Power Amplifier Run On Timer (PAROT)

Valve power amplifiers should not be put into 'transmit' mode until the cathode is at sufficient temperature, especially for coated cathode tubes (most commonly used on indirectly heated cathodes), as damage to the cathode surface can occur It can take from just a few seconds to five minutes or more for the cathode to reach suitable temperature.

Many valves benefit after full power operation from a cool down time to facilitate removal of heat energy from the PA enclosure before shut down of the cooling fan, and accordingly this feature is included in some amplifiers. For example the Kenwood TL-922 has a 140s cool down timer after filament is shut down and the service manual claims it will measurably extend tube life. Note that during PAROT's cool down phase, filaments / heaters are still powered and a longer cool down period is appropriate.

A common design for valve low end RF Power Amplifiers is that they are entirely powered from a single transformer, and often the cooling fan is dependent on that transformer. Again, low PAs often lack some form of protection against drawing current from a low temperature cathode.

Another flaw of some of these low end designs is that the mains fuses are inclined to blow if the power is cycled rapidly, just due to the residual heat in the fuses as a result of prior operation at high power, or possibly exacerbated by a soft start system that may be slow in recycling.

This article describes a simple external device to:

The device is inserted in the mains power connection between power outlet and amplifier, and between the transceiver's tx control line (referred to as PTT in this article) and PA's tx control. No alteration to the power amplifier is required, it is operated as normal except its power switch is left in the ON position and PAROT used exclusively for switch on / switch off.

Logical design

The design uses a small microcontroller to interpret presses of a single button to operate a mains switching relay, a PTT relay and a LED for system status indication.

Fig 1: 

The system lends itself to a simple Finite State Machine design. Fig 1 shows the updated state diagram, Long and Short mean long and short button presses respectively, and rct, hdt and tot are three programmable timers. (Diagram produced with Lucidchart.)

The most common sequence of states would be from power on which starts at the RST state, automatically moving to the OFF state after RCT (typically 2s) and awaiting button input. A short button press moves the state to HEAT which powers the PA and after hdt (say 180s for an indirectly heated ceramic valve) moves to ON which enables PTT so the amplifier can be put into transmit mode. When operations are complete, a short press of the button moves the system to COOL state which blocks PTT and after a programmed cool down time (say 900s) moves to RST which after rct moves to OFF where it waits again for button input.

At any point, a long button push does an instant shutdown and moves to the RST state, then after rct to the OFF state waiting for input.

A cool down can be aborted and status returned to ON status at any time during the COOL state. Similarly, during the HEAT state, a short (or long) push will abort the start up and moves to the RST state, then after rct to the OFF state waiting for input.

If a short mains interruption occurs, the PAROT should reset (so long as its power supply filters don't hold too much charge) and start in the RST state to enforce a minimum power off time before moving to the OFF state where a short button press can reapply power to the PA.

Key to the logical design is separation of algorithms and operating parameters. Key operating parameters are stored in EEPROM and are field programmable to adapt the system to specific needs. The program is stored in flash memory.

The firmware will run in an ATTiny25, 45, or 85. The firmware is compatible with ATB - a bootloader for AVR Tiny microcontrollers.


The EEPROM layout is version dependent. The current flash program will ignore EEPROM data that does not match its EEPROM version level.

Table 1: EEPROM structure
Variable Bytes Comment
ver 1 EEPROM version
opt 2 Options, see Table 2.
sl 1 maximum duration of short button press in 10ms increments
hdt 2 heater delay time in seconds
tot 2 ON time out time in seconds
cdt 2 cool down time in seconds
rct 2 recycle time in seconds
ckadj 1 signed clock adjustment (~0.8%pu)

Table 1 shows the EEPROM variables, their size and usage. All variables are unsigned binary integers.

Note that any timer set to zero will instantly expire, and any timer set to 0xffff will never expire. (Setting rct or hdt to 0xffff will prevent normal operation as it will prevent exit from the associated states.)

The MCU RC oscillator is specified to be within +/- 10%, and they are often much closer than that. For convenience, ckadj can be used to compensate for an MCU RC oscillator that is too fast or slow. Intervals are increased by ckadj/1.22%, so if the intervals were 5% slow, make ckadj=-4 for more accurate intervals. For most applications, most MCU clocks will be sufficiently accurate without using ckadj.

Table 2: Option bits
Option Bit Comment
  15:0 Reserved

Table 2 shows the use of the bits in the Option byte of EEPROM

Fig 2: 

Fig 2 shows an example binary EEPROM image in Hex Editor Neo. Hex Editor Neo Professional contains a really useful facility to bind a structure definition to the binary file contents, allowing exploration of the structure using the structure element types, sizes, labels etc. Isn't that clever!

In this example, the maximum short button press is set for 50x10ms=500ms (a good choice), heater time is 180s (for an indirectly heated cathode requiring 3min), tot is disabled, 900s (15min) cool down time, and 2s recycle time.


Power supply filters are designed to collapse quickly (with the mains relay operated) on loss of power to facilitate the recycle timing feature of the PAROT. Excessive filter capacitance is not an improvement!

Fig 3: 

Fig 3 shows the circuit schematic for a 230VAC implementation. RL1 needs to be resized for a 120V application.

RL1 must be a heavy duty relay to stand a prospect of surviving a PA fault (double the rated current of the PA is not overkill). RL2 is a DIP relay. Relays are used over electronic switches to improve robustness.

V1.03 of the firmware uses only the G LED, the R LED (was on pin 6 has been dropped).

V1.04 of the firmware connects the LED from pin 5 to ground with a current limiting resistor.

Physical implementation

The prototype was constructed in a Hammond 1550C die cast aluminium box.

The mains input and output cables were made from an extension cord cut at an appropriate point to reach the power point and provide a convenient tail to the socket for the PA.

Fig 4: 

Fig 4 shows the completed PAROT in operation on top of an AL80B.


Fig 5: 

Fig 5 shows the interior of the PAROT. The DIP14 is the PTT relay, the DIP8 is the ATTiny25.

Fig 6: 

Fig 6 shows the solder side of the Veroboard.

The spacing between Veroboard tracks is 0.64mm, and there are many inconsistent and seemingly unjustifiable recommendations for voltage breakdown. I have measured the breakdown between adjacent tracks, between two tracks with intervening floating track, and across us track breaking hole as seen above. Tracks reliably broke down just over 1400Vpk,  two tracks with intervening floating track at 2800Vpk, and a track breaking hole was stronger than the between track gap. The board above was sprayed with a heavy coat of acrylic lacquer which would further strength the breakdown ratings. Experience shows that the board works fine on a nominal 230VAC supply.


Though the state diagram at Fig 1 might seem complicated, only a few of the paths are exercised in normal operation.

Fig 7: Simplified state diagram 

Fig 7 shows the state diagram with the most commonly used paths highlighted, blue for automatic (timed) transitions, and green for manual triggers.

In normal operation:

In the simplest case, only two button pushes are used for an operating session, one short one to start the PA, and one short one to shut it down.

Post implementation review

The PAROT has been in use for a month now, and with not changes needed or desired even. The 15min cooldown time is quite adequate for the AL80B which if it has been running AM where the anode gets quite bright red, the exhaust air stream has cooled almost to the heaters only condition in 10min.

One could use a longer cool down time to extend the time available for a change of mind and cancelling the cool down cycle to return immediately to operating condition and without cycling the filaments... it would be a personal choice.

Measured power consumption of the PAROT alone with relays released is 3W, and relays operated is 5W.



Version Date Description
1.01 30/03/2013 Initial.
1.02 09/05/2016 v1.02
1.03 25/05/16 v1.043

© Copyright: Owen Duffy 1995, 2021. All rights reserved. Disclaimer.