HP 8566B Spectrum analyzer, trouble with band C (5.8 to 12.5 GHz)

[rastro]

Suggestion:
You need to go back and objectively confirm you don't have problems with the other bands.  You should really baseline all your upper bands.  I know band-C is going to look bad but take a base line regardless.
1. Go to factory default for the pre-filter DAC (value 32 in all bands)
2. For each band pick 10 or 12 equally spaced frequency points.
3. Go to each of these frequencies on you signal generator and use your 'power meter' to measure the level at the end of the cable you are using to test.  Don't adjust the power level - use the same power setting for all frequencies - yes there will be drop off at higher frequencies.  Just duly record each frequencies corresponding 'power meter' reading at the end of the test cable. Now you have a table of power levels for each test frequency that also reflects the true cable losses for each test frequency.  Basically this is what you are feeding to the SA.  No guessing how must the cable is changing the signal level! 
4. Now unhook you power meter from the end of the cable and hook it to the SA.
5. Go through the frequency points and record SA's power level next to that frequencies 'power meter' reading.  The delta of these 2 values says how far off the SA level is for that frequency.  The deltas of 12 frequency will give you a pretty good idea the flatness for that band.

With this procedure it's important you use the same cable and power level on the signal generator.  The only thing changing is the frequency.

[G0HZU]

That sounds like a good strategy for getting predictable test signal levels. In my case I was able to rely on the flat response of the HP83752A across this range and the Gore cable only has about 0.7dB loss up at 15GHz. If you look in the image below, I ran the generator at about +6.5dBm into the Gore cable and there's a decent 6dB SMA attenuator at the end of the Gore cable. I checked the levels with the power meter to make sure there was nothing wrong with the way this all interacted up at 15GHz.

The HP83752A has a neat user calibration where you can calibrate the ALC (using a lookup table) to normalise/level at the end of your test cable. However, I've never tried using this feature and I'm not using it here. Therefore, the setup in the image below does have a slight rolloff across the band due to the cable and the attenuator but it's very minor when measured with the power meter. It's a small fraction of a dB although it's hard to measure small variations like this accurately over a wide bandwidth with the gear I have here.

My main RF bench is currently crammed with gear as I'm working on some filter and oscillator designs plus I've been doing some research work for someone down at lower frequencies. So I had to quickly clear a space for the sweeper. But you can see the Gore cable and the SMA pad that I used.

[rastro]

Hello GOHZU,

Looks like you have a very nice RF bench.  I really like your HP83752A.  I made a few feeble bid's on eBay for one of these - but they typically close well outside my budget. 
I have an HP 8340B as my main RF generator that I'm pretty happy with it.  I'm currently plugging away at using GPIB to control my HP 8340B and HP 438A power meter so I can characterize my signal source at cables-end.  I'd also like to document a baseline on my spectrum analyzers in case I have problems in the future.

-rastro

[Berni]

Nice RF bench there indeed. Like the look of that network analyzer.

I did quickly figure out that preselect peak works on peaking whatever is under the marker so what i tend to do is press peak search beforehand to precisely move the marker to where it should be.

Sorry haven't been able to do much work on my sick analyzer since we have some big time bathroom renovation going on here.