Pictures of the build, which is the result of a complete lack of planning:

http://www.diystompboxes.com/analogalchemy/tmp/amp3.jpg
http://www.diystompboxes.com/analogalchemy/tmp/amp4.jpg

I didn't use a feedback loop. A guitar is enough to make it distort just slightly when playing hard, but does so in a crunchy kinda way (the point of using that particular input stage, it can't really hard-saturate).

I estimated 28W, and it sounds about that volume when maxed. The preamp is going to be a 4-stage Vulcan type circuit with 3-band + presence.

I made some changes you suggested, this is gonna be it...
http://www.diystompboxes.com/analogalchemy/sch/guitaramp.html

Thanks for all the great advice.

Possibly a candidate for the radio-shackable amp:
http://www.diystompboxes.com/analogalchemy/tmp/amptest.gif

The 400k is supposed to be a 1M trim. The amp only requires slight biasing changes for different supply voltages. Of course going beyond 40V would require heftier transistors/resistors/100V caps in some places. Will report back when it's built. RS doesn't sell insulator kits for the 2N3055 which it does sell. Is that stupid or what?

I'm almost done redesigning the power amp to use more appropriate parts for a 30-watter. Parts are radio shack-available too. Will check back here after I've tested the build, hopefully sometime this weekend.

The caps aren't sitting on the chassis They are on a terminal strip. The 1k decouples the 40v supply and lowers the current. (Most JFETs can't handle much current at all.) The actual voltage powering the fets is around 30-32.

I'm redesigning everything. The preamp is good for high gain, but that's about it. I think my original power amp design was better/cleaner/louder, so I'll probably work on that some more. I'm very indecisive

another idea:
http://www.diystompboxes.com/analogalchemy/tmp/fetamp1.gif

This is kinda neat because the biasing only varies a little with different supply voltages. So you can pick your wattage pretty easily. (Shows 80V, but should be 25-50 range).

http://www.diystompboxes.com/analogalchemy/tmp/

Pix of inside and front panel, and dirty socks. It's a recycled Fender combo.

This is not how to build an amp, it's a rough test built from junk. I'm not particularly thrilled with the power amp, it works alright but seems it could be better.

Did some more work on this. The power amp has a buffered input, and tons more volume. It sounds like 25-30W is about right after hearing it maxed:
http://www.diystompboxes.com/analogalchemy/sch/guitaramp.html

The preamp sounded slightly honky so I upped the bypass caps to 3.3uF:
http://www.diystompboxes.com/analogalchemy/sch/jfetpreamp.html

Overall it's real nice except my build quality (perfboard with stuff tacked on everywhere).

The biasing will change with the voltage somewhat, although 25-50 seems reasonable. The transistors themselves can handle as much as 80 I think, but not the resistors.

Here's the preamp:
http://www.diystompboxes.com/analogalchemy/sch/jfetpreamp.html

The completed amp has some volume, but should be louder. Guess it needs a buffer after the tone stack.

About 30 RMS @ 4-Ohms, guesstimated from Spice simulation. The 56k limits it somewhat, but also prevents the input transistor from being saturated, making overdrive more pleasent if it occurs. The accompanying JFET preamp will work similarly, and uses the same 40v supply.

I redesigned/built/tested this thing:
http://www.diystompboxes.com/analogalchemy/sch/guitaramp.html

It should fire up pretty close to the right bias level with the trimpots centered. Preamp coming soon!

The bass seems to be too high, didn't really notice it until plugging in pedals. I'm guessing the 1uF feedback cap is too high, but who knows.

I tested the clamping diodes. The downside is that you have to trim the gain/volume on pedals. Actually, the added distortion didn't sound bad. Shouldn't be TOO big of a deal, and protects from overloading. And it's still pretty loud before you get to that point.

Also attached the thermal switch, so hopefully it's safe enough now.

There's another design I've been stewing over, based on this:
http://depalma.pair.com/Analog/analog.html

It's been on the net for years. The simulated circuit:
http://www.diystompboxes.com/analogalchemy/temp/depalma1.gif
(subject to removal, don't link.)

I altered the feedback structure slightly, and used a Vbe multiplier. The interesting part is that it appears to soft-clip in a very rounded, ideal way, especially with little or no negative-feedback.

Thanks for the invite. Looks like a real nice board.