An end-to-end software validation has been performed to check that the heterodyning code used in the time domain analysis isn't the cause of spurious phase ofsets seen in the S3 pulsar injections. A signal (eq. 3.3 of the S1 paper) was generated at a sample rate of 16384 Hz with the following parameters:
| RA | 4.88670685367603 |
| dec | -0.217583646376765 |
| f0 | 100 Hz |
| f1 | 0.0 |
| fepoch | 751680013 |
| h0 | 1.0 |
| &phi0 | 2.5 |
| &psi | 0.0 |
| &iota | 1.570796327 |
The signal was heterodyned using the same software that was used in the S2
and S3 analyses. The heterodyned signal (which just varies with the beam
pattern) is shown below along with the model signal (eq. 4.10 of the S1
paper). These 2 signals match.
The parameters were then estimated from the heterodyned signal using the analysis software as used previously. These parameters match what was injected. (In the plots below &phi0 appears about 1-2o off its injected value, firstly i would say that this is nowhere near the size of the discrepancies seen in the hardware injections. It is also expected that the filtering would cause phase shifts of this order and I should have taken this into account by having larger variances when estimating the parameters, leading to wider pdfs that would include the injected value.)
| h0 | &phi0 | &psi | cos&iota |
| 
| 
| 
|
This software injection was repeated with different parameters which were again correctly extracted.