On Mon, 03 Dec 2007 14:07:52 GMT, Patrick Turner
wrote:



Don Pearce wrote:

On Mon, 03 Dec 2007 12:41:34 +0000, Nick Gorham
wrote:

Assuming that the decoupling is done with zero ESR caps, and there is no
inductance in either the cap or resistor, both of which will be false.

Given the assumed wonderful decoupling, maybe its worth thinking how the
switching spikes got to the valve?

Lack of star point grounding would be my number one candidate.

Nope. Star ground can mean many inconveniently long leads all running to
the same point.
Its OK in theory, but the OP killed the spikes with a 0.1uF across the
HT sec as I suggested.
The path of noise wasn't in the 0V rail or star point.

I often don't have a star point. I often have a 0V buss using 1.3mm
thick Cu solid wire
the length of the circuit layout, maybe 150mm long. Works fine.

Patrick Turner.

If you aren't too fussed about ultimate noise performance, then no
problem. When I was with Marconi Instruments, we made a doppler radar
speed meter for a train and the first iteration of the board failed to
read below about thirty mph, when theory said it should have been good
down to two. We had missed star-pointing every track around the input
by about two mm. That was all it took to allow massive 1/f ingress.

d

--
Pearce Consulting
http://www.pearce.uk.com