I finally got around to adding reverb to one of my Spitfire
clones. I used the standard Fender tank and transformer, drove
the tank with a 12AT7, and used half a 12AX7 for recovery. I
mixed the reverb signal back in at the unused phase inverter
input. Sounds pretty good, I must say...

Of course I can't resist the temptation to 'improve' the circuit. I
played around and reduced the circuit to one 12AT7 - half for
reverb drive and the other half for recovery. The drive half is
running at about 1W dissipation so I'm no where near the 2.5W
max for the AT7. It still sounds good to my ears... Anyone else
fooled around with minimalist reverbs?

I understand how you returned the reverb signal to be mixed in
the phase inverter (like some Vox & Matchless products did),
but was wondering how you picked off the dry signal and sent it
to the fender style reverb driver. Did you use any voltage
divider resistors, or a drive control pot, or just come off the
main signal path with a 500pf coupling capacitor?

I'm fairly certain that earlier Mesa Boogie amps used a single
section 12AX7 for driving the reverb transformer. Some of the
Crate vintage club stuff used 12AX7. It seemed to me that the
reverb driver in the boogies had higher voltage gain, but the
general sound was thinner, like the 12AX7 didn't have current
drive capability if you know what I mean.

Ted Weber pointed out to me years ago that this can have an
unintended beneficial consequence in the case of a small cap
strapped across the grid resistor in the reverb return circuit.
If you selected the cap to resonate with the inductance of the
reverb pickup transducer at the audio frequencies of interest,
then it would present a high impedance to ground to the
frequencies you're trying to amplify, while presenting a
shunting impedance to lower frequencies you don't want (like
60/120Hz hum). 

In summary, I suggest removing the cap across the reverb
transformer primary. But if there's a cap across the reverb
return, try unhooking it and see if the sound changes for
better or worse, or even try different values of small caps
there. 

Most reverb circuits take their input drive signal somewhere
after the volume control (I know Fender does). This has the
effect of driving the reverb harder as the volume control is
turned up, whether you like it or not, sometimes requiring the
Reverb level knob to be turned down for a more pleasing
overall sound. I was contemplating tapping off for the reverb
drive after the last fixed gain preamp stage, before any gain
pots, in order to drive the reverb at a constant proportion to the
guitar signal. For a home brew amp, a voltage divider pot
(Drive control) could be added between at the driver input grid,
allowing for greater overall reverb tuning, such as with the
6G15 stand-alone reverb box. The Reverb (return signal level)
knob could be set for a desired mix, and be left alone for a
while. Has anybody experimented with this method in their
amp?

I've done something similar to what you describe. I took the rev
drive signal after the 1st preamp stage, with the dry signal going
through a 100k or so resistor to the vol pot, fed it ( the rev
signal) to a 1 meg pot ( dwell) , with the wiper going to a
12AT7 rev driver tube. The return signal went to a 12AX7
recovery stage , then the rev level pot, then to another 12AX7
gain stage, and then to the volume control, where it was mixed
with the dry signal, via a mix resistor, about 470k I think. This
would be in a 5F6 Bassman type circuit. Still uses 2 tubes
though...

I picked off the signal for the reverb drive directly from the
wiper of the volume pot. The wiper connects directly to the
12AT7 grid. The traditional method uses a 500pf cap and
1Meg resistor to form a high pass filter at the input of the driver.
Unfortunately, this method adds some loading to the output of
the preamp. In a multi-stage preamp where you have tons of
gain on tap, this is not a problem. In the spitfire circuit there is a
limited amount of preamp gain and I did not want to affect the
tone of the amp in any way. I know that I'm probably being
obsessive, but Spice modeling and several hours of listening
confirmed that this was the way to go.

One thing - since the high pass at the input has been eliminated,
I still had to find a way to roll off the lows before they hit the
tank. Otherwise the reverb tends to sound muddy. I did this by
selecting a Ck/Rk combination for the driver that provided
some low-end roll off. Of course, as Randall Aiken pointed out
in his excellent article on triode gain calculations, this method
only provides 6dB of cut. This seems to be enough for my
purposes.

So, the driver consists of a 12AT7 with the grid connected
directly to the volume pot wiper, plate connected to reverb
tranny primary to B+ (about 300V), and 1.5K paralled with
1uF in the cathode circuit. I fooled around with different values
of Rk, dropping down as low as 750 ohms. I couldn't hear
much of a difference. I'm going to play with Ck a little more but
I'm basically satisfied with the current configuration.

In a multi-stage preamp, you could steal some signal prior to
the volume control and set the reverb up with a dwell control,
just as you suggest. The circuit that I'm experimenting with does
suffer from reverb level variations with different volume settings.
That doesn't bother me too much, so I'll probably leave it alone,
for now....

I've done similar things... using one half of an AT7 as a driver,
and the other half as a recovery amp or mixer. With careful
selection of the operating points of the driver, it works just fine
and gives plenty of reverb. 

I first attempted this when facing the challenge of installing an
FX loop in a friend's silverface Deluxe without adding tubes or
disabling the tremolo. I "stole" one half of the reverb driver tube
to act as the return mixer in the loop. After adjusting the Q
point of the reverb driver, the reverb sounded just as good as it
had before. 

It goes without saying (but I'll say it anyway) that when using a
single dual triode for both reverb drive and recovery, careful
layout is essential to avoid oscillation in the reverb circuit. 

For the 1/2 12AT7 driver, how did you vary the Q-point from
the original Fender design. The Deluxe Reverb used a paralled
12AT7 with Rk=2.2K and Ck = 22uF, if I'm not mistaken.
What did you change these values to?

Did you have to change the reverb drive transformer when
using half a 12A*7? Does the primary impedance need to
change when using half a tube or doesn't having an "ideal"
match matter for the reverb?

Yes, it is exactly like a low power, single-ended (class A) OT.
It doesn't really need a whole lot of low end bandwidth since
that would then reproduce the rumble the pan picks up. (The
response of the recovery circuit can be limited to control the
crap.) Like any transformer, it matched (in the Fender
example) the 12AT7 to the 8 ohm load. 

For best operating point, in terms of power output and
distortion, a triode wants to see a load impedance of twice the
internal plate resistance. A 12AT7 running at 250V has an
internal plate resistance of around 11K, so the best load
would be 22K. In the Fender type circuits, both halves of the
primary are paralleled, so the internal plate resistance is 5.5K,
and the best load would be 11K. BTW, the "standard" Fender
reverb tranny is 22K:8 ohms, if I'm not mistaken.

This load requirement is why you don't want to use a 12AX7
in this application. The internal plate resistance of a 12AX7 is
around 62.5K, which would require an optimum plate load of
125K. Such a large impedance would require many turns on
the primary, which would make the leakage inductance and
capacitance very large, killing the high frequency response.

Get yourself a PVC tube and end caps that is large enough to
mount the whole inner spring assembly/chassis inside. Make
fluid tight connections to bring out the electrical signals. Put the
spring assembly inside and fill the container with a light oil and
glue aluminum foil all over the outside. The oil absorbs and
damps the mechanical shocks. 

Hammond used to make oil filled spring reverb units for
organs, I understand. 

Pretty sure that by adjusting the amount of oil you can alter the
sound as well. Hammond had a series of springs in their oil
reverb units, and I believe one of them was designated for the
specific purpose of dialing in the 'depth' or 'dwell' or 'apparent
room size' or 'decay strength'. 

Don't quote me on this, but I think that's fairly accurate. Seems
like Doc had a pretty good handle on this technology and had
posted on this topic a while ago.