>
> Can anyone explain the behaviour?
>

It's hard to know what to do from a distance but here's what I think I'd look
at if I was faced with this problem.

I think failing under moderate load could be explained by one or more of the
following possibilities and probably others I haven't thought of:

- The power supply is not capable of producing sufficient current.  Check how
  the voltage across the main input smoothing capacitors which have been
  replaced varies while the load is applied.  If it dips severely, check input
  components such as filters, surge limiting devices, connectors and so on for
  breaking down under load.  If you can measure the ripple here while changing
  the load, an increase may indicate that one side of a fullwave rectifier is
  going high impedance or open circuit under load.  Also check for damage that
  might have occurred in the struggle to remove the capacitors.

- Overcurrent sensing is kicking in too soon.  Look for low value, moderate to
  high power resistors in the output current paths and check their values and
  how the voltage across them varies with applied load.  If they seem good,
  check associated small components.

- Regulation is not working correctly.  Try to figure out how the regulation
  is supposed to work and take measurements to see how it is behaving in
  reality.  Easy to say but may be difficult to do in practice.  If the PSU
  uses a chip to provide regulation and drive to a chopper device, the
  data sheet for the chip may provide some guidance on how it is supposed
  to work.  Be careful taking measurements as accidentally shorting something
  out could lead to big bangs.

- The PSU may be looking for feedback from other parts of the machine in the
  form of remote voltage sensing or remote current sensing or inputs which
  cause particular supply lines to be switched on or off or come up in a
  particular order.  If this is the case, the fault may be elsewhere in the
  machine or may be as result of operating the power supply without it being
  connected to the rest of the machine.

Hope this helps.

Regards,
Peter Coghlan.

Reply via email to