Sapphire amplifiers and the TDA2030 "phender"

My favorite amplifier is probably the Chempion I've built a couple of years ago. Since I use the amplifier mostly clean, relying on my pedals to give me a range of sounds, I was recently wondering if it was possible to recreate the same sound and output power in a smaller, less fragile solid-state amplifier. I had actually already laid out the bases for this with my line of "Sapphire" amplifiers, my alternative to the popular gem-themed Ruby 386 amplifier, so let's take a step back and look at those first.

The Sapphire

The Ruby is a very nice amplifier for how simple it is, but there were a couple of things I would have liked to change if I were to do it my own way. First, I don't want to use JFETs if possible, since they're either expensive or tiny these days; second, even when paired with a proper speaker, I feel like it's missing one thing that is present in so many guitar amplifiers that it's become part of the sound (also because it alters the timbre no matter what, and often not as intuitively as the controls would suggest): the FMV tone stack. Not all amplifiers have it of course, and this makes the Ruby more of a "Tweed" style amplifier.

I had to replace the JFET with something, lest I end up with the sad 50k input resistance of the 386 itself as in the old "Smokey", and since the tone stack has a big insertion loss and dual op-amps take up the same space as single ones, that's what I chose. This also allows me to add more gain and diode clipping instead of relying on the 386 internal gain for distortion.

This is the "original" Sapphire if there's one, and if it seems more complicated than it needs to be, don't worry, because we'll see that modularity is the key.

For example, the tone stack shown is Marshall-style, scaled down a bit. If one prefers a Fender-style FMV, you can swap that in, as I did for the version I built.

Same goes for the channel switch: I think it's fun to make your amplifier seem larger than it actually is by introducing two different sets of controls for gain and volume with a way to switch between the two, since the same topology can do all the clean and all the distortion you want, but that's of course optional. The gain switch wiring assures the negative feedback is kept at all times.

The preamplifier is wonderfully simple for how much you can get out of it:

• A non-inverting amplifier as input stage buffers the input and allows any desired amount of amplification, from unity gain up to ~46dB of gain for the middle frequencies. C2 and C3 are set so that you go from a flat response at low gain (thanks to the inability of this amplifier to go below unity gain) to a mid-focused response at high gain, as in many distortion circuits in both amplifiers and pedals. Of course the values can be changed to taste. The diodes allow for distortion at higher gains, but also limit the output of this stage to a defined peak level (well, it's soft clipping so it's not exactly defined, but close enough since the gain drops to unity past the threshold). This is necessary, so that the 386 itself isn't clipping and the source of clipping is the preamplifier only. If you turn the gain down, the gain not only drops enough that the diodes aren't clipping, but they get shorted out completely at minimum for the cleanest sound.

• Then we have the tone stack part, which I've already explained.

• The second op-amp section has two important roles, buffering the tone stack high-impedance output and giving enough recovery gain to compensate for the tone stack loss. This is made through some educated guesses, starting from the peak to peak level set by the diodes, multiplied by the loss of the tone stack (which is the less predictable part, being heavily dependent on the settings and the fundamental frequency of interest) and equated with the level required for full drive of the 386 at its minimum gain of 26dB, which is 300mV p-p with 9V and 8 ohm load according to the datasheet. Since I can't predict this fully, I've suggested a place for a Select-On-Test resistor on the schematic. I don't think this is necessary, but you can if you want to completely avoid overdriving the 386, you can set the preamp for the maximum output before clipping and use this resistor to limit the preamp volume to a level that doesn't introduce clipping. I think a little bit of clipping is fine when you turn up the volume, it adds post-tone stack "power amp" hard clipping like in any amp with the volume too high.

The power amplifier itself is as simple as only a 386 can be, pretty similar to the datasheet one, except for using AC coupling to ground on pin 2. Whatever miracle the 386 does to self-bias, this trick shouldn't mess with that by pulling this pin to ground.

I've nicknamed the preamplifier by itself "S.A.F. preamp", as in Simple As..., because I think it's a good basis for a no-nonsense, bang-for-buck preamplifier that gives you all the features you expect, to pair with any other commercial integrated power amplifier or premade board.

The other Sapphires

I've made other versions of the above schematic. This isn't an exhaustive list, but it shows some of the possibilities.

The standard Sapphire has too many knobs for you? The Bax version has a version of the Baxandall equalizer in it for bass and treble controls that actually do what they say. This version is probably very good for bass use.

The tilt version takes it further and lets you control both bass and treble with a single knob that boosts one while cutting the other.

The lite version just has a simple variable low-pass as tone control. If you prefer, you can use a rev-log pot or a log wired in reverse.

The Scapegoat is an honorary member of the family and probably the closest to the Ruby. It was meant as a testing amplifier to withstand some serious abuse. Thanks to the high-voltage input capacitor and diodes with series resistance, the amplifier is protected both from dangerous currents and voltages. You can probably poke around with a probe if you aren't silly about it, or use it to test your scary new prototype circuit.

A jellybean BJT like the 3904 still has enough current gain to not drop too much across the bias resistor without bootstrapping, but I guess you can increase the voltage divider output to compensate for the largest symmetrical output. The headroom of the circuit it probably severely limited anyway from the 386.

There are no tone or gain controls, the 386 is at full gain and you adapt to any input you're presented with by attenuating it with the volume control.

That said, if you use a standard capacitor and remove the diodes and resistor, this is still a completely usable guitar 386 amplifier.

Layouts

robots914 has kindly made vero layouts for the SAF, bax, lite and tilt preamps. You can pair these with any power amp of your liking.

The "phender" and the build

For myself, I went for yet another member of the Sapphire family. Since the goal was recreating the Chempion, an AA764 Champion, which I use almost exclusively clean, the most important factors were having a similar tone stack and output power.

• The tone stack is the same two-knob FMV, just scaled down to use common values and 100k pots. I have checked in simulation that this is indeed very close in response at all settings.

• U1B again gives enough gain to fully drive the power amplifier, given the diode-limited peak-to-peak voltage and the tone stack loss as before, and no more.

• The integrated amplifier of choice is this time the very common TDA2030 class AB chip amplifier. One reason I had chosen it was because I had a salvaged one from a broken small combo amplifier, where they are very common. Turns out it was probably the cause for the amplifier being broken, but thankfully these are cheap and easy to find. This part of the circuit is pretty much as in the datasheet, except for the gain being set to the safe minimum the amplifier can be set while being stable.

• How can I say with a straight face this is a phender amplifier? There's way too much gain in that first stage, and the distortion in those happens after the tone stack! Like I said, I use my amplifier mostly clean, so I thought to give myself some freedom on how to handle the distortion, since it's just a bonus and not something I intend to recreate as it's in the tube amplifier. The main differences are:

  1. Clipping before the tone stack. Using the tone stack as pre-emphasis is fun, but I wanted to exploit the treble control to tame the clipping harmonics instead.

  2. Amount of distortion. The tube amplifier only gets slightly overdriven unless you crank all the knobs. What I opted for instead, was an input amplifier that remains flat when clean (like a Fender amplifier), but is mid-focused at high gain. Think of this as having a free Tube Screamer inside your amplifier when you wish to. For what it's worth, my use of two series diodes means the input stage has the same input headroom as the tube in the Chempion, about 1.4V peak each way, so it will react similarly to large, boosted input signals.

About the supply voltage, the TDA2030 is happy with at least 12V, at which voltage it can reach an output power of about 2W with 8 ohms, according to the blurry datasheet curves, but both TL072 and 2030 can operate safely up to 30V, as long as the other components are rated accordingly, which gives about 10W with an 8 ohm load, 16 with a 4 ohm load! By that point, using a tiny box as heatsink as I'm doing is probably not the best idea, but you can fit an heatsink or a fan in a larger enclosure.

Pictures

Here are a few pictures of the amplifier build:

Demo

And a demo of it:

Update: muting circuit, more volume and bright switch

The 2030 circuit above works just fine and the chip deals with its own startup issues itself, but pops can still find a way from the preamp to the speaker when turned on with volume up. The updated schematic below shows a method I've devised that works very well and requires a minimum amount of parts that can be added on a second board:

This is similar to what is found in some amplifiers, even though I've come to it independently and uses a N-channel JFET. I've used a J113 because I had it, and its Vgs and Rdson characteristics are perfect for this use.

The idea behind the circuit is that the depletion mode FET conducts even when there's no power, so muting at startup is guaranteed. At that moment, the Vref capacitor is discharged, so Vgs is zero and the signal is shunted; as Vref rises, the gate stays grounded and the FET is smoothly turned off.

The second addition is a bright switch: some people love it, some people hate it, but it was something I was missing at certain times. Unlike old implementations, this is a fixed +6dB high shelf with a lower corner frequency of 1.9kHz, independent of volume settings, which is more useful in a small amp like this where you can use all of the volume travel safely. As it is, the "bright" mode is quieter than the "normal" one, because that one gets the full gain at all frequencies, but an equivalent network in parallel with R5 can be used instead, which will result in "bright" boosting up 6dB from "normal". If the bright switch isn't desired, it just means R5 goes straight to ground.

Finally, this new version has more gain in the recovery stage. The old version was conservative in avoiding power amp clipping at all possible settings, but that meant it might not have been possible to reach full output with the gain at minimum. The new version fixes that, accepting that high volume settings might result in clipping even with the gain down. This is a sign of having reached the maximum output from the power amp at that supply, and something that happens in almost all amplifiers as a result of a not well defined input level.

Layout

This layout has the updated recovery gain, but was made before the muting circuit and the bright switch. I added them in a functional but not pretty way, but if you're interested in them, a new layout might be better. Fitting a 2030 on vero is possible, but requires some very careful pin bending.