I'd really like to own a spectrum analyser and network analyser! These days it's possible to buy USB-connected versions of both that have great specs and aren't furiously expensive, but old-fashioned standalone types are a bit more appealing. I don't need either often enough to really be able to justify spending thousands of dollars, especially when I can go to my university and have access to both.
Naturally, the solution is to waste hundreds of hours and build my own uncalibrated one with dubious specs!
SDR-Kits previously sold a kit network analyser, but it is now available only in pre-fabricated form. I finally broke down and ordered one. It is currently in the mail. Now it's on my desk and I haven't had a chance to look at it yet.
Scotty Spectrum Analyzer is a set of plans for building a 0-3 GHz spectrum and network analyser for only a few hundred dollars. Unfortunately the information is a little scattered and it's not possible to buy it in a kit (even short form), so I was pleased to discover that bg6khc is building his own version and selling all of the PCBs for a bargain price. All of the modules can be assembled and tested independently, so it's tempting as a side project! This will probably go on the back burner now that I've bought a commercial SA and VNA, but I've ordered the PCBs and all of the bits required from Mini-Circuits and will start putting the modules together as I get time. The current design uses an LPT port and Windows software, but I have vague plans about using a cheap Android tablet or a Digilent Atlys (and HDMI monitor) as the user interface. Construction of a cavity filter looks fun but time consuming. A SAW filter replacement is in the works!
Advantest TR4131 - I ended up getting one of these for a price just on the edge of my threshold of pain. So far it has proven to be a valuable tool! The only problem with it is that it has a rather severe frequency drift issue - while I watch, peaks generated by otherwise fairly stable oscillators move off the screen. Could be a bad oscillator or something. Unfortunately I don't have any calibrated test equipment to attempt to resolve this at the moment, but fortunately it's not a big deal for what I'm doing right now.
Apparently this is a known design issue with the TR4131. Rather than using a single reference oscillator (e.g. a 10 MHz OCXO), it has a couple of independent non-compensated oscillators. I have a grand plan to build an OCXO-based master oscillator that replaces all of the bits. A quick look at the schematics shows that this should be reasonably straightforward to integrate, though I haven't cracked the box open to see if it can be done in an easily reversible fashion.
Home Brew Radio Projects has some very well written information and designs for a range of RF topics.