[SI-LIST] Re: Package Frequency?
- From: Pat Diao <Pat_Diao@xxxxxxxx>
- To: "'eric@xxxxxxxxxxxx'" <eric@xxxxxxxxxxxx>,SI list <si-list@xxxxxxxxxxxxx>, Pat Diao <Pat_Diao@xxxxxxxx>
- Date: Wed, 16 Jan 2002 10:31:36 -0800
Thanks all for the responses.
Eric I'll be at the upcoming netseminar. I also found the past seminar
materials on your website very useful.
Pat
Patrick Diao, Ph.D.
ASAT Inc.
Fremont, CA
phone: (510) 249-1227
cell: (408) 666-2285
pat_diao@xxxxxxxx
-----Original Message-----
From: Eric Bogatin [mailto:eric@xxxxxxxxxxxx]
Sent: Wednesday, January 16, 2002 8:28 AM
To: SI list; Pat_Diao@xxxxxxxx
Cc: Eric Bogatin; Gary Otonari
Subject: Package Frequency?
Pat-
You ask an interesting question, which I also get asked quite
often: What is the bandwidth of a package? or, what's the highest
clock frequency at which I can use a package?
Often times, an ok answer now is more useful than a great answer
late. In this context, I'll offer you a simple, first step, rule
of thumb to evaluate the highest useable frequency of a package.
Some of the details of this analysis are described in app notes
on the GigaTest Web site, www.gigatest.com, app note # 81 and
#72. Some aspects of this topic are also reviewed in the upcoming
web caste I'm doing with Agilent that is free to everyone. You
can register for it at
http://www.tmintl.agilent.com/us/SigInt.shtml
Obviously, to evaluate the limit of a package, you have to know
what breaks first, as you increase clock frequency. I find, in
most low frequency packages, the first effect to arise is too
much switching noise.
Packages, and connectors, are typically the largest sources of
simultaneous switching noise (SSN) in the entire system. It's all
about the mutual inductance between adjacent signal paths (and of
course, their returns). For a fixed mutual inductance, the amount
of switching noise will increase as the rise time decreases. This
means that if there is a max allowable SSN, there is a simple
relationship between the mutual inductance between signal pairs
and the minimum rise time at which the package can be used.
If the mutual inductance between two signal paths is M, and they
are in a 50 Ohm environment, and the noise budget allocated to
switching noise is 5% of Vcc, then with a little bit of algebra,
which is outlined in app note #81, we can estimate the
relationship between the minimum usable rise time as: rise time >
M/2.5, with M in nH and rise time in nsec. For example, if M is
2.5 nH, the shortest usable rise time is 1 nsec. A shorter rise
time and the SSN will be greater than 5% Vcc.
We can roughly relate the minimum rise time usable for the
package to the effective bandwidth of the package as BW =
0.35/rise time = 0.9/M. Since this is a rough estimate, I prefer
to use BW ~ 1/M, with M in nH and BW in GHz.
Of course, this is often times a high estimate, since we were
assuming the SSN was due to coupling between just two signal
paths. In practice, it is more typically 2-8 signal paths that
couple, depending on the number of return paths per signal paths.
If there are really n signal paths that couple together, each
with roughly the same mutual inductance of M, then the
relationship is BW ~ 1/((n-1) x M).
In a 100 pin PLCC package, for example, the mutual inductance
between adjacent signal paths, with good return pin selection, is
typically about 0.4 nH. If there are 2 leads that couple and can
switch simultaneously, then the BW of the package would be about
1/((2-1) x 2.5 nH) or BW ~ 400 MHz. This package would be
suitable for clocks up to about ~100 MHz.
Keep in mind this is a rough rule of thumb. It's meant to give an
answer quickly, not accurately. If you want to know if the max
operating clock frequency for a specific package is 110 MHz or is
it 120 MHz, don't use this rule of thumb. But if you want to know
will the max clock freq be 100 MHz or 500 MHz, this is a very
useful tool that will give you an answer in a few seconds of
effort.
The only way to know whether a particular package is going to
work in a specific application is to create a model for the
package and introduce it in a system simulation. However, you're
looking at a 2-10 day task, requiring a lot of expertise and the
right tools. This rule of thumb helps feed your intuition that it
is mutual inductance that's important, and roughly how much is
too much. It lets you glance at a package and evaluate if this
package is going to have a chance to work for 50 MHz, 200 MHz or
1 GHz applications.
As may be obvious, if the mutual inductance is the limitation to
a package's high speed performance, this says when you design the
pin assignments for the package, you really want to be looking at
the impact on the mutual inductance. That's why understanding the
physical basis of mutual inductance is so important. It also says
knowing the mutual inductance of a package is one of the most
important terms that characterize it. You can get the mutual
inductance between signal paths either from measurements or
calculations.
If you are interested in these topics, we review these techniques
in our short courses. GTL 122, SI 101, reviews the basic
principles of inductance, GTL 260 reviews how to use measurements
to extract the mutual inductance in packages and connectors and
GTL250 reviews how to reduce mutual inductance in your system.
Info is posted on our web site. Our next class coming up is
GTL260, Creating high bandwidth models of interconnects from
measurements. It is offered right before DesignCon in Sunnyvale.
I'm also doing a web cast with Agilent on Jan 22 at 9 am PST that
reviews a little bit about measuring the mutual inductance
between two SMT resistors. What I show is the measured mutual
inductance for this particular case is about 0.28 nH. The
bandwidth these terminating resistors might be useful up to is
about 3.5 GHz. We show the result of a transient simulation using
this model, extracted from the measurements, with a predicted
switching noise close to what is estimated. The web cast is free
for anyone to participate. You can register for it by going to:
http://www.tmintl.agilent.com/us/SigInt.shtml
see you there.
--eric
**************************************
Eric Bogatin
CTO, Giga Test Labs
v: 913-393-1305
f: 913-393-1306
e: eric@xxxxxxxxxxxx
26235 W. 110th Terr. Olathe, KS 66061
corporate office:
408-524-2700
134 S. Wolfe Rd Sunnyvale, CA 94086
web: www.gigatest.com
**************************************
From: Pat Diao <Pat_Diao@xxxxxxxx>
Subject: [SI-LIST] Package Frequency?
Date: Tue, 15 Jan 2002 17:14:17 -0800
Hi All,
I'm in the electronic packaging industry. One of the questions
frequently
asked by my customers is how high a frequency a particular
package type can
handle. Note it is asking about the frequency handling capability
of the
package, not the die/chip.
Personally I think there should be no limit on the frequency a
package can
handle, as long as after whatever distortion applied by the
package to the
signal, the signal sent/received at the chip is still bearable to
the chip
designer. In other word, a package works fine at 10 GHz, it may
still works
fine at 50 GHz just because the signal is still recognizable by
the
particular chip.
But is there any other concerns at high frequency? Can the
materials
degrade at high frequency?
Since I am new in SI, I like to have your input or comment on
this. Thanks,
Pat Diao
ASAT Inc.
Fremont, CA
phone: (510) 249-1227
cell: (408) 666-2285
pat_diao@xxxxxxxx
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
