C. Palomba - DA Meeting Update on the analysis of VSR1 data, focusing attention on the presence of spectral disturbances. More data more details can be seen We are interested in identifying disturbances of clear or likely instrumental origin in order to remove them. This brings to a reduction in the number of candidates found in the analysis (and possibly allows for a reduction in the threshold for candidate selection). Commissioning and noise people can be also interested. Analysis of spectral lines in VSR1 pulsar search F. Antonucci, P. Astone, S. D’Antonio, S. Frasca, C. Palomba
2 calibrated data Data quality SFDB Average spect rum estimation peak map hough transf. candidates peak map candidatescoincidences coherent step events Data quality SFDB Average spect rum estimation hough transf. Scheme of the DA pipeline The analysis/cleaning of the spectral disturbances is done at the level of peak maps. We will also see how they reflect in candidate distribution The peak map is built from the ratio between periodograms and the corresponding estimations of the average spectrum, selecting local maxima above a given threshold We look at the frequency distribution of peaks
3 Three main categories of disturbances: - ‘known’ Virgo lines (i.e present in the list); - lines of likely instrumental origin (mainly harmonics of a set of frequencies); - lines of unknown origin. persistency
4 Some examples of ‘known’ Virgo lines Mainly violin modes and calibration lines
Hz.584Hz Hz Small disturbances ‘accumulates’ in time and clearly emerges in the total peak frequency distribution Details not visible looking at a few days of data
6.2Hz Hz.585Hz 1.0Hz Hz
7 Harmonics of 0.333Hz: up to ~60 Hz (sidebands of 1Hz lines?) 1.0Hz: up to ~200 Hz Hz: mainly in the ranges Hz, Hz, Hz, Hz Hz: mainly in the range Hz 10Hz: nearly everywhere in 0-2kHz (but with decreasing persistency) Hz Hz: 55 harmonics spread over the whole band Hz Hz Hz
8 Even after removing lines from the Virgo known line list, there are residual disturbances
9 This happens because lines with rather small amplitude are not detected by the line monitor, but if they are persistent enough we anyway find them in the peak frequency distribution. Then, a further cut has been applied at the level of each peakmap, but still there are small but clear residuals in the total peak map, which produce candidate excess.
Hz Hz Hz Hz 498.8Hz Hz ~1171Hz (4 lines) Hz ~1726Hz (triplet separated by.596Hz) Hz Hz residuals of violin modes residuals of 1Hz and harmonics of 0.5Hz (not always present)
11 Higher order violin modes have become visible BS 9 th order? BS 11 th order?
12
13 Even a small peak excess can produce a large excess in the number of candidates. This is due to the fact that each disturbed frequency bin affects all the search frequency within a Doppler band range around it.
14 Searching for non-zero spin-down candidates slightly reduces the effect of narrow disturbances Better cleaning by cutting bands around violin modes and calibration lines Find lines to be cleaned on the total peak map histogram Suggestions by commissioning/noise people very welcome Conclusions