Tube DIY Asylum: Thinking out loud about HF filament supplies by Lynn Olson

Not long ago, a friend in Sweden wrote me an e-mail about using 5V computer switching supplies for filament heating. Although the economics are attractive, there are problems:

A) Audio circuits are have fairly high distortion above the audio band, thanks to assorted capacitive loads from Miller capacitance and stray C's in transformers. If there is no ultrasonic component in the incoming signal, no problem, but modern CD/SACD/DVD-A sources have significant ultrasonic residues going out 10MHz or higher. Plus ambient RF pickup from assorted RF sources in the home is now much worse than the "old days" in the 1950's.

Again, by itself, not too troubling, but *very* troublesome if some element in the power amp add its own harmonics in the same frequency range. Then we can expect RF crossmodulation and multiple low-level IM sidebands that extend down into the audio range. These randomly drifting tones are extremely undesirable, even at levels below the noise floor of -80dB.

B) So switching power-supply harmonics are extremely undesirable with modern signal sources, which are already contaminated with ultrasonic tones from digital sources. We must not assume audio circuits are linear up to 10MHz- instead, they'll act like a RF mixer, and generate many sidetones, some in the audio band.

C) But - if the supply can be phase-locked to the digital source, the sideband-generating problems go away! Not only that, phase-locking doesn't need to be ultralow jitter, as it is in normal external-DAC applications, so a inexpensive TOSLINK fiber-optic connection is good enough to achieve phase-lock for the RF circuit. All it has to do is more or less track the digital source. When an analog source is used, it can free-run at the same frequencies.

So, looking at the problem from a different perspective, instead of an off-the-shelf computer switching supply, a super-clean RF filament supply could be a quite interesting approach to filament heating.

The requirements would be:

1) Low or very low harmonic distortion sinewave, well under 1% distortion.

2) Resistance to AM or FM modulation by the audio signal, requiring VCXO crystal control and a very quiet, low-impedance power supply for the RF power amplifier.

3) Phase-locked to CD/SACD/DVD-A source at 44.1/48 kHz or twice that, 88.2/96kHz. However, since jitter is not too important, TOSLINK optical is probably good enough. If the source is analog, then it could free-run at 48 or 96 kHz.

Not too different than a late-generation HF bias for a mag-tape record head, except for the part about phase-locking to an external source. The load is also low-impedance and fairly reactive, which needs to be taken into account.

If these requirements could be met - and this isn't really too demanding, since all you're talking about is a phase-locked VCXO, heavy bandpass filtering, a small RF power amp, and HF transformer that feeds its balanced output into twinax shielded cable that goes right to the filament pins.

Since it would be desirable for the RF power-supply to float (thus retaining cathode bias for best tube life and sonics), it would desirable to transformer-couple the RF power to the tube pins instead of a building a complex floating power supply with lots of stray capacitance between the chassis and the audio circuits. In fact, the RF power supply really needs to have it's own metal can and independent AC supply isolated from the rest of the audio circuits.

A smart power-supply designer could build this. I'm not a solid-state guy, so I wouldn't even try - but those are the requirements for anyone who'd like to try it. I'm pretty sure it would sound good given the above requirements - it would avoid harmonics generated between CD residue and it's own operation thanks to the phase-lock feature, and balanced transformer-coupling would minimize filament-imbalance leakthrough to the main audio signal. You don't want non-audio HF being amplified by the DHT, so taking advantage of filament balance is very desirable.

A stable, transformer-coupled, accurately-balanced 88.2/96 kHz filament supply might in fact be very good sonically, without the subtle problems of 50/60 Hz AC or DC heating.