Date: Thu, 03 Jun 1999 17:15:23 -0400
From: Ken Gilbert
Subject: Re: [tubetechnical] Digest Number 77

IME, as a guitar amp tweaker / builder / player, you've got three basic characteristics of a "good" guitar amp:

a) There is a certain amount of frequency shaping going on. In the Fender/Marshall case I boiled it down to mainly a matter of midrange, which is over-simplified. It is also a matter of WHEN i.e. at what stage is it implemented. In fact, it may also be (and most likely is) actually divided across a few stages. A typical LF rolloff is inherent in the RC coupling, which may have vastly different break frequencies. In one order, it will sound a certain way. Flip the two time constants around and you've got a much different tone. A good example of this is the placement of the tone stack in Mesa Boogie amps--relatively early on in the stages. As a result, you can't really change the tone of the amp very much, since stages subsequent to the tone stack will tend to distort the same way, regardless of the tone knob settings. The more distortion you add, the less the controls matter at all.

Some interesting observations about tone shaping... If you increase the lower frequencies, you will be lowering the amount of THD, since you will be increasing the magnitude of the fundamental. Likewise, if the higher frequencies are boosted, you will simultaneously be increasing the THD of the amp's tone, since most of the frequencies up there are harmonics generated in the amp itself.

I myself prefer a scooped type tone, where much of the mids are removed. Usually I accomplish this with a "detuned" parallel-T type filter. The filter is detuned to some extent because the notch of a perfect ||-T is theoretically infinite in attenuation and very narrow in bandwidth. By carefully removing these mid frequencies, I am left with LF (which gives one a solid reference to the fundamental--like "what note is it," as well as a percussive "thump" as the strings are whacked, for a nice sense of headroom) and HF (which gives one a sense of the distortion introduced, as well as a nice quick slew rate to accompany that LF "thump" to give it a distinctive transient). Of course, remove too much midrange and the guitar sounds like a heavy metal tone, with all buzz and thump and no guitar left. This is a delicate balance that must be made on a subjective basis, naturally.

b) The amount each stage is overdriven by the previous. Too much gain from one stage, fed unattenuated into the input of the next, will lead to "grid blocking" due to the conduction of grid current.  The grid circuit is a high impedance node that normally does not have to deal with the quantity of electrons which appear as the voltage exceeds the cathode's. These electrons collect on one terminal of the coupling cap, which retains a negative charge. This will tend to pinch off the tube.

I have found that lower values of coupling caps work well here to limit the amount of stored charge, making the response to overdriven transients faster. I have used values as low as 2n7 as coupling caps for this reason. Another obvious way to limit this effect is to simply attenuate the signal with a voltage divider. The sound of "grid blocking" can readily be heard in most Fender amps if they are cranked all the way up. It is a farty, displeasing sort of distortion, and on an output stage will serve to bias them colder, introducing crossover distortion. This is due to Fender's use of large coupling caps from the phase splitter to the output grids, usually about 100n. The problem is made worse by the large amount of LF present in the Fender design that is simply not there in the Marshall's. This is just another reason the Marshalls are typically run wide-open, and still sound good (if you've got any hearing left).

In addition to the grid blocking effect, too much drive to one particular stage will remove what I consider one of the great unique qualities of tube amps--their "touch sensitivity." When players use this term, it is usually describing the ability to lessen the attack on the strings, and get less distortion from the amp. A good design will "clean up" in this way without too much of a change in final SPL. This allows the player to really adjust the tone of the amp merely by changing the way they pick--slam the strings in the chorus section, and the upper harmonics really come forward, opening up the tone. Back off on the attack during the verses, and the tone gets darker and less in your face. This can also be accomplished by adjusting the volume pot on the guitar itself, but I usually don't have time to screw with that. ;-)

The best touch sensitivity is achieved by overdriving the stages somewhat equally, so that a very light touch, even with all gain knobs cranked, will produce a clean output, without noticeable distortion creeping in.

c) There is a "complexity" factor involved, which I've used to describe various non-linearities in the amp. Some of these non-linearities stem from a deliberately undersized power supply, adding a compression effect where the initial transient is allowed to pass relatively unscathed, but sustained output causes the supply rail to collapse. This effect manifests itself as "pick dynamics" where the swish of the pick can be heard hitting the strings REGARDLESS of the amount of overdrive or distortion used.

There is also the non-linearity of any signal transformers in the amp, particularly the output tranny. Because of the magnetic properties of the core, the inductance of the OPT is lower with small signals, reaches a maximum with medium signals, and then trails off again with large signals. This ultimately represents a signal dependant change of frequency response. If there are any inductors used for tone controls (I swear by them for these very reasons) they will also exhibit a varying degree of inductance WRT signal level. I am currently looking into using interstage transformers to capture more of this complexity.

Of course there is the OPT itself, which is wired right to these electro-mechanical devices called speakers, usually flapping around inside a resonant system with a Q that's much too high to expect predictable response from. This produces the well-known "back-emf" which is forced back through the OPT and reflected to the primary, increasing and decreasing the flux through the core as well as the effective primary impedance. If there is a NFB network in place it will also impress itself through that.

Personally I have eschewed all forms of loop NFB on the output stage, preferring instead to use triode and UL topologies to achieve reasonable damping factors. With guitar amps, they are too often driven into heavy distortion, even in the output stages, and as we all know NFB falls apart quite un-gracefully when the loop cannot respond to too much error correction. Some of the best amps I've heard, including my own, run "open loop" and sound MUCH more dynamic and present.

When these non-linearities are present in an amp, it is almost as if the amp were "alive," responding to the way it is pushed and played. The complexity factor seems to be proportional to the amount of iron in the amp. Anyone who has perused texts on magnetics and transformers knows it to be a damned complex subject.