Thanks; that is indeed the paper.
But "superior" has to be in context. I have (say) a 80MPH minivan and a 160MPH supercar. I have to go 3 miles on 35-45MPH roads to get milk. Which car is superior? I maintain that a 13Meg impedance is moot when the load is 100K as stated. But still depends on context. The supercar may be able to get 0-35MPH a few seconds quicker than the minivan and save 30 seconds on the milk-run. When I get home my milk is 30 seconds fresher! So if aiming for point-zero-zero-zero-zero THD, the 13Meg source may get you part of that fourth zero.
...but their claim that the u of a cascode is based on the product of the mu factors of the triodes used to build the cascode is incomplete at best. The cascode gain should be calculated with the gm of the lower device times the plate load....just like a pentode......
Mu is not gain. Mu*Mu is very-nearly the effective Mu of the compound, minus many leakages which would not matter for reasonable Mu but dominate when Mu looks like 100*100= 10,000. The *gain* is, as you say, usually much more about the DC feed and the load resistance. For Mu much over a hundred the actual value of Mu hardly matters. If you don't lose it in a DC feed resistor, you lose it in any real-world load.
Take the despised 12AX7. Gm=1,600uMho or 1/625r. Mu=100. Compounded would be 10,000. But the DC feed for adequate current (especially for driving audio into capacitances) won't be much over 100K. So the gain rises from ~~50 single tube to 160 cascode with resistor DC feed. 3X not 100X. (Serious cascodes start with higher-current tubes for higher Gm, accepting the lower initial Mu because cascoded the Mu will be more than we can harness.)
I use cascode error amplifiers in regulator designs with the very ill suited 12AX7 and practically speaking have been able to realize gains of around 46 - 48dB with them.. (It's common and works reasonably well at surprisingly high voltages)
I also have used them with 6S3P and have been able to realize gains of up to 60dB or more in a single stage.
I've always treated them as Bandersnatch suggests although I find the gain predictions to be just a bit optimistic in some cases. (Most likely because my estimate of transconductance in the lower tube is actually a bit too high) No special factors required - I'd not heard the mu * mu thing before and it would seem to predict gains far beyond what is true in practice or have I missed something here. (In theory it might also predict less than the actual possible gain provided a large enough plate load resistor and suitably high supply voltage)
I also have used them with 6S3P and have been able to realize gains of up to 60dB or more in a single stage.
I've always treated them as Bandersnatch suggests although I find the gain predictions to be just a bit optimistic in some cases. (Most likely because my estimate of transconductance in the lower tube is actually a bit too high) No special factors required - I'd not heard the mu * mu thing before and it would seem to predict gains far beyond what is true in practice or have I missed something here. (In theory it might also predict less than the actual possible gain provided a large enough plate load resistor and suitably high supply voltage)
I have used triode on the bottom, and MOSFET on the top. This on two halves of a LTP phase splitter. Another amp I built on the same topology ran EL84's and another ran 6V6GTA's. The 6V6's got taken out and swapped with 6CL6's to harness more gain. They both sounded good...
cheers,
Douglas
I don't think a KT88 grid is a real heavy load. It is a EUR 6550 which is just a double-wide 6L6, which was designed to be "easy" to drive (relative to older power bottles, and to likely driver topologies).
Because the KT88/6550/6L6 has Mu(g2) near 10 (maybe 8), in SE (class A) the peak G1 voltage will never be over 1/10th of Vg2. The Vg2 supply is often a handy place to get driver supply. A plate-loaded resistance-coupled triode can usually do a peak voltage of 1/5th of its supply, _if_ we select the triode and resistors to the load. Use a number between 2 and 5; "3" is a fair starting point.
Load is said to be 100K. For KT88, higher G1 resistance is allowed if not max dissipation, or especially with cathode bias (which is almost always the best plan in SE). Input capacitance is about 25pFd, add strays (your coupling caps may be physically huge) and call it 50pFd, which crosses 100K at 33KHz. So it is indeed a 100K load over the audio band even if you go higher on Rg1.
100K divided by 3 is 33K, your driver plate DC feed resistor. 33K/3 is 10K, your desired triode plate resistance. And you will likely have ~~150V drop across those 33K resistors, so the triode must go to 5mA and some more.
12AU7 (or 6J5/6J7) is right in there. (That's why these tubes used to be "general purpose".)
12AT7 will come very close and give a little more voltage gain.
(12AX7 straight is not well qualified in this scheme. As cathode follower it may do, but the gain has to be made-up elsewhere. Mac did OK by bootstrapping a 12AX7 from its octopus transformer windings, but that is heavy DIY.)
Cathode follower is not needed; just a complication.
I'm not sure a driver needs a heap of voltage gain, if it means any compromise elsewhere. We divide the total task into Gain and Power. Power tubes have low voltage gain to get huge current and low impedance. Voltage gain is easy in the low-level low-current stages. The driver is an in-between case. It should have gain (and generally voltage gain) so we don't just move the voltage-swing problem one bottle to the left. However a gain of 10 or even 3 makes a lot less work for the stage before the driver, and allows a high-current driver tube for authoritative driving.
Certainly the 6J5/12AU7 class drivers worked for Williamson which have high-voltage triode-strap finals.
Now you pose an interesting thesis there PRR. Why would you want a lower voltage supply for the driver stages? What is served by this? Higher voltage leaves you able to take more plate load resistance, and drive more current through it...and then run it further from either saturation or cut off...resulting in lower distortion in the output.
cheers,
Douglas
cheers,
Douglas
The mu*mu thing is some sort of mistake. I am quite certain that if I build a cascode from a pair of 6S19P I can achieve voltage gain significantly greater than 4( this being the square of an optimistically reported mu of 2). It is fairly easy to show where the theory falls apart for that mu*mu thing...LOL
cheers,
Douglas
Al Hart of Grommes used 12ax7 cascode in a preamp with a 100K plate load - that was probably only a gain of ~100 (?) does 100-200 gain sound reasonable in a lightly loaded 12ax7 cascode 1st stage with plate load 100-220K ?
fwiw I did like the "sound" of the Precision Fidelity C7 with a Zener reference cap multiplier and RC 1st stage decoupler (did not like the stock regulator - C7 had a tube plate load instead of a resistor)
how well do anode followers with gain set ~2-3 simulate vs cathode follower when both use medium mu tubes?
fwiw I did like the "sound" of the Precision Fidelity C7 with a Zener reference cap multiplier and RC 1st stage decoupler (did not like the stock regulator - C7 had a tube plate load instead of a resistor)
how well do anode followers with gain set ~2-3 simulate vs cathode follower when both use medium mu tubes?
I typically use 162K - 274K as plate load resistor gains of up to 48dB or so should be possible at the top end of the range if the cascode is not loaded down by the following stage.
The 12AX7A though is not an ideal tube for this use, I would probably look at some others. (higher transconductance and lower rp)
The 12AX7A though is not an ideal tube for this use, I would probably look at some others. (higher transconductance and lower rp)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.