This argument about DC effects on filament has been around for more than 30 years - You take some and leave some. Some talk about "cathode poisoning" or some ion migration in heaters of IDH tubes and cathodes of DH tubes. It doesn't discredit the high frequency AC filament heating method. I have tried regulated DC before embarking on the HFAC and did find it was a bit "muffled", but I'm not endorsing the "need to switch current" mumbo jumbo. Interaction with stray heater/cathode/grid capacitances maybe?
At any rate HFAC is certainly more rational than sticking a class D amp operating at HF for heater duty IMO.
Last edited:
Could you expand on this? I'm unclear about the meaning of the third paragraph. Also, could you comment on the reduction of IM products of AC heating and DC heating* by paralleling identical valves with filaments in opposite polarity?
Thanks, as always, very much,
Chris
*edit: by this I mean the "0th" order products of the DC gradient
Last edited:
I don't take any of those arguments. None of them is based on evidence. I have presented some evidence (published on a peer reviewed journal) which debunks the "technical" myths about DC heating. A properly made DC supply doesn't muffle anything. I'd rather say that 99% of people have a flawed stereo system, starting from the listening room. But don't want to start an argument on this and hijack the thread.
A proper DC supply can probably be made but is not as simple as might seem. Morgan Jones' books Valve Amplifiers and Building Valve Amplifiers have many examples which I used for my LM317 implementation but it may have been operating too close to its limits current-wise or maybe stray capacitances from wiring came into play. Anyways the HFAC supplies were simpler to build and bypassed the whole DC argument altogether. Nobody argues about AC heating virtues after all apart from the hum problem which is inaudible at 50KHz.
I made my listening tests using electrostatic headphones btw - No listening room artifacts. But I agree, let's not hijack the thread.
Last edited:
I apologise for going off topic somewhat. Leon Theremin used RF to control the filament voltage of a triode to adjust the volume in his 1927 Theremin. BTW I have often wondered about RF heating of filaments.
https://en.wikipedia.org/wiki/Leon_Theremin
https://www.theremin.us/RCA/rca_theremin.html
https://www.theremin.us/RCA/rca_schematic2.gif
BTW I love the video.
https://en.wikipedia.org/wiki/Leon_Theremin
https://www.theremin.us/RCA/rca_theremin.html
https://www.theremin.us/RCA/rca_schematic2.gif
I just don't like the idea of running quite high power RF currents around an amplifier. Passing CE conducted emissions regulations could get challenging.
A 1 Hz square wave with smoothed rise and fall would be a better heater drive and absolutely incapable of making any audible hum. Easy to generate too.
A 1 Hz square wave with smoothed rise and fall would be a better heater drive and absolutely incapable of making any audible hum. Easy to generate too.
David maybe you are correct.
I was thinking of using an ultrasonic sine wave as no audio frequencies would be involved and the fact that a pure sine wave has no harmonics unlike a square wave. David you did suggest a smoothed rise and fall that would remove a lot of the high order harmonics.
This is a very interesting topic that could use some experimentation.
Apart from the CE EMC certification hazards, the low frequency RF drive risks intermodulation with the audio down into audibility. Valves are not perfect amplifiers and operate at high impedances, so capacitive coupling from the heaters to the signal pins is hard to avoid.
Also dht tubes the heater is the cathode so is directly in the signal path. another spanner in the works.
The third paragraph:
> Connecting PP sides & L and R channels all together corrupts the signal (anode current swing causes a differential voltage across the filament).
In each filament, the anode current signal passes through some portion of the length of the filament, and exits at one end towards ground.
Given that the filament is resistive (ca. 3.5Ω for 300B) the signal generates a small differential voltage (I'm assuming, of course that a high-impedance source is driving the filament heating).
If we connect the filaments of a PP pair together, the antiphase differential voltages are effectively cancelled by the flow of anode current from one PP half to the other. This flow of alien signal current where it does not belong is what I mean by corruption of the signal.
A similar mechanism applies with connecting L and R channel filaments together, and corrupts the channel separation.
I do - AC heat generates IMD sideband on the music signal that degrade the performance. See post #22 on this very thread!
https://www.diyaudio.com/community/...at-filaments-high-freq-ac.411444/post-7654579
If they were truly identical, and used separate cathode resistors, it may well work for Parallel-SE; with PP the phase of the sidebands will be also driven by the music signal, (antiphase for PP) so I don't believe cancellation for PP would work.
I never think about cancellation without Lynnn Olson's statement coming to mind (to the effect) the Gotcha with signal cancellation is the dirty nature of the residual. So I would measure and listen carefully to a variety of real valves.
Rod I think for PP amp with a CT sinewave drive on each tube you will just get 2nd harmonic both outputs in phase which will cancel at the OPT. I don't see an issue with driving the heaters from say 100KHz as any IMD will now fall outside the audio band and will be at least 30dB down. This is no worse than the operation of a class D amp. DC is easier with one tube but if you have 4 to drive then maybe AC is less component count. Driving with 100KHz square wave will work too but is more likely to cause radiation down the speaker cables. I quite like the idea to driving 4 tubes straight from a transformer with multiple secondaries tapped in the centre for each cathode resistor. The EMC issue is probably more easier to contain than the class D amp as the 100KHz drive is differential and harmonics less likely to get through the OPT.