Remember Nyquist Zones from your Signal Processing class? To summarize the sampling criteria, Nyquist zones subdivide the spectrum into regions, spaced uniformly at intervals of Fs/2. Each Nyquist zone contains a copy of the spectrum of the desired signal or a mirror image of it known as an alias. The signals below and above the sample rate, by an equal amount, fold on top of each other as aliases at the analog to digital converter (ADC) output.
Radios reject interference from other radios using filters. The choice of core RF architecture in the radio unit makes this either somewhat hard to solve, or very hard to solve. In this context, hard means expensive. If the radio architecture uses particular sampling rates, the sensitivity to aliases increases, resulting in heavier, more expensive filters. Unfortunately, the sensitivity problem may not be identified until the back half of the design cycle, when core architecture decisions have been made.
Zero IF radios reduce co-location problems by converting only the band of interest, while Direct RF architectures convert all the bandwidth and use filters to capture band of interest. A common Direct RF analog to digital sampling rate is between 3GHz and 4GHz. For C-band, this means that there’s a Nyquist boundary at around the desired band, which means that signals below and above the sample rate by the same amount fold on top of each other at the ADC output. All frequencies that could alias on top of the desired signal need to be filtered enough to not affect the receiver’s sensitivity. The stronger the signals at these frequencies, the larger, more expensive, and heavier the filter. The worst – and most expensive case – is when the source of the interfering signal is a co-located transmitter. As it turns out, C-band frequencies and some of the most commonly used FDD bands interfere with each other when using devices based on these ADCs.