Date post: | 14-Apr-2018 |
Category: |
Documents |
Upload: | bbteenager |
View: | 216 times |
Download: | 0 times |
of 25
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
1/25
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
2/25
X-ray observations of the Sun by RHESSI
background
Signal
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
3/25
T ~ 3K
The CMBR is very uniform
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
4/25
CMBR: all-sky temperature map from WMAP satellite (Feb 2003).
Temperature fluctuations of about ,
~380,000 years after the Big Bang
K10 5
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
5/25
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
6/25
Poisson statistics
For much of the E-M spectrum astronomical observations involve countingphotons. However, the number of photons arriving at our telescope from a
given source will fluctuate.
We can treat the arrival rate of photons statistically, which (roughlyspeaking) means that we can calculate the average number of photons
which we expect to arrive in a given time interval.
We make certain assumptions (axioms):
If our observed photons satisfy these axioms, then they are said to follow
a Poisson distribution
1. Photons arrive independently in time
2. Average photon arrival rate is a constant
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
7/25
Suppose the (assumed constant) mean photon
arrival rate is photons per second.
If we observe for an exposure time seconds,
then we expect to receive photons in that time.
We refer to this as the expectation value of the
number of photons, written as
R
R
RNNE ==)( (4.1)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
8/25
Suppose the (assumed constant) mean photon
arrival rate is photons per second.
If we observe for an exposure time seconds,
then we expect to receive photons in that time.
We refer to this as the expectation value of the
number of photons, written as
If we made a series of observations, each of time seconds, wewouldnt expect to receive photons every time, but the
average number of counts should equal
(in fact this is how we can estimate the value of the rate )
R
R
RNNE ==)( (4.1)
RN =
N
R
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
9/25
Given the two Poisson axioms, we can show that the
probability of receiving photons in time is
given byN
( )!
)(N
eRNp
RN
= (4.2)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
10/25
( )!
)(N
eRNpRN
=
N
)(Np
Poisson statistics
5.0=R
0.1=
R
0.5=R
As increases, the shape of the Poisson distribution becomes
more symmetrical (it tends to a normal, or Gaussian, distribution)
R
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
11/25
Poisson statistics
For the purposes of A2, all we need to work with is the mean or
expectation value of , and its variance.
We already defined the expectation value as
We can compute the mean value of using eq. (4.2),
and this confirmseq. (4.1).
N
RNE =)(
N
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
12/25
For the purposes of A2, all we need to work with is the mean or
expectation value of , and its variance.
We already defined the expectation value as
We can compute the mean value of using eq. (4.2),
and this confirmseq. (4.1).
We can also define the varianceof , which is a measure of the
spread in the distribution:
Poisson statistics
N
RNE =)(
N
N
( ) [ ]22 )(var NENEN == (4.3)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
13/25
For the purposes of A2, all we need to work with is the mean or
expectation value of , and its variance.
We already defined the expectation value as
We can compute the mean value of using eq. (4.2),
and this confirmseq. (4.1).
We can also define the varianceof , which is a measure of the
spread in the distribution:
We can think of the variance as the mean squared error in
Poisson statistics
N
RNE =)(
N
N
( ) [ ]22 )(var NENEN == (4.3)
N
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
14/25
For a Poisson distribution, the variance of can be shown to be
and the standard deviation of is
In practice we usually only observe for one period of (say)
seconds, during which time we receive (say) a count ofphotons.
We estimate the arrival rate as
Poisson statistics
N
RN =)(var (4.4)
N R=
obsN
obs N
R=
(4.5)
(4.6)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
15/25
obsN
We take as our best estimate for with error
i.e we quote our experimental result for the number count ofphotons in time interval as
Poisson statistics
obsN N obsN
obsN
(4.7)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
16/25
Usually there are several sources of noise in our observation,
each with a different variance.
Probability theory tells us that, if the sources of noise are allindependent, then we work out the total noise by adding
together the variances:
Adding Noise
2
other
2
Poisson
2
total += (4.8)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
17/25
Usually there are several sources of noise in our observation,
each with a different variance.
Probability theory tells us that, if the sources of noise are allindependent, then we work out the total noise by adding
together the variances:
Sources of Poisson noise:
1) fluctuations in photon count from the sky
2) dark current: thermal fluctuations in a CCD
Adding Noise
2
other
2
Poisson
2
total += (4.8)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
18/25
Usually there are several sources of noise in our observation,
each with a different variance.
Probability theory tells us that, if the sources of noise are allindependent, then we work out the total noise by adding
together the variances:
Sources of non-Poisson noise:
1) Readout noise. e.g. a CCD can gain/lose electrons
during readout. Usually constant
Adding Noise
2
other
2
Poisson
2
total += (4.8)
=Readout
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
19/25
Correcting for combined quantum efficiency of telescope and
detector
Noise and Telescope / Detector Design
0tot
h
ASN
(4.11)
Fraction of incident photons that
produce a response in the detector
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
20/25
Suppose we observe a point source, of flux density ,
through a telescope with collecting area for time , in
bandwidth centred on .
Total energy collected by detector (ignoring any absorption)
No. of photons collected is
Noise and Telescope / Detector Design
S
0
= ASEtot (4.9)
0
tot
h
ASN
(4.10)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
21/25
Correcting for combined quantum efficiency of telescope and
detector
Thus and
Noise and Telescope / Detector Design
0tot
h
ASN
(4.11)
Fraction of incident photons that
produce a response in the detector
totPoisson N= ( )2/1
ASNR
(4.12)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
22/25
Correcting for combined quantum efficiency of telescope and
detector
Thus and
Same result true for radio observations , even though we dont
count photons. Here, noise from source and detector electronics
Noise and Telescope / Detector Design
0tot
h
ASN
(4.11)
Fraction of incident photons that
produce a response in the detector
totPoisson N= ( )2/1
ASNR
(4.12)
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
23/25
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
24/25
Eq. (4.12) suggests that we can increase the signal-to-noise-ratio
by increasing the bandwidth of our observation.
Line Sources
This is not necessarily true if weare observing a source which emits
only over a narrow frequency range
e.g. a spectral line.
Increasing beyond the line width
will increase the amount of noise
(from the background continuum)without further increasing the
amount of signal (from the line).
See A2 Theoretical Astrophysics for more on linewidths
frequency
int
ensity
continuum
emission line
line width
7/30/2019 a2oa_sec4aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
25/25
The 2-D image of a faint galaxy observed by a CCD covers 50 pixels. For an
exposure of 5 seconds a total of 104 photo-electrons are recorded by the CCD
from these pixels. An adjacent section of the CCD, covering 2500 pixels, records
the background sky count. During the same exposure time a total of 105 photo-
electrons are recorded from the adjacent section. Show that, after subtractingthe background sky count, the signal-to-noise ratio for the detection of the
galaxy is estimated to be 73.
Calculate the length of exposure required to increase the signal-to-noise ratio to
100.
Example