Least-Squares

class ptp.ls.Ls(N, data, T_ns=inf)

Least-squares Time Offset Estimator

Parameters

• N – observation window length (number of measurements per window)

• data – Array of objects with simulation data

• T_ns – nominal time offset measurement period in nanoseconds, used for debugging only. It is used to obtain the fractional frequency offset y (drift in sec/sec) when using the efficient LS implementation, since the latter only estimates y*T_ns (drif in nanoseconds/measurement). In the end, this is used for plotting the frequency offset.

process(impl='eff', batch_mode=True, batch_size=4096)

Process the observations

Using the raw time offset offset measurements and potentially also the Sync arrival/departure timestamps, estimate the time and frequency offset corresponding to each window of samples.

Parameters

• impl –

  Least-squares implementation. Choose between the following:

  ”t2”Uses timestamp “t2” when forming

  the observation matrix H.

  ”t1”Uses timestamp “t1” when forming

  the observation matrix H.

  ”eff” (default)Computationally-efficient

  implementation.

  ”eff-vec”Computationally-efficient

  and vectorized implementation.

• batch_mode – Whether to process observation windows in batches, rather than trying to process all windows at once. Especially important for the vectorized efficient (eff-vec) implementation.

• batch_size – Number of observation windows that compose a batch.