From: Gregory R Edlund ([email protected])
Date: Mon Dec 18 2000 - 14:27:49 PST
Scott, Ken, and others:
I heartily concur with your observations that we speak with our purchase
orders. When two engineers talk to each other about this problem, they can
usually see eye to eye. But the engineer who works for the semiconductor
company still has to convince his or her management to spend money making
high-quality models - and they ain't free. I would even go as far as to
say that the SPICE or IBIS model data is every bit as important as the
information contained in a component datahsheet. How do we convince our
vendors of this? It's more complicated than just putting a clause in a
purchase agreement. There's the question of what you mean by high-quality
model data. There's also the issue of education. Furthermore, how do we
convince our own purchasing people?
Someone touched on the topic of lab verification, which is near and dear to
my heart. I have wandered down this road with two kinds of parts: 1)
simple buffers, such as are used to drive memory address nets, and 2)
SDRAMs. My observation is that it's very difficult to get the outputs of a
SDRAM in the right state to make the measurements you'd like to make. I'd
imagine this is true for the majority of parts we care about. Has anyone
tried to solve this problem? What we need is some test pin on the device
that allows us to control the outputs for model verification testing...
IBM Server Technology Development
3605 Hwy. 52 N, Dept. HDC
Rochester, MN 55901
Actually, the problem is much simpler than you make it.
The problem lays with the silicon vendor and the companies which
use their silicon. If you "need" simulation models in order to
perform a good engineering design ... and you cannot obtain
them from the vendor of the silicon ... then you should put pressure
on them to provide, or contract out the model development.
As the purchaser of the device, you have the only leverage available.
If designers start making analog simulation models a part of the
contractural obligation for the purchase of silicon, then this problem
would go away rather quickly.
Can anyone imagine performing an ASIC design, synthesis and
verification without having accurate models and specifications of
the cell libraries?
We execute 20 to 40 SI evaluations for customers each year and
have the same issues as you. When our customers are unable
to obtain the data necessary to perform accurate analysis and
design, we counsel them to:
1) place leverage on the vendor to obtain the data
2) switch vendors
3) add ibis/spice modeling to purchasing contracts.
If you can't obtain the information that you need from Toshiba,
what makes you think that a consulting house or a software
Ultimately, if you need a particular piece of data before executing
a design with a specific part and are unable to acquire it, then your
design is at risk. I often ask my customers to change parts in
their designs for this reason alone.
**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected] 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 : Tue May 08 2001 - 14:30:28 PDT