(2) |

When a telescope is quite displaced from the other two telescopes (i.e., an unbalanced triangle), the two longest baselines become nearly parallel give essentially identical phases, but with opposite sign. Their contributions therefore almost cancel in the closure equation. The only remaining contribution is that of the shortest baseline; when the source is unresolved on that baseline, the result is a closure phase whose magnitude is only a few degrees.

As shown by Massi (1989), the contribution of the longest baselines
to the closure phase is a function of the ratio *s*/*L*
between the shortest (*s*) and the longest (*L*) leg of the closure
phase triangle. For a very small ratio (i.e., )the contribution may drop to a few degrees.
This can be seen, for example, in the plots of closure phase
shown in Fig. 8.
The triangle at the top of the figure makes use of the intermediate
length baseline MK to KP (the one flagged in Test-2 of Sect. 2)
and has .
The closure phase for this triangle reaches a value of about ,significantly larger than the and reached for
the other two triangles shown, with ratios of and 0.05
respectively.

In conclusion in the case of missing intermediate spacing information
the observed closure phase is a small quantity.
While at first glance, this would imply a quite slow convergence,
instead it gives rise, as shown in Sect. 2, to a very
serious problem when the adopted starting model
is a point source and Loop "A'' is followed. In the next section we
show why for cases of uniform *u*-*v* coverage
Loops "A'' and "B'' are equivalent, but for cases where
only low closure phase values are present (associated with unbalanced
triangles), differences become apparent.

Copyright The European Southern Observatory (ESO)