Measuring P1dB - Avoiding instrument inaccuracies

Question

I'm writing a universal test set program and I'm having problems with measuring P1dB. In the end I will have an HP8757 connected to either an HP8350 or HP83752 along with a power meter. The problem lies with the HP8350 and HP83752. The way I wrote the test is as follows...
Take 10 points by increasing 1 dB each time and measuring the output power.
After I have the 10 points I take the output power - input power to get the gain.
If any of the points varies .2 from the first point back up 2 dB and start over.
I continue increasing 1 dB until I find the normalized gain - 1 dB. When I find this point I decrease 1 dB and change the increment to 0.1 dB.
I was originally using a signal generator which had a pretty accurate 1 dB step, but now a few of the 8350s have anywhere from 0.7-1.2 dB step (instead of 1 dB).
Does anyone have any suggestions? I know I could use two power meter heads and search for a 1 dB difference, but buying an extra power meter head isn't an option I have.

Answer 1

I use a cubic spline to curve fit to find the 1 dB compression point.
If you have a choice between the HP8350 and a synthesizer, use the synthesizer. A dirty signal with frequency components out of the calibrated frequency your power head expects will cause accuracy problems.
Also, if you add an amp between the source and your DUT, you'll have a lengthy calibration without two power heads because you shouldn't count on the test set amp being linear at all of your source power settings, but maybe you know that.
Hope I helped a little, good luck.