New Driver

[kingneb]

I am looking into making a new driver circuit as part of a second revision for my current KT88 PPP amplifier. I though about Morgan Jones's book and about how he stresses there should be only a single capacitive coupled stage in the whole amplifier. I remember him saying this mainly eliminates low frequency instability.

I designed a new driver with two direct coupled differential stages that capacitive couple to the output stage. Will this, plus the fact my new circuit is simpler, help improve natural stability, even for high frequencies? Will I still need step networks and the local feedback loop Waveborn suggested I install in my old design? Overall I am trying to eliminate the need for some of that extra stability circuitry, as it has some side unwanted effects.

Overall, I learned alot but I still am far from being an expert on the stability issue.

[DF96]

Removing coupling caps may help LF stability, but won't help HF stability. You may still need HF compensation networks. To a first approximation, LF and HF stability for a typical amp are unrelated. The best way to help loop stability (apart from understanding it) is to have a very high performance OPT.

An amplifier loop with just two LF poles (e.g. coupling cap, plus OPT) cannot oscillate, but it can still produce an LF peak so you still need to get the design right. Introducing a third LF pole, like many amps, does not suddenly turn a good circuit into a bad one. It just needs more careful design.

[MelB]

Here is my KT88 amp circuit. I can go see what I did for the values on the EF86 if you are interested. I can't remember how I hooked up the grids on the EF86 either. It's running as a triode and so are the KT88's (KT120's) I am not using any negative feedback. Using Hammond 1650N outputs. Next project is upgrading the power transformer from a Hammond 272JX to a 290HX.

[kingneb]

Someone suggested determining stability by building a phase bode plot. This would be by measuring, with a scope, the voltage at the feedback output relative to the input while loading the feedback with the resistance it will see.

Once I see unity gain at a 180 degree phase shift will that mean the frequency it will oscillate at?

[DF96]

Yes, you need to break the feedback network but then load both the network and the feedback point with equivalent impedances so that the rest of the circuit is unchanged. In practice an approsimation may be good enough.

180 deg feedback at unity or more gain means instability (roughly - ignoring conditional stability). Less than this may still mean a response peak.

[kingneb]

OK, lets say I find a unity gain, 180 degree phase shift point at 100k.

I am a bit rusty, but to bring that back into phase would the step network you would use some equation with complex numbers to represent the angle? I cannot remember what it is. Something to move the angle -180 degrees.

[DF96]

No, you don't need to change the phase by 180 degrees. Just enough to stop the loop gain locus from curling around the (1,0) point. The normal lead-lag compensation network in an anode circuit works by reducing high frequency gain, although it increases phase shift over a band of frequencies. The capacitor across the feedback resistor can sometimes help by reducing phase shift, although it increases loop gain.

[chrish]

The driver section of my 6L6 amp is two 6SN7 in DC coupled diff amp config. Designed by Tubelab, it works fine. May be of assistance to you...