[Prev][Next][Index][Thread]

Re: Coil form




Malcolm and all:

Malcolm commented

TL>I have built a couple of PCB plate caps using double-sided PCB with
TL>borders removed. They worked reasonably well, but it is difficult to 
TL>quantify as I was just getting going on a new system (rotten tune and
TL>all that). The major failure mode was arcing around the ends. I was
TL>running this cap in engine oil (never again - it carbonizes much too 
TL>easily). I've often wondered just what the merits of fibreglass are.
TL>Plenty of low powered RF circuitry is constructed on these. Even
TL>stuff in the 100's of MHz range. In fact stripline filters and 
TL>transformers are etched in as part of the circuits. The losses can't
TL>be too bad (at least at low power). I might have another crack at
TL>some stage but this time in transformer oil when I have a better
TL>controlled system running.

TL>Malcolm

TL>** If it's under 10 Amps it's leakage current **

I say that using fibreglass as a dielectric is reasonably sound in terms 
of the loss factor of the material: you are looking at tan delta of 
0.005 to 0.01 for FR4 fibreglass. For 1/16" PCB material the breakdown 
voltage would be high and the values of Er would be reasonably well 
defined (typically 4.5) but this assumes there are no voids in the 
material. FR4 fibreglass PCB material varies a lot between manufacturers 
and between batches.
Manufacturers such as Rogers, who make PTFE based PCBs, impose careful 
quality control on those products intended for high frequency (GHz) 
boards and the likelihood of voids is reduced as is the Er (typically 
2.2) but it costs a lot more.
You can often find big sheets of double sided FR4 that have been water 
damaged such that the Cu is discoloured. You could make up a load of 
boards whereby the top plate is brought through to an isolated area of 
the bottom plate so that the boards could be stacked. If you used some 
silicon oil and thin PE sheeting, I would suggest that the edge failures 
that Malcolm commented on could be minimised. It might be worth trying a 
bit of resistive field grading using blotting paper soaked in copper 
sulphate.
The main problem is that the capacitance per unit area for 1/16 FR4 PCB 
is small: approx. 25nF per square metre.
A quick and easy capacitor is a big reel of URM67 or RG213 coax or 
similar. Then you have a capacitance of 30pF per foot near enough so a 
500m drum would result in around 0.05uF. The problem is that you are 
sourcing out of 50R Zo and hence have an associated inductance of 75nH 
per metre which means for a 500m reel, nearly 40uH which is a lot of 
unwanted "Leakage L". You could do with having access to both ends of 
the reel so that the braid can be cut back and the ends connected 
together. If you have a big reel knocking about, but you don't have any 
capacitors to try , this would do to get you off the ground (NPI if 
there is one).


Richard Craven
---
 CMPQwk #1.42 UNREGISTERED EVALUATION COPY