Mu Follower to drive a Push-Pull Amplifier?

[Kashmire]

So I've been working on my KT88 Push-Pull power amplifier. I started with a cathode follower driver. The DC-coupled cathode follower buffers a high-gain anode (i.e. 12AX7 or 12AT7) with a low impedance cathode (12BH7) to drive the KT88 grids. The grids have 16pF of capacitance, which computes to 63kOhms at 100kHz. This is pretty high, so the driver is presented with an easy load. The schematic is classic and looks similar to below. Don't pick on the exact parts values. Schematic is representative and pretty close.

I look at this, and think, "there is a big resistor on top of the 12AT7, and a big resistor below the 12BH7, why not remove the resistors stack the valves?" Stacking the valves and moving the tap to the top tube's cathode results in an SRPP (note: shown with 12AT7 valves. This arrangement needs work to keep the 12AT7/12BH7 combination). The SRPP functions far differently than the original configuration, and easily drives the KT88 grids. Note the shared cathode resistor, which can be replaced with individual 330 Ohm resistors if wanted.

So I go back to the cathode follower, but add AC bootstrapping to linearlize the 12AT7 (edit: the 12AT7 plate resistors should be 25k:50k for a total of 75k).

But I'm still looking at the same situation as before. Why not blend the two stacks into a single stack? Moving some parts around, this results in the familiar mu follower.

Now the top tube's grid DC bias is a slave (through the 1M resistor) to the stack's bias point. One more tweak to decouple and stabilize the top tube's grid's bias point:

My question is as follows: why are SRPP or mu-followers rarely found on push-pull amplifiers? The cathode follower is ubiquitous. Why not the other two?

[kevinkr]

Economics, complexity, excessively high supply voltages required, etc.

I have experimented with several of the topologies in power amplifiers, and your last one is quite familiar also as single ended it is the basis of the gyrators I use in my phono stages. I take the output from the plate of that stage not the cathode. This is obviously a fixed bias mu-follower, it does present a load on the plate of the lower tube beyond what would be the case with a mu-follower, however it makes using dissimilar tube types quite easy as it gives you control of the voltage division across the upper and lower tube particularly if you insert a pot in the divider chain. (Note you can bootstrap the 1M resistor if you feel so inclined, but with low rp types this isn't of much benefit, and I think the additional complexity is not justifiable.)

I also like the voltage amp bootstrapped by the CF as that should give you gain approaching mu, and better linearity as you indicated.. Just watch out where you place the bootstrap LF pole if you wrap the entire amplifier in global feedback. IMHO it would need to be much lower than the dominant LF pole to prevent issues with LF stability. (big cap)

I used the differential SRPP in line stages and in a 6BQ5 based PP amp.. Works well particularly if you substitute a good CCS for the rather short tail resistor in your implementation.

[Kashmire]

excessively high supply voltages required

I don't see it. In the first schematic (the standard configuration), the 12AT7 plate resides at 1/2 of the preamp's B+ (in this case, 350V). The 12BH7 is DC coupled at the first stage's plate potential (175V). Each tube has that amount of voltage swing (up or down, depending on which tube). The first leg (12AT7) has voltage headroom in one direction (dictated by the "top" 175V established over its plate resistor), and the second leg (12BH7) has it in the opposite direction (dictated by the "bottom" 175V established over its cathode resistor).

With the Mu Follower (or SRPP), the same thing occurs. The middle bias point can be 1/2 of B+. The bottom tube "sees" its plate voltage imposed by the top tube. As the bottom tube conducts more current, its voltage drops, and that excess voltage appears across the top tube, just as if it was a plate resistor. Reciprocity says a similar (but opposite) thing happens when the bottom tube pinches current, making it look like a big cathode resistor to the top tube.

The only difference is a resistor can drop to 0V at 0A, whereas a tube is nonlinear and still drops voltage as current approaches zero.

[kevinkr]

Depending on tubes used and the value of cathode bias resistors you might swing 3/4 of the available supply voltage with reasonable linearity. I've found in several instances that I needed higher supply voltages to achieve the same general level of linearity in an SRPP as compared to a VAS driving a CF.. I use SRPPs to drive the 300B based output stages in my current amplifier and I find their output current capability significantly limited as compared to a comparable CF, and on a 400V rail my 10mA SRPP driver can swing just 200Vpp before it gives up due to load current required by the 100K grid resistors, and just a little beyond that grid current begins to flow.. The source impedance of this SRPP is probably on the order of a couple of K ohms, whereas the CF would probably be in the range of 200 - 300 ohms.. My experience is limited however to 6SN7/6SL7/12AX7/12AU7.. I never seen one of these even unloaded swing close the voltages implied by the supply rail, but there may well be types (low perveance types) that do far better.

A choke or transformer loaded driver will swing far greater voltages on lower rails, which since this is what I am doing these days partly prompted my comment about the supply voltage.

I think that any of your proposed designs could be very interesting to develop into a working amplifier.

[DF96]

The grounded-cathode with cathode-follower bootstrapped load is logically almost identical to the mu-follower, and can be analysed in the same way. The differences are that the quiescent current in the two valves no longer have to be the same, so the CF can run higher current, and there is a small built-in load applied by the upper anode resistor in parallel with the CF cathode resistor. The SRPP is simply a special case of the mu-follower.

Your original circuit was different, in that it had no bootstrap. This makes it different in three ways: lower output impedance, higher distortion, lower voltage gain. Whether this is better or not for the whole circuit depends on the rest of the circuit. For example, if you were driving the output stage into grid current then low impedance is important so 'no bootstrap' might be better.

The snag with mu-follower and SRPP is that they work best into a specific load impedance. In the case of the SRPP this can be much lower than the output impedance so gain suffers. This may be why they are rarely seen as drivers. The bootstrapped circuit gives more flexibility.

[SY]

Kash, if you RC couple between the input voltage amp and the CF, return the CF cathode resistors to -150-200V, return the CF grid leaks to -50V (adjustable), and direct couple the CF to the output tube grids, you'll get very effective antiblocking. You may even be able to squeeze more watts out.

[Kashmire]

You may even be able to squeeze more watts out.

...driving the output stage into grid current...

My parts runs out of voltage first. The output transformers (Plitron 4070CFB) are 4000 Ohm and the power supply is 450V. The transformers only have a single secondary, so this can't be adjusted (i.e. use the 16 Ohm tap). I can't squeeze any more watts without more B+ (or am I missing something?) I would have probably been better with the 2000 Ohm OPTs. Lower turns ratio, too.

The grounded-cathode with cathode-follower bootstrapped load is logically almost identical to the mu-follower.

Exactly. That's why I was pondering why it isn't used more often. The reactive load of the KT88 grids (16.5pF) is about 63k at 100kHz, going to more than 200k in the audible range. Since this is a frequency-dependent (complex) impedance, is the mu-follower inappropriate?

[Kashmire]

Something like this SY?

[SY]

The plate resistors on the 12AU7 are a problem. You also want those plates at AC ground. Ideally, you'd have a +100 to +150 supply running it. Lacking that, tie the two plates together, use a 15k or so dropping resistor from the +450, then bypass the plates to ground.

[DF96]

Originally Posted by Kashmire

is the mu-follower inappropriate?

It could be appropriate, but the bootstrap circuit gives more flexibility. However, a simple grounded cathode with no follower may also be appropriate. Why spend money on a fancy driver when a simple driver will do, and most of the distortion is going to come from the output stage anyway? (Note that a simple driver won't always do, but it often will). A poorly designed fancy driver could be worse than a well designed simple driver.