We are using a HMC700 to lock a microwave oscillator to a 10 MHz reference in integer mode. I am looking at the Allan deviation of the combination and the effects of the PLL noise floor at very low offset frequencies: 0.0001Hz to 0.01Hz, using a test system based around a Microsemi 5125A. The noise floor rises unexpectedly below about 0.001 Hz and I suspect thermal effects within the PLL are responsible. The HMC700 appears to be the most temperature sensitive of all the PLL components. Applying some heating or cooling to the top of the HMC700 package produces something like 30ps of phase shift per degree C. Is there any information on the effects of temperature on the HMC700 RF dividers and/or phase detector that would be consistent with this? My current guess is a dependency of the charge pump gain on temperature. Are any of the later-generation PLL devices known to be better in this regard? Charge pump current is set to 2mA in this design and, while the loop filter is passive, the charge pump output voltage stays between about 2V and 3V.