance, take a look at the statement 'a small vgs means a small id' - in terms of a non EE, it would appear to contradict the fact that when V_GS=0, I_D = I_DSS - though again, that is my fault.)

Thanks!

To those following along: because the cap transfers the same CHANGE in voltage at both the source and the gate, V_GS (the DC bias value) for the upper JFET remains the same, as the same amount of signal is added/subtracted from both V_G and V_S via the cap, keeping the current out of the upper JFET the same. [For a similar example, take a look at John Linsley Hood's 10W Class A Amplifier]

A couple of points: the absolute value of the amount of current from a current source does not relate to its impedence - its impedence is determined by its ability to resist any change in the current out due to variations in the voltage across it. And here is where the situation of a 9V effects circuit starts to come into play, may hint at why the muamp has a good tone, and I think also contradicts the assumption that the type of JFET in the upper position doesn't matter:

While V_GS of the upper JFET remains the same, its V_DS does not, and as the voltage across the upper JFET moves, it goes into and out of the linear and saturation regions, changing it's impedence because the current out changes with V_ds (I honestly don't know whether to use uppercase or lowercase here) - as the output signal swings down, increasing V_ds, it becomes more and more of a 'true' current source, raising it's impedence and thus the gain, but as the output signal swings positive, V_ds of the upper JFET is reduced, bringing it into it's linear region and reducing the current it can provide, making it less and less of a current source, thereby reducing it's impedence, and therefore the gain.

So for the negative swing you have high, constantly increasing, gain, that goes into hard clipping. For the positive swing, once the upper JFET's V_ds gets into the linear region, the gain is constantly reducing - thereby squashing the positive swing. To get more squashing, use a JFET with a high Vp. A Vp of roughly 4.5V would mean that the gain is constantly reducing for the positive swing and constantly increasing for the negative swing (this is actually always the case due to the slope of the saturation region portion of the transfer curve not being constant, but a Vp that is the same as the bias voltage at the lower JFETs drain would make it most pronounced). A low Vp JFET would mean high gain throughout most of the swing with squashing only at the extreme end of the positive swing (with a V+ of 9V, the squashing region would be [9V-Vp] to 9V). A higher V+ would have the same effect as a lower Vp upper JFET, a lower V+ would be like a high Vp upper JFET.

While this is still a JFET doing it, and I am extremely wary of writing what I am about to, so please, please, PLEASE, don't infer anything into what I am about to say (because a JFET is and always will be a JFET, and a tube is and always will be a tube), but this type of distortion, in the general sense, is how a triode stage distorts.

Again, alanalan - Thank you! Especially for your patience. I would rather learn than be right.

Regards,

Jay Doyle

[earthtonesaudio]

A nice mod that RG has illustrated on the "foolin with fets" page is to stick a 1k resistor between the drain of the bottom JFET and the source of the top one.  Everything else stays but it's important where it is attached: the cap goes from drain of the bottom JFET to gate of the top one, and the output is taken from the source of the top JFET.  This changes it from a "mu-amp" to a "SRPP" stage.

I believe this lets you have lower output impedance in addition to higher gain, because the AC impedance seen by the bottom JFET is multiplied by that resistor.  You can't make the resistor too large, though, or you start to mess with the bias of both JFETs.

This would allow you to use smaller resistors in the mu-amp stages, and have more flexibility with the tonestack.

[dschwartz]

SRPP? is that something like push pull?? if so, maybe can be really usefull to approximate to a AB power amp, with it´s crossover distortion and cancelling of 2nd harmonic?

[earthtonesaudio]

Both the SRPP and Mu-Amp are Class A push-pull.  So, no crossover distortion, but 2nd harmonic cancellation is possible if the JFETs are distorting.

Very nice write up here:
http://www.tubecad.com/may2000/

[dschwartz]

bummer, just when i was starting to feel like i know some electronics, i realize i don´t know nothing at all..

[alanlan]

Alanalan,

It took me a while to figure out what you were saying, and I get it, and you are are completely right. However, in the future if you could in some way differentiate when you are meaning DC vs. AC variables, or point out that your lowercase lettering means just that, it would help a great deal. I realize that it isn't always the easiest to do. I also realize that it is my fault, and that my confusion is based upon the fact that, as I'm not an EE, I don't immediately see 'vgs' as being a different quantity from V_GS - at least when written on the internet. (for instance, take a look at the statement 'a small vgs means a small id' - in terms of a non EE, it would appear to contradict the fact that when V_GS=0, I_D = I_DSS - though again, that is my fault.)

Thanks!

To those following along: because the cap transfers the same CHANGE in voltage at both the source and the gate, V_GS (the DC bias value) for the upper JFET remains the same, as the same amount of signal is added/subtracted from both V_G and V_S via the cap, keeping the current out of the upper JFET the same. [For a similar example, take a look at John Linsley Hood's 10W Class A Amplifier]

A couple of points: the absolute value of the amount of current from a current source does not relate to its impedence - its impedence is determined by its ability to resist any change in the current out due to variations in the voltage across it. And here is where the situation of a 9V effects circuit starts to come into play, may hint at why the muamp has a good tone, and I think also contradicts the assumption that the type of JFET in the upper position doesn't matter:

While V_GS of the upper JFET remains the same, its V_DS does not, and as the voltage across the upper JFET moves, it goes into and out of the linear and saturation regions, changing it's impedence because the current out changes with V_ds (I honestly don't know whether to use uppercase or lowercase here) - as the output signal swings down, increasing V_ds, it becomes more and more of a 'true' current source, raising it's impedence and thus the gain, but as the output signal swings positive, V_ds of the upper JFET is reduced, bringing it into it's linear region and reducing the current it can provide, making it less and less of a current source, thereby reducing it's impedence, and therefore the gain.

So for the negative swing you have high, constantly increasing, gain, that goes into hard clipping. For the positive swing, once the upper JFET's V_ds gets into the linear region, the gain is constantly reducing - thereby squashing the positive swing. To get more squashing, use a JFET with a high Vp. A Vp of roughly 4.5V would mean that the gain is constantly reducing for the positive swing and constantly increasing for the negative swing (this is actually always the case due to the slope of the saturation region portion of the transfer curve not being constant, but a Vp that is the same as the bias voltage at the lower JFETs drain would make it most pronounced). A low Vp JFET would mean high gain throughout most of the swing with squashing only at the extreme end of the positive swing (with a V+ of 9V, the squashing region would be [9V-Vp] to 9V). A higher V+ would have the same effect as a lower Vp upper JFET, a lower V+ would be like a high Vp upper JFET.

While this is still a JFET doing it, and I am extremely wary of writing what I am about to, so please, please, PLEASE, don't infer anything into what I am about to say (because a JFET is and always will be a JFET, and a tube is and always will be a tube), but this type of distortion, in the general sense, is how a triode stage distorts.

Again, alanalan - Thank you! Especially for your patience. I would rather learn than be right.

Regards,

Jay Doyle

No problem Jay!

One of the things I often forget though is that the nice linear small signal models (which are great for analysing and understanding transistor circuits) do not explain the large signal situations, as you have clearly described.  With FETs in general, this is where I think the magic (if you can call it that) happens and why people like them i.e. all that even order harmonic stuff.  To be honest, I'm not sure my ears are good enough to tell the difference, so I just tag along and assume that FETs will sound nicer.

[snap]

more mu-amp philosophy: http://www7.taosnet.com/f10/mustage.html by Alan Kimmel.

[earthtonesaudio]

bummer, just when i was starting to feel like i know some electronics, i realize i don´t know nothing at all..

Don't be too hard on yourself.  The mu-amp/SRPP/whatever-you-call-it gain stage is a very difficult circuit to understand.  Push-pull AND Class A, bootstrapping, quasi-constant-current-source, etc, etc.  You don't often see a circuit that contains ANY of these, let alone all of them.

And hey, the DUI Distortion sounds good, has a catchy (if a bit controversial) name, and you designed it!  Feel good and keep learning, and by all means keep posting circuits like this.  I always look forward to your designs.

[dschwartz]

Thanks For the support , Alex..When i design this stuff i feel like i know what i´m doing, but it´s a feeling from the guts, my mental language is based on waveshapes, clipping shapes, freq responses and impedances..and i validate that with simulations and breadboards..I´m not really sure if the hard technical knowledge can help me. For me, designing is pretty much like building with LEGO parts. Hart technical info may help to make new lego parts, but for now i feel comfortable with the parts available nowadays.

I have to inform that the schematic has a little error.. the clipping leds must be directly in the feedback loop of the opamp..it should work as it is now, but the gain control may be too extreme..

i´m experimenting now with a baxandall type of tone control stage, at least on simulations. The current BMP tone control feels weird, and has too much of swing..

[JDoyle]

No problem Jay!

One of the things I often forget though is that the nice linear small signal models (which are great for analysing and understanding transistor circuits) do not explain the large signal situations, as you have clearly described.  With FETs in general, this is where I think the magic (if you can call it that) happens and why people like them i.e. all that even order harmonic stuff.  To be honest, I'm not sure my ears are good enough to tell the difference, so I just tag along and assume that FETs will sound nicer.

Yeah - seeing as I got into this specifically FOR distortion tone, I tend to always start thinking about the large signal response, small signal models are really great for understanding as you mention, but if you think about it, very rarely, and only in circuits that are very well controlled, do we ever deal with a 'linear' portion of ANY transfer curve. Thank God.

I also agree with you that it is the large signal response that make JFETs popular - it lies in that trasition from saturated to linear regions. Which is why I think taking a look at transfer curves, if you can find them, is so important. A side note to that is JFETs designed to be audio amplifiers, unfortunately for us, will have the SMALLEST zone of linear region operation so as to increase the nondistorted headroom of the saturation portion.

In my opinion, I really don't think that there is a whole lot to the belief that JFETs are so much superior to other options for any other reason than the fact that the gain available from a single stage is so much less than any other type of transistor, even without a source resistor (which admittedly makes it a hell of a lot harder to bias without trickery). That may seem contradictory to what many may believe, but I think the best tone comes from cascading several low gain stages rather than beating the crap out of the signal with one - which with single BJT or MOSFET stages is hard to avoid because reducing the gain requires a emitter/source resistor, which then forms a hard wall below which the signal cannot travel, so you have the same hard clipping problem as before and didn't really solve anything...

If you also think about it, the other popular transistor type, germanium BJTs, suffer the same 'problem' of low gain in comparison to Si BJTs and MOSFETs......

Anyway, thanks again man...

Jay Doyle

[alanlan]

Good points again,
I'm tempted to try cascading say 3 stages of "average" gain JFET and comparing against a single stage with the same gain.  I guess that both could be calibrated with a very small signal so in theory comparing like with like (gain and bandwidth) - we've got a Prism test set at work so I can do that test.  If I manage to do it I'll try to post measured results and audio samples.

Daniel, I hope we're not diverting from your post too much, but these JFET discussions have a habit of getting interesting...perhaps if i do the experiment I'll start a new thread.

[dschwartz]

Good points again,
I'm tempted to try cascading say 3 stages of "average" gain JFET and comparing against a single stage with the same gain.  I guess that both could be calibrated with a very small signal so in theory comparing like with like (gain and bandwidth) - we've got a Prism test set at work so I can do that test.  If I manage to do it I'll try to post measured results and audio samples.

Daniel, I hope we're not diverting from your post too much, but these JFET discussions have a habit of getting interesting...perhaps if i do the experiment I'll start a new thread.

well..maybe it´s the proof that this thing sounds great..

[dschwartz]

did anyone gave a shot on this one?

[kurtlives]

In your Double Shot pre-amp thread the schematic got replaced with your DUI schem. Noticed today as I was looking at pre-amp designs...

[dschwartz]

uuups, i overwrote the schem..

google pages doesnt warn you if you overwrite..well.. i have it here..send me an email if you want a copy..