Here's a short list of lessons learned while designing and building my first guitar amp.
Sag and compression are normally associated with tube rectifiers. The problem is that a tube rectifier can't work into a large capacitance, so you also get ripple (which causes ghost notes) when the amp is pushed hard.
Screen compression gets you the same effect independent of whether the power supply is loose. Screen current increases along with plate current, and plate current decreases along with screen voltage. So if you arrange to feed the screen through a relatively high-value resistor bypassed with a cap, you'll get natural compression. The time constant of the screen resistor and bypass cap affect the response of the tube to transient signals: the cap holds the screen voltage up for a brief time, giving a punchy attack.
A cathode-biased output stage also contributes a bit of compression. However, screen compression has a much wider range of control without the adverse effects of bias shift and the resulting crossover distortion.
So many builders use tag boards that you'd think it was the only way to construct an amplifier. While the linear 2D layout certainly looks pretty, it really couldn't be much worse for purposes of stability. You have to add wire from the tubes to the tag board, thereby creating opportunities for cross-coupling. And the components themselves are all neatly laid out in a manner which further encourages coupling.
Point-to-point wiring between the tube sockets and terminal strips is the best way to minimize lead lengths and to separate inputs from outputs. It's the way all electronics gear was constructed until the economics of mass production produced the tag board as a labor reducing optimization.
There are still a lot of tools I should have and don't, especially with respect to metalworking and woodworking. But even the little things make a difference. I'm very glad that I invested in a set of chassis punches. And the bent-nose pliers are really handy for wrapping leads around solder lugs at odd angles.
I was surprised at how little I really needed from an oscilloscope, though. My boss gave me a very old Tektronix scope that just barely hangs together. I've learned to work around its quirks, such as the dead spots on certain controls. In using the `scope during tuning, the most valuable thing it showed me was the shape of the waveforms.
I put most of my effort into the design of the circuitry. After that, I did the chassis layout. Finally, I did the design of the enclosure. It all turned out OK in the end, but it could be better. I'd probably have made a few different design decisions if my initial design had included the enclosure. For example, had I known at the start that I'd be mounting the chassis vertically, I wouldn't have put the input jack above the volume control.
A lot of people simply repeat what they hear, turning fallacies into fact through repetition. That's really sad, because a lot of the information available about guitar amp design is wrong.
Case in point: ultralinear output connections. If I believed the common wisdom, I would have expected the mere use of an ultralinear output connection to give horrible results. In reality, it's easier on the tubes, simplifies the power supply and sounds just fine with a guitar. The people who have damned the ultralinear connection to obscurity do so based upon the high-powered Fender amps which were designed to deliver tons of clean power with low distortion. Today it seems obvious that those were not the best design decisions. However, there are several great modern guitar amplifiers (certain Dr. Z amps, for example) that use an ultralinear output connection.
If you look, you can still find good electrical engineering textbooks from the 1940s and 1950s. The information in these books is not predigested for guitar amp designers, but it's correct.
Unless you want an amp that only performs its best with clean tones, don't use negative feedback. It does nasty things to the character of the the distortion, especially right at the threshold of distortion.
Attempting to charge a capacitor with a current-limited source will generate a ramp, which is rich in odd harmonics.
Standby switches are not necessary for guitar amps. They add complexity and an additional point of failure. As applied to guitar amps, cathode-stripping is a myth.
This should be obvious, but it's not. Designers have a tendency to correct flaws by adding components. The proper approach to improving a design is to take components away until you're left with only what's necessary.