Difference between revisions of "Subpixel HFI"

From Planck Legacy Archive Wiki
Jump to: navigation, search
Line 1: Line 1:
 
{{DISPLAYTITLE: Very bright regions where the sub-pixel effect in the SRoll CO foreground templates prevent them to be used for cosmology or astrophysics analysis}}
 
{{DISPLAYTITLE: Very bright regions where the sub-pixel effect in the SRoll CO foreground templates prevent them to be used for cosmology or astrophysics analysis}}
  
<span style="color:red;background:yellow"> editing in progress </span>
 
  
 
The bandpass corrections have been optimized for high latitude regions which implied to reduce the noise of the CO and dust bandpass templates to avoid the introduction of significant correlated noise. The effect is negligible for dust but not for CO in bright regions. We simulated the effect to evaluate its level. The difference between the downgraded (at N<sub>side</sub>=128) CO simulated map and the input full resolution (at N<sub>side</sub>=2048) CO simulated map gives the simulated bias introduced by this downgrading for each bolometer. For each frequency channel, these bias maps are converted to HPRs, multiplied by the corresponding CO bandpass coefficients computed by SRoll using data, and projected onto a frequency map of the bias for the frequency. Those frequency averaged bias maps are displayed in the table ''(a)''.
 
The bandpass corrections have been optimized for high latitude regions which implied to reduce the noise of the CO and dust bandpass templates to avoid the introduction of significant correlated noise. The effect is negligible for dust but not for CO in bright regions. We simulated the effect to evaluate its level. The difference between the downgraded (at N<sub>side</sub>=128) CO simulated map and the input full resolution (at N<sub>side</sub>=2048) CO simulated map gives the simulated bias introduced by this downgrading for each bolometer. For each frequency channel, these bias maps are converted to HPRs, multiplied by the corresponding CO bandpass coefficients computed by SRoll using data, and projected onto a frequency map of the bias for the frequency. Those frequency averaged bias maps are displayed in the table ''(a)''.

Revision as of 13:17, 11 December 2017


The bandpass corrections have been optimized for high latitude regions which implied to reduce the noise of the CO and dust bandpass templates to avoid the introduction of significant correlated noise. The effect is negligible for dust but not for CO in bright regions. We simulated the effect to evaluate its level. The difference between the downgraded (at Nside=128) CO simulated map and the input full resolution (at Nside=2048) CO simulated map gives the simulated bias introduced by this downgrading for each bolometer. For each frequency channel, these bias maps are converted to HPRs, multiplied by the corresponding CO bandpass coefficients computed by SRoll using data, and projected onto a frequency map of the bias for the frequency. Those frequency averaged bias maps are displayed in the table (a).

For convenience, we also provide (b) the maps of the logarithm of the ratio of the bias variance, in each Nside=128 pixel, and the variance of the noise level estimated from the End-to-End simulations. The level of the bias is well determined and specific masks adapted to each scientific analysis can be computed at the appropriate level. The table also gives the sky fraction to be masked for thresholding the bias at 100, 10, and 1 % of the rms of the noise at Nside=2048 inside the pixel at Nside=128 referred to the rms of the noise at Nside=2048 inside the same pixel at Nside=128.

We also provide (c) the ratio of the variance bias maps to the variance of the full intensity signal.

The effects are largest in intensity and at 100 GHz. It can also be noted that, for polarization, and at all frequencies, only less than 0.04% of the sky is affected by this bias at 10% of the noise level. For intensity, the bias is a factor of 3 lower. If the user wants to reduce the bias at 1% of the noise in polarization, less than 2% of the sky will be affected. Users can also refer to the ratio of the absolute value of the bias with respect to the full intensity maps, which is 2 orders of magnitude lower, to build the relevant mask. For polarization, the two criteria are comparable.

As recommended in Planck-2020-A3[1], Galactic science work with the HFI data in these very bright regions should then be done starting from the HPRs (see HERE) at the full resolution.

CO templates
link to PLA 100 GHz 143 GHz 217 GHz 353 GHz
I Q U . I Q U I Q U
(a): frequency averaged bias maps [μK] LVLV 100-I CO Subpix Val.png 100-Q CO Subpix Val.png 100-U CO Subpix Val.png . 217-I CO Subpix Val.png 217-Q CO Subpix Val.png 217-U CO Subpix Val.png 353-I CO Subpix Val.png 353-Q CO Subpix Val.png 353-U CO Subpix Val.png
(b): [math]\log_{10} \left(\frac{\lt \textrm{bias}\gt \textrm{map}}{\lt \textrm{noise}\gt \textrm{map}}\right)[/math] LVLV 100-I CO Subpix.png 100-Q CO Subpix.png 100-U CO Subpix.png . 217-I CO Subpix.png 217-Q CO Subpix.png 217-U CO Subpix.png 353-I CO Subpix.png 353-Q CO Subpix.png 353-U CO Subpix.png
sky fraction above 100% of (b) . <10-4 % <10-4 % 0.0005 % . <10-4 % <10-4 % <10-4 % <10-4 % <10-4 % <10-4 %
sky fraction above 10% of (b) . 0.0112 % 0.0092 % 0.0346 % . <10-4 % 0.0061 % 0.0056 % <10-4 % <10-4 % 0.0005 %
sky fraction above 1% of (b) . 0.4186 % 0.7370 % 1.4318 % . 0.0270 % 0.1643 % 0.1094 % 0.0071 % 0.0361 % 0.0478 %
(c): [math]\log_{10} \left(\frac{\lt \textrm{bias}\gt \textrm{map}}{\lt \textrm{full intensity}\gt \textrm{map}}\right)[/math] LVLV 100-I CO Subpix Val DIV Signal.png 100-Q CO Subpix Val DIV Signal.png 100-U CO Subpix Val DIV Signal.png . 217-I CO Subpix Val DIV Signal.png 217-Q CO Subpix Val DIV Signal.png 217-U CO Subpix Val DIV Signal.png 353-I CO Subpix Val DIV Signal.png 353-Q CO Subpix Val DIV Signal.png 353-U CO Subpix Val DIV Signal.png

References[edit]

(Planck) High Frequency Instrument

Planck Legacy Archive