[Radio Wrangler]

There's a 1Hz RBW in the 8590A, and that uses crystals in its 100kHz IF. "RF" range is only 0-50kH, though.

The 859x analyser family have narrow (below 1kHz) RBWs done with a mix-down to a lower IF. 1kHz and wider is crystal, then LC at 21.4 MHz. Up-down keys will get you to 3MHz. But you can type in 5MHz and get it. Shape is not specified above 3MHz, but it's useful.

The higher performance 856x family can only go to 1MHz. They too switch in an extra conversion for the narrowest bandwidths, but later on, they were changed so the very narrowest IFs put the signal directly into what was otherwise the ADC after the log detector, and did the narrowest bandwidths as well as logging in DSP. This went on into the ESA family. I used the ESA as the basis of the noise figure analyser, but it has a completely different IF hardware and all RBW is DSP, aimed at 4MHz bandwidth for standard Y-factor measurements, but subsequent DSP can take it down to 100kHz if needed, but noise factor gets a bit slow to average down there. It was intended to be able to measure multiple 100kHz buckets at once and speed up measurement, but this didn't make it to the final product chiefly because of software development time for having to handle separate cal factors for each bucket. It was super-critical that the different IF gain stages had steps which combined arithmetically without combinatorial errors which were a limitation in the old 8970. The NFA has an extra mix-down from 21.4 to 6.25MHz. Partly, this helps with the ADC/DSP, but it also breaks up a lot of switched gain all at 20MHz in the 8970, and it's leakage around this chain which spoils the arithmetic combination of the gain steps.

One whole lotta fun (apologies to Robert Plant)

David