Effects of AC Heating Power Applied to Directly Heated Triodes

(c) Copyright 1999, Steve Bench. All Rights reserved. Revision 2.

This report examines the effects of using AC to heat the emitting structure (filamentary cathode) of a Directly Heated Triode. I will concentrate on these effects as applied to a single ended amplifier stage, as some of these effects are automatically cancelled in a push-pull arrangement. I will use an 801 thoriated tungsten filament device as a working example. The principles discussed in this report are applicable to all filamentary cathode structures.

Three major effects occur:

  1. Line frequency hum, primarily induced by imbalance in the filament circuit wiring, the transformer used to power the filament, coupling into the signal leads, etc. This is typically reduced or eliminated using a "hum balancing pot". This is a potentiometer connected across the filament whose wiper arm connects to the DC return of the system. This is fairly commonly utilized, and affects the line frequency component (and odd harmonics). No more about this will be mentioned in this report, as the principles are well established. However, the schematics included herein include this technique.
  2. Hum produced from within the filamentary structure. This is primarily caused by the lack of "infinite mass" of the filament. Since heating power is symmetric for each half cycle of applied voltage, the principle component of this hum is twice the line frequency (and even harmonics of the line frequency). Physically, the filament minutely heats up and cools down at twice the line rate, due to AC applied to the filament. The next section of this report covers this effect and its reduction.
  3. Signal modulation of the hum component. The effect is an intermodulation of the applied signal at twice the line frequency. This effect applies equally to single ended and push-pull directly heated devices. The effect is not a hum (low level) effect; rather, it is a high level distortion effect. For example, with a 60Hz line frequency, if you were to introduce a pure 1000 Hz signal into the amplifier, you will get not only an amplified 1000 Hz output (and some level of harmonics at 2000 Hz, 3000 Hz etc) but also a hum modulation set of components. The primary components in this example are at 880Hz and 1120Hz. As this is an intermodulation effect, there will also be components at many other frequencies, for example 760 Hz (1000-240 Hz), 1640Hz (2*1000-3*120) etc. Note that IM products, unlike harmonic components, are not sonically related to the signal. The third section of this report discusses the magnitude of this effect, and the conditions needed to minimize it.

The principles discussed in this report have been experimentally verified on the following devices:

801: Low MU Thoriated Tungsten filamentary heating structure.

841: High MU Thoriated Tungsten filamentary heating structure.

6B4G: Low MU Oxide Coated filamentary heating structure.

1U4 (triode connected): Medium MU small mass filamentary heating structure.

2. DHT AC Powered Hum Control

This section covers low level hum reduction associated with line powered DHT filaments. I will assume the line frequency component has been balanced out with the standard "hum pot". However, even when you have done this, the left over hum, primarily at twice the line frequency, may still be objectionable. Since it is at twice the line frequency, it is often assumed to be associated with the bias or HT circuitry. Attempts to better filter these sources sometimes makes things worst. In this section, we will see why that is so.

As mentioned above, the slight heating and cooling every half cycle of the AC input frequency is the source of this hum component. As it turns out, the phase of the hum components is very fortuitous. Namely, a full wave rectified signal producing a negative voltage (such as would be used for bias) injected into the grid reduces the hum, and a full wave rectified producing a positive voltage (such as used for the HT) injected into the plate also reduces the hum. This suggests that a properly phased and "inadequately" filtered source is actually beneficial in reducing this hum component. More on this later.

We will cover the magnitude of the effect in this paragraph. To investigate this, I used a filter bank to obtain a set of signals locked to the incoming line frequency. I then adjusted the phase and amplitude of each signal to see what hum reduction could be obtained. My "mains" frequency is 60 Hz. Those with other mains frequencies should translate the frequencies accordingly. I set up an experimental 801 amplifier. This was an RC coupled circuit. I also slightly "starved" the filament by running it on 6.3V instead of 7.5V. (Actual filament voltage was 6.62VAC). The HT was 600VDC VERY well filtered (residual AC component was 10 microvolts). Bias source was obtained from a 5651 filtered for a residual AC component of about 3 microvolts. This was done so that I could independently study the effect of the DHT hum components. The circuit was biased at about -26.5V, and the plate load was 27k, producing a circuit gain of about 6. The filament was balanced with a hum pot. The residual hum was 342 millivolts at the plate. (Note that if this were an output stage, thru the transformer impedance transformation and load, this would translate to a few mV, which is typical for SE DHT stages). Injecting a properly phased 120Hz component allowed a hum reduction of 18 dB. Adding in a properly phased 240Hz component increased this hum rejection to 25 dB. Adding in a properly phased component at  360Hz increased this rejection to 29dB. At this point, that major hum component left was 180Hz, as there is a phase difference due to the mass (capacitance) of the filament structure) that causes the 60 and 180 components to be slightly out of phase with each other. Adding in properly phased 180Hz increased the rejection to about 31 dB. This rejection remained stable over time and with cycling of power on/off. As long as the inserted cancellation components tracked the magnitude of the "line", the rejection stayed within a dB or so from a line input of 110VAC to 130VAC. Using the filter bank approach, the best I was able to achieve was 36 dB (120, 180, 240, 360, 480, 540, 600 Hz filters).

A practical circuit to do this would consist of perhaps 3 LC bandpass filters: tuned to 120Hz, 240Hz and 360Hz. Then, by slightly adjusting the resonant frequency, the phase can be altered, and by adjusting a "summing node" resistor the amplitude could be controlled.

There is a MUCH simpler "compromise" network that can provide very effective control. As I mentioned above, the phase of the induced hum components is such that a negative full wave rectifier injected on the grid or a positive full wave rectifier injected on the plate is of the correct phase to cancel this hum. Therefore by adding THREE COMPONENTS to the typical circuit, the DHT hum introduction can be cancelled. Here's the circuit:

The "left half" of the schematic is rather conventional. It does include the standard "hum pot". The "right half" circuit shows the introduction of two small diodes and a second pot to correct for the DHT induced hum. Using this circuit, I was able to reduce the output side hum from 342 mV to 29 mV, for almost 22 dB reduction in overall hum level. Notice that this compromise circuit provides most of the advantage of a very complex circuit with almost no added complexity. For the crowd who shun the use of the small silicon signal diode, a 6AL5 or 6H6 could be used instead. By the way, I used a "high quality" 10 turn pot for the second hum reduction potentiometer. There did not seem to be much difference with the type of diode, although depending on the tube and "layout" sometimes 0.02uF across each of the diodes helped.

This technique is also applicable to "cathode" bias techniques. Here's a sample circuit, again using an 801 as the example tube:

Notice that this is no more complicated than the fixed bias example. The cathode bypass should be "effective" well below the "mains" frequency", which is usually the case anyway. If the filament transformer has a center tap, don't use it.

This technique can be extended to include the entire power amp "front end". Here is an example of my entire driver amp with hum compensation:

In this example I added one more LC, roughly tuned to 240 Hz in a high pass configuration. The compensation was "injected" into the 6SN7 stage. Notice that the "phasing" is the same. Since this is additional injection is band pass (filtering 240 Hz, 360 Hz etc), I could bypass the first canceller with a 220 nF capacitor, providing some additional higher frequency line noise rejection. The overall output ripple from this circuit is about 12(!) millivolts, for a overall hum reduction of 340/12 or 29 dB reduction.

BTW, this driver provides about 40 dB of gain, and makes a very nice sounding power amplifier driver.

Other DHT Types

The 841 has essentially the same filamentary structure as the 801, and compensated almost exactly the same, when you account for the difference in gain (and difference in load, bias etc). The 6B4 has a slightly less massive filament structure, and the phase and amplitude components needed for cancellation were different, but about the same reduction was able to be effected. The 1U4, with its tiny filament didn't compensate quite as well, allowing only a 15 dB improvement.  I did not try the filter bank approach with the 1U4; this may allow more improvement. In case it wasn't obvious, DON'T EVEN THINK about applying 600 volts to a 1U4. The phase differences can be handled by either changing the "zero" (changing the 2.2uF in the circuit above), or adding a "pole" by adding a capacitor to ground at the wiper of the pot etc.

Other Methods

Also as mentioned above, an inadequately filtered HT and/or bias supply is actually beneficial. However, the phasing is such that the required filter is a power supply consisting of ONLY an RC filter. LC or CLC filters produce excessive phase shift and won't properly compensate. Using an RC filter, simply adjust the output C for minimum hum! There will be a definite "minimum"; increasing the C above this point will cause the output hum to increase, as the supply "ripple" is too low to compensate. Notice that a full wave section, not half wave nor doubler should be used.

3. HUM MODULATION Effects of the DHT

As mentioned in the introductory section, there is also an intermodulation effect caused by the minute temperature changes in the filament. This incrementally changes the transconductance of the device, leading to a modulation effect. This effect will be observable in SE and PP applications. In this section, we will investigate the magnitude of this effect and determine how it might be possible to minimize the effect.

To study this effect, I used the same circuit as shown in the schematic above. I applied various levels of 1000 Hz signal to the input of the amplifier, and measured the resulting component at 880 Hz as an indicator of this effect. The bias voltage was carefully adjusted to 26.5V bias, and the Filament voltage was 6.6V

For this test, I will "table" the INPUT peak signal level vs % modulation.

Input Peak Level vs % Modulation
Input Level (Pk) Hum IMD %
10 0.12%
15 0.13
20 0.14
21 0.142
22 0.144
23 0.16
24 0.20
25 0.24
26 0.28
27 0.31
28 0.34
29 0.35
30 0.35

The peak filament AC is 4.6V peak (half of 6.6*1.4). If you subtract this from the 26.5V bias you get 21.9V. Notice that in the table, the distortion more or less "linearly" increases with signal level (which, incidentally, is NOT normally characteristic of IM) until you get past 22V peak signal, then the hum modulation increases rather rapidly, reaching a maximum at about clip point (which is evident at 28-30V peak input).

This suggests that the way to keep hum "modulation" effects under control is to establish the bias conditions such that the bias is greater than the signal level + the peak of the AC filament voltage.

The hum cancellation discussed in section 2 did not affect the hum modulation by more than 1 dB, when cancelled in either the grid or the plate circuit of the 801. Not shown on the schematic above, but this stage is driven by a 6SN7 "input stage". When the hum cancelling signal is injected into the a small 10 ohm "sampling" resistor in the cathode of the 6SN7, there was no effect of the hum cancellation on the hum modulation IMD. This suggests that the small effect observed in the 801 grid is due to the slight variation in plate resistance of the 6SN7, and the slight effect when injected at the 801 plate is due to the slight variation in plate resistance of the 801.