Mail*Link(r) SMTP               FWD>Delay with feedthrus

Bilal-

The following is in response to you request for the electrical characteistics
of "feedthru" vias.

Propagation delay is a function of the sqrt(of the dielectric constant) and
the porpagation velocity, therefore, your "feedthrough" delays should be the
same as you conductors (though the impedance may be reduced, due to the next
comment).

Many years ago, we were concerned about the effects of via's and right-angled
corners on transmission line signal integrity.  At that time, we did some
literature research and found some mathematical modeling and limited hands-on
TDR like measurements had been performed.  We manufactured some 50 and 100 ohm
impedance transmission line test specimens in/on some 1.5 mm thick 4- and
6-layer MLB's and conducted some TDR and other measurements.  At that time,
lands and holes were respectively about 1.5 mm and 0.6-0.7 mm in diameter.  We
found each via added about 1-1.5 pF of capacitance, and right-angled corners
were not an electrical problem for clock rates to about 200 MHz with pulse
rise/fall times greater than 200 ps.  So being very conservative, we 45-degree
angled conductive pattern signal corners over a few MHz to ensure there would
be no electrical problems, and engineering design and manufacturing personnel
were happier not to see right-angled corners.  We have used this as design
requirements for almost 15 years.  About 4-years ago, Lee Ritchey (then at
Shared Resources, now at 3-Com) conducted some studies to evaluate the design
requirements for transmission lines in/on printed boards.  This effort was
presented at NEPCON West about 3-years ago, and I belive various version have
appeared in some trade journals.   

The via (feedthrough) you mention adds a little less than 1 pF of capacitance
to you conductor and for all practical does not cause a delay.  There are two
good references (that I don't have here at my finger tips); one is a
mathematical model like determination (I believe it was in an IEEE journal)
and the second is Lee's "hands-on" test.  In Lee's technical report, he shows
a transmission line with varying numbers of vias on several pitches.  As
expected, at each via, there is a reduction in impedance

Likewise, there have been three good studies (I know of) on the need for
"right- or 90-degree angled" corners on transmission line conductive patterns.
 Two of there were IEEE journal theoretical studies/models and the third was
Lee's hands-on TDR studies.  In summary, right angled corners are
"electrically" not necessary as long as the pulse rise/fall time is greater
than 100-200 ps (as I recall, I believe Lee uses a 100-125 ps limit). Avoiding
right-angle turns is more of a DfM, and therefore is a good practice to do
anyway.

If you need them, I'll look up the individual references.

Ralph Hersey
email:  [log in to unmask]

-----------------original message follows ---------------------
Date: 5/20/96 12:09 PM
From: Bilal Khalaf -Applications

Dear techies,

I need some help in calculating delays due to feedthrus.  I am not sure
what's involved in the equation but here's what I have:

FR-4 0.062 (50 ohm impedence)
FHS is 0.026, Pad 0.040
MCLK frequency is 60MHz (Testing a 7 ns SRAM)
Solder mask SR-1000
Tin-nickle (1oz Copper) plating
Solder type 96/4

SHould I worry about delay's at that frequency??  Any references would be
appreciated.  thanks in advance.


Please feel free to contact me if you have any questions.



Regards,

Bilal Khalaf
Applications Engineer
email: [log in to unmask]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
TYECIN Systems, Inc.
Phone:  415.949.8501     Fax:  415.917.1901
"Software Tools for Manufacturing Management"


------------------ RFC822 Header Follows ------------------
snip-snip all the old header stuff.