I'm curious how the demand factors in the table 220.84(B) are derived/what's the thought process behind them? Particularly that they get smaller as more units are added.
Short answer: the actual current draw from each unit will vary with time, and the calculated load is some fairly conservative (almost) bound for that load (almost because it would be fine if it were exceeded occasionally, like minutes per year, although in practice the Article 220 load may never be exceeded).
So for N units, taking that calculated load and multiplying it by N would cover the case that all N units are near their maximum current draws simultaneously. As long as the units are fairly uncorrelated, that is very unlikely to happen. The demand factor, which decreases as N increases, is a correction for this.
Cheers, Wayne
P.S. As an analogy, consider a normal distribution of mean 1 and standard deviation 1. Say we want only a 0.14% chance that the value from the distribution will exceed some threshold; in that case the threshold is the mean plus 3 standard deviations, or 4. Now if we sample this distribution twice and add the results, the result is a normal distribution of mean 2 and standard deviation sqrt(2) (the variances add). The equivalent threshold is now 2 + 3*sqrt(2) = 6.24, which is less than 2 * 4 = 8.
