We present 25 new eclipse times of the white dwarf binary NN Ser taken with the high-speed camera ULTRACAM on the WHT and NTT , the RISE camera on the Liverpool Telescope , and HAWK-I on the VLT to test the two-planet model proposed to explain variations in its eclipse times measured over the last 25 years .
The planetary model survives the test with flying colours , correctly predicting a progressive lag in eclipse times of 36 seconds that has set in since 2010 compared to the previous 8 years of precise times .
Allowing both orbits to be eccentric , we find orbital periods of 7.9 \pm 0.5 yr and 15.3 \pm 0.3 yr , and masses of 2.3 \pm 0.5 \mathrm { M } _ { \mathrm { J } } and 7.3 \pm 0.3 \mathrm { M } _ { \mathrm { J } } .
We also find dynamically long-lived orbits consistent with the data , associated with 2:1 and 5:2 period ratios .
The data scatter by 0.07 seconds relative to the best-fit model , by some margin the most precise of any of the proposed eclipsing compact object planet hosts .
Despite the high precision , degeneracy in the orbit fits prevents a significant measurement of a period change of the binary and of N -body effects .
Finally , we point out a major flaw with a previous dynamical stability analysis of NN Ser , and by extension , with a number of analyses of similar systems .