Fecundity and sound
The title of this piece is a bit of word play regarding observations that certain gear produces a "sterile" guitar sound. Clearly the term "sterile" has negative connotations, but what does it really mean? And what's the desirable "opposite" of sterile?
I've used the term myself on numerous occasions. In the linked articles, I describe a "sterile" soud as one having no distortion on clean sounds (wait... what?!), which isn't as crazy as it sounds. What I mean is that a good clean sound - which I have in the past referred to as a "Fender clean" - has a bit of distortion that both responds to pick attack (a stronger attack gives more of a "bite" to the sound) and "evolves" or changes character during the note decay. (I recall having read an article in The ToneQuest Report that the guys at Fender have an entire lexicon of of terms that they use to describe the way they want an amp to sound.)
As guitarists we have a lot of words do describe these phenomena. How the amp responds to pick attack is frequently referred to as "touch sensitivity" or "touch response". The amp's behavior during the note's decay has has a number of facets; we use different terms - like "bloom" and "swirl" - to describe each facet.
In addition to the dynamic behaviors there may be an element of EQ involved, but I haven't convinced myself that I can articulate that just yet.
I'll note that my personal universe of guitar tones is almost entirely at the clean end of the spectrum, bounded at the upper (distorted) end by a Tubescreamer. My model of what makes a "sterile" or - to coin a term because "non-sterile" seems awkward - "fertile" tone makes sense to me within the universe of clean guitar tones. I have no clue as to what "sterile" might mean in the context of a high-gain tone.
If you delve into the technolgies involved in producing guitar tones - both via tube amps and digital modelers - you'll learn a few things that seem to support the above observations:
- Tubes are dependent upon a bias voltage (we've all heard this in terms of getting proper behavior from output tubes; even preamp tubes have a bias voltage, even though it's not user-adjustable). The really interesting thing about a tube circuit - at least from the point of time-variant behavior (i.e. "evolution") is that the bias voltages (there's one per tube) are constantly shifting according to a lot of factors, the most important of which is the magnitude of the signal with which that part of the circuit just got hit.
- Older modeling technologies - if one is to believe patent filings, which are pretty much the only window we have into a manufacturer's technology - did in fact seem to model a static transfer function. IOW, while you still had the nonlinearities that are responsible for generating harmonics not present in the input signal, the nonlinearities remained constant no matter the magnitude of the input signal.
Putting those things together and noting that a changing bias voltage in a tube circuit will in fact change the nature of the tube's nonlinearities (a fact which can be confirmed by looking at any electronics textbook from the `40s or `50s) explains how some modelers might sound sterile.
Note that I said some - not all - modelers.
The DSP guys aren't ignorant; they're just working within a budget (both development time and production cost) and doing the best they can under the constraints they're given. While early modelers were in fact "sterile sounding" for the reasons I've noted, the technology has improved since the mid-`90s.
There are a number of approaches documented in the literature (and by that I mean university research, not marketing broadsides) to get behavior that "evolves". One such approach is to measure the transfer response for various signal levels, model each of those level-dependent responses, and add a mechanism to switch from one transfer function to another based upon signal level. If you're familiar with velocity-sensitive layers in sampled keyboards, the concept is similar.
Another approach is to digitally model the behaviors of the components in the actual tube-amp circuits. In this case, the model should behave exactly as the tube amp behaves under all circumstances. In practice, the DSP designers have probably had to make some simplifying assumptions in order to meet constraints imposed upon them. The real question becomes: did the designer simplify things in such a way that that you can't hear the difference between "the real thing" and the model?
Unfortunately, there's no universal "yes" or "no" answer to that question. Let's ignore the likelihood that you and I hear things differently and focus on the certainty that we each have a unique approach to playing our instrument. What that means is that - even given the same modeler with the same settings and the same listening arrangement - the time-varying input to the modeler will be different. Even if we play the same notes, the attack and decay and timing will still be different; this means that each note will evoke slightly different behaviors from the modeler. Also, our intonation (based upon fretboard mechanics and pick attack) will also be different, which means that the harmonics of the intervals we play will line up in a slightly different manner. (BTW: the same arguments apply equally to each of us playing the same tube amp.)
All of these things contribute to the way we perceive the playing experience. What works for me may not work for you and vice versa. This, I think, is exactly why you can read a post from one player saying that "brand X modeler does well for cleans, but not for heavy distortion" while another player posts that "brand X modeler does well for heavy distortion, but not for cleans"...