I've noticed that various data sheet application circuits for items such as RF
switches, amplifiers, etc. (e.g.,
http://www.hittite.com/product_info/.../hmc349lp4c.pd f )
use what I would call "microstrip traces surrounded by a ground pour 'guard'
to reduce coupling to adjacent traces." Someone else, however, has suggested
that the application circuits were really designed as co-planar waveguides
(with grounds). Anyone else want to venture an opinion?

For a 62.5 mil board, a 50 ohm microstrip's width is around 115 mils using
FR-4 (k=4.7). Using TxLine 2003, for a CPW w/ground I have to reduce the gap
to ~50 mils before the signal trace width reduces ~10% to ~104 mils. I take
this to imply that there's not a lot of coupling between the signal trace and
the copper pours (instead of the ground plane) until the gap width is
comparable to the board thickness.

In general, my impression has been that the use of copper flooding is more to
provide isolation between adjacent traces than to change the form of the
transmission line, and the use of the picket fenced vias was to insure that a
large chunk of copper pour didn't suddently turn into a resonator. Does that
sound correct?

---Joel Kolstad

Joel Kolstad wrote:
I've noticed that various data sheet application circuits for items such as RF
switches, amplifiers, etc. (e.g.,
http://www.hittite.com/product_info/.../hmc349lp4c.pd f )
use what I would call "microstrip traces surrounded by a ground pour 'guard'
to reduce coupling to adjacent traces." Someone else, however, has suggested
that the application circuits were really designed as co-planar waveguides
(with grounds). Anyone else want to venture an opinion?

Their Application Note 17 ("Design techniques enhance isolation in
switch assemblies") talks explicitly about grounded-coplanar-waveguide
techniques. The evaluation boards for the "349" switches certainly
appear to use the CPW techniques discussed in AppNote 17, including
closely-spaced plated thru-holes to tie top and bottom ground planes
together.

For a 62.5 mil board, a 50 ohm microstrip's width is around 115 mils using
FR-4 (k=4.7). Using TxLine 2003, for a CPW w/ground I have to reduce the gap
to ~50 mils before the signal trace width reduces ~10% to ~104 mils. I take
this to imply that there's not a lot of coupling between the signal trace and
the copper pours (instead of the ground plane) until the gap width is
comparable to the board thickness.

In general, my impression has been that the use of copper flooding is more to
provide isolation between adjacent traces than to change the form of the
transmission line, and the use of the picket fenced vias was to insure that a
large chunk of copper pour didn't suddently turn into a resonator. Does that
sound correct?

If by "picket fenced vias" you mean the projections outward from the
signal vias on the evaluation boards, I take those to be the plated
thru-holesdiscussed in AppNote 17.

Are you seeing something I'm not?

_Interesting_ gadgets. Awfully small, but interesting; may have to
ask for samples and eval boards.

--
Mike Andrews, W5EGO

Tired old sysadmin

Hi Mike,

Thanks for the response -- I'll go and read that app note.

If by "picket fenced vias" you mean the projections outward from the
signal vias on the evaluation boards

I mean all the vias going from the top side copper to the ground plane.
Viewed from the side -- and with a little imagination -- those vias kinda look
like a picket fense. OK, maybe not so much, but I *have* heard this term
before; I'm not making it up myself! :-)

the most important reason for the "pouring" is to reduce the inductance
of the upper side gorund plane.
most of the relevant field in the microstrip is confined in the volume
between the microstrip itself and the gorun plane beneath. there is a
phemomenon called fring effect, whereby fields generated on the upper
side of the microstrip come into play, but that occurs only at
frequencies of about 5GHZ and above, not at the usual frequencies we
talk about. it belong more to the microwaves' people realm.
leave it alone under 1GHz. you don't have to worry about that effect at
lower frequencies.
Saandy 4Z5KS

"Saandy , 4Z5KS" wrote in message
oups.com...
,,,,,,this is PRECISELY what a CPW is NOT!

It's a "coplanar waveguide with ground plane." This seems more popular than
strict CPWs, as far as I can tell.

This appears to be a blend of co-planar and microstrip. Co-planar has no
ground plane, that is, everything is in one plane thus the name "co-planar".
A "microstrip-like" line is flanked by two ground surfaces and there is no
ground plane under it.
Like this
---- - ----

Microstrip is a flat line over a ground plane. Like this
-
-----------

The vias along both sides of this "CPW" T-line appear to be used to
constrain the fields and

The Vias most certainly must enter into the characteristic impedance
equation since the spacing to the line is so close. Therefore the line must
be narrower than pure coplanar.

I was part of a program to design a family of his same type of switch on
GaAs, back in the early 90's and the co-planar line was used.

73, Steve, k9DCi




"Mike Andrews" wrote in message
...
Joel Kolstad wrote:
I've noticed that various data sheet application circuits for items such
as RF
switches, amplifiers, etc. (e.g.,
http://www.hittite.com/product_info/.../hmc349lp4c.pd
f )
use what I would call "microstrip traces surrounded by a ground pour
'guard'
to reduce coupling to adjacent traces." Someone else, however, has
suggested
that the application circuits were really designed as co-planar
waveguides
(with grounds). Anyone else want to venture an opinion?

Their Application Note 17 ("Design techniques enhance isolation in
switch assemblies") talks explicitly about grounded-coplanar-waveguide
techniques. The evaluation boards for the "349" switches certainly
appear to use the CPW techniques discussed in AppNote 17, including
closely-spaced plated thru-holes to tie top and bottom ground planes
together.

For a 62.5 mil board, a 50 ohm microstrip's width is around 115 mils
using
FR-4 (k=4.7). Using TxLine 2003, for a CPW w/ground I have to reduce
the gap
to ~50 mils before the signal trace width reduces ~10% to ~104 mils. I
take
this to imply that there's not a lot of coupling between the signal
trace and
the copper pours (instead of the ground plane) until the gap width is
comparable to the board thickness.

In general, my impression has been that the use of copper flooding is
more to
provide isolation between adjacent traces than to change the form of the
transmission line, and the use of the picket fenced vias was to insure
that a
large chunk of copper pour didn't suddently turn into a resonator. Does
that
sound correct?

If by "picket fenced vias" you mean the projections outward from the
signal vias on the evaluation boards, I take those to be the plated
thru-holesdiscussed in AppNote 17.

Are you seeing something I'm not?

_Interesting_ gadgets. Awfully small, but interesting; may have to
ask for samples and eval boards.

--
Mike Andrews, W5EGO

Tired old sysadmin

Hi Steve,

"Steve Nosko" wrote in message
...
This appears to be a blend of co-planar and microstrip.

The freebie program TxLine 2003 "knows" about (can compute dimensions from
impedances, etc.) CPWs, CPWs w/ground planes, microstrips, etc. and you can
get some feel for how wide you can make a CPW w/GP gap before it's effectively
a microstrip -- when the signal trace widths became nearly the same. I dug up
some more Hittite app notes, and it does seem as though they're clearly
playing in the CPW w/GP arena. They do point out that a major downside is
that you eat up a fair number of routing channels with all those vias.

I was part of a program to design a family of his same type of switch on
GaAs, back in the early 90's and the co-planar line was used.

I'm surprised just how much "chaos" there seems to be in the RF IC (primarily
MMIC) arena... there are plenty of old standbys like Hittite, MA/COM (now
Tyco), and Watkins-Johnson, but they seem to have a pretty well stocked
"stable" of components and only introduce a handful of new ones every year.
On the other hand, the little guys and mergers like Freecell, California
Eastern, and RF Micro Devices seems to have lots of good parts that suddenly
get discontinued, product introductions that turn out to be vaporware (a
friend claims that RFMD is great for this -- their data sheets are really part
of their marketing department, with specs drawn strictly from simulation --
they initially quote everyone a lead time of, say, 3 months, and if they don't
get orders for large quantities, they just never fab the chip in the first
place), and otherwise seem a little difficult to rely on at times!

---Joel