[time-nuts] Fwd: cheap frequency extension for timepod

[John Miles]

> -----Original Message-----
> From: time-nuts [mailto:time-nuts-bounces at lists.febo.com] On Behalf Of
> Gerhard Hoffmann
> Sent: Saturday, October 17, 2020 5:59 AM
> To: time-nuts at lists.febo.com
> Subject: [time-nuts] Fwd: cheap frequency extension for timepod
> . . .
> I have 2 17 dBm SRA1-WH, but it would require slaughtering something
> unimportant.

I prefer using high-level mixers because a level-7 mixer will be hopelessly crunched by most signal sources.  Note that a "+7 dBm" or  "level 7" mixer is meant to work with +7 dBm at the LO port, not the RF port.  For quadrature phase detection it's common to saturate both ports, but in this case the mixers are being used as actual mixers rather than phase detectors.  You will get better port isolation and conversion loss if you stay within shouting distance of the datasheet specs.  

I got a good deal on a bunch of connectorized ZFM-2H mixers on eBay a while back, and they've come in handy many times.   The rated 1-dB compression point with the ZFM-2H is +14 dBm with a +17 dBm LO, while SRA-1 is rated +1 dBm with a +7 dBm LO.  The extra +10 dB of LO injection buys a disproportionate performance improvement.  

In principle, I/Q mixers have an advantage in that they can be set up to reject noise at the image frequency.  This is worth thinking about for low-noise measurements, because you're going to be driving the two mixers at Ch0 and Ch2 with a signal from a splitter that will exhibit both common- and differential-mode noise over the whole spectrum of interest.  Those are conditions under which weird cross-spectral artifacts tend to show up.

Right now I'm working on an app note for downconversion measurements on the 53100A that will explore all of these questions to some extent, but there's nothing concrete to share just yet.  As a sneak preview, though, the app note will emphasize the importance of (broadband!) isolation and termination of all mixer and splitter ports in measurements with independent downconverter channels.  Because Wilkinson-style splitters work best at frequencies where their input ports are properly terminated, a splitter with degraded isolation due to input mismatch may end up coupling the two LOs to each other through their respective mixer RF ports.

Given the compact hardware layout in your photo, I'm concerned that good isolation may be tough to achieve.  The two oscillators may even try to injection-lock to each other.  Not necessarily a showstopper as long as the lock bandwidth is low enough, but if that aspect of operation is out of your control, it's not a Good Thing.  Empirically, I've found it best to run the LOs at a frequency separation less than 10% of the minimum offset of interest.  This isn't a concern on the 53100A, but the TimePod has only one DDS core for both of its input channels.

-- john