A nice little oscillator assembly came my way, supposed to generate about 17 dBm at 30 MHz. Nothing special at first glance, but after checking out its internals, it appeared to be worth a more careful look.
A hand-made box, and even more labor intensive assembly work inside. All build by point-to-point wiring, using only the best components available, glass trimmer caps, filters, mica caps – most of these parts are still available today – about 100 USD bom, at least.
After a bit of reverse engineering, here the schematic, a modified Colpitts oscillator. Note: base resistor of 2N5109 is 150 Ohms.
To measure phase noise, connected it to my HP 3585A spectrum analyzer (this is really a great piece of equipment, a bit heavy, but still best of class noise performance and holding this title for the last 35 years….). Connected the oscillator via a 6 dB attenuator, to provide a clean load to the output, rather than dealing with the imperfections of cables, adapters, and the analyzer input.
Quite shocking, all this noise. The green trace shows the analyzer noise floor. Check, and re-check, still a lot of noise. Too much to be true. After 3 hours of tests, found the issue: a defective BNC cable. Center connector was fine, but both shields were non-connected.
A bit more examination of these cable shows their lousy construction. Not bad for 2 dollars a piece, but you get what you pay for…. the shield is not even reaching to the plug – there is a 5 mm gap from the screen end, to the actual plug. So even if all would have been connected fine, the would still be a lot of leaking, from inside out, and outside in.
Notice the BNC plugs – these have a somewhat uncommon construction, the dielectric is covered at the front… not quite according to BNC standard.
Clearly visible, the cold solder joint…. Turns out, both ends were open-circuit at the shield.
Finally, using a good quality BNC cable (also, using LMR-195 double-screened cable). Looking much better. Noise is down -115 dBc at 10 kHz from carrier. It’s good, but not great. I think one could do better, especially, considering all the pricy parts, and high-quality construction. A good target for a Colpitts osciallator would be better than -130 dBc, at 10 kHz separation.
Note the pink trace – this is the bad cable, terminated with a 50 Ohm resistor (with the shield broken at both sides, it is actually a 1 meter wire antenna, with an open-circuit 50 ohm resistor at the end).