mu amp/SRPP bias voltages? filter formulas?

[cspar]

I lucked out and got a handful of jfets so I've been playing around with some mu amp and srpp booster configurations and trying to wrap my head around them.

There's only a little scattered information out there on jfet srpp circuits and I haven't seen anything technical about voltages for the two transistors either mu amp or srpp.

What I've been doing is biasing the bottom transistor to 4.5v at the drain with a source resistor and biasing the top resistor 4.5v at the gate. I have no idea if this is "correct". It definitely mellows out the gain without losing too much and doesn't seem to make much headroom loss but I don't have a scope.

In order to get the 4.5v for the top transistor I'm doing a few different things like making a voltage divider out of say a 10k's with a 100p to ground then from there a 51k and then the gate.

It seems that the capacitor in the above example doesn't respond like I'd think it would. What would be the formula for the filter?

This is all in place of the common 1M based voltage divider.

This brings me to the big bias question. What is the "correct" biasing of the top jfet? By what parameter and how? Should my gate voltage be based on a current reading at a test point or the current based on the gate voltage?

Additionally, I've been exploring using Miller integrators with mu circuits. I'm getting good sounds on the breadboard by following examples but not understanding exactly why and how as they seem a little different then BJTs.

If anyone could point me in the right direction for the filter formulas Miller integrators (rc between gd) that directly relate to jfets I'd much appreciate it.

I'm generally trying to wrap my head around the several rc filter points and wishing there was a simple schematic with a component by component analysis like some of Jon Patton's or the Beavis lpb-1 analysis.

Goals...

[radio]

http://www.geofex.com/Article_Folders/modmuamp/modmuamp.htm

Does this help?

[cspar]

Totally familiar with R.G's article as well as a few other of his that deal with the jfet mu topologies. They've been inspirational and definitely get the ball rolling in understanding the circuits. But still leave a lot of questions.

I'm also familiar with quite a few Catlinbread, Wampler and DIY schematics that use various mu/srpp stages.

There's quite a few ways that people have made it "work" with different voltages in different places. What are the virtues/detriment of different voltages?

Due to the impedence difference between jfets and bjts it seems rc filters work much differently.

I'm writing a schematic for what I have on the breadboard right now and will post it soon.

What I was working on originally was a Hollis Fallstaff that sounds good but is set aside and still on a breadboard. That led to me building this srpp which is a little on the treble side and heading towards an amp emulator type thing when combined.

I'm not satisfied with it but it works allright.

[cspar]

here's what I have on my breadboard for reference.

[amptramp]

If you want some additional reading material, try this:

https://www.tubecad.com/articles_2002/SRPP_Deconstructed/index.html

Consider tubes to be JFET's with a heater.  One of the points made in the article is if R5 (the upper JFET source resistor) is equal to the inverse of the transconductance of the upper JFET, the upper and lower transistors will have equal current swing.  It also points out the SRPP stage must be treated as a power amp since it has two current sources in series and it requires a load to remain linear.  As the upper lower stage pulls more current, the upper stage pulls less and vice-versa but there is always a difference in current delivered to the load and if the load is high, the signal will slam against the rails.  If the lower stage pulls 2 mA and the upper stage sources 4 mA, there has to be a place for the difference current to go in the load.  If you got this working right, you could make a great fuzz from an SRPP stage by using a variable load to create a limit for the difference in current - a high value gives you fuzz whereas a low value gives you linear operation.

[cspar]

Kind of silly but I haven't read the tubecad article even though I've know it exists. I've been tunnel visioned looking for stuff directly related to jfets because I want to understand filter response within the circuit too.

I'll read it and see what I can gleen from it.

In the above schematic if I swap the voltage divider and 51k for the commonly used 1M×2 based voltage divider on the top transistor it's gate voltage is 3v and the bottom transistor sounds best biased at 6.2v on the drain to my ears.

Undoubtedly it could be altered to be more fuzzy or distorted.

My current design goal is to add a srpp to a Fallstaff that is a little extra dirty to begin with. So instead of just putting in a "standard" stage, I'm trying to get a better understanding of the interplay between bias, filtering and time constant.

[cspar]

I read the tubecad article. Although it explains alot it left me with more questions and inspiration than answers.

I followed the ratio jist of the basic srpp voltages from the article so that q1 vd is half of q2  vd and the voltage at q2's gate is half of q1's vd.

To help get the gate bias I built a chain of voltage dividers/filters. I don't understand the balance between tone shaping-push pull and current sink-time constant so it could use some adjustment. Probably a whole new redesign in that section. It works and sounds pretty good.



Next I guess I should figure out fine tuning R6 to the transducance of q1?

[iainpunk]

i just want to say that you are working with a Mu amp design, not an SRPP. for it to qualify as an SRPP the gate of the top load transistor needs to go directly to the collector of the other one.

sorry for being nit-picky
cheers

[amptramp]

He does have capacitive coupling from the drain of the lower transistor to the gate of the upper transistor, so it does qualify as SRPP.  Otherwise, I agree with your definitions.  The µ-amp is just a JFET amp with a current source for a load, giving it high gain.  SRPP actively changes the current in the upper transistor in the opposite direction to the changes in current in the lower direction and the coupling capacitor does this.  It may not be as simple as with a tube amp where the upper grid can be taken to the lower plate and the biasing remains usable but AC-coupled signal still changes the upper transistor current in the right direction in the audio range.

[cspar]

In my understanding the capacitor between q1's drain and q2's gate makes the srpp connection and the time constant of that capacitor is the prodominant control for the phase switching of the push pull action in a textbook jfet srpp.

Srpp work work if the route between q2's drain and gate is dc coupled. That's why in the above schematic r10 and c3 are parallel. It's not like I'm just trying to simulate a leaky 2n2 capacitor there.

I currently have a version 2 on the breadboard where I changed out q2's feedback loop. I kept the 10k divider and followed it with something based off a shredmaster's contour section with different values and a resistor added for the dc flow.

I'm getting tone control that seems to be working mostly through phase cancelation but I am unsure of that and I'm working on a better response.

It's interesting how by using capacitors and resistors in parallel you can get both slight variations of phase and impedence variations that are frequency dependent. There's quite an effect by it in the feedback loop of a srpp.

By bridging the capacitors leading to the contour pot and such I've been able to get sounds that get towards slapback/vibrato and even slightly phaserish and shimmery.

The slapback/vibrato thing is cool and I'm wanting to refine that and hopefully get it on a switch.

The phaser/shimmer seems like I'm making the 1st part of an exciter and not cleaning it up after that or something. If I have it so it's doing that the guitar's tone knob might as well be a switch. Although interesting, it's not seeming to work out with what I'm trying to do.

I'm not sure how the Miller integrators work into the equation. I haven't found much information on them pertaining to a jfet ac coupled integrator.

It's not that I'm just curious about the gain and filtering aspects of the Miller. The effect the time constant of the Miller capacitor is part of the equation as would be a capacitor on q1's source if you had one, right?

I'm not using a scope but to me it seems that the slapback/vibrato that I'm getting is a combination of phasing that is sequenced by the time constants of the capacitors and the frequency dependent impedence variations sequences due to the use of capacitors and resistors in parallel.

There hasn't been any smoke or component failures so far and I've had some interesting results balancing phasing and filtering.

I wonder how much sequencing control you can get by turning the filter section into a bigger delta and if with a simple sidechain you could control the speed of the phasing?

Can you get a 2 or 3 transistor sequenced phaser by warbling the push pull of a jfet srpp and not have a lfo at all?

Some questions worth asking, but I'm just gonna focus on the shredmaster like filtering that's on the breadboard for now.

[amptramp]

One suggestion I would have is to take the feedback from the source of the upper transistor which is the output of the stage, rather than the drain of the lower JFET.  After all, you want feedback from the output to the input, not some intermediate point inside the stage.

By all means, if you get phaser and slapback operation that is useful to you, use the schematic you have.  Also, the transconductance of the upper stage varies with the operating point whereas the resistor from the lower drain to the upper source remains fixed, so the balance of the circuit is only maintained over a limited part of the operating range.  Beyond that, it may get more interesting and musically useful.

Miller capacitance in a JFET is a much bigger part of the story than it would be with a tube.  With a tube, you have several pF of capacitance and it doesn't change with the bias point.  With semiconductor devices, the drain-to-gate capacitance is not only orders of magnitude larger, it varies with the operating point.  If you have a FET power amplifier, you have to drive the gates of the output devices with a strong enough signal to swamp the effects of input and Miller capacitance and it would be difficult to use anything other than a source follower output stage - a common-source output would have tremendous variations in capacitance due to the change in operating point, enough that taming the capacitances would be a bigger issue than DC bias considerations.

[cspar]

To try to be clear, despite any schematic I've posted to get the most drastic effects I'm splitting the signal and then remixing it within the filter so that I have a balance of both ac and dc coupling. There is always a dc path between q2's drain and gate despite any dc coupling that occurs.

This seems to be keeping it in srpp while having a bandpass filter on q2 in theory as far as I understand. As much as I've gleaned from the tubecad article I still feel lacking in grasping it fully pertaining to jfets. It's apples and oranges and this is not fruit salad.

I'm using the same biasing formula. I've found that filtering before the rail voltage plays a big part in the equation too. Makes sense that the V+ ripple current is a fundamental part of the equation. It would probably be wise for me to add a dedicated regulator to supply the rail for better consistency and stability. It's like your last comment Amptramp, taming the capacitance is the issue.

So far I haven't had the best success with balancing the tone control of the filter, the guitar's tone control knob and the slapback effect. One of them is always wonky and working on one part puts another out of balance. That's compounded by the fact that I've switched around the filter topology a few times and haven't been taking the best notes breadboarding. There's been lots of online calculations, crossing fingers and random swaps.

I'll have some time later to work with it on the breadboard. I'm gonna focus on just the tone controls without much slapback and then work on increasing the wobble from there.

OK, so to sum up as I'm understanding it, the slapback type effect is a balance of;

A)the frequency dependent capacitance of the jfet
B)the frequency dependent capacitance of the Miller integrators
C)the frequency dependent impedence of parallel components
D)the time constant of the phase shifting of the push pull action

In relation to ;

☆)the phase shift and time constant of parallel components

And all those values are based in relation to these 5 things;

1)V+
2) the ripple current of V+
3)Idss
4)Vgs(off)
5)variations of the input signal

I am unsure at this point about the effect of mismatched transistors among the other variables and am going with as a tight a matched pair as I can for now moving forward.

On a sidenote, what I was thinking about for feedback, which I haven't tried, was to inject the dc coupled output into q1's Miller integrator in a way that blocks the feedback from q1's gate. I'm not sure if that would really work at all but the goal is to influence the phasing speed and vibrato more than the compression.

By no means am I an EE but in the last week I've breadboarded larger filters that are variable, had various results, there are no failed components or smoke and yet I can't find a srpp like schematic with anything more complicated than a basic 3 component voltage divider and the capacitor from q1's drain feeding q2's gate to reference... 

Am I just trying to refine something unruly that's considered a "bad" circuit which student EE's are taught to avoid in some textbook I haven't read?

Isn't there something with a bigger filter out there, even if it isn't variable? Maybe I'm not seeing the forest through the trees because the jfets aren't blatantly stacked the way it gets written usually?

I truly find it hard to believe that more filtering has ever been done in a published srpp/mu amp circuit in a book or on a forum. I only see examples of less filtering.

I've never used Lspice before but i guess that's how I'll end up with the best understanding of how turn the top side of an srpp into a fine tuned variable modulating bandpass filter.

[cspar]

It seems like the shimmery effect is the filtering of q2 creating a phase shift oscillator during parts of the push pull cycling. Perhaps shimmery isn't the right word but I'm not sure what to call it drone isn't the right word.

I don't really know what going on, just theorizing. I guess it's time to get an oscilloscope so I can feel like I have a flashlight probing in the dark and not just the metaphorical flick of a lighter from my multimeter.

I'll learned so much through pedal forums in the last few years. Often I've found insight from part of a discussion that relates to a different design or topologies.

Quite a few things have been brought up that are worth looking into and studying on their own without  using a srpp or mu amp. Jfet Miller integrators, parallel components and phase shift oscillators.

Even if I'm misunderstanding things I hope that my public brainstorming provides some inspiration and useful knowledge for some folks.

[amptramp]

One thing I am visualizing now is a fuzz that uses an SRPP stage feeding a variable resistor to ground.  When the resistance goes below a certain point, the signal is linear because there is a sink for the difference current between the upper and lower stages.  If the resistance goes higher, there would not be an adequate sink and the output would stay near the power rails or ground.

In this configuration, I would use Darlington transistors for predictable bias voltages and high input impedance.  Transistors have a high transconductance with an exponential value with respect to bias, so the resistor between the upper and lower stages would not be a fixed resistor but a diode-connected Darlington whose resistance drops in the same exponential fashion as the increase in transconductance of the upper Darlington.  That would give a linear fuzz over a wide range of operating currents.  And you would have an output buffer to permit the resistance range of the control to be constant no matter what the load and this could be another Darlington.

This sounds like it would be just imperfect enough to have a recognizable "voice" but excellent enough to be a viable fuzz effect and the high impedance would make it suitable as a bass fuzz as well since the coupling cap from the lower collector to the upper base could be quite low and still remain usable below the bass low note of 41 Hz.

[PRR]

> bright-tonesucker

Where is the "filter"?

What is all that extra stuff in the upper gate's bias?

[cspar]

> bright-tonesucker

Where is the "filter"?

What is all that extra stuff in the upper gate's bias?

Probably not, there is no posted schematic of the filtering used in the comments after that.

I had a filtered 10k voltage divider going into a modified version of a shredmaster contour section where the 33k's where 180k's and the pot was a c500k and then a bunch of weirdness with parallel components.

[cspar]

That extra stuff was a bad attempt at some kind of filtering where the "contour section" was.

It boils down to me having some interesting anomalies in a unruly area of a mysterious but common circuit that I don't ever personally see modded.

At times I was able to get a sequenced time constant phase relationship that was unruly but not random and I could get a slapback type effect that was mostly on the treble end. These two things independent of each other.

I've used parallel chains in the filtering so that I can use capacitors and resistors in parallel in one of the sides. Using different value capacitors on opposing sides of the same chain link led and some resistors and capacitors parallel to ground I got some sequencing in the response.

This has led me to believe that in a srpp the connection between the top transistor's drain and gate when it's in amplifier mode is both bias and a miller integrator but when it's off and sinking current to the bottom transistor that same connection is a closed loop and thus the bias/filter/sequencer of an oscillating current source.

Is that correct?

I've never seen a phaser design where the amplifier was a jfet so getting a bit of musical modulation intrigues me.

So I've been twisting my head in a knot trying to grasp an understanding of phase and filter functioning without an oscilloscope.

I use online calculators. It's beyond me to try and harness the modulation at this point and I feel like I'm just fumbling with it.

My original intention was to put an srpp or mu amp after a modded Fallstaff that I have on a breadboard. I haven't touched the Fallstaff since I started on this.

At this point I probably just will put something standard after the Fallstaff and look into an affordable oscilloscope.

It's just weird to me that I only see a few very basic variations in the upper transistors filtering and it seems there is no firm gate bias voltage standard of it.

[iainpunk]

having a filter to the upper JFET partially defeats the purpose of the design principle behind an SRPP. its specifically a symmetrical push pull amplifier, and filtering inside this makes it asymmetrical (turns the 'upper' transistor or tube in to a bad current source). there are good/great reasons not to mod SRPP stages, (you can't improve on a local minimum).

why do you want to use a SRPP/MuAmp design anyway? is there a particularly heavy load you are trying to drive?

cheers