[SI-LIST] : curve tracers and Output Impedance

Sandy Taylor (ataylor@rolyat)
Fri, 3 Apr 1998 09:04:27 -0800 (PST)

Continuing on the subject of determining the output impedance.

Curve Tracer on a real device:

When I have had real devices in the lab to test a great way
to determine the output impedance is to use a curve tracer.
You force the output to a known state, then sweep voltage
of the curve tracer over the range of interest and monitor
the current. The drive is current limited, so you can overdrive
above VDD and below VSS to verify latch up immunity.
There you have I vs V for the output driver, repeat
with the output in the other state, and you have the
DC characteristics of the driver.

Curve Tracer on a spice netlist:

Just like above, except you use a SLOWLY varying time dependent
voltage source to force the output while monitoring the current.
Play with parameters a bit and you can plot the DC equivalent
resistance: Vout/Iout, or you can plot the "AC resistance": dVout/dIout.

A (very) good driver will look like:

| -------
| /
| /
| /
| /
| /
| /
--------------- Vout

The flattening out is due to the diode clamping.

To get such a nice looking Iout vs Vout curve with CMOS devices
is not trivial, and involves active circuits. For completeness
such a circuit should also be analyzed in the frequency domain
to ensure stability and consistent characteristics over frequency.
Bode analysis can be performed with the circuit connected to
an external current source to drag the output voltage into the
region of interest.

One could argue that with this analysis you would not have
to do the transmission line "step" simulation, but it should
be done for completeness.

My $0.02 worth.

Sandy Taylor Sun Microsystems

Standard disclaimer: blah blah blah.