[TowerTalk] Circuit Board Question

[Jim Lux]

On 8/1/14, 9:22 AM, Richard (Rick) Karlquist wrote:
>
> FR-4 dielectric is quite lossy, comparable to the copper
> loss.

At HF?
I worry about it "a little" at 3GHz, but not much, but then, I'm doing 
fairly casual low performance designs.
I found one reference that gives 0.008 loss tangent at 100-3000 MHz.
An article has a graph that shows something like 0.25 dB/inch at 3GHz, 
but roughly linear with frequency, so at 30 MHz, it would work out 
0.0025dB/inch  (or is that 250 microBel?)
http://www.parkelectro.com/parkelectro/images/leysfinal.pdf

I've also seen loss tangent numbers like 0.05, which is pretty bad, but 
I don't know that loss at HF would be all that big a deal. I'd bet that 
ohmic loss would dominate. (just like in coax, at HF the ohmic loss 
dominates over dielectric loss).

Also, there's a significant variation in FR-4 dielectric properties from 
mfr to mfr and lot to lot, as well as substantial variations with 
humidity.  Then again, folks do build things like interdigital stripline 
filters and microstrip patches on FR-4


  Making the trace wider neither increases nor decreases
> dielectric loss, but it does spread the heat over a larger
> area.

And change the Z, if you care about that.

  What many designers do is print the trace on the
> bottom of the PC board in an area with no ground plane
> and then use the chassis underneath the PC board as the
> ground plane.  This eliminates most dielectric loss
> (there is still a little fringing within the FR-4).
> Dielectric loss is proportional to frequency while
> copper loss is proportion to the square root of frequency.
> Having air dielectric also approximately halves the copper
> loss.
>