On my last post I covered how the gyrator PCB can be used in a pentode driver. The pentode driver is the best candidate in a “plate to plate” / shunt feedback or “Schade” feedback amplifier which is the name typically used in the DIYAudio world. The triode doesn’t work well here as you need high gain and low distortion with a load which can get quite low (due to the feedback effect of the feedback resistor). I’m not going to cover the subject as it has been covered (and discussed) extensively before by many people, so I suggest you do a bit of research yourself if you are interested in the subject and want to learn more.
A Study example
I chose one of my two favourite valves for this circuit. The output stage is a GU-50. It can deliver good power, whilst having a good transconductance and gain, both are good advantages as we will see. The other valve is the 12HL7 that has been discovered recently and praised by many. I have tested it with great success in triode-strapped mode.
Here is the circuit for analysis:
Don’t get scared, it’s not that complex at all. Ok, let’s start with the driver first. In figure 1 we can see that the 12HL7 is biased at about 170V/28mA but has cathode degeneration with R12 being 100Ω which isn’t bypassed on purpose. This helps linearising the stage as the 12HL7 has lots of gain as well as transconductance (I measured μ=290 and gm=22mA/V). The gain of the stage is reduced due to this degeneration which reduces the effective transconductance. The simulated gain is about 38dB (or 79.4 times). The maximum gain that could be achieved on this stage without degeneration would be around gm⋅(RA||rp)or about 170 (we ignore Rf as is much higher than the RA || rp pair). The disadvantage of this topology is that you will burn some current through RA at idle. The higher is the anode quiescent voltage the more current is needed so you get the top FET(M2) and RA dissipating heat. The gate of GU-50 is driven from the low impedance output from the mu-follower (gyrator) circuit. This will provide sufficient current to drive the gate and its capacitance.
The “plate to plate” feedback is taken from the anode of the GU-50 and to the anode of the driver via Rf. The feedback ratio (B) is given by the ratio of (RA||rp)/(RA||rp +Rf). We typically want to keep this ratio around 10-20%. The more feedback, the lower the output impedance, but the harmonic profile of the stage changes as well. In this case the feedback is close to 9%. The output stage is biased at nearly full-tilt. 70mA and 33W of dissipation on the GU-50 (which has maximum of 40W). You can increase and get a bit more out of juice, but this stage already provides about 11-12W with a maximum input of 1.2V peak. The gain of the output stage is about 0dB. The GU-50 provides about 26x gain as gm⋅(Zaa||rp) = 7.5mA/V ⋅ 6KΩ//8k5Ω = 7.5 * 3.5 = 26.25. This is because the anode resistance is about 8k5Ω and the transconductance I measured was about 7.5mA/V. The x26 gain reduction is provided by the output transformer which is has a gain of 26.45 or 5k6:1 (without copper losses). The net gain is 1 (0dB) which matches with the simulation.
What is interesting to point out is that (ideally) the output resistance of the “Schaded” pentode is close to 1/gm. The higher the transconductance the better. In reality, the output pentode has finite and small gain (about 65 in this case) so the output resistance is closer to 1kΩ rather than the ideal 133Ω. The more feedback the better as well as the higher gm. So you probably want to look at other pentodes with high gm (vertical TV sweep valves are ideal here).
The gain with feedback is reduced down to 22dB. Not too shabby for producing 8W at 1.5% in single ended mode! Well, you need a 500V supply (ouch). There are other options to look into if you don’t want to use the GU-50 which likes the high volts.
Hope this post generates some thinking and debate. Looking forward to the comments.
And so Tim was right….. (Update I)
You don’t need Ra, when rp can do the job:
After playing a bit with my favourite Russian pentode (6Z49P-DR), I increased the drive on the GU-50, increased HT to get maximum output from this bottle. The RA has to be included as RF is high 100KΩ and at least Spice didn’t like it. I suspect there will be stability issues given the value of Rp. Safer is to add the RA to stabilise the gyrator. Now distortion is reduced significantly.
Here is the distortion profile from the simulation: