I am doing some phase noise analysis on a new design using the AD9915. It seems to me that the phase noise curves provided in the datasheet (Rev F) don't add up. I would expect the absolute phase noise to be the sum of the scaled reference source and the residual phase noise. This is not the case according to the graphs (Figures 14, 16, 17). As an example, consider the phase noise (PN) values at 1 MHz:
Reference source PN @ 2.5 GHz: -144 dBc/Hz, read from Figure 14
Reference source PN, scaled to 978 MHz: -152.2 dBc/Hz, calculated and confirmed by Figure 16
DDS residual PN @ 978 MHz: -150 dBc/Hz, read from FIgure 17
Calculated absolute PN @ 978 MHz (sum of the above): -147.9 dBc/Hz
Absolute PN @ 978 MHz according to datasheet: -144 dBc/Hz, read from FIgure 16
The difference is almost 4 dB, which can't be right. As a check, I performed the same calculations on the AD9914 (Datasheet Rev. F). In that case, the figures seemed correct.
It is important to me to used the correct values of the residual phase noise in my analysis, but I can't trust the values in the datasheet. Please confirm which are the correct values.