[Edited from original post]

The 'ps' part is the r.m.s time jitter. The 'mV' sounds like the r.m.s noise voltage at the VCO input, but over what bandwidth? My gut feeling is to take this noise bandwidth to be equal to the bandpass jitter bandwidth of the PLL.

For a first order loop i.e; loop filter H(s)=1, the quantity of interest works out to 1/K_pω_c. For a f_c=1 GHz and Kp=0.5 rad/V, the jitter per unit voltage works out to 0.3 ps/mV.

The number is O(mrad/mV) because K_p is typically O(Volt/rad).

M.G.Rajan
www.eecalc.com