The all-sky survey would benefit tremendously from every possible improvement that can be made to the E configuration sensitivity. For example construction of a ground screen and interference fence under and around this compact configuration could redirect a significant fraction of the far side lobe power onto the cold sky. Cooling of the polarizer assembly should also receive a high priority to improve system performance. High dynamic range receivers must be employed to minimize data loss due to saturation by RFI. Low loss, preferably tunable, RF filters will probably also be necessary to allow operation in the hostile RFI environment. Minimizing data loss by shadowing and cross talk in the tightly packed E configuration might also dictate an elliptical configuration with a NS to EW aspect ratio of about 1.5:1.
The surveying speed is directly proportional to the instantaneous velocity coverage that can be used, this drives the request for implementing the possibility to use 500 MHz and 2048 channels. Since interference will be a major problem at the lower frequencies, schemes of online rejection of interference need be investigated.
As pointed out earlier, the three parameters that determine the
surveying speed are field of view, sensitivity and instantaneous
bandwidth. The GMRT has a larger collecting area than the VLA and will
be more sensitive at the lower frequencies. The currently planned
correlator will allow for an instantaneous bandwidth of 32 MHz at
most. Thus the enhanced VLA will be a factor of ten faster. Note also
that the GMRT has no plans to install feeds for the frequency range
between 760 and 1000 MHz (0.4 
1), currently thought to be one of
the most interesting regimes for galaxy evolution. The WSRT will
cover this frequency range but its collecting area is half of the
VLA's and (more importantly) to get sufficient spectral resolution (66 km/s)
one can at most use a bandwidth of 80 MHz, so the VLA
will be 5 times faster. Arecibo gains by a factor five in collecting
area, so will play a major rôle in directed studies of known
high-z objects. As a surveying instrument however, it loses by a
large factor (about 25) relative to the VLA, because of its much
smaller field of view. Synthesis arrays are also much less sensitive
to interference, a major restriction on work at those redshifts.