HP 8662A Signal Generator

The HP 8662A has been a gold standard for low noise RF synthesizers. Unfortunately, the way HP presents performance data in the 8662A manual makes it a little confusing to see the whole picture. This table pulls together the various phase noise information contained in the manual, including corrections published in the errata.

The "Absolute" and "Residual" numbers come from the "Typical Performance" part of the specifications. The "Absolute" data is interpolated from a graph, so it has at least a couple of dB ambiguity.

Offset Specification
320-640MHz		 Absolute
320-640 MHz		 Residual
320-640 MHz
10-100-90-107
100-112-111-117
1 kHz-121-128-128
10 kHz-131-135-136
100 kHz-132-134-136

I measured the phase noise of the 8662A at several HF frequencies. Here's a table showing my results, as well as the HP-measured residual noise for the 0.01 to 120 MHz range.

-
Offset Residual
0.01-120 MHz		 Absolute
5 MHz		 Absolute
10 MHz		 Absolute
15 MHz
10-115-100-105-103
100-126-118-117-116
1 kHz-133-129-127-126
10 kHz-137-135-134-133
100 kHz-137-137-137-136

Here's a plot of the full results. I also included the oddball 17 MHz frequency to see whether the noise was significantly different with a random frequency that exercied all the digits of resolution; it's interesting that it shows significantly lower noise than the round numbers do.

Here is a comparison of the phase noise at 14 MHz using the internal 10 MHz oscillator and using an external oscillator with extremely low close-in phase noise (though its noise floor is similar to the internal oscillator). I'm looking for an explanation of why the external reference is so much worse than the internal one...