[time-nuts] Keysight N5511A - phase noise measurements down to theoretical-177 dBm/Hz

[Magnus Danielson]

Hi,

On 2019-08-16 08:30, John Miles wrote:
>> My guess is well into 6 figure$.
> It's a modular system like its 3048A and E5500-series forebears, so the only limits are your imagination and your bank account. :)  Very cool hardware, noting that it has some competition from all-in-one instruments like the FSWP that the earlier 'doomsday machines' didn't have to face.  
>
> I'm sure you could spend a quarter of a million dollars on either the FSWP or the N5511A if you checked all the boxes.
Indeed. You get some pretty good stuff when you do.
>
>> BTW, I don't know what -177 dBm/Hz has to do with phase noise.
>> The relevant units are dBc/Hz as any time-nut knows.
> I haven't seen PN analyzers rated that way before, but it makes sense to a certain extent.  They are basically saying they can measure the noise on a 0 dBm signal down to -177 dBc/Hz, with signals above 0 dBm being measurable with a commensurately-lower floor.  Their rated power limit is +20 dBm, so -197 dBc/Hz would be the theoretical limit.
Well, assuming the internal noise becomes a limit, which does not
necessarily need to be, but it is the traditional limit.
Cross-correlation can measure below it, but you then need to avoid
spectral collapse as well as AM-to-PM conversion which is another issue.
> The brochure goes into that in some detail, but there's no mention of cross-spectral collapse, which I thought was interesting.

Most hand-wave around that question. I've only see two reasonable
approaches, the onces we two provided, each with it's benefit and
drawback. A third approach has been proposed, but I have not seen a
realization of that.

>   The white-noise floor in their example plots is enviably flat, with no divots, suspicious-looking valleys or other artifacts.  If they are able to avoid that problem in the general case, it would be interesting to hear more about the strategy being used.

Agree. It's a tricky one as one try to reach that level. The approach I
proposed is so far impractical even if it works.

Cheers,
Magnus