Optimal reference frequency for dedispersion

[mhvk]

@quantumfx pointed to a paper by Pennucci (https://iopscience.iop.org/article/10.1088/0004-637X/790/2/93), which includes a discussion of the optimal reference frequency for dedispersion, which minimizes covariance between DM and time shifts:

1/f_ref**2 = sum weight 1/f**2 / sum weight

Here, the sum is over the frequency bands involved. In principle, the weight includes the spectral shape of the pulsar and gains, etc., which may make this perhaps not so easy to use in practice, though even with constant weight, it should be better than just taking the mean frequency.