Well, i guess both statements are correct. Most of us (Poddar and I ), and the leading Crystal Oscillator houses ( Axtal, Morion and Wenzel ) and others know the limits imposed by the crystal and the semiconductor. I have several patents how to do it better (noise canceling technique) but we are victims of the crystal noise (Operating Q at 100 MHz about 70 000) and the transistors which are getting worse as the production is aimed towards more profit The low aging 10 MHz oscillators use 50uW at most in the crystal, phase noise not much better then -160 dB down,and the locked 100, 120, 125 and 128 MHz Crystal oscillators with calculated phase noise approaching -190 dBm. The result will be to have at lest three custom made transistors up to 100 Ghz and low flicker noise. One transistor for 150 MHz Oscillators, then one 400 to 2000 MHz (SAW) and then finally 5000 to 40 Ghz (DRO). Here you need up to 18 dBm output power, KF < 10E-10, and almost 1 W + dissipation . The low noise large signal AM/PM conversion circuits are not yet fully understood . I am supervising 3 PhD's, one on crystal oscillators , one on Meta Material Structure DRO's and one on power amplifier with an additive noise at -170 down relative to the large signal input. So there is money and problems from 400 MHz to 40 Ghz, not easy ... 73 de Ulrich In a message dated 10/30/2016 11:50:20 A.M. Eastern Daylight Time, attila@kinali.ch writes: That's something that has been bothering me lately: Most of your publications are about the noise in UHF and GHz applications, hardly any for the area where most of time metrology happens: at 10MHz and 100MHz. Is this because the sub-100MHz range oscillators are a solved problem and hit the physical limits of what noise optimization can do? Or is it because there is more money to be made in the >100MHz range?