Building Coaxial transmission line on PCB?

[CBFalconer]

He said "coax"...

Which is just one more way of implementing a line. As is a piece
of wire in some medium.

Ridiculous cross-post level reduced by setting follow-ups.

--
<http://www.cs.auckland.ac.nz/~pgut001/pubs/vista_cost.txt>
<http://www.securityfocus.com/columnists/423>

"A man who is right every time is not likely to do very much."
-- Francis Crick, co-discover of DNA
"There is nothing more amazing than stupidity in action."
-- Thomas Matthews

 

[werty]

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.

----------------------------------------------------------

Boxed ! the wavelength is far greater than
your dimensions , thus higher modes can not
exist , thus you do NOT need sides .
When you reach 10 Ghz , then maybe
you need sides in ur boxed "coax" .

But the big joke , is in the real world ,
they use cheap PCB to xmit 2.5 Ghz .
No strip line , no microstrip , nada ..
It works well , so quit arguing reality .

BTW , i saw some novice , trying to
use juice cans to launch WiFi .
He figured the more cans , the more
gain . He had 3 cans , T'd .
to divide the power .
Gain is not in cans , its in size of
the dish .

Another book worm said all i needed
was $26 for 100 meters of blah blah
coax at 2.5 Ghz ..

10 times that price !
and 1.8" dia hard line !

At these wavelengths , its lower loss
to send it TEM and thru the air ,
not thru a coax .

This is goin to FPGA ? Do those relics
still exist ?! Oh well , i supose ya gotta
try to "protect" your firmware by reinventing
the CPU !

 

[Bob]

How DARE you not know what he had in mind!

Bob

 

[Tom Bruhns]

----------------------------------------------------------

Boxed ! the wavelength is far greater than
your dimensions , thus higher modes can not
exist , thus you do NOT need sides .
When you reach 10 Ghz , then maybe
you need sides in ur boxed "coax" .

But the big joke , is in the real world ,
they use cheap PCB to xmit 2.5 Ghz .
No strip line , no microstrip , nada ..
It works well , so quit arguing reality .

BTW , i saw some novice , trying to
use juice cans to launch WiFi .
He figured the more cans , the more
gain . He had 3 cans , T'd .
to divide the power .
Gain is not in cans , its in size of
the dish .

Another book worm said all i needed
was $26 for 100 meters of blah blah
coax at 2.5 Ghz ..

10 times that price !
and 1.8" dia hard line !

At these wavelengths , its lower loss
to send it TEM and thru the air ,
not thru a coax .

This is goin to FPGA ? Do those relics
still exist ?! Oh well , i supose ya gotta
try to "protect" your firmware by reinventing
the CPU !

Ah, yes, reality. The reality is that we do need the vias for our
work. I can imagine that John J. may well also. When you're aiming
at 100+dB isolation among traces, you do have to be pretty careful,
even at "low" frequencies.

Cheers,
Tom

 

[Jim Granville]

----------------------------------------------------------

Boxed ! the wavelength is far greater than
your dimensions , thus higher modes can not
exist , thus you do NOT need sides .
When you reach 10 Ghz , then maybe
you need sides in ur boxed "coax" .

<snip>

Seems video distribution is not one of your areas of expertise ?

Note that John was talking about crosstalk, in the analog domain,
and yes, what he did certainly will have a measurable improvement, and
is somewhat 'industry practise' when minimising crosstalk.
Note he says 'cured', that means he is on both sides of the problem,
and it is a brave (or something else?) person that counters such direct
experience

I imagine in extreme digital cases, such as where you are worried not
only about sending the clock, but also about the ps/fs of jitter, then
this design approach would also help : it's not hard to do.

-jg

 

[John Fields]

He said "coax"...

 

[Uwe Hercksen]

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?

Hello,

the technology used to build multilayer PCB does not allow to build a
round coaxial transmission line.
You may have structured planes of copper foil separated by insulation
material, but nothing like the shield of a coaxial cable. Even a
rectangular shield around a center conductor is not possible with the
existing methods to build multilayer PCB.

bye

 

[Uwe Hercksen]

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.

Hello,

but how about a real closed square shield around the center conductor?

Bye

 

[John Larkin]

This is goin to FPGA ? Do those relics
still exist ?! Oh well , i supose ya gotta
try to "protect" your firmware by reinventing
the CPU !

When you need, say, 40x the crunch power of a decent DSP processor,
those relics come in handy.

John

 

[John Larkin]

<snip>

Seems video distribution is not one of your areas of expertise ?

Note that John was talking about crosstalk, in the analog domain,
and yes, what he did certainly will have a measurable improvement, and
is somewhat 'industry practise' when minimising crosstalk.
Note he says 'cured', that means he is on both sides of the problem,
and it is a brave (or something else?) person that counters such direct
experience

I imagine in extreme digital cases, such as where you are worried not
only about sending the clock, but also about the ps/fs of jitter, then
this design approach would also help : it's not hard to do.

-jg

I do stuff down to a few ps RMS jitter on a regular 6 or 8-layer
board, microstrip traces, with switching supplies and uPs and display
drivers on the same board. Picoseconds aren't tough these days.

John

 

[Geronimo Stempovski]

Thanks for your help, so far. Really appreciate it. I'm already a step
further in my considerations.

Now I'm looking for a diagram like frequency (some MHz to 10 GHz for
example) versus loss tangent and / or epsilon R for FR4. I only found a poor
black-and-white copy from 1991 in a paper which I searched with Google. I
wouldn't have thought it to be so hard to find a graph... Does anybody know
where I can find that?

 

[john jardine]

Hello,

but how about a real closed square shield around the center conductor?

Bye

Would have been ideal. At the time was thinking about a method to do this
and sorted a setup that may have been worth talking to the PCB people about
but a customer was paying to clear down his urgent problem and not to start
up a research project .
No ... I don't remember what I figured out. Ideas are easy, it's the
implementation that's a problem .
No doubt it'll surface again if I'm I'm under pressure.
john

 

[Jim Granville]

Hello,

but how about a real closed square shield around the center conductor?

To do that would need plated slots, and slots have the PCB fabs
luke-warm at best : hard to do cleanly, they also weaken the PCB if
long, and also have minimum router sizes, plus machine time......

A better direction would be thinner laminates, and using the space you
would have lost to the slot anyway, as wider GND webs on the same plane,
coupled with stitching vias (which can be smaller dia than slots)

-jg

 

[john jardine]

[...]

----------------------------------------------------------

Boxed ! the wavelength is far greater than
your dimensions , thus higher modes can not
exist , thus you do NOT need sides .
When you reach 10 Ghz , then maybe
you need sides in ur boxed "coax" .

But the big joke , is in the real world ,
they use cheap PCB to xmit 2.5 Ghz .
No strip line , no microstrip , nada ..
It works well , so quit arguing reality .

BTW , i saw some novice , trying to
use juice cans to launch WiFi .
He figured the more cans , the more
gain . He had 3 cans , T'd .
to divide the power .
Gain is not in cans , its in size of
the dish .

Another book worm said all i needed
was $26 for 100 meters of blah blah
coax at 2.5 Ghz ..

10 times that price !
and 1.8" dia hard line !

At these wavelengths , its lower loss
to send it TEM and thru the air ,
not thru a coax .

This is goin to FPGA ? Do those relics
still exist ?! Oh well , i supose ya gotta
try to "protect" your firmware by reinventing
the CPU !

Que?.
Don't listen to the worms.
Video comes in forms other than digital.

 

[Uwe Hercksen]

Now I'm looking for a diagram like frequency (some MHz to 10 GHz for
example) versus loss tangent and / or epsilon R for FR4. I only found a
poor
black-and-white copy from 1991 in a paper which I searched with Google. I
wouldn't have thought it to be so hard to find a graph... Does anybody
know
where I can find that?

Hello,

I am sorry to tell that, but for frequencies of 1 to 10 GHz, FR4 ist not
the right material, there are other PCB materials which are better for
high frequencies, take a look here
http://www.andus.de/Leiterplatten/Impedanz/hfmat.htm
They write there that FR4 may be used up to 4 GHz.

Bye

 

[John Larkin]

Hello,

I am sorry to tell that, but for frequencies of 1 to 10 GHz, FR4 ist not
the right material, there are other PCB materials which are better for
high frequencies, take a look here
http://www.andus.de/Leiterplatten/Impedanz/hfmat.htm
They write there that FR4 may be used up to 4 GHz.

Bye

FR4 can be used at 20 GHz, depending on what you're trying to do.

John

 

[Del Cecchi]

FR4 can be used at 20 GHz, depending on what you're trying to do.

John

Also you need to narrow down which variety of FR4 you are interested in.

 

[John Larkin]

Also you need to narrow down which variety of FR4 you are interested in.

FR4 is a commodity that's not well controlled. There really aren't
controlled varieties, and most pcb houses don't guarantee they'll
always use the same stuff. If you want loss or Er consistancy, you
usually have to call out something more specific. Or design so that it
doesn't matter.

John

 

[Geronimo Stempovski]

Again, I'm looking for a diagram like frequency (some MHz to 10 GHz for
example) versus loss tangent and / or epsilon R for FR4 or other usual PCB
material. I only found a poor black-and-white copy from 1991 in a paper
which I searched with Google. I wouldn't have thought it to be so hard to
find a graph but as noone replied to my previous question so far it does
seem to be hard!

Does anybody know where I can find that?

Regards Gero

 

[John Larkin]

Again, I'm looking for a diagram like frequency (some MHz to 10 GHz for
example) versus loss tangent and / or epsilon R for FR4 or other usual PCB
material. I only found a poor black-and-white copy from 1991 in a paper
which I searched with Google. I wouldn't have thought it to be so hard to
find a graph but as noone replied to my previous question so far it does
seem to be hard!

Does anybody know where I can find that?

Regards Gero

Google gives lots of hits on stuff like "loss tangent fr4 frequency"

But FR-4 varies a lot, so there's no definitive data.

What are you trying to do?

John