Annu. Rev. Astron. Astrophys. 1992. 30:
653-703 Copyright © 1993 by Annual Reviews. All rights reserved |
3.2.2
Two different observational programs were carried out on this
telescope, designated the ``North Celestial Pole (NCP) experiment''
and the ``RING experiment''. These are described separately below.
The NCP program
which is consistent with zero - the expected mean value. It is
remarkable that this result has been obtained from the raw data
without any subtraction of means, allowances for drifts, etc.
This gives great confidence that systematic errors have been
successfully removed down to about 10 µK.
The results
of these observations are shown in
Figure 1b. It was clear from
the first series of observations that one of the fields (NCP 7) was
badly contaminated by a confusing discrete source, but observations
were continued on this field since they gave some confidence that the
system was working correctly. The confusing source in this field has
been identified with observations at the VLA. For the purposes of
estimating the intrinsic anisotropy this field was discarded and only
the remaining seven fields were used. Readhead et al carried out a
number of tests to ascertain the reliability of their data, and
concluded that there were no serious problems with the data set. As
the dataset has 2
1, statistical analyses
based on Bayesian methods and on the likelihood ratio test yield
similar upper limits, which are given in
Table 3. These are the most
stringent limits obtained to date on angular scales of a few arc
minutes. They definitively rule out adiabatic fluctuations in
baryonic matter as the primary agents of galaxy formation with
standard recombination. They also place interesting limits on many
theories of galaxy formation
(Readhead et al 1989,
Vittorio & Muciaccia
1991,
Bond & Myers 1991a).
The RING program
The 96 fields were observed with the same setup as the NCP program,
except that here the individual fields were observed for twelve
minutes centered on upper culmination each day. The results of the
observations on these 96 fields are shown in
Figure 1c. A number
of fields stand out. Convolution with a matched filter, which affords
the maximum signal-to-noise-ratio for detecting signals with a given
signature in noise, results in seven fields which differ from zero by
more than 3 . VLA observations
at 1.4 GHz and 8.5 GHz
revealed that four of these fields are contaminated by discrete
sources. The remaining three fields contain no known sources which
could be responsible for the observed signal. It is possible,
however, that these fields are contaminated by sources with flat or
inverted spectra, or variable sources. A detailed analysis of the
level of the ``excess variance'' in the fields shows that it is
unlikely, but by no means impossible, that this is entirely due to
contamination by discrete sources
(Myers et al 1991,
1992).
For this reason the RING results are interpreted as upper limits rather than
detections of intrinsic anisotropy. There is clearly a danger, in
this approach, of misidentifying real intrinsic fluctuations as due to
foreground objects, and the fields in which significant ``signals''
were detected will be studied further to ascertain the origin of these
signals.
The upper limit derived from the RING data after
subtraction of the effects of the identified confusing sources is
given in Table 3. This is about a
factor 2.5 higher than the limit
derived from the NCP program for Gaussian fluctuations, but it
provides more stringent limits on models with non-Gaussian
fluctuations.
The design of the interlocking fields in the
RING program was chosen specifically to provide an independent test of
the experimental procedure. Whereas in the case of the NCP program we
expect the average of a sufficient number of fields to approach zero
in the case of the RING program the true sky mean must be zero.
Thus a stringent test of the observing procedure is provided by
averaging all 96 fields. The result of this average is:
The fact that the value of < T/T >
for the raw data lies within 1
of zero when no zero
levels or drifts have been subtracted is a powerful demonstration of
freedom from systematic errors.