20GHz is pretty ambitious, though I've done a few things at 10GHz without fancy test equipment. It helps that there were other folks around doing similar things.
For example, I've built two 10GHz transverters from surplus building blocks. The commercial mixers were found at swap meets, now Ebay has lots of them. Those rigs used rather expensive microwave sources for the LO that required 24V and lots of tender care, but now there are credible alternatives that use cheap mmics and micro-strip filters to multiply up from a 1GHz synthesizer. The preamps were taken from surplus downconverters.
But these radios were built entirely with discrete building blocks and semirigid coax. (Even the plumbing was surplus.)
My current 10GHz transverter was built from a kit.
Most of the transverters I've built, both from kits and from scratch were done without a spectrum analyzer and certainly without a network analyzer. I had a power meter (HP 432A) and a sweeper. That's it. With those two tools you can line up filters, test out amplifiers, and do some rudimentary testing.
Today you can build a useful power meter from parts you have around the house (if you live at Digikey
. (See
http://www.w1ghz.org/new/portable_powermeter.pdf.)
The last transverter build was helped along with a 1970's vintage spectrum analyzer, but it wasn't absolutely necessary.
As for simulation -- certainly no "for hire" designs should rely on "cut-and-try" and if you're doing this for a living you should be using simulation and analysis tools. But the notion that simulation is the only path to success for "one off" or hobby devices is not well supported. Building block approaches work just fine, and between Ebay finds and stuff from catalogs you can build some pretty capable systems without spending a whole lot of money on test equipment. If you build from kits, the test equipment requirements are minimal.
Beyond that, we should remember that there were a whole lot of microwave systems designed and built long before simulators became useful. Hand calculation and rough paper and pencil modeling can go a long way when it comes to "just making something that works." Beyond that, this is hobby time, cut-and-try-and-recycle can be both educational and practical.
So, 20GHz? No firsthand experience here. But for 10GHz you can get along fine without the fancy gear. It isn't easy. It isn't simple. But there is a lot of good stuff on the web to help and there are even a few good kits and suitable-for-home-built designs.