Request for comments on a low-voltage hybrid preamp/amp

[Gnobuddy]

...but why is that so? Did you discern that from the datasheet somehow?

I just used Kirchoff's voltage law. If you run at 1mA drain current, and anticipate a maximum signal swing of +/- 1mA, then you have a max drain current of 2 mA.

You have a (nearly) 4k drain resistor, which will drop about 8 volts at 2 mA. The 1k source resistor will drop another 2 volts (if un-bypassed), or 1 volt (if bypassed). With your 11.75V supply, that leaves about 2 - 3 volts across the MOSFET, which should be just about enough for a typical small-signal FET to operate properly.

Depending on your exact MOSFET, you might find that a little bit more or less than 1 mA is optimum. But tiny changes of a few percent aren't audibly significant - a decibel is about the smallest change we can detect by ear, and that's about a +/- 12% change. Anything less than that isn't even worth wasting your time with, as it's of no practical use.

In this case I've kept the bypass capacitor since this stage is right after the tone stack (lossy).

Sounds good, since you're after output headroom rather than input headroom for this stage.

I suggest re-connecting the source bypass cap as shown in the attached figure. This will make the MOSFET stage quieter, as ripple and noise voltage on the positive supply rail will now be shorted to ground, rather than fed into the MOSFET source and amplified (!)

This is because you can't really treat both the (-) and the (+) supply rails as AC ground, even though the textbooks say so. In practice, there will be some ripple and noise voltage between the two rails - so both of them can't be treated as true zero volts AC. You have to pick one to be zero volts, and assume the other one will be noisy. 🙂

Incidentally, you need a rather large 100uF cap at the source to get a bandwidth down to 80 Hz at -3 dB.

The lowest note you can play on a normal 6-string guitar in standard tuning is around 83 Hz, so you don't need the circuit to be able to go lower than that.

this is letting me tune the biasing exactly how i had wanted, brilliant!

Excellent! 🙂

-Gnobuddy

 

[andyholmes]

Last time on Andy's Amp Adventure...

When I started this project I was aiming for a 12V Fender AA764 Champ that could be used as a pedal or a mini-head. I've given up on replicating the Champ in most regards, but I've kept the basic topology because I like the simplicity. I spent much time experimenting with different compound amplifier stages including CCSs, bootstrapped followers and even super triode(ish) stages. I also spent some time comparing JFETs, MOSFETs, BJTs and opamps for use in these stages. I didn't spend much time on the tone stack, because I'm pretty easy to please in that regard and decided on a single-knob tilt-style filter.

An externally hosted image should be here but it was not working when we last tested it.

The bandpass filter on the input is rather aggressive but I'm in an electrically noisy environment, thoughts on that are welcome.

I chose a MOSFET follower input, with tons of input swing such that the tube will most certainly clip before the MOSFET does. The tube has it's cathode straight to ground and anode bootstrapped with another MOSFET follower. Even though the output impedance is increased with bootstrapping, the stage can drive the tonestack much harder, and could probably accommodate the standard Fender tonestack without trouble. The tonestack I chose here is pretty self explanatory; it's low loss and the extra capacitor there keeps more treble in the mix when the knob is leaning towards bass.

The second stage is identical to the first, however I've taken advantage of the above-the-rail voltage swing at the bootstrapped anode by using it as an alternate channel. The output from the MOSFET follower sounds a lot more modern and crunchy, while the output at the anode has a far more vintage overdrive feel and sound to it IMO.

It became obvious that to be used as an effect pedal (or at all actually) I'd need a Master control and a 1Meg pot doesn't seem to be a problem for the bootstrapped anode to drive (this was not the case with the tonestack).

Sorry the titles are BELOW the scope traces, the quality of which I'm also sorry for

An externally hosted image should be here but it was not working when we last tested it.

Some questions however (aha):

• If bootstrapping brings the stage gain up to the mu of the valve, do the anode resistors only affect the DC biasing? Or is that just the case once the resistors are large enough for the stage gain to reach the mu of the valve? Or should that be ignored and the anode resistor(s) chosen as usual?
• The bootstrap capacitor and upper anode resistor form a HPF so it's rather large already, but does the capacitance also relate to the amount or length of time that the voltage can exceed the rails, or is the upper limit here actually the triode's mu being realized where it would normally clip?
• When the second stage is driven really hard, there is a very noticeable dip and "bloom" sometimes over 2 seconds. I assume that this is what blocking distortion sounds like, and assume further it's caused by the capacitor/resistor in the MOSFET->Valve biasing scheme?
• In Rod Elliott's bootstrapped preamp he uses a feedback resistor I've not seen in other versions of this topology. Could this help me with blocking distortion (assuming it is that)?
• I've been running the bootstrapping MOSFETs without gate resistors, so far without problems, but is this something that might come back and bite me later?
• Since I'm using a 6111 tube I've just been using the heater's voltage regulator (~6V) as a reference, but I may replace that with a resistor or switch to a 12AU7. If a slightly different 12V adapter results in 11V or 12V instead of the ~11.5V I currently get, could a MOSFET biased with say a zener or tl431 maintain it's Vgs and the valve's Vgk without adjusting the trimpot? If not, would a 50/50 voltage divider be the best bet?

Thanks for following along with my adventure, I hope someone else gets something out of this.

 

[Gnobuddy]

Last time on Andy's Amp Adventure...

Welcome back! 🙂

I've kept the basic topology because I like the simplicity.

And the million Belarusian Ruble* question: How does it sound?

( *Roughly equal to $72.53 CAD; see XE: Convert CAD/BYR. Canada Dollar to Belarus Ruble )

thoughts on that (aggressive input bandpass filtering) are welcome.

If there isn't enough treble loss to bother your ears, and you're happy with it, it's fine. (You might try an A/B comparison with a flat frequency response preamp set to the same voltage gain, just to give your ears a reference.)

Even though the output impedance is increased with bootstrapping,

The impedance at the valve anode is increased. Not so at the MOSFET source.

...the extra capacitor there keeps more treble in the mix when the knob is leaning towards bass.

The reactance of those two caps in series, at 5 kHz, is only about 3.8 kilo ohms. Pretty heavy loading on the preceding MOSFET, but if it sounds okay, it is okay.

The output from the MOSFET follower sounds a lot more modern and crunchy, while the output at the anode has a far more vintage overdrive feel and sound to it IMO.

The whole concept of bootstrapping is that the (AC signal) voltage at the anode, and the MOSFET source, are nearly identical, thus increasing the apparent value of the anode resistor.

However, the fact that you are hearing very different timbres at the anode and the source tell you that those two signals are not in fact anything close to "nearly identical"!

Which in turn means you're not really dealing with proper bootstrapping as such, but rather with a form of weird, heavily nonlinear, positive feedback from MOSFET source back to the anode driving the gate.

However, as always with guitar amplification, if it sounds right, it is right. Doesn't matter how "textbook" it is (or isn't!)

If bootstrapping brings the stage gain up to the mu of the valve, do the anode resistors only affect the DC biasing?

Ideal bootstrapping (exactly zero voltage across the anode resistor, due to a perfect unity gain follower) would indeed remove the anode resistor entirely from consideration when it comes to AC analysis. And then it would only affect the DC biasing, as you say.

In your case, the "follower" isn't really a follower, as it's being overdriven, and the signal at the source isn't very much like the signal at the gate (and anode). So, during each cycle of the signal, the amount of "bootstrapping" is varying all over the place, and therefore, so is the effective value of the anode load.

So what you've got, in effect, is a quite nonlinear anode load on the triode. It's value and detailed behaviour will depend on the value of the anode resistor itself, as well as the detailed behaviour of the source-follower. Complex to analyse, as it depends on the non-linear behaviour of the overdriven MOSFET wannabe-follower.

But the bottom line is that it appears to be making noises that you like, so thumbs-up! Success!

should that be ignored and the anode resistor(s) chosen as usual?

In normal audio engineering circa 1950, anode resistors would be chosen based on logical things like power supply voltage, desired anode current, intended quiescent anode voltage, et cetera.

But guitar amps, once again, are an entirely different animal. You still need to be aware of such things as maximum permissible power dissipations, of course. But, for a deliberately non-linear guitar effect or amplifier, the "normal" way to choose the anode resistor is by how it sounds.

This is a lesson I am still learning, as I have many years of the traditional solid-state design-by-equations approach under my belt. But for valve guitar amps, very often, the best way to "design" is with a signal generator, a potentiometer each on the anode, cathode, and screen grid (if any), and a capacitor decade-box or three, if you are lucky enough to own them, or a few clip-leads and a box full of high-voltage caps if not.

The bootstrap capacitor and upper anode resistor form a HPF so it's rather large already, but does the capacitance also relate to the amount or length of time that the voltage can exceed the rails, or is the upper limit here actually the triode's mu being realized where it would normally clip?

LTSpice will probably give you the most honest answer to this, at least for the small-signal (linear) case. Run some frequency response simulations with different value bootstrap caps, and see what happens.

Bootstrapping is a form of positive feedback, and can have some strange side-effects to trap the unwary. (For instance, if you had AC-coupled your MOSFET to the triode, bootstrapping can cause a huge "hump" in the frequency response, usually at sub-audio frequencies.)

...there is a very noticeable dip and "bloom" sometimes over 2 seconds. I assume that this is what blocking distortion sounds like

You're hearing some sort of time-dependent, sliding-bias behaviour. In more extreme form, it can indeed lead to blocking distortion.

But full-on blocking distortion usually is much more pronounced, and much less pleasant - the sound may rapidly "gate" on and off, sounding like your amp is blowing a raspberry, or may even cut out completely for short periods of time, preceded or followed by ugly-sounding distortion.

Depending on your preferences as a guitarist, that "dip and swell" effect may be highly desirable, or extremely annoying.

assume further it's caused by the capacitor/resistor in the MOSFET->Valve biasing scheme?

Probably. Sliding-bias effects are caused by asymmetrical current flow through a capacitor, which charges it up over several cycles, with the resulting DC voltage then shifting the operating point (bias) of some active device in the circuit.

In Rod Elliott's bootstrapped preamp he uses a feedback resistor I've not seen in other versions of this topology.

It appears to be standard negative feedback, just like the textbook inverting opamp circuit I'm sure you're familiar with. It has the effect of lowering the input impedance of the circuit, something we do not usually want from a valve circuit.

That entire circuit is based on the very quixotic concept of attempting to wring Hi-Fi levels of distortion out of an ancient, inefficient, low-gain audio amplification technology, otherwise known as vacuum tubes. It is very much like attempting to build a bridge across a river using ripe cucumbers as your only structural element - utterly pointless.

Could this help me with blocking distortion (assuming it is that)?

Almost certainly not. Negative feedback usually worsens blocking distortion; when blocking occurs, the output voltage drops, so there is little or no voltage being fed back to the input. That means no negative feedback, so the signal amplitude near the input stages of the amp rises, overdrives the amp harder, and worsens the sliding-bias problem that caused the blocking distortion in the first place.

I've been running the bootstrapping MOSFETs without gate resistors, so far without problems, but is this something that might come back and bite me later?

You can potentially have continuous RF oscillation going on, maybe at frequencies as high as hundreds of MHz. There may be no audible symptoms in some cases.

In Alaska or the Yukon, you never leave your house in winter without at least one spare pair of socks, because if your feet get wet, you'll get frostbite within minutes, and a spare pair of dry socks may be all that stands between you and having your feet amputated.

When using MOSFETs, you never go without a gate resistor, unless you're prepared to potentially create a pirate RF oscillator that interferes with legitimate communication signals around you. Worst case (unlikely, but potentially very serious), you could cause interference for emergency response personnel like firefighters or paramedics, or interfere with tower-to-aircraft communications.

could a MOSFET biased with say a zener or tl431 maintain it's Vgs and the valve's Vgk without adjusting the trimpot?

I don't understand your question, please clarify with example schematics.

I hope someone else gets something out of this.

Someone already did, I've enjoyed trying to help you get to your goal. 🙂

Beyond that, if I may get philosophical for a second, you (and I, and almost everyone else, of course) represent the tenth decimal place in a world with 7.5 billion human beings in it. Most likely, any and all of your activities for your entire life will have virtually zero influence on the world at large. Ten decimal places downstream is too small to matter.

So don't do stuff for the benefit of "society" or "the world", because it usually makes no difference whatsoever. But by all means, do stuff for your own personal benefit and enjoyment. Or for your small inner circle - you probably represent the first decimal place in your own family, or collection of close friends, for example, and there you do have a chance to actually make a significant difference. 🙂

-Gnobuddy

 

[andyholmes]

Welcome back! 🙂


And the million Belarusian Ruble* question: How does it sound?

( *Roughly equal to $72.53 CAD; see XE: Convert CAD/BYR. Canada Dollar to Belarus Ruble )

Actually pretty great! Thanks for you extensive reply, I got distracted with another project for about a week here. Running it through my peavey rage like pedal sounds pretty good, but the frequency response of two sequential amps requires some adjustments. Running it into the speaker in an old combo amp via the TDA2030 module sounds amazingly rich. Overdriving the input just sounds gorgeous, it's hard not to just stand there and let power chords ring out, over and over.

The reactance of those two caps in series, at 5 kHz, is only about 3.8 kilo ohms. Pretty heavy loading on the preceding MOSFET, but if it sounds okay, it is okay.

...LTSpice will probably give you the most honest answer to this, at least for the small-signal (linear) case. Run some frequency response simulations with different value bootstrap caps, and see what happens.

I have to admit this is the first time I haven't just crammed someone else's tonestack in my circuit verbatim. Do you happen to have any solid bookmarks on tone stack creation? Or do you just sort of look at it as an amplitude bleed with impedance/frequency response?

I spent a week (or two) trying with various curve capturing and curve plotting tools to get a usably accurate low-voltage tube working in SPICE, but I just don't think any of the current "models" of modelling tubes (eg Koren) are going to hack it. In any case I've really only had usable results simulating portions of my circuit and I haven't the experience yet to catch things like that ahead of time.

The impedance at the valve anode is increased. Not so at the MOSFET source.

The whole concept of bootstrapping is that the (AC signal) voltage at the anode, and the MOSFET source, are nearly identical, thus increasing the apparent value of the anode resistor.

However, the fact that you are hearing very different timbres at the anode and the source tell you that those two signals are not in fact anything close to "nearly identical"!

Which in turn means you're not really dealing with proper bootstrapping as such, but rather with a form of weird, heavily nonlinear, positive feedback from MOSFET source back to the anode driving the gate.

However, as always with guitar amplification, if it sounds right, it is right. Doesn't matter how "textbook" it is (or isn't!),

My impression from overdriving at different levels is that being overdriven through the first stage, it always comes through the MOSFET follower and I've scoped it being quite squared, then coming out the second stage anode is when swing at the anode exceeds the rails, the MOSFET can't keep up at the source and the tube clips more gradually as the capacitor drains. This is my impression, but negative/positive feedback is one of those concepts I'm aware of, but don't completely grok yet.

...And then it would only affect the DC biasing, as you say.

...But the bottom line is that it appears to be making noises that you like, so thumbs-up! Success!

...But, for a deliberately non-linear guitar effect or amplifier, the "normal" way to choose the anode resistor is by how it sounds.

This is a lesson I am still learning, as I have many years of the traditional solid-state design-by-equations approach under my belt. But for valve guitar amps, very often, the best way to "design" is with a signal generator, a potentiometer each on the anode, cathode, and screen grid (if any), and a capacitor decade-box or three, if you are lucky enough to own them, or a few clip-leads and a box full of high-voltage caps if not.

That's good to hear because that's pretty much how it went for me! I spent some time crunching numbers and reading various Valve introductions and biasing summaries, but in the end a sine wave and my $16 oscilloscope beat out my calculator and LTSpice for results.

You're hearing some sort of time-dependent, sliding-bias behaviour. In more extreme form, it can indeed lead to blocking distortion.

...

Depending on your preferences as a guitarist, that "dip and swell" effect may be highly desirable, or extremely annoying.

...

Almost certainly not. Negative feedback usually worsens blocking distortion...

That's what I figured, I did spend some time at one point chasing that effect in some effect pedals (that are over there on perf boards not it boxes) but the effect is a little strong. I'm going to try lowering the capacitance to see if it helps and if so, maybe I can just make up for the high-pass effect elsewhere.

I did some cursory testing of negative feedback there and the results surely agree with that.

You can potentially have continuous RF oscillation going on...When using MOSFETs, you never go without a gate resistor...you could cause interference for emergency response personnel like firefighters or paramedics...

No good, with my electrical intuition (and luck) it will probably be me they're coming for anyways! Gate resistors it is!

I don't understand your question, please clarify with example schematics.

My question was a little vague, mostly because I was trying to imagine a convoluted biasing solution where I could direct couple the MOSFET buffer to the grid of the valve. I think your approach with the cap over a mini-divider is really the best though, ingenious little subcircuit 🙂

I did have a bit of fun on the input tying a JFET source directly to grid, with no resistor to ground. Self-biasing the JFET, I found the source was pulling the grid up to about 300mV, or maybe it was the grid pushing the JFET source up, but unfortunately +/- 300mV on the input buffer doesn't suit my needs.

Someone already did, I've enjoyed trying to help you get to your goal. 🙂

Most likely, any and all of your activities for your entire life will have virtually zero influence on the world at large.... do stuff for your own personal benefit and enjoyment. Or for your small inner circle - you probably represent the first decimal place in your own family, or collection of close friends, for example, and there you do have a chance to actually make a significant difference. 🙂

Too true. Coming to that realization a few years back is when I started doing all the things I wanted to, but was afraid wouldn't matter or I couldn't do perfectly. Bought a ukulele, started programming python, chased a girl to Winnipeg...well not everything works out 😉

I find your comments super helpful, you seem to have experience giving just enough of a solution not to impede a persons education, much appreciated!

 

[Gnobuddy]

Actually pretty great!

Congratulations, one can't ask for more from a DIY project!

Do you happen to have any solid bookmarks on tone stack creation?

No, and honestly, I doubt if such a thing exists. For Hi-Fi, the active Baxandall tone control was so good that it replaced everything that had come before, and became almost universal for many decades. For guitar, Leo Fender's cheapness won out, and he settled on a barely functional circuit that used a minimum of components; because of Fender's commercial success, that circuit then became widely adopted by Fender-copiers around the world, including Marshall and Vox.

That popularity in turn led to tens of thousands of guitarists spending far too many hours learning to coax their desired sound out of that horrid Fender tone control circuit, until eventually they became accustomed to it, and eventually, decided that this was the proper way any guitar amp tone control should work.

I'm glad that the Internet, and the various guitar forums that followed, have since given birth to a variety of alternatives to that FMV tone stack. The curious can at least find alternatives to explore, now!

I spent a week (or two) trying with various curve capturing and curve plotting tools to get a usably accurate low-voltage tube working in SPICE, but I just don't think any of the current "models" of modelling tubes (eg Koren) are going to hack it.

There is a supposedly more realistic mathematical model by Dirk Reefman, though I know little more than that.

My own few attempts to use existing LTSpice valve models produced such poor results that I lost interest in the whole concept. As you say, the existing models leave a lot to be desired.

negative/positive feedback is one of those concepts I'm aware of, but don't completely grok yet.

For our purposes, you can think of it this way: negative feedback reduces distortion, makes a circuit more linear up until clipping begins, and then causes harsh and abrupt onset of clipping.

Positive feedback usually results in oscillation, and is usually avoided in amplifiers. But very small amounts of positive feedback cause many of the opposite characteristics that negative feedback do - i.e. it increases distortion, makes a circuit more nonlinear, and softens the onset of clipping.

All these are bad things for Hi-Fi, but are also exactly the sort of things that make a guitar amp designers ears perk up!

Interestingly enough, because a source-follower has a voltage gain of less than unity, it is a good platform to experiment with small amounts of positive feedback, without quite as much risk of bursting into outright oscillation.

I think you may have stumbled across a very useful guitar amp building block when you added bootstrapping to your triode/source follower combination!

I think your approach with the cap over a mini-divider is really the best though, ingenious little subcircuit

I'm glad that worked for you, but I can't help but wonder - why not just place a pot across the supply rails, and feed the triode grid from the wiper? Capacitor couple the signal, just as my little circuit does. And you'll have Dial-a-bias (TM) without involving the MOSFET at all...

Bought a ukulele, started programming python, chased a girl to Winnipeg...well not everything works out

Sorry that didn't work out - but at least you gave it a good solid try. That's the best any of us can ever do!

I find your comments super helpful, you seem to have experience giving just enough of a solution not to impede a persons education, much appreciated!

Thank you, I very much appreciate the comment!

I have indeed spent a lot of years as a teacher. I've always thought an educators job was to try to assist the learner to think his or her own way to a solution. Give a girl a fish, teach a girl to fish, you know!

-Gnobuddy