Building Coaxial transmission line on PCB?

I think transmitting high-speed signals is very easy when you have a
360-degree ground reference, round conductors,
and no other nearby signals like in coaxial cables. My aim is to design PCB
tracks as much like a coaxial cable as
possible. Anyone tried this before? Is it possible with regular FR4, anyway?
Thanks for your help.

Gero

 

Microstrip has absolutely nothing to do with the coaxial structure I had in
mind.

 

But a microstrip will achieve the same results without the technical
difficulty of a real coaxial line. It is even easier to use
balanced/differential connections.

Meindert

 

Nope- in all the decades of high speed PC circuit design, you are the
first to think of it!

Not even close, the "phase velocity," or speed to you, will be less than
that of free space by a factor of 1/sqrt(epsilon-sub-r), so go figure.

 

http://en.wikipedia.org/wiki/Stripline

You won't get any more TEM-like on a PCB.

regards
Henning

F'up2 d.s.e. (Da darfst Du dann auch wieder deutsch sprechen.)

 

You can build a wave guide out of a multi-layer board with lots of
micro vias. The performance gain was non-existent vs traditional micro/
strip/line, and cost a lot of money. These days you can satisfy
yourself of this reality with a 3D field solver. Back when we tried
this, workstations were as slow as building it "for real", and more
expensive.
Just break out to a connector, use a coax assembly, and connect back
in. Easy peasy.

 

There is such a thing as microcoax, so he can chisel out a little groove
in a thick pc and stuff that in there.

 

Semi-rigid is the term, and if it is in coax, it doesn't need to be
IN the PCB from point to point.

I wish you guys would stop cross posting all over the place. Many
ISPs have group inclusion limits.

It is also not considered proper Usenet practice.

 

Sure. Microstrip, stripline, coplanar waveguide, or even the very
strange slotline.

John

 

If you're such an expert, why are you asking here?

In my experience working with stuff in the low end of the
microwave region (~ 1GHz), microstrip is pretty much what
everybody uses as a PC board alternative to coax.

 

(Trimming the number of groups to one...no need to post it so
broadly.)

Have a look at some PCBs that route critical analog signals around.
You'll likely find stripline, which is indeed TEM, that's guarded by
closely-spaced vias on either side. The stripline can be single-ended
or balanced. This is done much more for signal integrity (extremely
low cross-coupling to other signals) than for speed.

Transmitting high-speed signals doesn't necessarily need such fancy
lines, but it does need low loss, and that is a problem with FR4 as
the frequency goes up. Thus, people use Teflon, Duroid and the like.
Have a look at PCB designs that deal with 10GHz and above.

Cheers,
Tom

 

And then there's the crude method - put in a coax jumper wire. Worth a
thought, depending on the application.

 

I haven't tried it myself, but it's not entirely uncommon in microwave designs
to do something like this -- it's essentially a stripline design with a
"picket fence" worth of vias to serve as the sidewalls. I suspect the reason
it isn't particular popular is that the performance isn't that much better
than a stripline, the models for it aren't found in ADS/Microwave
Office/Ansoft Designer, and the manufacturing costs may be higher. ("Maybe"
because I've seen a lot of people who've started transitioning from microstrip
to a co-planar waveguide, which requires a bazillion drill hits as well. You
trade off the number of drill hits for isolation...)

---Joel

 

Gero,

Why?

Seems others have already progressed to the "abuse" phase, however I am
curious.

If microstrip, or stripline constrains the electric fields such that for
all practical purposes the matched line does not radiate, why is a
coaxial line any better? The only benefit of a coaxial line is that
unmatched, it can not radiate (all the RF energy has to come out of the
ends).

Is this a concern for radiated emissions?

Again, unmatched microstrip or stripline structures can be engineered
with adjacent shielding such that EMI/RFI should not be an issue.

Trying to create a coaxial guide by placing many vias and metal lines is
just too ugly to even think about without a valid reason. Once the
reason is known, the first or second approximation to the structure is
probably completely adequate.

Austin

 

Had trouble with crosstalk on a mass of video signals. Cured with a
multilayer board where each signal was 'boxed in' by ground plane to the
sides, above and below. Sort of square coax.

 

I would think that would be difficult and/or costly.
Working from bottom to top:
Start with (say) 6 mil laminate 0.5/0.5 at bottom; bottom layer has a
narrow stripe to emulate the lowest part of a coax, and top part is a
wider stripe: both ground (shield).
Next layer (#2) is (say) 6 mil laminate 0/0.5 (ie bottom has no
copper and top is 0.5 ounce); stripe is wider and is ground.
"Middle" or next layer is (say) 6 mil laminate 0/0.5 (ie bottom has
no copper and top is 0.5 ounce); 3 stripes: ground / center conductor /
ground.
Next layer is (say) 6 mil laminate 0/0.5 (ie bottom has no copper and
top is 0.5 ounce); stripe is as wide as layer #2.
Then use (say) 6 mil laminate 0/0.5 at top; where the top copper has
a "wide" stripe same as first laminate top stripe as ground.
Finish with (say) 6 mil laminate 0/0.5 at top; where the top copper
has a "narrow" stripe same as first laminate bottom stripe as ground.

Use more layers if they are thinner.
Use vias liberally for tying the ground stripes together.
Note the 6 mils is a wild guess.

 

He said "coax"...

 

Gee, coax cables, even those that use spiral teflon seperators, are
like that...