*Filter Effects on TS PSDs*
I've plotted the psd of 5 good channels after each filter step in the XMMJ1230 reduction. I then average the signal in two bands; the knee band ( 1 Hz < knee band < 3 Hz) and the narrowband ( 12 Hz < narrow < 18 Hz). Reminder:
1. The polysecant function simultaneously fits a 30th order polynomial and a secant to the time stream data for a set of 20 circles (TS data spans 80 seconds). This filter acts on a single timestream
2. The removeannularmean function acts at each time step. For each bolometer, the mean signal outside a 2.5' and within a 5.5' radius is subtracted from the signal. The bolometer signals are gain-weighted using the daily planet calibration data. This function uses information from multiple bolometers and is a possible point of contamination from noisy channels.
Figure 1 Raw Timestream Data. Narrowband noise: 3.15e-11 A/sqrt(HZ)
Figure 2 Before 30th order polysecant. Narrowband noise: 3.15e-11 A/sqrt(HZ)
Figure 3 After 30th order polysecant. Narrowband noise: 2.57e-11 A/sqrt(HZ)
Figure 4 Before 1st iteration of removeannulartemplate. Narrowband noise: 2.56e-11 A/sqrt(HZ)
Figure 5 1st iteration of removeannulartemplate. Narrowband noise: 2.37e-11 A/sqrt(HZ)
Figure 6 Before 2nd iteration of removeannulartemplate. 2.37e-11 A/sqrt(HZ)
Figure 7 After 2nd iteration of removeannulartemplate. 2.82e-11 A/sqrt(HZ)
Conclusions:
Figures 4,5,6,7 show that some noise features are being mixed between channels. In figures 4 and 5 the 21Hz line gets mixed from neighbouring channels to the good channels. In figures 6 and 7, other frequency peaks get mixed in ( 7 Hz for example ). Also in figure 7 the blue channel shows a slight increase in the 1/f region.
A look at Scan 89752 This scan was taken on Dec 23, with a PWV of 4.5mm (compared to the above scan 118 taken on Dec. 31 with a PWV of 0.4 mm) Figure 8 Before 1st removeannulartemplate. Narrowband noise A/sqrt(Hz)
Figure 9 After 1st removeannulartemplate. Narrowband noise A/sqrt(Hz)
Figure 10 After 2nd removeannulartemplate. Narrowband noise A/sqrt(Hz)
Conclusions:
Figures 8,9,10 show an increase in the psd amplitude after the 2nd removeannular template. Here there doesn't appear to be any specific features introduced into the spectrum (as shown above), but the noise level does increase. The second iteration of removeannulartemplate increases the average narrowband noise between 10 - 15% for these two scans.
The second iteration of removeannulartemplate has the potential to have less channels in the annular calculation. This second iteration could be more sensitive to noise in the timestreams (It could possibly be more sensitive to small noise than the first iteration is to large noise in channels) if the number of channels in the fit is drastically lowered. Removeannulartemplate is currently hard-coded to cut any channels with less than 6 channels in its fit. It appears that the number of channels don't change much between the two iterations of the removeannulartemplate Figure 11 Number of channels included in the 1st removeannulartemplate. (channel 179, scan 89752) 34 channels in the fit.
Figure 12 Number of channels included in the 2nd removeannulartemplate. (channel 179, scan 89752). 31 channels in the fit.
This topic: APEX_SZ > WebHome > XMMJ1230 > FilterEffectsOnTSPSDs Topic revision: r6 - 2010-02-01 - JamesKennedy
I've plotted the psd of 5 good channels after each filter step in the XMMJ1230 reduction. I then average the signal in two bands; the knee band ( 1 Hz < knee band < 3 Hz) and the narrowband ( 12 Hz < narrow < 18 Hz). Reminder:
1. The polysecant function simultaneously fits a 30th order polynomial and a secant to the time stream data for a set of 20 circles (TS data spans 80 seconds). This filter acts on a single timestream
2. The removeannularmean function acts at each time step. For each bolometer, the mean signal outside a 2.5' and within a 5.5' radius is subtracted from the signal. The bolometer signals are gain-weighted using the daily planet calibration data. This function uses information from multiple bolometers and is a possible point of contamination from noisy channels.
Figure 1 Raw Timestream Data. Narrowband noise: 3.15e-11 A/sqrt(HZ)
Figure 2 Before 30th order polysecant. Narrowband noise: 3.15e-11 A/sqrt(HZ)
Figure 3 After 30th order polysecant. Narrowband noise: 2.57e-11 A/sqrt(HZ)
Figure 4 Before 1st iteration of removeannulartemplate. Narrowband noise: 2.56e-11 A/sqrt(HZ)
Figure 5 1st iteration of removeannulartemplate. Narrowband noise: 2.37e-11 A/sqrt(HZ)
Figure 6 Before 2nd iteration of removeannulartemplate. 2.37e-11 A/sqrt(HZ)
Figure 7 After 2nd iteration of removeannulartemplate. 2.82e-11 A/sqrt(HZ)
Conclusions:
Figures 4,5,6,7 show that some noise features are being mixed between channels. In figures 4 and 5 the 21Hz line gets mixed from neighbouring channels to the good channels. In figures 6 and 7, other frequency peaks get mixed in ( 7 Hz for example ). Also in figure 7 the blue channel shows a slight increase in the 1/f region.
A look at Scan 89752 This scan was taken on Dec 23, with a PWV of 4.5mm (compared to the above scan 118 taken on Dec. 31 with a PWV of 0.4 mm) Figure 8 Before 1st removeannulartemplate. Narrowband noise A/sqrt(Hz)
Figure 9 After 1st removeannulartemplate. Narrowband noise A/sqrt(Hz)
Figure 10 After 2nd removeannulartemplate. Narrowband noise A/sqrt(Hz)
Conclusions:
Figures 8,9,10 show an increase in the psd amplitude after the 2nd removeannular template. Here there doesn't appear to be any specific features introduced into the spectrum (as shown above), but the noise level does increase. The second iteration of removeannulartemplate increases the average narrowband noise between 10 - 15% for these two scans.
The second iteration of removeannulartemplate has the potential to have less channels in the annular calculation. This second iteration could be more sensitive to noise in the timestreams (It could possibly be more sensitive to small noise than the first iteration is to large noise in channels) if the number of channels in the fit is drastically lowered. Removeannulartemplate is currently hard-coded to cut any channels with less than 6 channels in its fit. It appears that the number of channels don't change much between the two iterations of the removeannulartemplate Figure 11 Number of channels included in the 1st removeannulartemplate. (channel 179, scan 89752) 34 channels in the fit.
Figure 12 Number of channels included in the 2nd removeannulartemplate. (channel 179, scan 89752). 31 channels in the fit.
This topic: APEX_SZ > WebHome > XMMJ1230 > FilterEffectsOnTSPSDs Topic revision: r6 - 2010-02-01 - JamesKennedy
© 2020 Winterland Cosmology Lab, McGill University, Montréal, Québec, Canada