Thanks all!
One thing I "cheated" on a little is the tube Mu triode rating, strictly speaking it would be
MUt = Gmg/(Gms+Gmp) as normally given in data sheets. I simplified it to Gmg/Gms since Gmp is typically less than 1% of Gms.
Setting U% to 1 in the UL Mu formula to get the triode connected case points this out:
Mu = Gmg/(U%Gms + Gmp) ==> MUt = Gmg/(Gms+Gmp)
Also, just to clarify, the previous derivations use Vg,Vs, & Vp in the sense of electrode to cathode voltages.
Looking at my last CFB derivation, I think I still have an error yet in this:
has: Gmg(dVin+C%*dVout)+Gms*C%*dVout+Gmp*dVout = 0
but should be:
Gmg(dVin+C%*dVout)+Gms*C%*dVout+Gmp*(1-C%)dVout = 0
this needs to use Gmp*(1-C%)*dVout for the last term, since only (1-C%) of the output variation is on the plate electrode. But since I simplifyed by dropping the small Gmp term later on, it dropped out anyway.
Next problem to solve I think is the output impedance formulas for UL and CFB. One would start with the general formula:
1) Iout = k(Gmg*Vg+Gsm*Vs+Gmp*Vp)^1.5
2) substitute in the relations for the UL or CFB case
UL: dVs = U%*dVp
CFB: dVg = dVin + C%*dVout, dVs = C%*dVout (or 0 for a bootstrapped screen), & dVp = (1-C%)*dVout
3) set dVin = 0 and take the derivative for dIout/dVout and algebraically solve for dVout/dIout to get Rout = dVout/dIout
Don
One thing I "cheated" on a little is the tube Mu triode rating, strictly speaking it would be
MUt = Gmg/(Gms+Gmp) as normally given in data sheets. I simplified it to Gmg/Gms since Gmp is typically less than 1% of Gms.
Setting U% to 1 in the UL Mu formula to get the triode connected case points this out:
Mu = Gmg/(U%Gms + Gmp) ==> MUt = Gmg/(Gms+Gmp)
Also, just to clarify, the previous derivations use Vg,Vs, & Vp in the sense of electrode to cathode voltages.
Looking at my last CFB derivation, I think I still have an error yet in this:
has: Gmg(dVin+C%*dVout)+Gms*C%*dVout+Gmp*dVout = 0
but should be:
Gmg(dVin+C%*dVout)+Gms*C%*dVout+Gmp*(1-C%)dVout = 0
this needs to use Gmp*(1-C%)*dVout for the last term, since only (1-C%) of the output variation is on the plate electrode. But since I simplifyed by dropping the small Gmp term later on, it dropped out anyway.
Next problem to solve I think is the output impedance formulas for UL and CFB. One would start with the general formula:
1) Iout = k(Gmg*Vg+Gsm*Vs+Gmp*Vp)^1.5
2) substitute in the relations for the UL or CFB case
UL: dVs = U%*dVp
CFB: dVg = dVin + C%*dVout, dVs = C%*dVout (or 0 for a bootstrapped screen), & dVp = (1-C%)*dVout
3) set dVin = 0 and take the derivative for dIout/dVout and algebraically solve for dVout/dIout to get Rout = dVout/dIout
Don
UL-CCS circuit
Here is an interesting circuit that puts the pentode in a different light I think. I now think of them as triodes with an auxilliary control output.
The MOSFET follower could also be replaced with a cathode follower, but I wanted to point out the new Fairchild FQP1N50 (and also the near complement P-channel FQP1P50 ) 500V MOSFETs. These parts are very low capacitance parts, and in follower mode have entirely negligible input capacitance. They have about 1/3 the capacitance of the IRF820 parts I usually see used for plate CCS or tube hybrids. For 600V or 900V there are the new ST P2NK60Z and ST P2NK90Z also, but not as good on capacitance as the Fairchild parts.
This circuit allows the tube to run in constant current "triode" mode with very low output impedance loads.
R2 could easily be replaced by a P-channel FQP1P50 to make for a full totem pole output stage. Needs some crossover gate biasing then.
Full hybrid possibilities are to use the new Motorola NJL1302DG and NJL3281DG bipolar outputs if the tube is big enough to drive them directly (or keep the MOSFETs as drivers for them) . (these have the thermal biasing diodes for totem pole config. built right onto the transistor dies, no more "transistor sounds" with these)
If one can find a pentode with a MU triode rating up around 20 one could try using it as the amplifier input stage, with the screen grid used for global feedback from the speaker output terminal (cap coupled). No attenuator network needed then.
Don
Here is an interesting circuit that puts the pentode in a different light I think. I now think of them as triodes with an auxilliary control output.
The MOSFET follower could also be replaced with a cathode follower, but I wanted to point out the new Fairchild FQP1N50 (and also the near complement P-channel FQP1P50 ) 500V MOSFETs. These parts are very low capacitance parts, and in follower mode have entirely negligible input capacitance. They have about 1/3 the capacitance of the IRF820 parts I usually see used for plate CCS or tube hybrids. For 600V or 900V there are the new ST P2NK60Z and ST P2NK90Z also, but not as good on capacitance as the Fairchild parts.
This circuit allows the tube to run in constant current "triode" mode with very low output impedance loads.
R2 could easily be replaced by a P-channel FQP1P50 to make for a full totem pole output stage. Needs some crossover gate biasing then.
Full hybrid possibilities are to use the new Motorola NJL1302DG and NJL3281DG bipolar outputs if the tube is big enough to drive them directly (or keep the MOSFETs as drivers for them) . (these have the thermal biasing diodes for totem pole config. built right onto the transistor dies, no more "transistor sounds" with these)
If one can find a pentode with a MU triode rating up around 20 one could try using it as the amplifier input stage, with the screen grid used for global feedback from the speaker output terminal (cap coupled). No attenuator network needed then.
Don
Attachments
Whups! The pentode circuit above needs a resistor, around 5k, across the zener diode, so the MOSFET can turn off too. (and for the NE-2 neon bulb bias current too.)
The circuit, by the way, provides an exact gain of Mu(triode) = Gmg/Gms at its output.
Just using a CCS loaded ordinary triode followed by a follower would introduce the gain loss of the follower, which some say they can hear.
This circuit also gives a very low output impedance due to the high local loop gain available at the CCS loaded plate. (ie, better than just the output Z of the follower alone)
Don
The circuit, by the way, provides an exact gain of Mu(triode) = Gmg/Gms at its output.
Just using a CCS loaded ordinary triode followed by a follower would introduce the gain loss of the follower, which some say they can hear.
This circuit also gives a very low output impedance due to the high local loop gain available at the CCS loaded plate. (ie, better than just the output Z of the follower alone)
Don
Going back to the split-load (Quad) topology, I am one of the guilty ones who never really cared too much about Walker's maths - remember in those days one had mainly slide rules! All of that was of comparative limited practical use to me because the output transformer characteristics were still not taken into account. What the maths showed could be guestimated comparatively easy with little use for refined mathematical gymnastics.
I started by experimenting with several topologies of output transformer and found no serious differences in ultimate performance between them. Dr Walker would not then divulge the details of his output transformer - so we simply took a (somewhat later) burnt-out one apart! With truer UL taps of about 20% - 25% (the Quad's are only 10%) some h.f. compensation was always required, e.g. as a series R-C element from anode-g2. This mainly concerned frequencies quite higher than audio, so was no great worry. I wondered at the all of 6 secondary sections used by the Quad, which raised the cost with no mentionable advantages over, say, 4 sections. I nowadays settle for G2 "taps" round about 20%. Some 30% would yield lower distortion, but which must be discounted against the higher G1 signal driving voltage required.
As said before, it is cost effective here to have output transformers wound locally; that offers an advantage.
I started by experimenting with several topologies of output transformer and found no serious differences in ultimate performance between them. Dr Walker would not then divulge the details of his output transformer - so we simply took a (somewhat later) burnt-out one apart! With truer UL taps of about 20% - 25% (the Quad's are only 10%) some h.f. compensation was always required, e.g. as a series R-C element from anode-g2. This mainly concerned frequencies quite higher than audio, so was no great worry. I wondered at the all of 6 secondary sections used by the Quad, which raised the cost with no mentionable advantages over, say, 4 sections. I nowadays settle for G2 "taps" round about 20%. Some 30% would yield lower distortion, but which must be discounted against the higher G1 signal driving voltage required.
As said before, it is cost effective here to have output transformers wound locally; that offers an advantage.
EC8010,
I had an article on Peter Walker and Quad (one of many) where it was stated that round about the time of the Queens Award (1978) he received an honourary doctorate for his pioneering work in audio. I cannot find that now, neither can I get it back on the internet (where you may have tried too). I will keep looking for my print; for now that was where I got it, so "memory serving" etc.
Regards.
I had an article on Peter Walker and Quad (one of many) where it was stated that round about the time of the Queens Award (1978) he received an honourary doctorate for his pioneering work in audio. I cannot find that now, neither can I get it back on the internet (where you may have tried too). I will keep looking for my print; for now that was where I got it, so "memory serving" etc.
Regards.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.