On Tue, 07 Dec 2004 18:57:25 GMT, "Frank"
wrote:

|Hi Richard,
|
|With solid state power amplifier design; the criteria was always that you
|must present an impedance, to the output devices, such that the desired
|output power is delivered to the load (while not exceeding device
|dissipation). Any attempt to optimally match the load to the source
|impedance will result in over-dissipation, and probable destruction of the
|source device -- probably by excess collector/drain current. If you
|remember, Motorola used to publish Smith charts of the output impedance for
|their power amplifier devices. Talking to one of Motorola's design
|engineers; I asked "How do you derive these Charts". His answer was; "We
|use a matching network and adjust it for the required output power, then
|measure the input impedance of the network. The complex conjugate of this
|impedance is then defined as the source Z". The fact is these data are not
|the actual source Z of the device, but are probably considerable higher.

The data were useful as presented. When using a Smith chart for
matching network design, the published data could be used as the
"starting point" for the network and the rotations were made toward
the load; the opposite from the usual case of matching a load to a 50
ohm source.

This method was really an early example of load pull characterization.
Maury Microwave app. note 5C-041 is one reference for this. Another
is "A New Load Pull Measurement Technique Eases GaAs
Characterization", Microwave Journal, Nov, 1980, pp. 63-67.

|I don't remember anybody actually trying to measure the large signal S
|parameters of solid state devices.

|I seem to remember that tube amplifiers were designed based on the source
|impedance calculated as 2Vp/Ip, (Where Vp is the plate voltage, and Ip the
|plate current), and have no idea how, or if, it relates to the actual source
|Z of the device.
|Anyway, I am not convinced that source Z is important.
|Where I think some confusion may have come from is Hewlett Packard's 12 term
|error correction analysis derived for vector network analyzers. Here source
|Z is important because measurements are made in both directions.

It's not the error correction that is confusing. The error correction
simply removes systematic errors from the measurement(s). The
parameters usually measured when 12-term correction is called for are
the small-signal S-parameters. If I understand you correctly, "source
Z" is actually the output reflection coefficient (s22), the signal
exiting port 2 due to an input to port 2.

The amplifier output reflection coefficient can be very important,
even in high power amplifiers, when non-dissipative filters are used
for harmonic or spurious rejection. Such filters are commonly
specified and measured in 50 ohm systems and function by reflecting,
not dissipating, the out-of-band energy. When driven by other than a
50 ohm source the rejection will be other than what is measured in a
matched condition.