Amps with warm biased gain stages? Clipping theories?

[teemuk]

I would not regard the early 1970's Russell Hamm whitepaper as too trustworthy source of information about this topic. Most of its content was debunked already when the paper was released and there are other major issues as well, biggest one that the paper is written according to principles of high fidelity sound (re)production, not according to principles of generating distortion deliberately and as a musical effect. Hamm also studied characteristics of mic preamps, which are totally different to guitar preamps. His “referenced material” on that part was extremely limited as well as he made all his conclusions based on reviewing only four different preamps. Coincidentally all tube units had single-ended architecture while solid-state units were push-pull, which is about the sole reason Hamm determined that tubes produce even order harmonics and solid-state produces odd. (A theory that was debunked instantly).

---

Back to topic. In my experience, there are not many (if any at all) high-gain design relying solely on asymmetric clipping, as in one lobe of the waveform is retained undistorted in the process. Several ones are, however, derived from the classic Marshall/Soldano format where the waveform is initially clipped in very asymmetric manner (one lobe clean, the other one almost completely clipped off), after which the signal overdrives a gain stage with more symmetric clipping characteristics. This produces hard clipping coupled to distinct duty cycle modulation, which varies dynamically.

If intermodulation distortion is factored in, the most beneficial clipping method for “high gain” is actually indeed symmetric clipping because it produces mainly odd order harmonic distortion (and its intermodulation products) instead of both even and odd harmonic distortion (and their intermodulation products). The more IMD the more unintelligble the signal becomes. Some clipping schemes actually dynamically shift from producing asymmetric clipping at lower signal levels into producing symmetric clipping at higher signal levels.

Such dynamic characteristics, IME, are not overly important for “tone” but for “feel”. In my experience, all distortion (even vs. odd, soft vs. hard) more or less "sounds the same" but if it lacks dynamic reaction to one’s playing then the distorted sound is easily perceived as sterile and boring. For instance, typical clipping of a push-pull output stage is not only symmetric but often has subtle nuances and effects such as dynamic crossover distortion (“swirly” characteristics added to timbre of distortion at higher levels of overdrive) and sagging effects (clipping threshold varies in accordance to signal envelope). This is why it can’t be accurately faked with just a pair of clipping diodes.

In addition, most competent designers of distortion circuits have noticed that the distortion actually plays a secondary role in comparison to pre- and post-distortion emphasis. If you look at it, “high gain” circuits often follow certain pre-established concepts of “voicing”; what frequencies to filter out before distorting the signal, and what frequencies to emphasize or de-emphasize after that. This is extremely important aspect in design if one wishes to create the archetypal smooth, singing sustain and focused “high gain” tone instead of just unintelligible mess of farty, gated buzz.

By the way, regarding clipping distortion, I would not concentrate on -all- gain stages of an amp, just to ones that actually produce the said clipping distortion (and there are usually no more than perhaps two or three of them in an amp). For example, a certain classic Marshall design may have a 2K7 bias on the first gain stage but from perspect of clipping it is largely inconsequential because input signal amplitudes under normal operating conditions will never be high enough to clip the particular stage. The effect of the bias conditions is therefore merely reduced to that of changing gain ratio of the stage, which, BTW, also reduces “shelving” effect of the hi-pass filter circuit implemented to the cathode circuit, which I think was the most likely purpose for Marshall designers selecting 2K7 cathode resistance instead of a “standard” value of 1K or 1K5.

[Littlewyan]

This is excellent information. Thank you all for taking the time to type it out.

[wpaulvogel]

I think power tube clipping is the best sound. Unfortunately it’s loud. I like the sounds of 6K6GT overdriven and they put out a bearable 15-16 watts. But the problem is clean headroom. It’s always a trade off.

[tubeswell]

I vote for iterative band pass filtering to accentuate mids, and hot>cold>hot>cold biasing through 4 or more stages (to clip top and bottom peaks etc) with heaps of inter-stage attenuation and local feedback, and plate bypass caps, and the occasional DC-coupled pair (or DC-coupled CF for added compression), and a really cold biased finishing stage (Rk=39k unbypassed) with a balanced LTP with a 220-470nF decoupling cap on the slave triode, into a bevy of EL34s or KT77s. (SLO100, PV5150 etc). At least 2 channels with opto-isolated switching and different ‘voicings’. One of the channels needs a FMV tone stack with a switchable 33k/100k slope resistor and a 470nF bass cap. Try a grid leak biased input stage on one of the channels (and experiment with input blocking cap values and HF shunting caps. A parallel FX loop is mandatory for delay FX. For the power supply use SS diode rectifiers and 50-100uF reservoir capacitance with a similarly beefy screen supply, or even better, a separate screen supply or a regulated screen supply. YMMV