idea for a simple bias servo for AB PP

[vladn]

The purpose of this servo is to maintain steady *combined* bias current for a PP output stage in AB. Balancing bias current between the tubes has to be done manually. But it is simple enough to be used where dedicated fixed bias winding is unavailable and therefore proportional g2 voltage and bias voltage is difficult to obtain. Stabilization of g2 voltage is not required with this circuit.

First the simplified block diagram illustrating the idea. D1/C1/R2 captures the bottom of each current cycle (the trough of the 2nd harmonic of the signal) which corresponds to the quiescent current in both steady state, class A and class AB operation. C1/R2 time constant is set to sub-audible frequency range. The a simple servo uses this voltage to set the bias voltage



And the possible real implementation with a single pnp transistor servo and all necessary power supply circuitry. R6 trim pot used to manually balance bias current. Note that wiper disconnect event on R6/R7 lowers the bias voltage so it is safe. Zeners provide stabilized positive rail for reference voltage and negative rail for bias voltage (negative rail is stabilized only to limit max collector voltage). A current-limited AC feed from B+ or some other suitable winding provides the power for the servo.

[Michael Koster]

I built a circuit like this last year and showed it at VSAC this spring.
One difference is that I applied it to class A using 300B tubes,
but the principle is the same. I track the troughs of the
combined cathode current waveform using a negative peak
follower circuit and fed that to an error amplifier with a LPF,
and back to control the cathode voltage. I also tried a balancer
circuit in the loop.

http://db.audioasylum.com/cgi/m.mpl?...ediy&n=142992&

My goal is to provide the "stiffness" of fixed bias in class A with
the stability of cathode bias or even CCS bias.

I had trouble with stability at large signal conditions, and I have
not completed the investigation into why. Some could have been
implementation issues with my opamp based circuit (basic debug)
but also there were some things with the behavior of the cathode
current that I didn't anticipate. Basically, I went back to ruminate
on it some more and switched the amp to resistor bias for more
measurements.

I like your approach of a simpler circuit, though I haven't thought
through your particular circuit through yet.

I'll dig back through my notes and see if there's anything that might
help. One suggestion I do have is to make a lot of observations
under different signal conditions so you really understand what it
is you're controlling ;-). I haven't looked at class AB for a while but
it could be a whole different set of issues.

Anyway, it sounds like a great idea and project and I look forward
to hearing about your experiences.

Michael

PS some of the problem is class A was due to signal current unbalance

[Michael Koster]

I remember one problem with class A being related to the fact
that it is possible for the cathode current to decrease as signal
increases due to unbalanced signal between tubes. You should
not have this problem in class AB.

[vladn]

Any AB biased amp has a region of class A operation. So I did some simulations introducing deliberate imbalance in both gain and static bias running at different signal levels (including class A operation). Imbalance leads to imperfect suppression of the main tone at the common cathode resistor and some interesting trough shapes. However the resulting bias estimation error was really small (below few percent).

I think what you may have seen is an interaction between your bias balancing servo and common bias control servo if you had both loops running simultaneously.

[Michael Koster]

I thought that early on, so I disabled the balancer and it still had
the same instability. At some level of signal, the controlled current
would oscillate (signal generator input) or jump up (music peaks)
at the time constant of the servo LPF.

As I recall, the sensed cathode current would briefly drop to near
zero on one side only, leading to the servo over reacting and driving
the current up on both sides by about 1.5X until the servo settled
again. I could observe this drop in cathode current across the cathode
resistor, so I concluded that my MOSFET voltage regulator was
probably glitching for some reason. I tried the usual remedies
like feedback caps etc. but never really tracked it down.

Michael

[vladn]

I've done some more simulations with the loop closed (needed to tweak some values to make it work). I do not see signs of instability with closed loop either (again trying different levels and imbalances). There is some 2nd harmonic feed into grids but it is symmetrical so it does not affect amp output much. But maybe this feed has a potential of creating some instability via non-linear effects when the servo gain is too high.

[Wavebourn]

Here was mine, from Alligator project (SE, of course class A):



Edit: for class AB I was thinking about a similar device with switched time constant, like a S/H: shorter time constant on low signal levels, and very long one with some signal applied. It should work well. Also, I saw some guys implemented PIC device that used to "reset" bias when the amp is switched on each time.

[kenpeter]

R1 is a resistor. Held at constant current (by servo or whatever)
means a constant voltage drop. Replace with a string of LEDs
and be done with it.

If you wanna get fancy, an adjustible current source bypassed
with a good audio cap will get you the same result.

I'm not sure why the need to return your servo to the grids
rather than to the cathode? Just seems rather complicated...

[vladn]

kenpeter - I think you are missing few points. First of all your suggestion of shunted CCS is not suitable for AB operation. As for your suggestion of a fixed drop in cathode circuit - there are a lot of ways to generate stabilized fixed bias, all of them require stabilized g2 voltage as well for a pentode design which this is all about.

BTW Michael Koster circuit does exactly the same thing (I had little doubts that I am reinventing the wheel here since I am new to the tubes) and the servo feedback goes to the cathodes but more complex servo amp is required.

[kenpeter]

Sorry, most my designs have twin parallel current sources with
some sort weird cathode bridging or bypassing. Got blindsided
by the not-so-weird single tail.

I agree, with only one common tail for DC, the single bypassed
current source does not bias optimally for AB.

--------------------------------------------------------------------------

You have been less than clear in some details. How exactly does
G2 supposedly figure into it?