What if I actually want my power amp to distort?

[ampetrosillo]

I'm a bit frustrated at the available material on the 'net for building MI amplifiers: I want to build a solid state bass/guitar amp, which does not require hi-fi levels of linearity (like, average power could be stated @ 2-3% THD and it would still be "clean"), can clip gracefully and softly, can interact with the speaker if desired (variable mixed-mode feedback then it is) and can be used as an instrument (as you do with electric guitar and bass) but essentially what you can find documentation-wise is either hi-fi-oriented stuff (where the notion of power amp distortion is dismissed a priori, and of course, it makes sense) or tube stuff, with nothing in between. Even the solid state guitar amplifier people suggest moving everything clip-wise to the preamp, apparently ignoring the fact that what makes tube amps so highly sought after, by guitarists especially, is their interactivity with the speaker (therefore clipping is really a function of speaker impedance, therefore in order to be "authentic" it should happen at least in the feedback path from the output to the power amp's input) as well as their relative ease of use (no thinking about clip lights, correct gain staging, etc. just turn that pot up and play).

(I do realise that this means that a 50W amp needs to be cranked to get that tone, and this is often inconvenient - which is why many amps incorporate lower power modes these days, by the way - or, conversely, a 15W amp can't be expected to sound clean at higher pressure levels, but there still is value in this).

Therefore, are there any materials I could peruse to gain a better insight over this?

 

[DVDdoug]

That's a good question... You might want to study transistor & MOSFET curves to see if you can make them more non-linear in a way where they more-slowly start to saturate before hard-clipping. I don't know, but the output stage may not be the best place to do that. It might require something "special"... maybe something with op-amps & diodes... something similar to a compressor-limiter circuit.

Negative feedback ("corrective feedback") tends to linearize and correct everything so you'll want to avoid or minimize feedback, at least in the stage that's generating the saturation.

Solid stage electronics do tend to hard-clip more than tubes & transformers.

You can add resistance (or "effective resistance" in front of the main capacitor in in the power supply to make the voltage "sag" with sustained output. That's a different effect (like compression or limiting with an attack time) but it's a common effect in amplifiers and it MIGHT be something guitar players like. Or you might get enough of that effect simply by using an unregulated power supply. (But power supply regulation helps a LOT with hum filtering and you may not want to eliminate it.)



...I read something once about a guy repairing a guitar amp and he noticed something about the circuit that was making it non-linear and "fixed it". When the guitar player got his amp back, it didn't sound right and the guy had to un-do his "fix".

 

[sergeauckland]

One of Marshall's Hybrid amps has all the distortion and effects in the pre-amp section, followed by a very linear SS power amp, good enough for HiFi. That way, you can still get the full effect even at low volume as it's all done before the power amp section.
S

 

[ampetrosillo]

One of Marshall's Hybrid amps has all the distortion and effects in the pre-amp section, followed by a very linear SS power amp, good enough for HiFi. That way, you can still get the full effect even at low volume as it's all done before the power amp section.
S

That's relatively easy, but it doesn't really do what I'm trying to achieve. I would like the amplifier to actually respond to playing dynamics with distortion. I realise you can sort of achieve the same thing with mimicking a high-impedance power amp's interaction with the speaker with symmetrical pre-post distortion filters, but that way you actually lose the possibility to make the amp sound different based on the speaker you use with it.

 

[kemmler3D]

That's relatively easy, but it doesn't really do what I'm trying to achieve. I would like the amplifier to actually respond to playing dynamics with distortion. I realise you can sort of achieve the same thing with mimicking a high-impedance power amp's interaction with the speaker with symmetrical pre-post distortion filters, but that way you actually lose the possibility to make the amp sound different based on the speaker you use with it.

This all makes sense... and I can accept "amp distortion, not other kinds of distortion" as a valid goal. But I have to imagine there are more straightforward ways to get a similar or even identical sound, no? I think the high power requirement is what throws me personally. Having your tone tied to SPL seems like it would be troublesome for practice as well as performance, sometimes.

 

[ampetrosillo]

That's a good question... You might want to study transistor & MOSFET curves to see if you can make them more non-linear in a way where they more-slowly start to saturate before hard-clipping. I don't know, but the output stage may not be the best place to do that. It might require something "special"... maybe something with op-amps & diodes... something similar to a compressor-limiter circuit.

Negative feedback ("corrective feedback") tends to linearize and correct everything so you'll want to avoid or minimize feedback, at least in the stage that's generating the saturation.

Solid stage electronics do tend to hard-clip more than tubes & transformers.

You can add resistance (or "effective resistance" in front of the main capacitor in in the power supply to make the voltage "sag" with sustained output. That's a different effect (like compression or limiting with an attack time) but it's a common effect in amplifiers and it MIGHT be something guitar players like. Or you might get enough of that effect simply by using an unregulated power supply. (But power supply regulation helps a LOT with hum filtering and you may not want to eliminate it.)



...I read something once about a guy repairing a guitar amp and he noticed something about the circuit that was making it non-linear and "fixed it". When the guitar player got his amp back, it didn't sound right and the guy had to un-do his "fix".

I realise what the problem is with having the output stage clip softly: to actually do that (and counteract the tendency of solid state electronics to clip hard, which is their natural behaviour) you need to increase feedback the closer you get to the rails, similarly to what you do with soft clipping on an opamp. The problem is that it is probably impossible to design a power amp that will, eventually, get to unity gain and remain stable. (You can just place diodes in the feedback of a non-inverting opamp because most audio opamps these days are unity gain stable). So you can get somewhere in the middle, I'm guessing, by applying some limiting on the input and then some soft clipping in the NFB path of the power amp which never drops the gain to zero (but keeps it at a healthy level, like a gain of 10). That means that NFB actually increases with level, tracking the output stage's distortion at lower levels so that it can remain clean when operating at low power levels but actually getting "cleaner" (but overall dirtier!) when power rises.

 

[ampetrosillo]

This all makes sense... and I can accept "amp distortion, not other kinds of distortion" as a valid goal. But I have to imagine there are more straightforward ways to get a similar or even identical sound, no? I think the high power requirement is what throws me personally. Having your tone tied to SPL seems like it would be troublesome for practice as well as performance, sometimes.

Yes, that's exactly how traditional amps operate. Which is why people either have different amps for different situations, or amps have lower power modes (usually something like Class A operation at 5W and Class AB at 50W or so, with some amps actually allowing a continuous power adjustment), or they use a dummy load together with their speaker.

 

[Zapper]

...I read something once about a guy repairing a guitar amp and he noticed something about the circuit that was making it non-linear and "fixed it". When the guitar player got his amp back, it didn't sound right and the guy had to un-do his "fix".

Something like that happened to me way back in high school, where I had an unprofitable side-gig fixing amps and such for broke musicians. This knucklehead got tired of fuses blowing in his amp (because the power transformer was failing) so he stuffed the fuse holder full of aluminum foil. The result was a smelly meltdown. While repairing the damage I concluded that a pair of solid state rectifiers would be a more efficient replacement than the tube rectifier. And that was true - the plate voltage increased and maximum power went up. But the sound got worse. So I had to fix my fix.

 

[kemmler3D]

Yes, that's exactly how traditional amps operate. Which is why people either have different amps for different situations, or amps have lower power modes (usually something like Class A operation at 5W and Class AB at 50W or so, with some amps actually allowing a continuous power adjustment), or they use a dummy load together with their speaker.

I've been an in-the-box guy for a long time so this sounds like a total pain to me... I also don't play guitar but I appreciate the drive for an authentic sound/experience. Sorry I don't have more to contribute here, but LMK if you ever start looking into DSP / VST stuff.

 

[Zapper]

There is a lot of info on modding guitar amps on the net. Reading through some of that material will give you some insights into design parameters that are considered desirable in MI amps.

 

[SSS]

To get powerstage saturation distortion it may be possible when using with transistors or mosfets an output transformer as tube amps do. At the early transistor days many power or also small power amps had a driver and output transformer.

 

[ampetrosillo]

I've been an in-the-box guy for a long time so this sounds like a total pain to me... I also don't play guitar but I appreciate the drive for an authentic sound/experience. Sorry I don't have more to contribute here, but LMK if you ever start looking into DSP / VST stuff.

This kind of DSP (guitar amp sims) is "not that hard". It's just relatively time-consuming and capital-intensive. It needs access to the original gear you want to emulate, some reasonably accurate measurements (and the right kind of measurements) and then it's all essentially a question of dynamic waveshaping (with the right amount of oversampling and aliasing mitigation techniques) + EQing, which even when done inefficiently takes very little computing power even for real-time operation.

 

[Zapper]

Even the solid state guitar amplifier people suggest moving everything clip-wise to the preamp, apparently ignoring the fact that what makes tube amps so highly sought after, by guitarists especially, is their interactivity with the speaker (therefore clipping is really a function of speaker impedance, therefore in order to be "authentic" it should happen at least in the feedback path from the output to the power amp's input) as well as their relative ease of use (no thinking about clip lights, correct gain staging, etc. just turn that pot up and play).

The tube amp output stage is a complicated system, comprising the high voltage supply, the tube output stage and the bias network for it, and the output transformer. It's hard to replicate in solid state.

Here is a suggestion for a solid state output that will interact strongly with the speaker. Use a Class-D output, with a fairly high impedance (droopy) power supply, and low (or maybe no) negative feedback. An open-loop PWM Class-D amp will output an average voltage that equals the duty cycle times the input voltage, so if the power supply droops, so does the output voltage. The damping factor will be low, just like a lot of tube amps. You can tune the time response dynamics of that by the impedance of the supply and the size of the filter cap. The output impedance can be tuned by the amount of negative feedback, as well as the frequency characteristics of the loop gain. You can use a nice big laminated steel core inductor in the output filter to get some magnetic nonlinearities too, just like those output transformers.

 

[ampetrosillo]

The tube amp output stage is a complicated system, comprising the high voltage supply, the tube output stage and the bias network for it, and the output transformer. It's hard to replicate in solid state.

Here is a suggestion for a solid state output that will interact strongly with the speaker. Use a Class-D output, with a fairly high impedance (droopy) power supply, and low (or maybe no) negative feedback. An open-loop PWM Class-D amp will output an average voltage that equals the duty cycle times the input voltage, so if the power supply droops, so does the output voltage. The damping factor will be low, just like a lot of tube amps. You can tune the time response dynamics of that by the impedance of the supply and the size of the filter cap. The output impedance can be tuned by the amount of negative feedback, as well as the frequency characteristics of the loop gain. You can use a nice big laminated steel core inductor in the output filter to get some magnetic nonlinearities too, just like those output transformers.

I would never want to replicate a tube amp with solid state electronics (it would also make no financial sense). What I would like to approximate closely enough is the end result. Class D makes me wary (I'm an amateur dabbler in electronics, and Class D essentially means buying an IcePower module and essentially set it up the way it wants to be set up, because there is no way you can design one at home).

That said, I did find a few days ago a design that supposedly does something similar to what I imagined earlier (I didn't really analyse the schematic, but it appears that the basic idea of low NFB then getting higher when clipping is there):

Patrick's kleine Homepage

http://xipix.de/elektro/gimos50/GIMOS50-schem.pdf

Although I don't understand why he went to the trouble of a *double* (symmetrical) input stage + current sinks, when I assume that a FET (or otherwise) phase splitter on a LTP would have been fine.

 

[KSTR]

I realise what the problem is with having the output stage clip softly

IMHO, soft-clipping, compression etc is best implemented before the complete power amp, at its input clipping the input voltage, so that the power amp never goes into voltage clipping. That includes any "overdriven tube power amp" style of more complex distortion including supply sag etc.

A high output impedance, OTOH, is best implement in the amp's feedback network by mixing in some additional voltage derived from the output current, sensed in a resistor.
Unfortunately, this goal is at odds with soft-clipping at the input as the effective voltage gain is now variable.

So some nice clipping is still needed in the power amp itself which is best implemented with a reduced feedback, going down to unity gain, by shorting out the feedback resistor with Zener diodes or similar means... which requires a unity-gain stable amplifier.

When those two distortion modes are balanced carefully I think this gives a good sounding guitar power amp that compares well to a tube output. A regulated power supply would surely be a benefit for predictable operation conditions.

Very conservative de-rating of everything is required as are protection features to make it bullet-proof and ready for any potential abuse on the road (if that is the goal, that is).

 

[KSTR]

Although I don't understand why he went to the trouble of a *double* (symmetrical) input stage + current sinks, when I assume that a FET (or otherwise) phase splitter on a LTP would have been fine

Me thinks that maybe it just worked, sounded good and was manageable complexity so the author may have settled with that without further desire to optimize things.

 

[Timcognito]

Audio distortion and feedback - Pass Labs

Audiophiles seem to revel in minor controversies – vinyl vs CD’s, tubes versus solid state, capacitor, wires, magic dots… and negative feedback. At one extreme, the position is that “feedback makes amplifiers perfect”. At the other extreme, “feedback is a … Continue reading →

www.passlabs.com

 

[Zapper]

I didn't really analyse the schematic, but it appears that the basic idea of low NFB then getting higher when clipping is there

Right, that should add some soft clipping.

Although I don't understand why he went to the trouble of a *double* (symmetrical) input stage + current sinks, when I assume that a FET (or otherwise) phase splitter on a LTP would have been fine.

It keeps the output symmetric when overdriven. A classic design has a single ended gain stage, usually a common emitter NPN. The pullup current is provided by a current source or a bootstrapped resistor. When overdriven, the NPN can pull down harder than the pullup circuit can pullup. The output gets pulled low and a negative DC component appears at the output. This will produce all sorts of bad sounds and maybe speaker damage. The circuit shown is (approximately) symmetrical in pullup and pulldown so should not develop that type of dc offset when overdriven.

 

[kemmler3D]

This kind of DSP (guitar amp sims) is "not that hard". It's just relatively time-consuming and capital-intensive. It needs access to the original gear you want to emulate, some reasonably accurate measurements (and the right kind of measurements) and then it's all essentially a question of dynamic waveshaping (with the right amount of oversampling and aliasing mitigation techniques) + EQing, which even when done inefficiently takes very little computing power even for real-time operation.

Sounds like you're well up to speed. Agree. Carry on sir, I think this should be an interesting project to follow if you stick with it.

 

[er|κzvio1in]

I used to design and build guitar tube amps when I was younger, and even though I never pay electric guitar anymore these days, I can still appreciate a good guitar sound and still know the schematics from the top of my head. I also fooled around in electronic stimulation software trying to grasp a more detailed understanding about where tone comes from.

The important thing is that it's more art than science: you need to work with your ears and hands and use that as basic guidance. In my opinion you're approaching this too theoretically.

When designing/building a guitar amp it starts with inspiration: which tone and feel are you after? Build a basis and start tweaking (simply changing out capacitors, resistors and pots) until you get it at the sweet spot.

Also: don't forget that a lot of the tone is made by the speaker cabinet, in case of distorted sounds, the cab even determines more of the tone than the amp.

Then my advice if you are looking at getting tube like distortion from solid state: base your design on overdriving FETs. Look for example at the schematic of the ZVEX Box of Rock: that's an overdrive pedal with tube like behaviour and a good sound (considering it's not real tubes).


Also watch this video if you haven't already