Sometimes things don't go as you planned. You try and try again but the issues persist. This project has been around in some form for longer than the blog, funnily enough, but it's finally in a state I'm happy enough with to share it.
Whether I want to record easily or practice without disturbing others, I'm fond of Guitar Rig. One amplifier model it features is the Hot Solo, a Soldano SLO (100). Since it's the foundation of my “no-frills metal” preset, I thought it would be fun to turn the preamplifier circuit into a DC-powered box I could plug into my interface or the power amp input/fx return of an amplifier, or even as distortion pedal in front, for no other reason that I could, and while I had already made a DC-powered tube amplifier, a preamplifier was an idea I wanted to try for a while. It's well known that the SLO-100, being an amplifier provided of master volume, at most times concentrates its distortion timbre in the preamplifier, making this a worthy endeavor.
I started from the schematic available on el34world. As many others did, I focused on the Overdrive channel, which is the most popular one and the one I was interested in, to keep things compact and use few tubes.
The circuit
The first thing I wanted to address is the B+. This, as many other amplifiers, uses very high voltages in the preamplifier, in this case quoted at 359V, 350V, 378V, probably because they're derived from the even higher voltages of the power amplifier. These voltages are above the maximum ratings of most 12AX7 tubes, and also near the limit of the boost converter I was using. Do they need to be? Of course too low voltages can cause issues too, but I chose to use the safe and standard 250V supply. This doesn't seem enough “fiery heavy metal high voltage” to you? Well, I'm sorry.
This doesn't mean I just carelessly changed the voltage and called it a day hoping for everything to fall into place. The most interesting part of this adaptation for me was to rescale everything to work the same at this voltage. The two things to address are output swing and bias, since I've checked with a load line tool that the voltage gain stays roughly the same between these voltages.
About the first point, I've realized two important factors just in time: for all stages, the peak-to-peak input headroom (from grid conduction to cutoff) gets scaled down by the same factor of 1.5, save some small variations due to the different B+ voltages. The output swing also is scaled down by the same amount, because the anode cannot swing as high with the lower supply voltage (this was also measured on load lines. So while I was tempted to scale the output of each stage down, proportionally to the reduced input headroom, this was already done by the same reduction of supply voltage and I didn't need to do anything more, no attenuation between the stages or changing anode resistors. In hindsight this was to be expected if gain had stayed the same. This doesn't apply completely to the first stage, since the input headroom is reduced but the input is determined by the external source. Thankfully though, as it's usually the case for a good, strong, input stage, this stage doesn't clip by itself under normal circumstances. You need at least a 1.1V positive peak (which is the one clipped by soft grid conduction) or 3.4V negative (for the harder cutoff) to make it clip, and if you boost into it that much it means that's what you wanted after all! In conclusion, after all these measurements I just had to account for the 1.5x larger signal coming from the first stage, but not coincidentally there's the gain control right there. Instead of tweaking the series resistor R5 while keeping the shelving filter the same, I left the extra amplitude in the hands of the gain pot: think of this as having a little “extra” 3dB at the end of the travel.
The second point is bias. Changing the B+ moved the bias point sometimes considerably. Since the peak-to-peak swing was taken care of in point 1, my goal was to keep the same bias “temperature” of the original, the relative balance between cold and hot bias, the proportion of positive to negative swing from the bias point. I achieved this with a load line tool and by changing the values of the cathode resistors and capacitors accordingly.
Another change from the original is the smaller grid stopper resistor on the input stage, according to the suggestions in Designing Tube Preamps for Guitar and Bass, 2nd edition for lower noise in the crucial input stage of such an high gain circuit. I did calculate the equivalent grid capacitance (130pF) but I went from a smaller capacitor to ground than the required 750pF to keep the same cutoff, to not affect the guitar pickup resonant peak too much, an effect not considered by the author.
After this it was time for a quick count. I had four common cathode amplifiers, which means a pair of tubes already. After checking that the fx send and return had complementary amounts of attenuation and amplification, and given that there's nothing else past that to clip before the master volume and the phase inverter, I just had to account for the cathode follower and another buffer after the tone stack and volume to give me a nice low output impedance fit for a preamplifier.
The cathode follower is where I had to make a judgment: being fashioned after Marshalls (as most of the circuit is) and the Fender Bassman before that, it has the same hot bias that results in it softly clipping. While this effect is present, I don't know its importance among the vast amounts of distortion this circuit can provide, so my choice was to replace it with a MOSFET follower which doesn't clip, neither nicely or badly, and just faithfully drives the tone stack, which allowed me to keep the tube count to two. You aren't forced to make the same choice.
The IRF710 I've used is a cheap and unimpressive high voltage MOSFET, with a generous 400V maximum Vds, completely at ease in the undemanding application. The 710 has smaller gate capacitances and higher Rds(on) than the others in its line, which seems suitable for the purpose.
There's not much to say about this stage, other than the additional zener diode to clamp Vgs and the resistors in series with gate and source to prevent oscillation.
The tone stack is the same as in the original. In the place of the master volume, since I needed to bring down the levels to line sooner or later, I put a fixed attenuator. I decided for this position instead of leaving the attenuation at the output because this way I could adjust it without worrying about the output impedance or the volume pot value. The amount of attenuation was an educated guess based on the peak-to-peak output voltage of the last stage which I later refined. It's easy to change it with R24. The load on the tone stack is roughly 500k, just as in the original with the two parallel 1M master volume pots.
The output buffer is another source follower, this one AC coupled and with a simple bias that works well in this position. The fixed bias and the source make me quite confident that I don't need another zener diode here. The 1k resistor isolates the output and limits current to the 12V zener clamps which are the last line of defense against output spikes. I've measured the maximum output voltage to be about 3V peak, so lower voltage zeners would work just as well, but they're just added safety since the signal is already attenuated.
While I had calculated the gain and cutoff of the presence control according to the transformer ratio and gain of the stages, which agrees with other sources online which have done the same, I couldn't find a place to add a shelving filter after the clipping without another active stage. The input to the tone stack needs to stay low impedance; the output of the tone stack would almost work if it weren't that its output impedance is highly variable, throwing off the filter completely; having it at the output would ruin my output impedance. I decided that while it is possible to have this control, it is not of crucial importance to the core timbre of the circuit, and I can achieve the same effect with an equalizer if the tone stack isn't enough.
Power supply
Another fun challenge while beginning this project was to make my own high voltage boost converter. A preamplifier would be the perfect candidate, since the voltage is high, but the current is just a few mA. There are a few such designs, many of them featuring a simple 555 oscillator driving a MOSFET, with voltage feedback. Here's one example, while my friend Ethan has come up with his own variation.
Since I had all the necessary components, I tried this kind of circuit, and it kind of worked but with poor performance, with the output never reaching too far above 100V even when tweaking the feedback. I've tried this both with some commercial inductors I had in that value range, both with homemade ones with wire around a ferrite core, which again measured a suitable inductance. The DIY ones weren't probably very good, but neither the commercial ones reached the desired performance before arcing over. Turns out small inductors very rarely have a voltage rating, not being intended for this application, so while mine clearly wasn't up to the task, I couldn't find any I was sure would, so I put this idea aside for now.
Instead, this build features a return of the “yellowboi” 8-32V to ±45V-390V, UC3843-based boost converter, with all its merits and flaws, even if it's honestly overkill for this power. The heater supply is again the MP1584 buck converter, which had proven itself ideal already.
Since I knew the output from the 'boi wasn't very clean, I was sure to provide plenty of filtering before hitting the sensitive, high gain stages. The supply for the whole circuit is first passed through a MPSA42 capacitor multiplier, with the high voltage low power BJT being fit for the purpose and then through a simple RC filter to the tube amplifiers. Since this drops a certain amount of voltage, the output of the boost converter was turned up to compensate, giving the tubes the 250V by design and a bit more to the relatively more current thirsty MOSFET followers that don't need as much filtering. The result is satisfactory, but I suggest trying a different, more suitable boost converter if possible.
The build
I've built the circuit in a tight 1590BB enclosure with a mix of point-to-point wiring, perfboard and the PCB of the boost converter boards. You can see it as the natural history of electronic circuits. The point-to-point includes the tube pins, the central socket pin used as ground point and two terminal strips as additional support and was done for space concerns more than anything. The perfboard includes the gyrator and the MOSFETs. Both perfboard and boost converter are secured in place with brass standoffs that screw to the bottom lid.
The enclosure design is made of a polished aluminum enclosure on which I tried caustic soda etching for my second time. The etch came cleanly but not very deep, so I had to forgo the planned filling of the etched areas with black paint because it wouldn't come out as I wished. I like the subtle look of it and it's sill readable.
Uh-oh...
If this had been it, the build would have been completed months ago. But things don't always go as planned. Usually I don't give up easily, and if I set on soldering together something it's because I like it enough to see it to the end. I don't take it apart just because it doesn't work. That's why I was both ashamed and disheartened that it spent a long time in a cupboard when I decided I needed a break from struggling with it.
The first small struggle was power supply noise, but I was expecting that from the 'boi, so I added more filtering until it was decent. The real issue was an impossible to ignore, loud oscillation at a couple of kHz with the gain past half. I wasn't surprised at first, this circuit has tons of gain, even more than a similar sounding distortion pedal because of the relatively larger voltages required to clip tubes. Those same tubes are in high impedance stages and their output swing is proportionally gigantic, so it takes little to start a fire.
My optimism went down and down as I checked off all my possible fixes: shielding the input wire, increasing the input capacitance, the grid stopper, rethinking grounding more carefully, mixing up tubes, redoing the perfboard, adding series resistors to the MOSFET and some other things I don't remember. Then I started breaking the circuit in various points to see how extensive was the problem, finding that I didn't have oscillation before the MOSFETs but I did after the tone stack. I tried taking apart the tone stack then to see if it was involved. I must have repeated this step a few times, forgetting the exact results after some days of break between the attempts. I also noticed that, other than being affected by the gain control, grounding the input or using a low impedance source, instead of terminating it with a resistor or pickup, prevented the oscillation. So the feedback loop enveloped almost the whole circuit, from the input to the DC follower.
This is where I left it for a long time. After finishing my last project, I pulled this out of the cupboard and set on finishing it once and for all. After a while, moving things around and powering it up, to avoid memorable shorts, I noticed that having the perfboard and the pots out of the box made the oscillation go away. I thought it was the wire going to the first gate but it turned out the oscillation was caused by the perfboard and pots coupling signal into the input jack. The wire going from it was shielded, but the jack itself was not. To solve this, or at least make the problem much less severe, I made a shielding bracket from an aluminum sheet that mounts to the input jack, generously padded with electrical tape. I still have a bit of oscillation in the last bit of the gain pot, which I don't use, unless I use a low impedance source like a boost pedal. I hope a different internal layout might solve the issue completely for someone else, but I'm happy to have a working, completely usable circuit anyway.
Demo
Here's a quick demo of what you should expect this to sound like.
Conclusions
I'm happy to have reached a state with this project that I can call done. It's a fun and beautiful box if I can say so, but does this mean I will never use the Hot Solo again? Probably not, since the convenience of having not only the amplifier but also any necessary tube screamer-like pre-emphasis in software is hard to beat, and while I always thought VST models gave their best with high-gain amplifiers, I can only confirm this for how close the two sound. The SLO preamplifier might be the choice in winter since it also works as space heater though.
コメント