Supplier eBooks

39 | the signal features. ADCs are limited in RF bandwidth to the Nyquist criterion, which states that to accurately recreate a signal, the sample rate must be at least twice the highest-frequency component that can be accurately sampled (i.e., the Nyquist rate). That means that a 1GSPS ADC has a Nyquist frequency of 500MHz, allowing it to capture signal components up to 500MHz when sampling baseband signals. It is important to note that the sampling rate, not maximum frequency, dictates total RF bandwidth. Additionally, sections of bandwidth at higher frequencies can be sampled using downconversion. For instance, if the signal to be converted has only 10MHz of bandwidth but resides at a center frequency of 100MHz, it requires a sampling rate of 20MSPS, not 200MSPS. However, for the ADC to be able to convert that 10MHz of bandwidth, it would need to be able to sample at 210MSPS (i.e., 100MHz + 5MHz of half bandwidth) and digitally extract the desired 10MHz of bandwidth or use downconversion to extract the 10MHz bandwidth and convert it at 20Msps. Sampling rates that exceed the signal bandwidth (a technique called oversampling) can provide some benefits, such as improved signal-to- noise ratio (SNR) in the digital domain (Figure 3). Essentially, the faster a signal is sampled, the lower the noise floor can be realized, as noise is typically spread widely over the entire sampling spectrum. With higher sampling rates, the total noise remains the same but is spread over a much wider frequency range, which can be further reduced using sampling techniques and digital filtering. The ideal noise floor (NF) of an ADC is defined as: This describes the level of noise quantization within the converter for a given number of bits (N) and how noise can be impacted by sampling rate (Fs) and bandwidth (BW). This equation shows that doubling the sampling rate can reduce the NF by 3dB. Combining oversampling with digital filtering can dramatically reduce the noise floor and enhance the SNR. Another concept is undersampling, which is sampling at a much lower frequency (typically less than half) than the actual signal frequency. Undersampling can perform a conversion very similar to mixing, where the undersampled signal has frequencies that are aliased to the baseband or first Nyquist zone, like they were originally in the baseband frequency range. However, with undersampling, a design must ensure that the wanted signals are aliased into the baseband and that the signals are in the proper spectral orientation. Undersampling can cause spectral reversal, as with mixers and lower sideband reversal. While designing a digital receiver, engineers should Figure 3: Digital sampling diagram. (Source: NASA) 2

• Cover