25 November 2011

Once again, the Cathode-Coupled Amplifier
Over the last 12 years, I have written much about the cathode-coupled amplifier. The reason is easy to find: I like the circuit a great deal and I believe that it is underappreciated. Two triodes and a few resistors are all that is required.

     1999/tubecircuits/Common-Cathode Amplifiers
     1999/tubecad.com/july99/page16
     2005/April/blog 42
     2007/05/blog 105
     2007/05/blog 106
     2007/05/blog 107
     2007/05/blog 108
     2008/11/blog 152
     2010/11/blog 194

I even have a PCB for the cathode-coupled amplifier, the GlassWare CCA design, which uses an Aikido cathode follower as an output stage and which scrubs away the power-supply noise riding on the cathode-coupled amplifier's output signal.

The CCA line amplifier presents no phase inversion and offers a very low input capacitance, as the typical Miller-effect capacitance is sidestepped; the input triode's plate does not move with the input signal. In addition, it produces little distortion and a low output impedance. Yet, few have bought the PCB or kit. (One day, the word will get out and I will instantly sell out of the remaining boards. I am often startled by how a certain PCB can languish on the shelf, then instantly sell out in a day or two, accompanied with breathless e-mails asking how many hours before the kits arrive. This happened recently with the H-PS-1 low-voltage regulated power supply. Word got how good the Squeezebox line sound with it in place of the cheap switching wallwart.)

So, if I have already covered the cathode-coupled amplifier, why am I returning to this topology? I have been corresponding with a reader who wants to modify an existing tube line stage he owns that holds three tube sockets mounted on the chassis top and is hardwired inside. He is keen to get the most performance possible from these three tubes, as he has just listened to a friend's octal Aikido All-in-One and now his line stage amplifier doesn't sound nearly as good. Good man.

In fact, a stereo Aikido line stage can be made from just three tubes.

The plan requires that the input triode's plate resistor split the B+ voltage (and by extension, the power-supply noise). Then the Aikido cathode follower will null the 50% of the power-supply noise that the input stage leaks. The plate resistor should be roughly equal to

       rp + (mu + 1)Rk.

I could stop right there, as such a line stage would sound much better than what he had been running before (a grounded-cathode amplifier followed by an SRPP stage and a global feedback loop), but but he wished to retain his super-expensive trio of NIB, NOS tubes: one 12AX7 and two 12AU7s. The problem with using a 12AX7 as the input tube in the three-tube Aikido is that it creates far too much gain, +34dB, as gain equals mu/2. Of course, replacing the 12AX7 with one extra 12AU7 would solve that problem. Still, his request for help got me thinking about my old friend the cathode-coupled amplifier as a possible candidate.

Aikido Cathode-Coupled Amplifier
I have presented this circuit before: it injects a portion of the power supply noise into the second triode's grid to cancel the power supply noise at its plate. The only new addition is the resistor in parallel with the capacitor, which introduces a positive grid-bias voltage for the second triode's grid, allowing both triodes to conduct an equal amount of current, in spite of the dissimilar cathode-to-plate voltages.

The right triode sees two input signals: the one at its cathode from the left triode and the one from the connection to the B+ through the two-resistor voltage divider and the 1µF capacitor. By carefully selecting resistor value, a deep null in power supply noise at the circuit's output can be achieved.

If the 1µF capacitor has shunted the bottom of the three resistors in series (the way 99.9% of tube gurus would do it), the cathode-coupled amplifier's poor PSRR would result in most of the power supply noise exiting with the output signal. But by injecting the optimal amount of power supply noise to the right triode's grid, we can provoke an anti-phase output signal just great enough to null the power supply noise at the plate, thereby producing a noise-free output signal. The exact ratio will depend on the tubes used the plate and cathode resistor values.

Now, let's fill in a design example's values. (Of course, this is just one of a near-infiite number of possible examples. So, yes, a 6SN7 could be used in this topology, as could two ECC99s or two 300Bs or two 834s. "Topology is separate from parts," which I would have tattooed on my arm, if I ever got that drunk, drug addled...or senile.)

Note how high the cathode voltage and the right triode's grid voltages are in this design example. With a cathode-to-plate voltage of 240V, the leftmost triode must see -10Vdc on its grid to establish an idle current flow of 5mA, whereas the rightmost triode's 115V cathode-to-plate voltage requires only -3.1Vdc of grid voltage (10V - 6.9V) to match the 5mA of current flow.

The next thing to note is the absence of a coupling capacitor. Where is the coupling capacitor? I left it off the schematic purposely, as an external load resistance will force a different set of resistor values for the two-resistor voltage divider. This makes for a poor line-stage amplifier, as a good tube-based line-stage amplifier should work well with a large range of load impedances, say from 10k to 100k. Of course, we could target the right values for 31.6k, relying on other load impedances being close enough for government work. (Speaking of which, a friend of mine was once coming back into consciousness and as the anesthesia wore off, he heard the army doctor say, "Okay, let's stitch him up: I'm getting hungry for lunch; besides that's good enough for government work." ) The better path is not to load down the cathode-coupled amplifier's output with any resistive load.

If the same 12AU7 & 12AX7 pairing is used, with the 12AU7 triode as the input triode (the left riode in the differential pair), the 12AX7's grid voltage will be much lower, as a result so, too, must be the feedback ratio, which means a higher gain at the output will result. This might just be feature in some applications, but for line-stage amplifier use, a better choice would be to use the same medium-mu triode throughout, say two 6CG7, 6SN7, 12AU7, 12BH7, 5687, ECC99 or, even triode-connected EL84 and EL86 tubes. Why? Using medium-mu triodes will develop a higher grid voltage for both cathode-coupled amplifier triodes, which will allow more feedback to be used, resulting in less gain at the output.

The two constant-current sources were added to make for the lower cathode-to-grid voltage on the cathode-coupled amplifier input stage, but a negative power supply could be used instead. In fact, -12Vdc might prove sufficient. I would use the -12V negative power supply to power the cathode-coupled amplifier's tube's heater and give the top two triodes a heater power supply referenced to 140Vdc. Yes, two heater power supplies is a pain, but so is exceeding a tube's maximum cathode-to-heater voltage, usually only 100Vdc.

I better stop before I catch my breath and write another section.

Next Time
I hope to have more news about the big PCB shipment and I will return to the current-output amplifier.

//JRB