The Hundred-Buck Amp Challenge

The channel in FETS (both JFET and MOSFET) behaves a bit like a voltage-variable resistor, controlled by the gate-source voltage. This property has been used in phasers (the audio effect, not the Star Trek weapon!), voltage controlled oscillators, audio compressors, et cetera.

The trouble is that this drain-source resistance is not very linear, except for very small voltage excursions. Usually it works fine if you keep the AC voltage across the FET to no more than a couple hundred millivolts. Beyond that, things get increasingly nonlinear.

I found this out in the 1980s, when I was fixated for a while on trying to make a guitar compressor to increase the sustain from my home-made electric guitar. I tried to use a JFET as the feedback resistor in an inverting op-amp circuit, and that caused more distortion than compression. But if I put the JFET ahead of the op-amp in the signal chain, where the signal was much smaller, it worked okay.

My guess is that this is the problem George ran into. The signal level at the cathodes of some of the later valves in that signal chain would be quite large - probably the FETs didn't cope with it very well.

Lately I have been chewing on a completely unrelated idea which requires electronically controllable resistors, so I've been looking at digipot data sheets recently. Microchip (the PIC company) has some new digipot chips that can cope with 36 volt (or +/- 18 volt) supply rails, and can cope with analog AC signals within that range. A few of those might cope better with the fairly large signal swings at the cathodes of a valve guitar preamp.

Digipots are digital devices, so George's simple analog compressor would give way to something with a microcontroller; read the output signal strength with an ADC, figure out the proper value to set the digipot to, send a digital signal to the digipot, rinse and repeat.

That's what I don't like so much about digital; it always seems to take a few million transistors in a digital circuit to do the job of one or two transistors in an analog circuit!

The bad news is that those new chips are only available in surface-mount packaging. A break-out board, magnifying lens, 3rd hand tool, and steady hands will be required. And for those of us with no-longer-young eyes, probably a bit of cursing, too!

A while ago I ran into another simple analog compressor scheme for a guitar amp. This one used the output signal strength to control the DC bias on the grid (yes, G1) voltage on the input valve, which was a pentode. That in turn varies the gm of the valve, so it acts as a compressor. The scheme can be seen here:
"Spunky" Tube Amplifier - How It Works

-Gnobuddy
 
I'm starting to get my head around mosfets in the anode circuit Tubelab goes and puts them in the cathode.......The channel in FETS (both JFET and MOSFET) behaves a bit like a voltage-variable resistor

Not my idea.....my "secret weapon" in the 80's was my ADA MP1 MIDI guitar preamp. It is a digitally controlled tube / opamp hybrid guitar preamp using mosfets in the cathodes of a pair of 12AX7's to control the gain. It worked rather well by 80's standards. I still have mine, but it has become rather microphonic.

I tried the mosfet in the cathode trick in that amp, but it wasn't implemented correctly and did not work. All of the cathode mosfets were removed and I never looked back. The mosfet buffers in the plate circuit are used to increase the gain by buffering the load of the next stage so that the tube only sees its plate resistor as a load. I used tubes that had a Mu of 33 so I added an extra tube, and all the mosfets. The amp had so much gain that it was highly microphonic, so some mosfets were removed and a stage or two was bypassed.

The mosfet in the cathode circuit can be made to work. The idea is to create a variable resistance in series with the bypass cap.

I have found a better way.....An LDR. It is a Light Dependent Resistor, basically an LED and a photoresistor sealed in a small plastic box. Drive the LED from an Arduino, you can tweak all sorts of things........

New project under development, time to retire the MP1.

Note from the burnt parts department.....those fancy digital pot chips blow up in tube amps. Even the "high voltage" versions don't eat the signal levels found in tube amps unless some limiting is applied which will cause distortion (not always a bad thing).

The schematics for the ADA MP1 are somewhere on this site....I think. I know I downloaded them somewhere, recently.

Mine can be seen in this, my most extreme (and crazy) guitar amp experiment. There is an 833A tube running in SE and making over 200 watts on 1500 volts. You can see a thin yellow Radio Shack quality clip lead running from the power supply to the high voltage capacitor behind the glowing 833A. It caries the main B+. The power supply is rated for 1500 volts at 1/2 Amp. That's why there is a Lexan shield between me and the amp, and a fire extinguisher on the floor nearby.

The MP1 is on the left bench. It's sitting on a book and a foam pad to prevent microphonic feedback. This "test amp" only existed for a week, but could be heard from two blocks away.
 
I have found a better way.....An LDR. It is a Light Dependent Resistor, basically an LED and a photoresistor sealed in a small plastic box.
I have used that approach too, and I too found it worked better than a JFET.

I also found some LDRs had really long recovery time constants, taking several seconds to return to their full dark resistance once they'd been briefly exposed to light.

We have incredibly efficient LEDs now, so what took 10 - 20 mA to drive to adequate brightness now only needs 1 mA or less. That would make the LDR/LED approach even more attractive now.

Unfortunately LDRs seem to be pretty thin on the ground nowadays. Perhaps because the cadmium in them is toxic, and no longer allowed under ROHS regulations?

Here in BC, I no longer see much 63/37 solder any more, either. Dunno what that's about, but it seems to have disappeared off the shelves around 2014 or so.

Note from the burnt parts department.....those fancy digital pot chips blow up in tube amps. Even the "high voltage" versions don't eat the signal levels found in tube amps unless some limiting is applied which will cause distortion (not always a bad thing).
Duly noted. FWIW, I was talking about using them at the cathode - where, at least in preamp stages, the peak signal levels are typically only a couple of volts or less.

I've been experimenting with a 6JW8 noval small-signal TV triode/pentode. To my surprise, the pentode is able to put out 120 volts peak-to-peak while running on about 140 volts B+, with an impressively low saturation voltage of only 20 volts. An older 6AG5 maxes out at around 90 volts peak-to-peak on the same B+.

Either of those voltages is far, far more than the +/- 18 volt limits of the "high voltage" Microchip digipots, and I wouldn't dream of letting a digipot come into contact with them.

Mine can be seen in this, my most extreme (and crazy) guitar amp experiment.
I saw that pic some years ago - linked by someone on the Australian guitar gearheads forum, if I'm not mistaken. That crazy extreme guitar amp was seen around the world! :D

-Gnobuddy
 
The mosfet in the cathode circuit can be made to work. The idea is to create a variable resistance in series with the bypass cap.
I forgot to mention, I notice there is only going to be a fairly limited range of control available in this way with many preamp valves.

Inside the valve there is an effective cathode resistance of roughly (1/gm), which is around 670 ohms for a typical 12AX7. Outside, you have a typical external cathode resistor of 1.5k. So the total resistance between ground and cathode can only be varied between the limits (670+1500 ohms) and (670 ohms). That's only about a 3:1 range, or 10 dB of control range.

I noticed this while trying to calculate the value needed for a cathode bypass cap. Almost every online reference I can find uses the typical high-pass RC filter formula, and relates it to the value of the external cathode resistance, which is quite sadly wrong.

In actuality, the impedance at the cathode is mostly determined by effective internal resistance (roughly 1/gm) of the valve), and not by the external resistor.

The external cathode resistor Rk does have an effect - to be more accurate, the Thevenin resistance at the cathode is (1/gm) in parallel with the external cathode resistor. But usually 1/gm is smaller than Rk, and is the dominant factor here.

Years ago I've seen the same mistake made - in print - about transistors. Once again, it's actually the gm of the device that primarily determines the size of the emitter bypass cap, and not the value of the external emitter resistor.

-Gnobuddy
 
"high voltage" Microchip digipots, and I wouldn't dream of letting a digipot come into contact with them

I found some that were good to 36 volts or so...don't remember exactly, or who made them....Analog Devices? Maxim? Either way I put a reasonable resistor in series to drop some level and attempted to use them for a "master gain" type control at several places in the preamp stages. I wouldn't have believed that a wimpy 12AX7 type tube could poof those things, but I killed two or three of them. Hey I'll try anything once...or twice.

That's only about a 3:1 range, or 10 dB of control range.

That's enough when you do it to all 4 stages in a high gain preamp.

That would make the LDR/LED approach even more attractive now.

Some LDR's do have a significant recovery period, especially if you hit then hard (full bright). Look at some of the Soldano schematics. They use LDRs as switches, full on or full off, avoiding the linear region. I got a bag of surplus LDR's from Electronic Goldmine when they had a sale. For 20 cents each, I don't mind blowing a few up.....they don't like being used as a plate load in a 12AX7 circuit, but they do work Soldano style for switching resistors in parallel with the existing plate resistor.

Here in BC, I no longer see much 63/37 solder any more, either.

I found one pound rolls for $5 at a ham radio show a few years ago, so I bought all the guy had. I'm headed out to the "worlds biggest hamfest" (Dayton Ohio) tomorrow.....I will come back with tubes, solder, parts, and ???? late Saturday night.
 
And only $1000 or more in man hours to scavenge/clean/reassemble the parts. LOL
I've only once actually run across vintage equipment to plunder / rebuild. An acquaintance gave me most of an old Bogen P.A. system, minus the turntable, which he kept.

After spending several hours on just one task - cutting the lopsided hexagonal hole for a modern IEC electrical inlet in the steel chassis - I burned out on the project and set it aside. Too much work for too little fun, at least at that point in my life!

So I hear you on the man-hours required to rejuvenate a piece of discarded scrap!

I guess I'll have to re-visit that project some time when I'm in a more settled frame of mind. I don't like starting things and not finishing them.

-Gnobuddy
 
Junk tube gear parts

I spend quite a bit of time searching Ebay for junk tube gear parts.

I try to buy items like broken reel to reel recorders, old test gear, and some items that actually are or were audio amplifiers.

Mostly I go for items I can get for $40 or less including shipping.

I have a favorite piece of HP gear I spend $75 or less including shipping for. It has premium Mullard/Amperex tubes and a beefy PT I use. Very good salvaged parts value from this particular item.

You really have to know what you are looking at, search for schematics/details online, to find items of good value.

There are lots of broken/junk amps offered at ridiculous prices for the uninformed to buy. You have to avoid those, and search fringe tube gear items to get premium Mullard/Amperex tubes and other high quality parts for cheap.

Then you have to be willing to spend untold man hours cleaning/extracting/testing those parts. It keeps me busy. LOL
 
Whatever works to keep your enthusiasm going!

I stumbled across this thread some years ago, and that's when I found out that there were plenty of high-quality, NOS valves out there for a buck each, not to mention other unpopular types that might be $3 or so.

That discovery was a wonderful one for me. It meant that if I was willing to design my own circuits, I could find all the (unpopular) valves I needed, very inexpensively, without having to compete on Ebay.

Add in voltage doublers and quadruplers for generating B+ (an idea I got from the very smart Roly Roper, of the Aussie Guitar Gearheads Forum), small switch-mode power supplies for heaters, and 70V audio line transformers for outputs, and I feel like a kid in a candy store!

The one really bad thing about this approach is George's discovery that someone is willing to buy up these old valves by the tens of thousands as soon as a good use for one is found, taking them off the market for good. And, my gut instincts say, quite probably crushing them into powder to prevent possible competition and loss of sales.

If someone makes $1500 valve guitar amps, and someone else shows the world how to make $100 valve guitar amps using $1 NOS valves, and the first person has no ethics whatsoever, then buying and crushing the entire world stock of those $1 NOS valves makes perfect business sense. Spend a few thousand to eliminate competition that might cost tens of thousands? Perfectly logical, if you're a snake disguised as a human. :bawling:

-Gnobuddy
 
Now how often does your local lumberyard bring in a pallet of reject guitar tops? Might not be good enough for the big guys but I managed to find some useful pieces for myself.

Went%20out%20for%20transistors%20-%20one%20hundred%20bucks%20later._zpstmfu205y.jpg
 
I'm not aware of spruce having been used on any of the commonly available electric guitar designs (not counting electrified acoustics). I wonder what a semi-hollow electric guitar with a thin spruce top would sound like? Maybe a little closer to an acoustic guitar than an ES-335, if the top was thin enough, and braced like an acoustic?

Personally I'd be quite interested in a thin-bodied electric guitar like that - one that sounded a lot like a traditional steel-string acoustic guitar, but without the big bulky body, high action, and limited number of frets.

I've read that the original design intent of the ES-335 was to split the difference between the sound of a solid-body electric guitar, and a hollow-body acoustic guitar. If that was true, IMO they ended up a lot closer to the solid-body-electric end of that spectrum. A good 335 sounds sweeter than a solid-body electric, but not very much like an acoustic guitar.

I know a few manufacturers have taken stabs at the other end of that sound spectrum (i.e. thin, but sound like an acoustic), but the ones I've tried have been uninspiring, lacking both the beautiful sound of a good steel-string acoustic guitar, and the easy playability of a good solid-body electric guitar. A friend has a Taylor T5, but the sounds he get from it are very much at the solid-body end of the spectrum.

-Gnobuddy
 
The theoretical voltage-to-current transfer function for a JFET is a square law, and a little paper-and-pencil math suggests that it should generate a good amount of 2nd harmonic distortion (and, within the limitations of the mathematical model, no higher order harmonics.)

Work has had me busy again lately, no time for soldering in the last few days, but I did throw a couple of JFET simulations together in LTSpice.

Simulation isn't reality, of course, but if LTSpice is even halfway accurate, it looks like the simplest of JFET common-source amplifiers (see attached pic) is all it would take to make a nice amount of almost pure 2nd harmonic distortion. This is the stuff that makes for a nice "clean tone" to a guitarist, based on my experience so far.

Drive the circuit a little harder (in the simulation), and the negative tips of the waveform start to compress as the JFET hits saturation. The Fourier transform then shows the usual picket-fence of harmonics emerging from the noise floor. Not just 2nd harmonic.

So there's a good chance that this circuit will transition from fairly triode-like cleans to slight overdrive as it's driven harder.

In LTSpice I cheated and biased the JFET by simply typing a DC offset voltage into the signal source; in reality one would need to add some additional circuitry. (The simulated circuit is entirely DC coupled, as it avoids having LTSpice simulate the relatively slow initial transient as all the caps charge up to operating voltage.)

BJTs behave nothing like this, by the way, in case anyone is wondering. Their exponential voltage/current transfer function makes nasty-sounding high-order distortion even at very low drive voltages. You can clean it up with negative feedback, but then you get the usual thin, sterile, cold, too-clean, solid-state guitar amp sound.

As I mentioned earlier, I am trying to make a guitar amp on a tight budget as a gift for a friend, and it's going to have to be solid state to fit in the budget. Hence the interest in a JFET preamp - my turn to hunt for the elusive "make a solid state device sound like a good valve" Holy Grail.

-Gnobuddy
 

Attachments

  • JFET_square_law_001.png
    JFET_square_law_001.png
    45.8 KB · Views: 227
I know from experience that BJTs sound bad (harsh) when they begin to distort, and I know from theory why that's true (tight exponential transfer function generates a host of harmonics even for small input voltages.)

But I thought I'd let LTSpice loose on a single-transistor common-emitter BJT amplifier stage, to see what showed up, and to compare it with the single common-source JFET amplifier stage in my previous post.

As before, for simulation purposes, DC bias is generated by the input source, simply to avoid dealing with capacitor charging time-constants in LTSpice.

The attached screen capture is for an input strength of only 22mV (I think that's peak value in LTSpice). Even with this small input voltage, you can see a forest of harmonics, all the way out to at least the tenth harmonic(!)

You can also see a forest of non-harmonic frequencies, including sub-harmonics of the 1kHz input. There is lots and lots of intermodulation distortion going on here, in other words.

IMO bipolar junction transistors, and electric guitars, are a match made not in heaven, but rather, in Dante's ninth circle of hell. The two just do not get along!

Run an electric guitar into a BJT without feedback, and there is too much signal voltage, and too little input impedance, so you get nasty distortion and bad frequency response.

Run an electric guitar into a BJT with feedback, and you have sterile "too clean" tone, bursting into abrupt "too harsh" clipping, and there is still difficulty getting adequate input impedance.

-Gnobuddy
 

Attachments

  • BJT_exponential_law_distortion_001.png
    BJT_exponential_law_distortion_001.png
    53.4 KB · Views: 189
I've played with a diode-resistor string in an op-amp's feedback look in an attempt to get more gradual clipping in a solid-state circuit. I posted a schematic somewhere on DIYaudio a few years ago.

With the re-release of the Minimoog (the current Moog Music company took pains to make the current model as much like the original as practical, even claiming to have original transistors remade, though I'm skeptical of that), there's lots of interest in it.

The Moog "ladder filter" has always had a unique sound, apparently because of the mild distortion the circuit gives. It's a remarkable design, the variable "resistance" in the filter is the dynamic base-emitter impedance of a transistor which varies in inverse proportion to the (DC) base current - the higher the current, the higher up the exponential curve of Vbe, and the lower the resistance. The signal levels are kept low, in the 20mV range, to keep distortion to a reasonable level.

I'm wondering if the circuit would give "good" distortion if it is used without the the four capacitors that make it a filter.