Current thinking regarding a new correlator and IF transmission scheme
are based on a 2 GHz total bandwidth capacity (1 GHz in each of two
polarizations). Since modern feeds can provide good performance over
a bandwidth 
of the center frequency in linear
polarization, it is clear that a great deal of potential sensitivity
is given up at the higher frequencies. An alternative is to consider
an ultra-wide bandwidth system, in which accurate polarization,
spectral resolution, and field of view are given up for maximizing the
instantaneous bandwidth.
An ultra-wide-bandwidth correlator for continuum observations above 18 GHz could be built as a separate instrument from the main 1 GHz correlator. The required spatial field of view is likely to be a cost driver, since it determines the frequency resolution needed to avoid bandwidth smearing. Preliminary thinking indicates that 8 GHz/polarization at 200 MHz resolution might be feasible, but no cost estimates have been made. An analog correlator should be considered, especially if a coarser frequency resolution (>100 MHz) might be acceptable. Economic and performance tradeoffs should be studied for three approaches: a pure analog correlator, with optical fiber delay lines (considerable testing will be needed to see if adequate passband stability can be achieved); digital delay followed by conversion back to analog for correlation; and an expanded pure digital correlator. It is likely that the digital approach would win if frequency resolution <100 MHz is required.
It should be recognized that this is an expensive item that will be useful only at the highest frequencies, and that various compromises will prevent it from achieving very high dynamic range. Assuming a 35% bandwidth, and the improvements mentioned elsewhere in system temperature and antenna surface accuracy, sensitivities of better than 1 microJy in 12 hrs are possible between 18 and 40 GHz.