John- I do agree with you. The pain is in the early stages of the LO chain. Since I'm planning to use QRSS CW (that's very slow speed Morse Code with very narrow demod bandwidths for non-hams on this reflector who may not be familiar with QRSS), the noise very close to the carrier become key.
A noisy carrier can't be detected very well on a waterfall display if it too, looks like noise. I have a several MSA-1105 MMICs in the chain to provide isolation and give gain while multiplying the 5MHz signal to 10MHz and then to 20MHz. The 20MHz acts as a reference for a 1320MHz PLL. The 1320MHz then is multiplied several more times on it's way to a sun-harmonic mixer for 630GHz. I am wondering if the MSA-1105s could be causing more close-in noise than I expected. The CW note sounds a bit "rough" by ear. The final 600GHz mixer that I'm using comes from some of the mm-wave boys that work with NRAO. But my LO noise requirements are a bit different than their needs. -------------- Original message from "John Miles" <[email protected]>: -------------- > The painful part is probably the first few stages, if you are starting at 5 > MHz. You probably want to do some HP 8662A-like tricks using crystal > filters to shave off the broadband noise below 1 GHz, and maybe SAW filters > above that. This will do nothing for noise within 1 kHz, though... do you > really need a clean signal that close to the carrier all the way up to 630 > GHz? > > The noise characteristics of the MMICs seems to depend a lot on the fab > technology. I can't seem to find my .PDF copy of it right now, but I have > one paper on microwave regenerative dividers where the authors measured the > residual PN of several contemporary parts driven to saturation. At 4.5 GHz, > the 10 dB/decade corner frequency wasn't reached until past 100 kHz for the > Stanford Microdevices SGA-4186, which didn't speak well for the PN > performance of SiGe HBT parts. They showed -143 dBc/Hz at 1 kHz for that > one. > > The GaAs HBT part (Mini-Circuits ERA-5SM) they tested was among the best > (-156 dBc/Hz at 1 kHz). Second-worst was an InGaP/GaAs HBT part (Stanford > NGA-489) at about -153 dBc/Hz at 1 kHz. Still much better than the SiGe > part. > > My understanding is that the newer GALI-series parts from Mini-Circuits are > InGaP HBT devices so they'd presumably perform about like the NGA-489. > You'd want to measure them to make sure, though, if your app is that > critical. > > Take a look at NRAO's recent publications, especially those associated with > the ALMA array (many of which are on their site). They're doing the real > bleeding-edge work at sub-mm these days. > > -- john, KE5FX > > > -----Original Message----- > > From: [email protected] [mailto:[email protected]]on > > Behalf Of [email protected] > > Sent: Monday, December 15, 2008 8:54 AM > > To: [email protected] > > Subject: [time-nuts] Close-in phase noise question... > > > > > > Looking for comment here... > > > > The background: > > I'm working on a sub mm-wave LO chain for > > a ham radio application. While chasing issues > > of close-in phase (ie: within 1KHz of RF > > carrier) by peeling the "layers of the onion", > > I'm starting to question the performance of > > the MMICs that are used as buffers and amps > > following my Wenzel reference OCXOs. > > > > Question(s): > > Should any MMIC be allowed to be driven > > close to compression or into compression > > when striving for best close-in noise? > > > > I know and have seen the NF of a MMIC > > degrade while in compression, but my > > target right now is close-in noise rather > > than broadband noise. > > > > My design, in summary, takes 5MHz up to 630GHz > > via several multipliers and PLL stages. > > > > -Brian > > > > > > > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
