Filament Power Supply and Common Noise

Almost every effort within the design of audio circuitry to increase immunity to B+ noise increases sensitivity to filament noise because it increases Rk, but heater supplies are different from B+ supplies, and cause considerable confusion. Filament supply ripple may be defined as differential-mode noise, since it is the difference in voltage between one heater pin and another. Provided that differential-mode noise is reasonably low, the considerable thermal inertia of an indirectly heated cathode filters it out completely, and differential RF noise is irrelevant.

Nevertheless, there has been a trend of using exotic and expensive heater regulators in the belief that this allows the finished amplifier to sound better. If tube filaments were designed to be supplied by 6.3 Vrms of AC ripple, it hardly seems likely that they will be materially affected by minor differences between filament regulators producing perhaps < 10 mV peak to peak, since this is essentially perfection anyway, so why are there believable reports of audible improvements caused by changing heater regulators?

Tubes may not be particularly sensitive to differential-mode noise of filament supplies, but they certainly are sensitive to common-mode noise. In the context of a filament supply, common-mode noise leaves the potential difference between the ends of the filament constant, but both voltages are bouncing up and down, and if we view the filament as a single conductor, its voltage with respect to the cathode is changing, rendering it perfectly capable of inducing noise into cathode.

Common-mode heater noise is a problem for small-signal tubes because the noise current is capacitively coupled from the heater directly to the enclosing cathode. Unless decoupled by a cathode capacitor, the noise current develops a noise voltage at the cathode primarily determined by Rk which is then added to the wanted signal and amplified by the tube. Cathode follower amplifiers have a low Rk while differential pair amplifiers have quite a high Rk, so it is clear that we need to be able to prevent common-mode noise from reaching the filament supply.

Because regulators are designed to address differential-mode noise they are normally ineffective against common-mode noise, although their support circuitry may inadvertently assist in rejecting common-mode noise. Useful common-mode filtering is only gained by employing RF techniques involving series RF chokes and shunt capacitors to chassis.

Because the filament is capacitively coupled to the cathode, and the reactance of filament cathode capacitance is low at RF, the cathode is likely to be particularly sensitive to RF for two reasons:

1. In a single-ended amplifier, the cathode will be decoupled to ground using a large capacitor. But the capacitor is not connected directly onto the cathode emissive surface, so inductance in the connecting wires reduces its effectiveness at RF. Additionally, the electrolytic capacitor has some inductance.

2. In a differential pair amplifier, the cathode unavoidably has quite a high resistance to ground, via the anode load resistors and not the cathode resistance, and cannot forma useful CR filter in conjunction with capacitance provided by cathode filament pair. We are forced to rely on the usually poor RF balance of the differential pair to reject RF, so a differential pair is likely to be more sensitive to filament-borne RF than a single-ended amplifier.