Analog Tips - ADC Decimation

Latest update time：2020-05-07

Reads：

Good articles should of course be shared. If you like this article, please contact the backstage to add it to the whitelist. Reprinting is welcome.

Wideband GSPS analog-to-digital converters (ADCs) offer numerous performance advantages for high-speed acquisition systems. These ADCs provide visibility across a wide frequency spectrum. However, while some applications require a wideband front end, others require the ability to filter and tune to a narrower frequency spectrum.

When narrowband is required, it's inefficient for ADCs to sample, process, and consume power to transmit wideband spectrum. There's no need to use numerous FPGA transceivers to decimate and filter wideband data in post-processing. High-performance GSPS ADCs enable digital downconversion (DDC) to be relocated inside the ADC. Reducing the number of JESD204B ADC output channels minimizes data rates and system layout complexity.

Decimation is a method of observing only the periodic portion of the ADC's samples and ignoring the rest. The result of decimation is a reduction in the ADC's sampling rate. For example, a 1/4 decimation mode means (total number of samples)/4, effectively discarding all other samples.

The ADC must also include a numerically controlled oscillator (NCO) and a filtering and mixing element (serving as a companion to the decimation function). Digital filtering effectively removes out-of-band noise within a narrow bandwidth set by the decimation rate. The digital tuning word of the NCO, acting as a local oscillator, provides fractional division of the sampling rate, providing precise positioning by the number of bits of resolution. The tuning word has the range and resolution to place the filter where it is needed spectrally.

The filter passband should match the effective spectral width of the converter after decimation. A significant advantage of using a DDC is the ability to locate the harmonics of the fundamental signal so that they fall outside the frequency band of interest.

The DDC filter's digital filtering removes noise outside a narrow bandwidth. The ideal ADC's SNR calculation must account for the processing gain of the filtered noise . With a perfect digital filter, the processing gain from filtering the noise increases by 3dB for every power-of-two reduction in bandwidth.

Ideal SNR (including processing gain) =

6.02 × N + 1.76 dB + 10log ₁₀ (f _s /(2 × BW))

Figure 1. Using a lowpass filter and an NCO to perform frequency translation implements a bandpass filter. Frequency planning ensures that unwanted harmonics and spurs fall outside the bandwidth.

AD9694