Special thanks to Claire Chandler who motivated this project, and may yet let her name be on the auther list. Also, praise and thanks to Michael Rupen and Tim Cornwell for nice but not yet implemented ideas. And thanks to Scott Foster for technical support.
References
Cornwell, T.J., and Evans, 1984. A&A,
Cornwell, T.J., Holdaway, M.A., and Uson, J., 1993, A&A.
Holdaway, M.A, 1991, MMA Memo 68, ``A Millimeter Wavelength Phase Stability Analysis of the South Baldy and Springerville Sites''.
Holdaway, M.A., 1992, MMA Memo 84, ``Possible Phase Calibration Schemes for the MMA''.
Holdaway, M.A., Owen, F.N., and Rupen, M.P., 1994, MMA Memo 123, ``Source Counts at 90 GHz''.
Holdaway, M.A., and Owen, F.N., 1995, MMA Memo 126, ``A Test of Fast Switching Phase Calibration with the VLA at 22 GHz''.
Holdaway, M.A., Radford, Simon J.E. , Owen, F.N., and Foster, Scott M., 1995, MMA Memo 129, ``Data Processing for Site Test Interferometers''.
Thompson, A. Richard, Moran, James M., and Swenson, George W., Interferometry and Synthesis in Radio Astronomy, John Wiley & Sons, New York, 1986.
Figure 1: Three different reconstruction techniques (columns) applied to
four different magnitudes of phase errors (rows).
Column 1: imaging without any decorrelation correction.
Column 2: imaging with the statistical phase deconvolution.
Column 3: imaging with correction of the amplitudes only.
Row 1: 17 degree rms phase errors. Row 2: 35 degree rms phase errors.
Row 3: 70 degree rms phase errors. Row 4: 105 degree rms phase errors.
The statistical phase deconvolution technique is superior for
large phase errors.
Figure 2: The phase errors will degrade the resolution as well as the
sensitivity. These figures illustrate how the resolution and
sensitivity at the highest resolution degrade with increasing phase
errors in our simulations. The sensitivity curve is consistent with 
when we consider that the phase errors
at the highest resolution are higher than the mean phase errors we plot.
Also, the beam fitting is ill-conditioned in the high phase error case,
which explains why the resolution seems to flatten out at the right
side of the graph.
Figure 3: Three measures of image quality for the three imaging methods compared
in this memo: (a) Dynamic Range, defined as the image peak divided by the
off-source rms, (b) Median Fidelity, defined in the text, and (c)
First Moment Fidelity, defined in the text.
Table 1: How high a frequency could you operate the MMA in Chile
if 30 degree rms phase errors were required, and if 70 degree rms
phase errors were required? We present here the very conservative
estimates based on the NRAO 300 m, 11.2 GHz site test interferometer data
taken for the month of June 1995. The phase at 11.2 GHz was better than
2.95 degrees 75% of the time, indicating that 113 GHz observations would
have phase errors of less than 30 degrees more than 75% of the time, and
266 GHz observations would have phase errors of less than 70 degrees more
than 75% of the time. This table does not consider any form of calibration
aside from the decorrelation corrections described in this memo.
Active phase calibration could increase the maximum frequencies quoted in
this table by a factor of 3.5.