By now some of you have already discovered new data on the data site. One of the newer sets is a capture on the "strongest" pulsar in the galaxy, called psrb0329+54.
This pulsar is very close by and is so strong that you can see the pulses right in the raw data! The pulses have a duration of ~0.71452 sec. If you perform Fourier analysis of this, you may see evidence of these pulses, but that is not the best approach (to see the pulsar).
The pulsar can most easily be seen by using the following algorithm:
Taking the first 0.71452 seconds worth of data, squaring the data values and store this in an array. Then take the next 0.71452 seconds worth of data, square it, and added it (element by element) to the first array. Continue this process of squaring and adding, until you see the pulsar peak rise up out of the noise. Shouldn't take long.
If you perform this processing over very long time periods you may find the pular peak will broaden because the duration (0.71452 sec) is not precisely correct. Can you obtain a better value?
Of course, we're looking for SETI signals here, so regular frequency power spectrum analysis is good for that.
Following my Frequency Domain analysis of this Pulsar, I used the figures regarding the Pulsar rate that I obtained using this method (see the Pulsar Detection region of the Wiki, for the description of this algorithm) to undertake an interesting study.
As previously noted that the data that I had used to do the Frequency Domain analysis had been lowpassed filtered at 128 cycles/Setisec (A Setisec is a unit of time which is defined to simply be 2^23 samples of the signal at the standard sample rate and is approximately 0.96 seconds); and then decimated down to 4096 samples per Setisec.
Having worked out that the Pulsar Period was 1.344 cycles per Setisec I again resampled the previous data set such that I had exactly 4096 samples in 1 pulsar period.
Thanks to the original low-pass filter there is no need of any fancy multirate processing here, when a new sample position lies somewhere between two original sample position, the new value can simply calculated by linear interpolation - nice and fast. In doing this I ended up with a data vector containing exactly 163 Pulsar cycles, I discarded any surplus samples.
I then rearranged the data into a 2 dimensional array each row being exactly one Pulsar cycle long (Really fast if you use the Matlab/Octave reshape function). This array is in an ideal form for undertaking Synchronous Integration, all we need do is find the mean value of each column, and we get the Profile of the mean Pulsar Pulse in the array. The results are shown below: -
Rather a good looking plot clearly showing the two side lobes. Bearing in mind that this was derived only using 1 of the files for this Pulsar, the results even astound me.
That is outstanding analysis! The pulsar profile plot is beautiful. It must have been fun when it first appeared with such high SNR.
Just like Rob, I'm impressed with the analysis. Nice work!