Boring Signal Generator Accuracy/Linearity Data

UPDATE: it turns out that my “best” power meter was actually about 0.5 dB off. See my later post for further details.

I want to calibrate my software defined radio receiver prior to doing a bunch of recordings of the ham bands for the upcoming solar eclipse (check out the HamSci web page for details). I have an HP 8642A signal generator which has high-precision attenuators and should be a very good reference to measure recorded amplitude vs. actual.

Unfortunately, this generator hasn’t seen a cal lab in a decade or two, so I thought I’d try to check its performance (amplitude accuracy, and linearity over a wide output range). To do that, I used several sources:

* HP 438A power meter with 8484A power sensor that’s sensitive down to about -60 dBm
* HP 3586C selective voltmeter(s) that are sensitive down to about -110 dBm
* HP 3585B spectrum analyzer that is sensitive down to about -100 dBm

The power meter and sensor are literally a few days past their calibration due date; the other gear has not been to a cal lab in decades.

Here’s what I found measuring at 10.100 MHz:

* At 10 dBm steps from -50 dBm to +10 dBm, the power meter read 1.0 dBm lower than the 8642A amplitude setting. At -60 dBm, the reading was 1.2 dBm low but was also noisy, indicating the threshold of the power sensor’s sensitivity. The LMR-240 jumper cable and coax adapters could account for the 1 dB of loss. So, we can say based on these readings that the generator is accurate at the measured frequency within 1 dB and linear within 0.2 dB. Actual performance is probably better than that at the low end.

Because the power meter’s sensitivity is limited, I used two other pieces of gear to measure lower amplitudes. Since all instruments covered the range from -60 to -20 dBm, we can guess that these units closely match the power meter in their common power ranges. They used the same jumper cable, but had one less adapter in line than for the power meter measurements.

* From -110 to 0 dBm measurements with the 3586C selective voltmeter were 0.3 dBm lower than nominal, plus or minus 0.1 dBm.

* From -110 to 0 dBm measurements with the 3585B spectrum analyzer were 0.6 dB lower than nominal, plus or minus 0.2 dBm.

I then did spot checks with the power meter at the other frequencies I plan to record, at values of -60, -50, -40, and -20 dBm. At -60 dBm the variation across frequencies was plus or minus 0.2 dBm (but the signal was clearly a bit noisy). At all other values, the readings varied by less than 0.1 dBm.

So, I think these measurements verified that the 8642A is well within specs in the HF frequency range, and that my three power measurement devices match within better than 1 dB. (I also learned that two of my 3586C units need repair or adjustment, but that’s another story…)