On Fri, 03 Jul 2009 15:11:10 -0700, John Larkin
wrote:

On Fri, 26 Jun 2009 09:46:01 +0100, Raveninghorde
raveninghorde@invalid wrote:

On Thu, 25 Jun 2009 17:22:21 -0700, Capt. Cave Man
wrote:

On Thu, 25 Jun 2009 07:11:19 -0400, Spehro Pefhany
wrote:

On Thu, 25 Jun 2009 11:40:39 +0100, Raveninghorde
raveninghorde@invalid wrote:


Here is my new light bulb, minus reflector.

168 112 lumen leds in 14 strings.

The main problem is the aluminium of the pcb is not earthed causing
problems with current control presumably due to much higher stray
capacitance than we had on FR4. Just got to find a neat way of
earthing it.

That is aluminum clad? There are holes through it? Are they isolated?


I think he is nuts. Looks like copper to me. What PCB houses use Al
clad media? What etching process works with AL cladding?

What solder sticks to Al cladding.

Current control should be easy and precise, even with ALL of them
floating without "earth" grounding.


The pcb is an aluminium plate with a very thin insulator and a copper
layer.

http://www.aismalibar.com/archivos/8...pdf?download=1

Current control was stable on FR4 prototype but the higher capacitance
is causing interaction between current controllers.

It's hard to imagine that the pcb adds much capacitance. You could
measure it.

Maybe the instability comes from a shared-ground problem.

John


Surface mount solder stripped and tinned segments of 16 Ga. SPC wire
along the trace runs between devices. Much better than mere clad media
alone.

I would have designed it that way. Bare 'pads' right on the traces
that allow one to add wire segments as those I described. It turns any 4
oz pwa into a proper, point-to-point wired device. That makes the pwa
just a backbone for the conductors, as it should be. It already is, but
it would also definitely be better with the added wires. Every milliohm
counts.