Grounded via "fences" between microstrips

traces

You seem to be comparing apples and pears - the Hittite guidline is talking about using a four layer stack to provide an environment where you can layout sensible constant-impedance traces - it doesn't say anything about using grounded track on the same layer as signal tracks to improve isolaton, while the Kyocera paper is very specifically about such grounded tracks, reinforced with grounded vias (a via fence).

Se if you can get hod of a copy of Howard W. Johson and Martin Graham's "High-Speed Digital Design" ISBN 0-13-395724-1. Section 5.6 - pages

Joel:

I agree with Bill....the two transmission line geometries are quite different, and neither behave exactly like microstrip.

One thing to keep in mind with stripline (strip conductor between two ground planes, as in the Kyocera paper) is that any assymetric structure, such as a via connecting to another layer, or connector protruding through one ground plane, excites an assymetric parallel-plate waveguide mode between the two ground planes. This propagates at low loss, without requiring the presence of the normal transmission line conductor. Vias between the two ground planes are normally needed to kill unwanted couplings due to this waveguide mode. In this case these vias are located near the assymetric object (connector or whatever) for greatest effect.

I imagine this probably happens in the CPW-with-ground plane structure as well (since the transmission line geometry is itself inherently assymetric with respect to the two ground planes), though I haven't experienced it personally to know how significant it is.

Hi Bill,

Sure it does:

"As a result of using CPWG it is necessary to have as many via holes connecting the top and bottom ground planes along the RF transmission lines. l/20 or less should separate these holes in order to achieve the modeled transmission line performance."

When they say "performance" I'm reading "constant impedance and isolation."

I realize that for CPWG -- especially with the dimensions Hittite is using -- there's a lot of coupling to the ground *plane* as well as the co-planar ground on the same layer as the signal trace, but I think there's still an interesting question as to how the via spacing in Hittite's case affects isolation -- and from the other paper, it doesn't seem as though "closer spacing" would always result in better isolation.

I do have a copy somewhere around here; my recollection of his advice was something like, "if you can afford the extra space needed to stick in a guard track, you probably would have enough isolation even without it." But... and this is a very big but... he's worried about signal integrity, whereas I'm worried about high-isolation RF switching -- 50dB isolation between two digital traces is great, whereas it's rather mediocre for RF.

Thanks for your input,

---Joel

Hi Steve,

I've read that adding extra vias where CPWGs have such transitions is a good idea, and the Hittite paper does allude to this:

"Critical factors for this type of CPWG construction are the via hole placement (mentioned earlier) and plating requirements.... Additional via holes around the ground plane, bias and control lines eliminates any hot spots due to leakage."

---Joel

As noted, the situations are different, if only because the Hittite CPW board is real and the stripline thing is simulation.

I certainly wouldn't want my cpw topside ground pours to float. They'd make excellent resonant couplers between the signal and bias traces. Lots of vias looks good to me. Personally, I'd pepper the interiors of the pours with vias, not just the edges, and go with bigger vias wherever possible.

But the Hittite layout looks fine. We did a similar thing recently with two cascaded Hittite '465 distributed amplifiers, 30 dB of net gain from DC to 20 GHz, driving a modulator for a rather large laser, and it worked. Well, we did slip a ferrite bead over one of the heatsink mounting bolts...

Wanna see a pic?

John

Hi John,

Yeah, apparently I was trying to read a little bit too much into the Kyocera results.

Agreed! I think it's a pretty safe statement that no copper pour is better than a floating pour between traces.

I just know the layout guys are going to scream when I tell them I want a via every 100 mils. :-) They'll get over it, though.

Certainly!

---Joel

OK, see a.b.s.e.

John