Date: Fri Jun 01 2001 - 12:42:38 PDT
In a message dated 6/1/01 9:32:37 AM Pacific Daylight Time,
> It is possible that at very high freequencies, 20 GHz and above, that a right
> angle corner might affect signal quality. If I or anyone else were
> working at
> those frequencies, it would be good engineering to retest to demonstrate
> or not a right angle corner might be problematical.
> A significant problem our profession has is applying blanket rules when they
> might not be appropriate. The right angle bend rule for logic is one of
> As an aside, there are often many right angle bends in the IC packages we
> use every day. Odd that no one worries about those!
Unfortunately, "good engineering" is subjective and depends on one's
experience. Here, I interpret that the comment only refers to the signal
"Blanket rules" do indeed need questioning; however, there are usually
long-forgotten reasons for the evolvement of most "rules" that may not be in
the users/designers toolkit or experience. Where commonly used design rules
are penalizing some aspect of the design (e.g., board space), intense
scrutiny is of course warranted to re-justify the need for implementation.
As regards the applicability of right angle (or any sharp) turns in PCB
traces, there are other factors (than just electrical signal propagation)
that should be considered. For me, the generation of even minor reflections
from multiple trace bends decreases signal design margins. More important is
the potential for reduced reliability in the field caused by thermal cycling
and vibration environments over the life of the end product. As traces have
become narrower to accommodate higher numbers of routing layers, the
potential for generation and propagation of cracks from the inner corners of
the traces increases. Such failures (mostly intermittent) are not uncommon in
normal day-to-day use of commercial products (especially portable ones) in
use today. The noted physical environments were the source of dictates (years
ago) in military and other rugged use applications for rounded trace corners,
and tapered (teardropped) trace interfaces at pads.
Mitered right-angle bends have been shown to improve signal transmission
quality (through reduced reflections) for the higher frequencies (and
harmonics of the lower operating frequencies). However, the necked-down trace
width immediately opposite the inside 90-degree bend substantially aggravates
the susceptibility to physical environments and a short field life should be
anticipated. How many of you have been plagued by an intermittent failure
that likely (if found) would be traced to an intermittent trace connection? I
had one that would only appear in the temperature range between -24 and -40
degrees Centigrade. And its existence might have cost a $150,000,000
satellite, had I not found it.
I prefer to implement designs that are as close to design targets as is
practical. This approach is particularly applicable to signal integrity as
the allowable design margins seem to be shrinking each year as speeds go up
and voltage levels go down. And I try to deliver designs that have > 10 years
expected operating life. To generate those designs takes a negligible bit of
extra engineering time when the techniques are known.
As regards right-angle turns in IC packages, they are mostly small, rigid
structures that will experience minimal physical flexure and will inherently
be less susceptible to physical failure than larger (PCB) assemblies. The
traces (particularly on silicon and ceramic) are generally much narrower (a
typical PGA ceramic uses trace widths on the order of 100 microns);
therefore, the reflections may not become a problem until frequencies well
above 100 GHz are involved. Bottom line, IC packages are less susceptible to
those problems that can arise on larger PCB assemblies; hence, there are
fewer (similar) problems and less discussion on the subject.
That said, I recommend that all designers take the time to research,
understand, and (hopefully) enhance all "design rules" encountered.
Innovation comes from such efforts, as well as design confidence in what
margins YOU are designing into the products of tomorrow.
I think my soap box just cracked, so I'll leave it at that. Have a good
Michael L. Conn
*** Serving Your Needs with Technical Excellence ***
**** 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 : Thu Jun 21 2001 - 10:12:13 PDT