Some time ago I finished a built an all tube amplifier system consisting of a modified version of Joe Curcio's "Daniel" preamplifier (TAA 2/85, p. 7) and a 40 W Class A power amplifier based on the legendary Williamson amplifier.
I needed a system in which I could turn on the power to all equipment with onlyone switch, located on the preamp front panel along with other controls. In principle this is not difficult to achieve with a bundle of extension cords and a master power switch. With a tube-based power amplifier drawing over 300 W from the power line, it's inconvenient to switch the system on in the morning and then switch it off in the evening, which is exactly what I did with my previous solid state amplifiers.
My tube preamp is on for an average about 40 hours per week and easily accumulates 2000 hours per year. Fitted with a set of Russian 6922/6DJ8s good for about 10,000 hours, the preamp will probably last for five years without any tube changes. The power amplifier, on the other hand, might need new power tubes once a year!
An audio operated power switch to promptly turn on or off a power amp without extra stress on the tubes would be convenient. Switching on could be activated by monitoring the input audio signal, while switching off could be delayed from 10 to 30 minutes after no input signal. Since the active use of the power amp is probably less that half of the total on-time, I would more than double the lifetime of the power amp tubes with this scheme.
My theory regarding the failure of tubes is as follows:
Convincing? Unfortunately I haven't uncoverd any published information about the effects of low filament voltage on tube life. My reference books (Eastman, Fundamentals of Vacuum Tubes; Millman, Vacuum Tube and Semiconductor Electronics; Valley & Wallman, Vacuum Tube Amplifiers) ignore the topic completely. Therefore, my "theory" is based solely on my own reasoning and some articles inGlass Audio Magazine.
The VTL Book (by David Manley, available through Old Colony) contains one of the few fragments of information on tube life. The data sheet for the GE's KT66 states that at full rated power the expected life time will be about 2,000h, and at derated power 10,000h, provided that the tube is not switched on more that 12 times per day! Trouble is that today's tubes are (among other things) not what they used to be.
I experimented on my prototype amp to meet the challenge of creating a fast switch-on for tubes. I attempted to determine the lowest filament voltage to keep the tubes warm enough for immediate operation after restoring normal filament and plate voltages (Table 1). The start-up time is the interval from switch-on to normal sound output. Above 2.6V the plate supply voltage turn-on time primarily determines the delay. If the filament voltage during the warm period is more than 2.6V, the amp will operate (you can hear the sound) virtually immediately after the switch-on from a low-level filament voltage to full supply voltages.
The cold current surge is a serious problem in a power amps using multiple output tubes. For example, the resistance of a cold 6L6-GC filament is roughly 3-4x that of a hot filament. A 6L6GC, normally drawing 0.9A, has a surge of about 3A during a cold start. This becomes even more hair-raising in my 12-tube amp (eight power tubes and four others, with a normal heater current of 9.6 A), with the surge current exceeding 30A.
Figure 1 shows how to achieve lowered filament voltage when the plate supply is switched off. When the power is first applied using the main power switch S1, the relay K1 remains unenergized. The filaments of the stereo channels are connected in series, along with an optional NTC resistor. Thus the filament current will have a very smooth and easy rise with absolutely no cold current surge.
In my amplifier the total cold filament current is initially about 1A, rising to a steady 3A level in two minutes. The voltage drop over the NTC resistor (Philips type 2322 644 90008 or Siemens type Q63036-S1509-M) was about 1.2V, leaving some 2.6V over the filaments of both channels. Without the thermistor the cold current peak is about 8A, dropping to 3.5A in 30 seconds, and the cold current surge is no larger than the normal full-power heater current. Even if you can't purchase a suitable thermistor, series connection of the stereo channels will help reduce tube stress during switch-on.
In the standby mode my prototype amplifier consumes about 20W of line power, which uses about 175 kWh energy per year, provided the standby power feature is enabled full-time. At my local power company's current energy rates, the additional costs is less than $20 per year.
The reduced filament power lets you switch the power amp on or off as necessary. To fully exploit this feature, I designed a control unit, which can be fitted inside the amplifier (Figure 2). Alternatively, you can build the unit as a separate add-on in its own case. See the parts lists in Table 2.
The operation is quite similar to "An Audio Activated Power Switch" in TAA 1/84, except a few details. This unit has an initial delay before power is applied after a cold start. The input buffers feature very high input impedance to minimize signal source loading. The power switching occurs with a relay, since a multi-pole switching, such as K1 in Figure 1, is necessary.
After a cold start the unit functions as a delay timer, gently powering up the tube filaments to a reduced operating voltage and then switching the amp on to full power. This will take 136s at 60Hz line frequency or 164s at 50Hz instead of the normal 10s (which is the price you pay for the soft start). After the power-up it will then act as a full-power/standby-power switch, monitoring the signal input. Switching from reduced standby power to full power is virtually immediate.
During the initial power-up R1, C7 and IC3B generate a short reset pulse that clears the counters IC2 and IC6. The flip-flop formed by gates IC5A and IC5D is also reset at power-up. IC4F keeps the counter IC6 cleared as long as the flip-flop stays reset. After a 136s delay, determined by counter IC2 output 13, the flip-flop IC5A/ IC5D is finally set. This enables counter IC6 and switches on the relay. Thus a warm-up delay of about two minutes is provided after a cold start.
The dual operational amplifier (IC1) monitors the input signals to the power amplifier. A noise-free input circuit is essential, since we must not inject any noise to the input signal. The diodes D1-D4 protect the op amp inputs from possible switching-on transients of a tube preamplifier. The capacitor C3 and diode D7 form simple peak-to-peak detector.
On positive peaks capacitors C3 is charged through diode D7. On negative peaks the entire peak-to-peak value (minus the diode voltage drop), referenced to VCC, is available at the junction of diode D7 and resistor R16. As soon as this voltage falls below the logic threshold of the Schmitt-trigger input inverters IC3F/IC3E, the wired or CLEAR line is pulled low, clearing the counter IC6.
If the peak-to-peak level of the input signal is above 20 mV p/p, counter IC6 will be repeatedly cleared, thus keeping relay K1 on. With no input signal, IC6 counts output pulses from IC2, enabled with gate IC5C. The frequency is about 1 Hz (actually, 60/64Hz or 50/64Hz). After a time period, the jumper selected output of IC6 will switch high. IC5C freezes the counter IC6 by blocking the clock signal. Gate IC5B switches off the relay.
The amp is now switched to standby mode with reduced filament power and no plate voltage, but the control unit is constantly monitoring the signal inputs. When the signal level again exceeds 20 mV p/p, counter IC6 is cleared and the amplifier is switched on to full filament power, now already warm and ready for immediate operation.
You can change the timings by selecting other outputs of IC6 with the jumper block J11 (Table 3). You can also connect an optional override switch in solder points J9 and J10. If the override switch is on, the amp will stay on full power regardless of the input signal level. The input signal sensitivity can be changed with resistors R11..R14.
| IC6 pin | Delay (60 Hz) | Delay (50 Hz) |
| 4 | 137 s | 164 s |
| 13 | 273 s | 328 s |
| 12 | 546 s | 655 s |
| 14 | 1092 s | 1310 s |
| 15 | 2185 s | 2621 s |
| 1 | 4369 s | 5243 s |
The power supply is very simple, so you can easily use any available transformer secondary in the range of 6-9 V ac. Keep in mind, however, that the ground of the power supply is connected to the input signal ground. Any noise or hum on that ground will be superimposed to the audio signal. I suggest you use a separate transformer for the control unit.
This concept is applicable to all tube power amplifiers. If you can't easily connect in series the filaments of the stereo channels, you can add a suitable series resistor, which is then by-passed in full-power mode with the relay contact. In my amp, this resistor would have to be a hefty 0.5 (/25W unit. The power consumed in the resistor is wasted, thus the power consumption in standby mode would increase to about 45...50 watts.
If you don't want to use the automatic switching feature, just replace the relay K1 in Fig. 1 with a 110 V coil unit and use a separate full-power switch for manual control. I designed the printed circuit board layout small enough to fit it inside an amplifier (Figure 3 and Figure 4). I needed to fit a few resistors vertically. Don't forget to install the seven jumper wires.
I hope these guidelines help you extend the tube lifetimes and also make your tube amplifiers more practical and user-friendly.
Figure 1: Lowered filament voltage schematic.
Figure 2: Schematic of the amplifier timer.
Figure 3: Printed circuit board (scale 2:1).
Figure 4: Stuffing guide (scale 2:1).
Back to Glass Audio Article List
Feedback is invited: