I was asked by a local station to help debug his problematic amplifier, this unit had been quite unreliable in the past.

The notes here represent my findings on this particular amplifier.

The amplifier is a GS35 grounded grid HF amp manufactured in the UK. It uses a GM3SEK Triode Control board. Previous failures included the destruction of the HT glitch resistors and a constant stream of cathode bias transistors going S/C. On its last visit to the manufacturer the amp was brought up to their "latest" specification, also a new valve was fitted.

The choice of components used by the manufacturer when modifying this particular amplifier suggests that they believe that the failure mode of the cathode bias transistor (Q1) was due to HV flashover, all of the modifications and guessed improvements focused on this type of failure, to no avail.

After the big refurb .. it lasted just 3 hours!

This time the damage was substantial. The amp had run for some time with the cathode bias transistor shorted. Failure of the bias transistor (Q1) which goes S/C causes the tube to conduct, sitting at about 1A Anode current. The tube had thus been dissipating several KW (3000V x 1A) until it was noticed. The copper of the heat exchanger was black! The owner was justifiably annoyed!

So what was the real failure mode?

The Triode Board is in use in countless amplifiers throughout the world and there are no reports of reliability issues. I have several in use here and would recommend them to anyone. The Triode board used in this commercial amplifier is a second generation product with some changes to how the bias is switched to cut-off the tube. It is however substantially the same circuit so it should share the same reliability.

I spoke to Ian, GM3SEK, the designer of the control board and we followed several lines of investigation. Ian pointed out an important note from the original Triode Board manual -

Quote: Heatsink for Q1: 4in x 3in x 1in, or equivalent (2°C/W or less). For low bias voltages and lower-power tubes, you can mount Q1 on a cool area of the chassis.

Of course the GS35 isn't a low power / low bias voltage tube at all, so following the note above you would expect to find a large heatsink for Q1. In this amplifier unfortunately the transistor is simply fitted to the corner of the chassis, the corner of the chassis used wasn't particularly cool either.

Time for some measurements, The nice thing about the circuit used to provide the bias control on the Triode Board is that you can make most of the tests without any HT connected. Its possible using a bench power supply to simulate Cathode and Grid currents without harming the valve or putting yourself at risk. This is all simple stuff explained in the manual.

Fitting a thermocouple probe to the tab of Q1 and simulating a few minutes of heavy usage showed an alarming rise in temperature. In less than 1 minute transmitting 50% duty cycle "dots" at 1A cathode current the temperature rose to 110C at which point I stopped the test!

Here is a chart showing this rate of rise.

Q1 is a TIP147, and the specification for this device suggests the transistor was being allowed to rise dangerously close to the maximum temperature of 125C.

Further measurements made by placing a Perspex sheet on the base of the amplifier with 10 thermocouple probes fitted, showed that the temperature in the lower rear left hand corner of the amplifier was not being adequately cooled. The air flow from the blower circulates through the chassis through two small holes near to the valve, the cooling of the control board and in particular Q1 is left to convection. As the wiring harness, transformers and relays create pockets of air that cannot move, the whole area just gets hotter and hotter. During test, even with the amplifier in standby the ambient temperature in this area was 40C when the workshop air temperature was between 13C and 16C.

Time for some radical re work- it was decided to fit Q1 on a sensible sized (50mmx50mmx60mm) finned heatsink and clamp the transistor to the sink with a (50mmx15mmx10mm) bar. This heatsink was to be fitted onto the chassis next to the valve. New cooling holes were drilled in the chassis to supplement the old ones, all arranged to force cold air through the new heatsink. These additional holes are visible in this photograph and they are very important. As well as blowing cold air right onto the fins of the heatsink, they also create enough turbulence to cool the corner of the amplifier that was previously getting too hot.

It was necessary to re locate the 240V step start resistor to make space for the heatsink. The step start resistor was placed where Q1 was originaly fitted.

This picture shows the layout described above. Its all self explanatory but the red hatching denotes the area that was getting too hot.

Back to testing, the probe was again coupled to the tab of Q1, simulated runs at 1A, 1.5A and eventually 2A Cathode current were used both with and without the cooling fan running. This produced the much more favourable results in the second chart note that even at 2A cathode current for a L O N G time the temperature of Q1 stays safe.

All these modifications were made with the strong approval of GM3SEK. Ian also says that if one of his boards has had the red VDRs removed and replaced with a small white gas tube, this mod was not approved by him and the VDRs should be put back immediately. There are two reasons for this, first that isn't where the problem was. The second reason, should the gas tube ever fire, it will short-circuit the bias supply, leaving the tube to overheat with zero bias until the mains is switched off.

It's an unfortunate feature of the GS35 that if the bias is removed, the anode current is still within the valid operating range, so the anode trip will not operate. The tube just sits there and stews, along with the glitch resistors and the transformer. If you don't push the tube hard on TX either, the trip may never operate at all.

After repairing the rest of the damage to the amplifier and returning the control board to "as designed" the amplifier was RF tested into a dummy load, it ran comfortably cool at 1000W out for 30minutes and then after running a 25wpm cw test message at 1500W for 30 minutes it was pronounced "fixed"

Since rework the amp has not failed despite being well used... Of course time will tell.