[jakeinspace]

I'd assume they used some pretty solid statistical methods for reducing the atmospheric noise to a relatively negligible amount, based on covariance of their dish readings.