Check also the FSB thread!
Peavey, Pavey, Pevy, or is it Pee-vee? One of my favorite amplifiers is still to this day the Peavey Bandit 112 I was gifted for a birthday. That was my first actual guitar amplifier, but even after having the chance of playing and owning fancier, more expensive, more tube amplifiers, I still come back to this one once in a while. Having made a cab sim, I thought it would be nice to have a pocket size version of the pre-amplifier to use with it without bringing the whole amplifier with me.
Let's be honest, the lead channel is nothing amazing, but one can't expect an amplifier to do everything to one's aesthetic requirements. People have uses for that one, but it's probably what they tend to remember better, discounting the rest of the package based on that memory.
I'm here to talk about the clean channel. In particular, I tend to use the "vintage" mode, even if my model offers three. Maybe some people will think there's not much to ask from a clean channel, or that the category begins and ends with the letter F; I think there's something that sets this aside from the usual examples, even if we will see that it's topologically very similar.
Analysis and principles
Here's the reference schematic to follow along. The basic building block of all the preamplifier stages is a (patented?) stage composed of a NPN darlington pair made of completely unremarkable transistors. Values might change, the emitter might be bypassed differently, but the common template is clear. The base is biased open-loop from a dedicated voltage reference, which would need to be different depending on the stage. A diode is connected from the base to one of the voltage references, anode to base.
Focusing on the clean channel, once we treat those stages as building blocks, we can see that the arrangement is pretty much the same found in Fender and other amplifiers from the 60s on: input stage, volume control with bright capacitor, a second stage driving the TMB tone stack and out.
The second stage deserves some extra words, because it's fully bypassed with negative feedback; this turns the input into a pretty good virtual ground, while the output has a deceptively low output impedance that has no problems driving the tone stack.
Making the circuit
My first step was to recreate the circuit in LTspice to analyze things like bias, clipping, frequency response and such. Since I had just done some work on the Bandit, I also had measured bias voltages to compare against.
The diodes
The first thing I noticed was that I couldn't find any influence of the base diodes comparing the circuit with and without them, no matter the input amplitude or frequency. Output waveforms looked identical and the diode current was nanoamps peak, too small to matter.
I did my tests with the clean channel, but since neither of them conducted ever and the stages are the same in this regard, the results can probably be extended to the lead channel as well. The diodes seemed clever, an attempt to perhaps emulate grid conduction on a vacuum tube, but in hindsight it would make sense that going one diode drop higher, a center-biased transistor would already be saturating for a while.
Although admittedly this could have been analyzed more in-depth, I'm confident in my choice of ignoring these diodes in my circuit. My goal was to recreate a clean preamplifier after all; so that was my first departure from the original.
The pot
The second challenge to face was the "treble" pot. In the original, it is 300k center tapped with 25% A taper. I had no interest in finding such a weird combination of specs, and it would have gone against my principles and my goal to make circuits that are accessible to everyone. So I had to do something about that: after all, center tap and switches aside, this is the same tone stack found in Marshall amps topology wise, and the goal seemed reasonable.
It's hard to find good information about tapped potentiometers, aside from their few uses in the past. It's even harder to find specific information about logarithmic ones. In essence, my dilemma was how to treat the resistors switched in parallel to the tapped half; in a linear center tapped potentiometer, the tap resistance to either end is half the potentiometer resistance. In a logarithmic one, given also how the tracks are made, it makes sense that the tap lies on the center-position resistance of the particular taper, in this case 25% of the total.
This way, the upper resistor being much larger makes sense, since it would have to, for the effect on tapering to be similar.
With this established, it's relatively easy to treat the two halves of the potentiometer to be separately tapered by the parallel resistors (or if you want, two potentiometers, the one where the wiper is lying being tapered, the other making up a fixed resistor equal to the parallel combination). In particular, tapering with a resistor in parallel to the whole potentiometer skews the taper towards a W taper, with the pot spending more of the travel in the middle region of the voltage divider.
What remains to be done is just to calculate the total resistance in both settings, to recreate, taper aside, an equivalent circuit. The calculations are:
Rvin=300*0.75+300*0.25*39/(300*0.25+39)=250k
Rclas=300*0.25+300*0.75*470/(300*0.75+470)=227kAs luck would have it, a common 250kA potentiometer seems a good substitute when both modes are considered!
The headroom and FR
At this point, I just made extensive use of LTspice to approximate the circuit with more standard values, but there's one other thing to consider: the original runs on a 28V supply. That's not unusual for an amplifier where those voltages are useful for other things, and within the reach of common charge pumps, but do we need that much?
Since the input stage has some gain (~16dB), I can instead scale that down by the ratio of the supplies (28/9=3.1) and everything else would follow. Even better, since that would leave the input stage with just 2x gain, I can substitute the darlington pair with a single transistor and the generous emitter degeneration allows me to keep the original, somewhat low, input impedance.
The bias
The original uses a resistor ladder to make multiple DC voltages from the supply for biasing the different stages. While this somewhat works, I quickly got tired of how fiddly, beta dependent, gain dependent and unreliable this method is. In fact, measuring the original amp showed that the bias isn't perfectly centered in all stages, something it can get away with given the large supply voltage. Furthermore, the resistor ladder uses the least components to make separate, decoupled DC references, but it was maddening to achieve the precise voltages required with the interaction of the voltages with any single resistor.
I quickly switched to feedback bias for both stages: DC-only feedback for the first, a favorite of mine that doesn't impact input impedance or gain; for the second, given the feedback already present, I just made that DC-coupled.
Arrived at this point, I can show you my first iteration of this idea, which is probably the most faithful to its inspiration:
The input stage has the same input network as the original, just with a different bias. As I said the gain has been reduced by the supply voltage scaling and that made a single BJT sufficient for the same job.
The first stage in the amp has a shelving network in the emitter bypass, but given the very small effect it had, I decided to adjust the "bright" networks around the "volume" control to account for that and keep the same response at all settings.
The second stage is pretty much as in the original, since I liked this discrete version of feedback inverting amplifier. The current in both stages has been kept similar to the original by scaling the collector resistances as well.
The tone stack keeps most of the characteristics of the original, but reduces the poles of the "mode" switch to just two: the treble control has been simplified as discussed before, and the pole switching the bright capacitor has been made into a separate switch for more flexibility. The response has been again tested against the original (with the same substitution of the "treble"pot).
My amplifier, and the schematic above, actually show three modes, the third being "warm". I've never been a big fan of this mode, and as further motivation the older models didn't have it either, so I decided to keep only "vintage" and "classic". The amplifier uses some large complicated 4P3T switches to achieve all the above, and I don't want to keep those.
It was fun to reverse-engineer the 21k and 47k resistor into a standard 68k just to find out that's what was used in the older model.
Finally, a common emitter buffer gives nice low output impedance with the same 220k load on the tone stack as the original.
Can I please have some headroom
Why didn't I stop there? the circuit above has very very close frequency response and headroom to the original at all settings, keeping the same discrete transistor nature.
The thing is, the original doesn't have much headroom after all.
It's not something you notice at first, since the amplifier gets very loud well before having the volume knob at half, but it definitely doesn't sound clean past that anymore, although the clipping seems a bit smoother, maybe because later stages come into play at that point.
What my almost perfect 9V emulation gave me was a not very loud preamp, which distorted somewhat unpleasantly on large peaks when set for a decent output. While the output could be boosted by rasing the gain of the mixer input, it struggled to match instrument level, let alone line level which was my original goal. I wasn't very interested to emulate the clipping behavior of my supposedly clean amplifier channel after all, so it was back to the drawing board.
To be clear, since headroom is a poorly defined and often misused term, by "headroom" here I mean the largest input signal that doesn't cause clipping at any given setting, or just at maximum volume setting.
Successive approximations
My first attempt at solving the issue was to decrease the gain on the second stage, so that "volume" would have to be set higher to make it clip, all other things being equal. This somewhat helped but in small measure, and only made the lack of output worse. Also, the input stage gain could be increased a bit before clipping, to improve SNR. Do you start to see where this is going?
To solve the issue completely, the second stage gain would have to be reduced to 1, and the tone stack output liberally amplified to match what my desired level would be. That would just leave me needing two clean, high input impedance amplifiers and a buffer, which was only necessary to isolate the volume control network from the tone stack.
Since I had already strayed pretty far from the original in terms of approximations (but not in frequency response!), and high input impedance, arbitrary gain stages isn't something discrete amplifiers are very good at, I reached out to the familiar TL072 dual op-amp with no so many regrets in leaving behind the original amplifier topologies.
The new circuit retains all the characteristics and good parts of the previous and its inspiration, but with very good output volume and gracious clipping limited to loud peaks when you crank the volume and hit the strings hard, if you want a bit of that.
One thing deserving some explanation is the use of LEDs. The TL072 is a very good choice as op-amp for guitar, because at its price it offers hard to beat noise performance with this particular signal source, being JFET input. The only downside is the hard to ignore crackly saturation behavior when the output hits the rails and recovers. Some other op-amps might behave a bit better, but a more complete solution is to prevent the saturation using feedback diodes. Red LEDs, in particular, come very close to the peak signal swing of a 072 at 9v, so they do this job without a large cost in output headroom, which was confirmed by a practical comparison.
While the pair on the input protects that stage from large or boosted input peaks, the pair on the output deals with the clipping due to the preamplifier itself. 11x is a substantial amount of clean gain, but the tone stack drops some amplitude and the volume control even more. In practice these LEDs will conduct only with the volume maxed and for brief peaks (I've tested in a dark room so you can trust me!), providing an emulation of "power amp clipping". Furthermore, they provide a limit to the output to levels that are larger than what you'll ever need.
The gain on the input stage is modest, just 2.5x, and was chosen to give a meaningful improvement to SNR without clipping with any normal signal level. The output gain is generous and achieves unity with the input signal with the volume at a bit more than a quarter.
Of course, while I think 9V operation has all the headroom one would need (after all, it's just a matter of decreasing the gain), using a 12V or 18V supply might skip the need for the input LEDs (keep the output ones, you'll never need more output) and possibly increase the input stage gain, with every dB going in the SNR spoils. In this case, a "for 18V" switch achieving this might be more useful than the often unnecessary switches that bypass an internal charge pump.
In hindsight, separating the bright and mode switches, at first done for practical considerations, has revealed to be a welcome choice, because the combinations "vintage no bright" and "classic bright" not possible in the original are very useful, and the latter might be my new favorite.
To conclude the analysis, plots comparing the original preamplifier and this circuit. The tone stack has all knobs at half and wouldn't influence the results anyway, since it's the same in both.
Demo
Layout
Update
Following up my own suggestion, I've made a little update to the schematic. This falls into the former scenario: "keeping the same input headroom (which is good), but improving SNR". There are two ways to achieve this: the schematic shows the first, which is to use a TLC2272 op-amp, removing the LEDs and increasing the gain (the gain increase roughly equals the output headroom increase = same input headroom). You pay for it, but this op-amp offers great noise performance, it's FET (so no current noise) and is even R2R.
The second option, which is the one I'm currently using, is to use a TL072 at 18 V instead, just as suggested in the original post. The schematic is still the same above. While a 2272 offers almost three times the headroom over the LEDs at 9 V (or almost twice over the bare 072, but that's not recommended), a 072 at 18 V has almost exactly three times the headroom of a 9 V 072 because of the additive quantities in the ratio. Again, this is done for the purpose of noise, not headroom for its own sake. You also get some more output from it, which is useful.