Re: [SI-LIST] : how to best validate edge rate and output impedance

About this list Date view Thread view Subject view Author view

From: Chuck Hill (chuckh@altaeng.com)
Date: Thu Nov 18 1999 - 12:42:22 PST


Pat,

See my comments below.

At 02:31 PM 11/17/99 -0600, Pat Zabinski wrote:
>
>We are having "conversations" with a vendor of macro libraries
>about how to specify/validate two key parameters for output buffers:
>1) output impedance; and 2) edge rate (full-swing, rail-to-rail
>CMOS).
>
>We have clearly specified these values, but we failed to specify
>how these values are determined (oops!). The approach we typically use
>produces much different results than the vendor uses, and I'm
>interested in getting input from folks on any "industry standard"
>or "best engineering" approach you use.
>
>For output impedance:
>
> * we validate by running a full spice simulation with
> the buffer driving an ideal transmission line and tweaking
> the line's impedance until we obtain proper waveforms
>
> * the vendor uses DC I-V curves to design the output
> impedance and stops there
>
> (for this particular project, their values are about
> 10-15% higher than ours)
>
>For edge rate:
>
> * we validate by simulating an output buffer driving an
> ideal transmission line terminated with an input buffer;
> we then measure the 20-80% edge rates at the far end.
>
> * the vendor places a lumped capacitor load directly on
> the output of the buffer and measures the edges (I'm
> not sure what %'s they use).
>
> (for this project, their edge rates are 3X-longer/slower
> than what we obtain!)

Model the output driver as a Thevenin equivalent valid during the switching
interval. If the output voltage is highly load dependent, the output
impedance is not low. When driving a long transmission line, the voltage
and current at the output are initially in-phase since the characteristic
impedance is real valued. Thus, dv/dt is proportional to di/dt. Now,
think of the output driver (during the switching interval) as a Norton
equivalent; the dv/dt when driving the transmission line is determined by
the di/dt of the driver and the line impedance; the dv/dt when driving the
lumped capacitor is proportional to the current i of the driver. The i and
di/dt are different. My conclusion is both loads are useful, each
providing information about the output, that is, when measured at the output.

When you measure the dv/dt at a capacitive load on the end of a long
transmission line you are indirectly measuring the current wave propagated
on the line. The current wave in the line is the result of the Norton
current source, a current divider between the output impedance, and the
characteristic impedance. Unfortunately, the current source and output
impedance are both functions of time. But there is more information
resulting from this test. This is also useful since it is closer to the
real case.

Charles Hill, consultant

>
>Is anyone aware of industry-standards for characterizing these
>values (in simulation, not measurement)? Outside of standards,
>does anyone have a particular approach they like to use?
>
>Thanks for the input,
>Pat
>
>
>
>**** To unsubscribe from si-list: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list, for more help, put HELP. si-list archives are accessible at
http://www.qsl.net/wb6tpu/si-list ****
>

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:38:59 PST