LMX2492: Spur performance.

I am testing the LMX2492 eval board using the USB2ANY interface and TICS-Pro Ver1.6.8.0 for control.

I am using an external Wenzel 100 MHz OCXO as the master reference. The onboard reference is disabled (powered down).

The observed phase noise is quite good with the external OCXO used. No problems here.

However, I am seeing quite modest spur performance. I have tested at 9.6, 9.8, and 10 GHz, observing the Fo/2 port on a spectrum analyser.

These frequencies were chosen deliberately to be at the "boundary spur" (N x Fpfd) frequencies

Of course, the spectrum is clean when Fo is to the exact 9.6, 9.8, and 10.0 GHz frequencies.

Using a 100 kHz step at Fo/2  (N div fraction = 500), the first two spurs at +/- 100 kHz and +/- 200 kHz are at -29 dBc and -23 dBc, at the Fo/2 port.

I have also tried using a larger N DIV Fractions of 5000 and 50000, to no great effect. Also, the FRAC Dither has essentially no effect.

This seems very modest spur behavior. I have also noticed that it can be quite sensitive to the REF Pin value. I am using +11 dBm (2.2Vpp) nominal.

I have tested with frac order 1, 2, 3, and 4. Frac order 1 is not usable, and order 2 and 3 look quite similar, with frac order 4 simply giving more spurs.

Am I missing something here? I have read through several of the TI application notes, with AN-1879 seeming to be most relevant.

Thank you.