Jack Olson <[log in to unmask]> wrote: 

In a previous discussion, I asked about designers calling out
specific material thickness/stackup versus just calling out the
desired finished board thickness. The general consensus was that
unless there is a good reason for specific materials and thicknesses
(for impedance control) it would be best to let the vendor decide
the best stackup for a particular design.

In another discussion about impedance control, the question was
whether the designer should take it upon himself to do all of the
calculations, or to just try for appropriate trace widths and call
out the desired impedance for those particular traces. The general
consensus was that if the designer did all of the work and told the
vendor exactly what to do and it came out wrong, the designer had
only himself to blame. If he calls out only the desired impedance
and lets the vendor "tune" his process for it, the vendor is to
blame if it comes out wrong.

Well, I'm not trying to illustrate how we as designers should do as
little as possible and leave responsibility with the vendor... and
there sure has been a lot of talk about "Design for Manufactur-
ability" and how we should do our best to learn what's going on
instead of just "throwing it over the wall". Yet when you try to get
educated, you are told to "leave it to the vendor".

What's a designer to do?

Just a semi-casual thought (not trying to start any wars or anything)

onward thru the fog,                                    Grasshopper

p.s. Have a nice holiday you guys, I'm outta here...
----------------------------------------------------------------------------------------------------
Grasshopper,

	You bring up some good questions. I see these issues being discussed 
over and over and again. I work with a company that has a very large
customer base. More than 50% of the products we build have Controlled
Impedance requirements. There are many different Impedance 
modeling or simulation software packages out there.  Most of these packages
use the formulas that originated from the Motorola MECL program (Or the
Mantech program). For their time, they were extremely accurate and 
predictable. Keep in mind that that was during the 70's and think back
about the type of technology being built back them. Most of the work
was for 50 ohms with wider than today's technology. Contributor's
supplied material samples out of PTFE. Test vehicles were built
and tested to prove out the algorithms'.  FR-4 material's were 
substituted and dielectric thickness' were adjusted to target the
desired ohmic values (again 50 ohm ranges). I've used these
formulae and found them fairly accurate at the 50 ohm range 
after I have calculated the actual Dk based on resin to glass ratio.
As you stray from that combination, so does your accuracy. 
Now couple this with resin to glass variation. (Resin is ~3.6 and
Glass is ~6.3)
	If your designs need controlled impedance and you meet the
critical line length (or greater) criteria, then your options may be
to work up the stackup and target the impedance based on your
software calculations. Or, you may desire to let your fabricator
adjust the PCB parameters to meet the desired Zo. Please keep
in mind that the fabricator works off "empirical data" from past
known good builds. They may not be aware of the Dk variations
of the material(s) within the stack, but they should have a decent
knowledge of what they have built successfully. That data should
prove highly advantages to "adjust" or "tweak" the software
results to target a "closer to actual" impedance finding.  If you
have been using software that has been yielding acceptable
results and you feel comfortable with this, then you should 
continue your direction. 
One thing I do want to bring up is that the smaller the linewidth
gets, the more sensitive the impedance gets to any minor variations.
(linewidths, spacings, dielectrics, and copper thickness') Sub 4 mil
is a challenge. The new emerging technologies pose to really give
us a "run for our money".

Toodles

Groovy	

***************************************************************************
* TechNet mail list is provided as a service by IPC using SmartList v3.05 *
***************************************************************************
* To unsubscribe from this list at any time, send a message to:           *
* [log in to unmask] with <subject: unsubscribe> and no text.        *
***************************************************************************