1. Apex scanning strategy & map-making
A. Amblard & M. White

2. We simulated the observation from Atacama site : longitude : - 67˚45’
latitude : - 23˚1’
We choose WFS
Observation time per day in equatorial coordinates for
an elevation above 30 degrees

3. Good time to observe : the beginning of the year because the sun is under the horizon when WFS is at least 30 degrees above
January 2004 : elevation of WFS (black), elevation of the sun (blue),
& angular distance (-100 degrees) between WFS and the sun

4. 3 scanning-strategy tried :
Pure drift scanning
Drift scanning + sweeping of secondary mirror
Drift scanning + sweeping of the telescope
24 minutes of drift then repointing
Speed on the sky 0.25’/s
10’ peak to peak at 0.5 Hz
Speed on the sky ~ 10’/s
2 degrees peak to peak at 50 mHz
Speed on the sky ~ 12’/s

5. Drift Scanning
The coverage is very homogenous, with a dispersion of about 1%

6. Fast sweeping
The coverage is very inhomogenous, with a dispersion of about 40%

7. Slow sweeping
The coverage is quite homogenous, with a dispersion of about 10%

8. 8 hours for 90 days simulated with a 10 Hz, about 25 millions
Simulations
Main characteristics :
8 hours for 90 days simulated with a 10 Hz, about 25 millions
samples, only about 20 millions (enough to cover a 6x6 degrees field around WFS center) used for map-making
2 components introduced :
SZ signal obtained with M. White simulations (2048x2048
pixels map of 0.2’ resolution
1/f gaussian noise with a spectrum model
with σ= 280 μK/√Hz, fknee=10 mHz, α=-2
the noise is generated in fourier space in 1 chunk
of 20 millions elements

9. SZ map input
The original map has been smoothed to 0.8’ and centered on WFS

10. Position of the SZ signal in frequency space
Beam effect
due to low speed on the sky

11. SZ signal in the timeline
Fast sweeping
Pure drift
Slow sweeping
After a 0.1 Hz higpass filter
the SZ signal has almost
disappeared

12. Map-making using MADmap
Pure drift :
The map-making converge to a striped solution, adding a CMB and dust component
or a second detector does not improve the convergence

13. Fast sweeping The map-making converge to a reasonable solution,
we recognize the principal clusters on the map

14. Slow sweeping The map-making converge also to a reasonable solution
(in fact faster by a factor 5),
& we recognize the principal clusters on the map

15. Noise properties
Using MADCAP, we performed again the map-making and obtain the pixel-pixel
noise correlation matrix for the sweeping scan-strategies
Fast sweeping
Slow sweeping
Histograms of the diagonal element of the noise correlation matrix
As seen on the observed time per pixel, the noise level is more
homogenous in the slow case (8% against 20% of dispersion)

16. Pixel-pixel noise correlation
Fast sweeping
Slow sweeping
3 rows of the noise correlation matrix (first,middle and last) showing how the pixels
are correlated with each other
The slow sweeping have the lowest level of correlation due to a better mixing of the
time correlation on the sky

17. Decrease of the correlation
auto-correlation
first neighbours
On average the correlation decrease faster for the slow sweeping than for the fast,
there is a factor 5 between their decrease on the first neighbours.

18. Conclusions avoid the pure drift scanning strategy, its speed on
the sky is slow (speed on the sky should bring the
signal in good frequency range);
slow sweeping is better than fast one : it seems
better to increase the amplitude than the frequency
but imply to move the telescope ;
with slow sweeping correlation could be small, but
what if the noise is more correlated ?