What does the covariance matrix look like?
date:14/01/09
Runs of Olivier (13-14 January 2009):
Comparison of the different nSide:
For the bolometer 143-5.
Different nSide = [128,256,512,1024,2048]
CONCLUSION:
It seems that the aberant values for high nSide is due to the parameter nPixelsRange set to 50000 for the results below.
While building the matrix, the program works out the contribution to the matrix of every pair of sample.
But, as the number of pixel may be too important (arrays would be to big for a proc),
the Pixel range is cut in pieces of nPixelsRange and the program loop over these pieces of the map.
It seems that the program is corrupted if nPixelsRange is smaller than the total number of pixel
(means that the cut into pieces isn't working for now).
The results are false for nSide=128 (and bigger value) because with nPixelsRange = 50000, it is smaller than 12*128*128 = 196608.
You can see that on the plot of the diagonal profil of the matrix which is not regular (see diagonal profil plot below).
If nPixelsRange is set to 200000. it should be OK with nSide=128 but not with higher value of nSide. (follow the link bellow.)
For other results see nPixRange
- Caracteristics of the data:
dmc_output_path = /data/dmc/dmc_logs/m2db/2009-01-13/30690859_proc_mrun.out
mpi_Proc = 49
Calibration1 = 2.26281416757e-13
polarizerAngle1 = 65.7399742019
endRing = 13098
beginRing = 580
BoloID1 = 10_143_5
IMOFile = 15710361
AstroFrame = GALACTIC
thresholdForStart = 5e-25
thresholdForConvergence = 1e-34
nPixelsRange = 50000
nSideForPolarizer = 128
nSideForIntensity = 128
number_of_Offsets = 1
number_of_pointingStruct = 1
number_of_orbitalDipole = 1
number_of_noiseWeights = 1
number_of_Signal = 1
Object name:
name = 'object:/space/dmc/m2db/DATA:bench6_VECT_Offs_Cov:143-5_covariance_580-13098_49__W_TauDeconv_m2db_NOV08_noPS_dest0128'
Covariance matrix will be diplayed for the different nSide
Caracteristic values of the matrix:
min= -5.5098573e-08
max = 2.2301586e-05
mean= 7.3862828e-10
| Covariance matrix between the ring 0 and 5000 | Covariance matrix between the ring 0 and 5000 |
 |  |
| The full Covariance matrix after a decrease of the resolution (by a factor 3 with the methode OnePoint) | The full Covariance matrix after a decrease of the resolution (by a factor 3 with the methode Avg) |
 |  |
| Profil of the diagonal of the Covariance matrix | Profil of the antidiagonal of the Covariance matrix (logscale) |
 |  |
| Profil of the antidiagonal of the Covariance matrix (logscale, zoom around the diagonal intersection) |
 |
Caracteristic values of the matrix:
min= -1.6251480e+29
max = 1.6707348e+29
mean= -2.1961203e+21
| Covariance matrix between the ring 0 and 5000 and scaled between -1 and 1 | Covariance matrix between the ring 0 and 5000 and scaled between -2e-7 and 2e-7 |
 |  |
| The full Covariance matrix after a decrease of the resolution (by a factor 3 with the methode OnePoint) and scaled between - 4e-7 and 4e-7 |
 |
the Average method give the same results
Caracteristic values of the matrix:
min= -9.1659052e+21
max = 8.9744066e+21
mean= 4.1663334e+13
| Full Covariance matrix after a decrease of the resolution (by 3 with method OnePoint) and scaled between -40e-8 and 40e-8 | Full Covariance matrix after a decrease of the resolution (by a factor 3 with the methode OnePoint) and scaled between - 4000e-8 and 4000e-8 |
 |  |
Caracteristic values of the matrix:
min= -1.7317642e+30
max = 1.7033970e+30
mean= 6.0627078e+21
The full Covariance matrix after a decrease of the resolution (by a factor 3 with the methode OnePoint)
and scaled between -1 and 1
Caracteristic values of the matrix:
min= -2.0993759e+50
max = 2.7784859e+50
mean= -8.6867845e+41
The full Covariance matrix after a decrease of the resolution (by a factor 3 with the methode OnePoint)
and scaled between -1 and 1
