I'm so dang proud of the Muntzing and tweak I came up with tonight, if I were vain I'd call it the "Broskie-van der Sluys Autobias"
As a continuation of this thread
, I pulled a little 4558 off a board laying around the shop and went to work.
So, what does it do? It's actually three circuits in one - DC servo, Broskie bias compensator and foldover circuit (think of "runaway" brakes, like on a train).DC servo.
The 1/2 of the 4558 that feeds the base of the A1380 shunt-regulator is the heart of the entire kaboodle. By itself, it's a DC servo. The "-" input (because of the transistor, the inputs are inverted) is fed with reference, through a 220K resistor. The "+" is fed with sense voltage, developed across the 10 ohm cathode resistor. The opamp tries to keep its differential inputs equalized and as a byproduct of that, we get a constant cathode current.Broskie bias compensator.
As John Broskie pointed out in his TubeCad articles here
, the failure of the typical DC servo is when the tube input signal drives the bias toward zero volts and is assymetrical, the tube starts coming out of class-A, due to the audio appearing across the cathode resistor that confuses the servo, making it push the bias more negative. With a seriously assymetrical waveform (what music is a sinewave anyway?), the servo can push the tube into bad non-linear conduction. Result, even before you are *near* and not *at* 0V on g1
, you have terrible sounding music.
What Broskie figued was that a DC servo worked quite excellent as long as the grid didn't conduct and the tube stayed in class-A. But as the tube approached the borderline, the circuit should up the bias to keep the tube class-A. It was a very simple task to accomplish - toss a diode in there so it turned on as the point was reached. That's what the 1N4148 in series with the 220K resistor does. As the tube starts developing odd audio-related votages across the cathode resistor, the diode contucts and "assists" the reference voltage, effectively increasing the bias, keeping the tube in the class-A area with large input signals.
This brings us to those two 0.1uF polycaps... the crossover/recovery time must be as balanced as possible from DC servo to bias-compensator, unless you want to introduce some distortion (it's a subsonic bass pulse, your woofer will hate you if you get it wrong
). Make the values too large and the circuit will be too slow on transients. And on recovery, the tube current shoots up for too long. Too small of values and audio will leak into the bias circuit. Since when this part of the circuit is active, the two 220K resistors will be acting in parallel, hence keeping the time constant equal with the 100K/0.1uF on the sense line.
The Broskie circuit does a wonderful job and can actually increase the output power of an amplifier. Its flaw though, is it doesn't know when to quit! If you keep pouring more signal on the input, it'll keep raising the bias through the tube. On my CV4060/S11E12 with mu-follower driver, as I turned the function genny to max, the current through the tube was an insane 130mA with 465V on the anode - twice its 60mA bias and well over Pd
So, we need a third circuit in here, which tells the device that "enough is enough" and cut back on the current. The other section of the 4558 does just that. There is a divider on the input that sets this "trip point" as to when to hit the brakes. The values shown (in my CV4060 circuit) will cut the current very abruptly at 72mA, where any higher input causes the current through the tube to fall lower and lower, until it hits the DC servo mode again, despite the waveform. Use the variable alternative to tweak the trip point. This is all simply accomplished as the foldover diode turns on, it places voltage on the sense input of the control opamp, making the chip realize there's too much current flowing and because its voltage will be stronger because the impedance on that diode is lower (thanks to the opamp driving it), the Broskie compensation is overridden.Caveats.
As with anything, nothing is free. The higher the trip point, the better the sound at full power and the more power you get (I get a solid 10.5W sinewave with the 72mA trip point). The disadvantage is shortened tube life and a power supply that may moan and whine.
Also, in this "Muntzed" version, there is some component interaction. If you change the 220K resistor on the Broskie compensation diode, your trip point will go all to heck. But this interaction is one-way... the trip point resistor doesn't change the Broskie section at all..
So, I got this all down with one dual opamp and some peripheral parts. Thanks to the PSRR of the opamp, a regulated supply isn't needed. I also don't see why most any dual opamp can't be used.
Because this is based on the Broskie circuit, Copyleft dictates it's available for use under the same conditions as the original - any and all use, including commercial, for attribution.