Observations were made with MSRT between January 1985 and December 1993.
The sky area north of declination 
had been divided into 156
fields of view. Basically all adjacent fields were separated by
angular distance. Fig. 1 (click here) shows the layout of the fields in the sky.
Only 6 dishes in the sub-array B are used during a single observation to
reduce the effect of shadowing between any two adjacent elements.
So a complete UV-coverage is obtained by two observations.
For each field of view four observations were performed
to reduce the effect of interference.
The preliminary
calibration for the constant parts of the phases and gains of the array were
made by observing Cyg A for 10 min. before or after the
observation (Nan 1986).
MSRT has been suffering
from various kinds of interference. Usually about 15 percent of the recorded
data were rejected because of the contamination by interference.
In some extreme cases, we found it necessary to repeat the observations.
Fields #54, #57, #115, and #116 have no data yet.
Figure 1: The arrangement of the fields in the Miyun 232
MHz survey
To check the stability of the receiving system, long duration calibration observations of Cyg A were made about once a month.
The description in general about data editing, mapping, CLEAN and self-calibration for each individual field of view was made in paper I. Here we just summarize as following and add some description in detail.
(1) A procedure was employed to remove the effect of interference and compensate the zero-offset errors. Two methods had been developed by Yang et al. (1985) for this purpose.
(2) Calibrations for the constant parts of phases and gains of the whole receiving system were made daily by means of the modeling software which was written by Zheng.
(3) Mapping, CLEAN and self-calibration were performed with the support of the AIPS package. Software had been developed by ourselves for interfacing Miyun data to the AIPS and modifying the CLEAN components list to ensure most sources within the wide primary beam of the MSRT are included (Zhang 1992).
(4) We leave the correction for the instabilities of the
instrument and ionosphere to the self-calibration stage.
Research on such instabilities have been done for the calibration
of MSRT data (Zheng 1988; Zhang et al 1989).The
accuracy of the calibration at this stage is better than 
for the phases and 10% for the gains.
Data of most fields of view were reduced according to the procedure mentioned above. The weighting function was changed from `NA'(natural, the visibilities of short spacings are given much more weight) to `UN'(uniform). With the `UN' weighting function, the background fluctuation of maps become much flatter than that of `NA' weighting maps. Despite of some loss of the ratio of signal to noise (Condon et al. 1995), especially for a field of view with some strong sources far from the center of the field, source searching could go much deeper without increasing the number of spurious sources.
In Paper I, we described the method of removing interference and compensating
for the zero-offset. This method could be used up to the north pole area of
sky.
In the data reduction of the whole sky survey, we
subtracted model-source first for all sky areas of declination larger than
. Though the effects of zero-offset were more serious at areas around
the north pole, with this method the background fluctuation was found to be no
worse than that of the areas at smaller declination.
Design of the pixel grid
allows a synthesized beam to contain at least 3 to 4 pixels, to ensure that
the CLEAN and self-calibration processes as well as the fitting program work
properly. Usually the cell size is 
. Some times it is
changed to ensure that strong sources far from the edges
of the map can be included.
After several cycles of CLEAN, data editing, and self-calibration, maps usually became stable and converge. At this stage, low level features of a set of straight lines can been seen sometimes in the background. This could be caused by the holes in the visibilities when the self-calibration is not perfect. A way to reduce the level of this kind of background is to add some extrapolated visibilities to fill the hole in the visibilities. We calculated the visibilities of these holes by modeling with the CLEAN components.