Newsgroups: sci.electronics
Path: cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!fs7.ece.cmu.edu!europa.eng.gtefsd.com!gatech!udel!news.intercon.com!psinntp!megatest!mithril!alung
From: alung@megatest.com (Aaron Lung)
Subject: Re: IC Packages
Message-ID: <C5LG71.5Ds@megatest.com>
Organization: Megatest Corporation
References: <1993Apr16.142715.12613@sol.ctr.columbia.edu>
Date: Fri, 16 Apr 1993 20:47:27 GMT
Lines: 58

In article <1993Apr16.142715.12613@sol.ctr.columbia.edu> seema@madvlsi.columbia.edu (Seema Varma) writes:
>Hi,
>	I am looking for some help in choosing a package
>for a high-speed silicon ADC (100Mhz) currently being 
>fabricated. This is a PhD research project and I have to test
>the chip at speed on a PCB. I expect to have roughly 100
>packaged circuits and will do DC, low-speed and high-speed
>testing using 3 different set-ups for the test chip.
> 	
>	I know for sure that a DIP will not work
>(the long lead lines have too high an inductance).
>Getting a custom-made package is too expensive, so
>I am trying to choose between a flatpak and a
>leadless chip carrier. The flatpack would be hard
>to test since it has to be soldered on to the test
>setup and I would spend loads of time soldering 
>as I kept changing the test chip. The leadless chip
>carrier sockets also have long lead lines and may
>not work at high speeds.
> 
>	Does anyone out there have experience/knowledge
>of this field ? I would greatly appreciate help! Any ideas/
>names of companies manufacturing holders/sockets/packages
>would help. 
> 
>P.S. The multi-layer fancy GaAs packages seem like a bit
>of overkill(?)
>	
> --- Seema Varma


You didn't mention whether or not cost is an issue.  Where
exactly are you running 100MHz??  THe digital side?  TTL?
ECL?  We run 200MHz and 100MHz all over our IC test equipment 
all day long in the ECL domain, and we use DIP's along
with PLCC's, 25mil and 20mil pitch 256pin QFP's to name
a few.  I don't see a problem in packaging as long
as you adhere to sound engineering practices.  

A good source of information is Motorola's MECL System Design
Handbook. The latest ed. is dated 2/88.  That is considered
to be one of the *bibles* in high-speed design.

The very fact that you need to build a test fixture means
you're most likely going to need a socket.  It in itself
has far more inductance per pin than the package you are
testing, not to mention any impedance discontinuities.  I
don't see the big concern over the packaging because it
probably isn't going to make that much difference

If you're trying to get TTL to run at 100MHz, have fun...
TTL was never designed to run in a 100MHz environment.
:-(

aaron

P.S. My opinions have nothing to do with my company...the
standard disclaimer applies.
