[Patrick Turner]

Recently I worked on a Fender Deluxe Re-issue which makes 21Watts at clipping with 2 x 6V6 in output driven by 12AT7 LTP.

The input for clipping at signal input g1 pin2 'AT7= 2.5Vrms, and then you have 0.7Vrms of GNFB from Vo applied to the NFB input g2 pin7 'AT7. The amount of NFB applied 3dB, and makes almost no change to the sound if the 820r is removed to disconnect the GNFB.

The owner wanted variable NFB. So, I installed a linear 1k0 pot on amp rear with small knob. pot bottom goes to 0V, wiper goes to top of junction between 820r and 47r NFB network. 47r was removed, and replaced with 100r. Top of pot taken to Vo, ie, the whole of the single secondary winding on OPT.
The NFB can now be varied between ZERO and about 10dB max.
The center of pot gives the original amp's amount of NFB. But a problem occurs where GNFB exceeds 6dB, with the cabinet speaker connected as a load. Above 6dB NFB the amp breaks into HF oscillation once past the class A to AB threshold. It was to be expected because the Fender OPT is not much interleaved like a Hi-Fi OPT. The oscillations only occur as HF waves appearing on positive going peaks of Vo, and the amplitude grows larger as clipping is reached and the oscillations have amplitude larger than the LF part of waves even when over driven.

So I can only guess that in class B during cut off for one 6V6, the circuit changes to that of an HF oscillator.

The solutions tried were 15r & 0.27uF across Vo, 680pF & 8k2 across anode to anode, placing small C from Vo to pot wiper for phase correction of NFB. None of these worked. Then I remembered the right solution, TWO R&C zobels across each 1/2 primary.
The RLa-a load on two 6V6 is about 7k nominally, so I put 4k7 & 2n2 in series network across each 1/2 primary. The pole between 4k7 and 2n2 = 15khz, so at the F which the amp caters for, 50Hz to 8kHz, the networks cause no loading loss of HF in the wanted region of BW.

So, no matter what sine wave or square wave F I applied at input, there was NO HF oscillations with 10dB FB, so problem solved.

HF oscillations in guitar amps may be more common in over driven output stages than anyone suspects.

With the RF present, a high level 200Hz tone becomes harsh, and you can hear the effective cut off and interrpution to sine waves while RF is present on wave peaks. With RF not present, high overdrive sounds "more grunty", as the amp is wrestling to force the speaker to comply with the applied signals.
with 20dB over drive, with NFB or none at all, sound is similar because it consists mainly of a series of square waves dominated by the lowest note played.
Unfortunately, there is no Master volume control to amplify the preamp signals which have been raised to a desirable level of distortion, and then be able t play the sound in small venues but with a distorty sound.

Now for a tube warmer.

Someone wanted a warmer sound for their guitar amp, so I suggested they have 3 cascaded stages with a 1/2 12AU7, so 3 gain stages and a CF at output is possible. You want warmth? Well, methinks that means chocolatizing the signal with high levels of harmonics produced **without any stage clipping**.
(( nobody has offered any funds for the warmer, but the recipe here is free ))

How? OK, have a normally biased 1/2 12AU7 with say 22k DC carrying anode load and B+ supply about 200V, so Ia = 3.0mAdc. Ea = 130Vdc. Bypass the cathode resistor. Following grid in next stage has 0.27uF and 390k coupling.
Input from a gain control can produce about 42Vrms at onset of grid current and cut off. At onset of g1 I and Co, 2H may be 5% and 3H 2% and other H in declining %.
The voltage gain g1 to a = 11 approx. The maximum g1 input could be approx 4.2Vrms, and easily produced usually by the guitar amp stage just previous to the power amp section.
Now consider you have THREE such stages in series, ( cascaded ). BUT, instead of ending up with gain = 11 x 11 x 11 = 1,331, you use the 390k in a divider network with 11 : 1 signal reduction, using 390k + 39k, and then the gain from g1 to top of 39k is about 1.0. You can have 3 stages in series and still get overall gain of 1. Having 10 stages might be tricky, because if gain is say 1.5, or 0.5, after 10 stages the V0 is either too large of too low.
One thing is for sure, the sound of un-clipped devices becomes very obvious,
and I leave you all to ponder just what that might be like.

Max Vo after 3 stages is taken from top of last 39k and buffered with CF. So max Vo = Vin and perhaps just an input control is all that is needed.

Patrick Turner.