From: Moran, Brian P (email@example.com)
Date: Mon May 01 2000 - 15:14:28 PDT
A little more detail. The normal board stack was a 16-18 layer 0.093. This
would normally yield 4 mil dielectric spacings and 4 mil trace width in a
dual stripline configuration (4/4/4 x 4), for about 50 ohms. We expanded to
6/9/6 x 6 mils, on two sets of routing layers,with the Speedboard, and were
able to retain the 50 ohm spec. This put pressure on board thickness but
improved high frequency performance. Forgive me if my numbers are off a
liitle, its been a year or so. Some vendors will also build with Cynate
Ester pre-preg which gives even better performance with acceptable cost.
From: Degerstrom, Michael J. [mailto:firstname.lastname@example.org]
Sent: Monday, May 01, 2000 2:43 PM
Subject: RE: [SI-LIST] : Losses in RF boards
Ray and Ron,
Yes, of course attenuation is on a per unit length basis. It
seems (especially when one refers to RF applications) that when
you are concerned with one only one high speed signal you have the freedom
to keep the lines wide to reduce attenuation and can therefore go
a "reasonable" distance without excessive losses. So for a .032"
thick laminate you can have a .05" wide microstrip (or whatever
it takes to get your desired Zo).
Contrast the scenario above to the one which I was faced with
where I had to accommodate ten 2 Gb/sec differential signals with
lines on a 12 mil pitch. I don't remember the details but the
laminate was probably 4 mil thick FR-4 with 4.5 mil wide microstrip
lines with a soldermask overcoat. So for this case I'm not
convinced that the general statement holds that "FR-4 dielectric
losses dominate over skin effect losses above 1 GHz".
By the way, I'm presently designing a 10 GHz (OC-192) board
that has one (differential) 10 GHz path. I'll keep this path
quite short (~1" long) with reasonably wide lines. The other
high speed paths are sixteen 622 MHz paths which are perhaps 5" long.
I'd like to stay with FR-4. Can anyone share their experience
where a better dielectric was required to meet your electrical
performance? If so, please include data rate, line pitch and
Mike Degerstrom Email: email@example.com
200 1st Street SW
Gugg. Bldg. RM 1042A Phone: (507) 284-3292
Rochester, MN 55905 FAX: (507) 284-9171
> -----Original Message-----
> From: WAUGH,RAY (A-SanJose,ex1) [mailto:firstname.lastname@example.org]
> Sent: Friday, April 28, 2000 5:20 PM
> To: 'email@example.com'
> Subject: [SI-LIST] : Losses in RF boards
> In open microstrip lines on common plastic substrates,
> dielectric losses
> dominate over conductor losses. In that event, the total loss (mostly
> dielectric loss) will measure as a straight line when plotted against
> frequency. Where conductor losses are important (as is the case with
> thin-film hybrid circuits on alumina substrates), total loss
> does not plot
> as a straight line against frequency. I make such wideband
> from time to time, just as a sanity check (and quality check
> on the material
> I buy).
> It can be shown from Schneider's work that dielectric losses can be
> expressed as a constant loss per wavelength (dB/wavelength).
> See attached.
> If one line is twice as long as another, it will have twice
> the losses at
> all frequencies.
> My own measurements of loss tangent on 0.032" thick FR4
> microstrip (half
> ounce copper lines and ground plane) is 0.030. For thinner
> material of
> uncertain heritage, 0.060 would not be out of line.
> Raymond W. Waugh - WSD Diode Applications
> E-mail: firstname.lastname@example.org
> USPS: Agilent Technologies
> Wireless Semiconductor Division
> 39201 Cherry Street, MS NK20
> Newark, California 94560
> -----Original Message-----
> From: Degerstrom, Michael J. [mailto:email@example.com]
> Sent: Friday, April 28, 2000 1:52 PM
> To: firstname.lastname@example.org
> Subject: RE: [SI-LIST] : RF boards
> Isn't your statement that PCB dielectric losses dominate above
> 1 Ghz only relavent for your particular line length and line
> width? For
> example, the few published comparisons of eye closure for
> various laminates
> tend to base their measurement on .012" wide 12" long lines. My
> experience is with .004" wide 2" long lines in FR-4. Simulations
> with a .02 loss tangent and sqrt(freq) skin-effect losses predicted
> 12 Gb/s. I measured 5 Gb/s. I believe the extra loss is attributed
> to surface roughness as I don't think a loss tangent of .06 at 5 Gb/s
> is realistic.
> By the way, with the lower loss laminates, how do
> you know that the lower loss is mostly attributed to the lower
> loss tangent? Couldn't the surface metal be smoother and/or of better
> quality than the metal typically used on FR-4 laminates?
> Mike Degerstrom Email: email@example.com
> Mayo Clinic
> 200 1st Street SW
> Gugg. Bldg. RM 1042A Phone: (507) 284-3292
> Rochester, MN 55905 FAX: (507) 284-9171
> WWW: http://www.mayo.edu/sppdg/sppdg_home_page.html
**** To unsubscribe from si-list or si-list-digest: send e-mail to
firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
This archive was generated by hypermail 2b29 : Wed Nov 22 2000 - 10:50:10 PST