Holdaway (1992) investigated image quality as a function of phase error
magnitude for point sources and concluded that even with 30 degree rms
phase errors, reasonable imaging with dynamic range of about 200:1 was
still possible. In considering a more complex source, a more
realistic atmospheric model with baseline dependent phase errors
should be employed, as in the atmospheric simulations of Holdaway
(1991). The particular atmospheric phase screen model used in the
simulations described below results in phase errors which increase as
the baseline raised to the 0.33 power, which is seen during good
conditions on the potential MMA sites. During poorer conditions, the
phase errors usually increase more steeply with baseline length, at
least out to baselines of 300 m, but the basic conclusions derived
from this work should be independent of the details of the phase
structure function. Simulations were performed with a random circular
array of 1 km maximum baseline. Samples on the 
tracks were
calculated for 5 s integrations, the standard M31 HII region model
image was Fourier transformed and degridded into the simulated 
points. We assumed no decorrelation occurred on time scales less than
5 s. The entire simulated data set was 18 minutes long, or ten
atmospheric crossing times of the array's longest baseline. The
amplitude of the phase screen was scaled as required, the phase screen
was ``blown'' over the array with frozen turbulence at a velocity of
10 m/s, and the phase errors were then applied to the antennas below,
thereby corrupting the phase of the visibilities. No other errors
were added to the visibilities.
For the purpose of representing the typical level of phase fluctuations graphically, we parameterize each of the scaled atmospheres in the rms phase error calculated over the full 18 minute observation, averaged over all baselines. Hence, a model atmosphere with mean rms phase of 35 degrees will have some baselines with phase errors as high as 50 degrees.
We have imaged the corrupted visibilities in three different ways: