[Moon] ON0EME beacon

charlie at sucklingfamily.free-online.co.uk charlie at sucklingfamily.free-online.co.uk
Sat Aug 24 19:42:02 CEST 2019


Hi Eddy

It looks like your recent failure is different to what happened to the
DL0SHF 50W GaN PA.  In that case the pcb failed midway between the device
and the output connector.  The device itself was fully functional.  The
more recent failure again seems to be a pcb failure.  I do not know yet if
the device is still functional.  There is no obvious reason why the pcb
substrate failed.  In operation the calculated temperature rise of the
microstrip carrying the output power is only 20-30C above chassis temp
which should be totally benign according to Rogers.  There is a Rogers
application note that discusses thermal cycling as a cause of cracking due
to work hardening of copper.  The stresses result from differential
thermal expansion of copper to the substrate.  Over time the constant
stretching and compression forces can lead to failure. I could not prove
this with the DL0SHF PA as all that was left was a charred mess on the
pcb.  The DL0SHF PA had accumulated just under 500 operational hours, with
some 30,000 temperature cycles.  This kind of operation is atypical of how
PAs tend to be used in normal applications, so we are definitely in a
learning mode here!

I have had some discussions with an LDMOS PA expert who  tells a tale of
amplifier failures in a digital broadcasting application where the PA ran
during the day and was switched off at night (to save energy) and failures
of the soldered joint between drain and pcb occurred after only a few
months. The solution was to run the PA 24/7, after which no more failures
occurred.  The theory was that the solder joint was being mechanically
stressed due to mechanical expansion/compression.  The stress on the
solder joint is worse for devices (such as LDMOS or discrete GaAs or GaN
transistors) since the joint has to carry both RF and DC currents.  The
GaN device in the DL0SHF PA has separate DC and RF connections, so the
solder joints are less stressed, and do not seem to have failed.

Looking at the temperature excursions in more detail, both LDMOS and GaN
are operated Class AB so when RF goes on the devices will heat up probably
in a few seconds and then cool down when RF goes off. The pcb temperature
will rise further due to losses (and probably faster than the device due
to the higher thermal resistance).  Additionally there will be larger
temperature changes when the beacon goes on at the start of the moon  pass
and off again at the end of the pass.

Another possible theory for failure could relate to environmental
conditions. If for example a chilled water cooling system is used to cool
the PA, when the beacon goes off at the end of a pass the amplifier
chassis would end up below ambient temperature and if the dew point of the
surrounding air is above that temperature,  water could condense on the
amplifier, possibly causing corrosion or some other kind of damage when
operation is resumed. I have seen such water condensation in the lab when
using a chilled water cooling system so it certainly can happen.  In a
professional lab, chilled water coolers have a case around the amplifier
which is flushed with a dry gas to prevent such condensation.

I think what we are learning is that beacon operation with PAs that are
built 'normally' is a whole different ballgame compared to typical amateur
operation! With multiple potential factors at play, and root causes not
yet identified I think users will have to accept that from time to time
failures are unfortunately  inevitable.

73

Charlie







More information about the Moon mailing list