Bandwidth is invaluable for communications and other systems. Whether applied for commercial communications or defense electronic-warfare (EW) and phased-array radar systems, the analog-to-digital converters (ADCs) in these systems often need to operate at higher-order Nyquist rate bands. In fact, next-generation gigasample-per-second (GSPS) ADCs allow sampling of GHz frequency signals well into the third or fourth Nyquist bands, with decimation options to provide the dynamic-range benefits of oversampling.
Given sufficient ADC input bandwidth, it is possible to downconvert high-frequency signals directly within the ADC by undersampling a system’s intermediate-frequency (IF) signals. The use of ADCs with wide input bandwidths and high sampling rates allow direct sampling of IF signals and possible elimination of an entire frequency-conversion stage within a receiver to save on power and complexity.
ADC undersampling is essentially a method of using a sampling frequency that is less than twice a maximum (harmonic) frequency component in the signal. This technique is also referred to as harmonic sampling, bandpass sampling, or super-Nyquist sampling. To reconstruct the original signal from the sampled version, the Nyquist-Shannon Sampling theorem states that the sample rate must be twice a signal bandwidth of interest. This should not be mistaken with a sample rate that is twice the maximum IF signal frequency component.
If BW refers to the signal bandwidth of interest, then a sampling frequency, Fs, of greater than twice the bandwidth is required, or Fs > 2BW. The signal bandwidth of interest can be from DC to BW or from A to B where BW = A – B. As long as the bandwidth of interest does not overlap an ADC’s Nyquist band, which is one-half the sample rate, or Fs/2, undersampling can work for higher-frequency bands using ADCs that have a wide full-power bandwidth (FPBW) relative to their respective sample rate (Fig. 1).
Secrecy and signal security are essential to military operations and many systems are designed to maintain security across a wide range of operating conditions. To reduce the probability of intercept or detection, the form and magnitude of a radar transmission is designed in many cases to spread energy over the widest possible frequency range. Low probability of intercept (LPI) and low probability of detection (LPD) are classes of radar systems with certain performance characteristics which make them nearly undetectable by modern intercept receivers. LPI features prevent the radar from tripping off alarm systems or passive radar-detection equipment.
To provide resistance to jamming, radar systems can be architected by intelligently randomizing and spreading radar pulses over a wide frequency band so that a limited amount of signal energy will be present in any one band, using a technique known as direct-sequence-spread-spectrum (DSSS) modulation (Fig. 2). Frequency-hopping-spread-spectrum (FHSS) is another method of moving signal energy around the available bandwidth to make it more difficult to jam the signals. In these cases, more bandwidth is consumed than would normally be needed for the signal of interest. As a result, a wider receiver bandwidth is required in support of such anti-jamming approaches.
One of the most important factors for success in an LPI system is the use of as wide a signal transmission bandwidth as possible to disguise complex waveforms as noise. This conversely provides a higher-order challenge for intercept receiver systems that seek to detect and decipher these wideband signals. Therefore, while this creates improvements towards LPI and LPD functions, it also increases radar transceiver complexity by mandating a system that can capture the entire wide transmission bandwidth at once.
The capability of an ADC to simultaneously digitize 500 MHz, 1 GHz, and larger portions of spectrum bandwidth in a single Nyquist band provides the means to tackle this system-level challenge. Moving these bands higher in frequency, beyond the first Nyquist band of an ADC, can be even more valuable.
Available wideband ADCs offer the potential to handle multiple wide Nyquist bands within an undersampling mode of operation. However, using a higher-order ADC Nyquist band to sample wideband signals requires precise receiver front-end anti-alias filtering and frequency planning to prevent spectral energy from one Nyquist band from leaking into other Nyquist zones. It also ensures that unwanted harmonics and other lower-frequency signals do not fall into the band of interest after that band has been folded down to the first Nyquist zone.
A bandpass filter (BPF) upstream of the ADC must filter any unwanted signals and noise that are not within the nominal bandwidth of interest. Newer GSPS ADCs, such as models AD9234 ( a dual 12-b, 1 Gsamples/s and 500 Msamples/s ADC), AD9680 (a dual 14-b, 1 Gsamples/s ADC), and AD9625 (a 12-b ADC that works to 2.6 Gsamples/s) offer multiple Nyquist-band sampling with high dynamic range maintained across wide input bandwidths.
Since direct sampling folds the signal energy from each Nyquist zone into the first Nyquist zone, there is no way to accurately differentiate the sources of the signal content. As a result, rogue energy can appear in the first Nyquist zone, which will degrade the receiver signal-to-noise ratio (SNR) and spurious free dynamic range (SFDR). Spectral issues can potentially plague government and military applications, both for sensing and communications.
Digital radio transceivers for military communications represent another example of the use of high-speed ADCs and digital-to-analog converters (DACs) that can potentially replace a traditional baseband frequency mixer stage. A digital-converter architecture has several advantages because tight filtering and adjacent-channel signal rejection can be performed in the digital domain for baseband conversion.
Direct RF/IF sampling in radar front ends offers several advantages compared to analog signal processing. First and foremost, it can enable a reduction in the number of components required, when an entire frequency downconversion stage can be eliminated (Fig. 3). Direct sampling also removes the need to design a frequency mixer for a uniquely tailored frequency plan. It can simplify the design of next-generation receivers for future signal bandwidths that become available as radar systems are modernized and updated.
All that may be needed to work with a new carrier frequency is to select an appropriate sampling rate and incorporate an appropriate bandpass filter. Direct sampling also makes it possible to apply a single RF front-end design for multiple frequency bands, to eliminate the need for multiple front-end architectures.
Current generation ADCs now offer a plurality of internal digital-downconversion (DDC) processing blocks. Each DDC can apply its own decimation rate and numerically controlled oscillator for tuning placement within a Nyquist band. Processing gain can be achieved within a narrower bandwidth to digitally filter out-of-band noise.
This reduces the ADC output data required and minimizes processing complexity in complementary field-programmable gate arrays (FPGAs) and digital signal processors (DSPs). However, additional channelizer signal processing can also be performed downstream of the ADC in a receiver for flexibility and performance enhancement.
Wideband communications and sensing systems require extremely high-speed data converters. State of the art GSPS ADCs such as the models AD9234, AD9680, and AD9625 not only offer high sample rates for wider instantaneous bandwidths, but also the capability to sample high-frequency inputs above the first Nyquist zone.
A single direct-sampling ADC used at a wide high-frequency bandwidth can potentially replace an entire IF-sampling or zero-IF-sampling subsystem consisting of frequency mixers, local oscillators (LO) or frequency synthesizers, amplifiers, and filters while achieving greater design and performance flexibility. This can significantly reduce a system’s bill of materials (BOM) and cost, design time, circuit-board size, weight, and power consumption.
Ian Beavers, Applications Engineer, High-Speed Analog-to-Digital Converters
Analog Devices, Inc., Greensboro, NC
For Further Reading
W. Kester, “What the Nyquist Criterion Means to your Sampled Data System Design,” Analog Devices Training Seminars, MT-002.
P. Poshala, “Why Oversample when Undersampling can do the Job?” EE Times India, 2013.
J. Shea, “Military Wireless Communications,” from University of Florida.
W. Stallings, Data and Computer Communications, Prentice-Hall, Upper Saddle River, NJ, 2010, pp. 286-294.
R. Zarr, “ADCs Feel the Need for Speed," Electronic Design, 2014.
Looking for parts? Go to SourceESB.