Microwave Journal

State-of-the-Art Spectrum Monitoring Technologies

September 14, 2015

Spectrum monitoring, that is, sensing for signal occupancy in the RF spectrum, constitutes one of the four key spectrum management functions – the others being spectrum planning, spectrum engineering and spectrum authorization. Spectrum monitoring helps spectrum managers identify utilized and underutilized radio bands. The results are then used to effectively plan and allocate frequencies, avoid incompatible usage and identify sources of harmful interference. As the number of connected devices continues to grow exponentially with the growth of 4G cellular, Wi-Fi and IoT technologies, spectrum monitoring plays an increasingly important role in commercial, regulatory and military applications. Real-time spectrum analysis (RTSA) is often considered one of the key enabling technologies for spectrum monitoring, with heavy emphasis on visualization aspects, such as persistence, waterfall displays and spectrograms. This article discusses additional powerful inline or post-processing spectrum monitoring algorithms, such as cyclostationary feature detection, frequency based event detection and intelligent signal identification. The article describes how these applications are enabled by the recent developments in software, processing units and high throughput data movement bus technology.

Figure 1

Figure 1 Three-tiered model adopted in the U.S. for sharing spectrum.

Due to the proliferation of portable wireless electronics and the bandwidth intensive applications that they enable, radio spectrum is becoming increasingly crowded. Today, wireless technologies such as cellular LTE, Bluetooth enabled wearable electronics, and Wi-Fi enabled first generation IoT devices are a big driver of economic growth in the commercial domain. E-commerce and social networking, and the economic benefits that come alongside, have been popularized due to the wide proliferation of always-on wireless portables. Similarly, in the public safety and military usage domains, newer video based applications require extensive wireless bandwidths to provide the necessary mission-critical performance. The RF spectrum, despite bringing so much value to the economy, is a finite, limited resource. Hence, the cost to access the spectrum itself has skyrocketed in recent years. In 2009, the auction of the 700 MHz band by the Federal Communications Commission (FCC) raised $19.5 billion, and the 2014 auction of the AWS-3 band netted $44.5 billion.1 Spectrum monitoring provides valuable data that policy makers can use to determine which frequency bands are underutilized and hence, can be reallocated or repurposed through auctions and/or policy changes. Particularly, data from long-term continuous spectrum monitoring stations is crucial in helping spectrum policy makers and planners make informed decisions.2 Spectrum monitoring is also important for enforcement purposes – to identify unauthorized users infringing on the expensive spectral resource, detect interference and ensure compliance with spectral masks.

Due to recent policy adoptions in Europe and in the U.S., the importance of continuous spectrum monitoring is set to increase with new spectrum sharing policy models.3 Long-term spectrum monitoring studies4,5 have shown that although the proverbial “spectrum crunch” exists in certain commercial bands, like the cellular and 2.4 GHz ISM bands in high population areas, most of the other bands are underutilized. Armed with this empirical knowledge of actual usage and the knowledge that resource reallocation is a time-consuming and expensive process, a paradigm shift in spectral policy has recently taken place to allow dynamic shared spectrum access. In a shared spectrum environment, secondary users can operate in the same band as the incumbent spectrum licensee, subject to interference constraints. To this end, the European Commission (EC) has recently identified Licensed Shared Access (LSA) as a regulatory approach that allows secondary users to access an incumbent user’s band and receive a certain Quality of Service (QoS), in accordance with sharing rules negotiated between them. The U.S. has adopted a different three-tiered hierarchical model for spectrum sharing (see Figure 1). The 3550 to 3650 MHz frequency region has been selected in the U.S. as a fast-track band to deploy the three-tier model. Spectrum sensing is a key enabling technology for updating the database that controls shared access to the band. Hence, spectrum monitors that permit such sensing are critical. The next few sections list signal processing techniques that enhance the capabilities and sensitivity of spectrum monitors.

Figure 2

Figure 2 Block diagram of the cyclostationary feature detection process.


Cyclostationary feature detection (CFD) uses the “spectral correlation function” signal processing technique to detect low power received signals that are often below the noise floor of the spectrum monitor. Modulated information carrying signals are typically modeled as a cyclostationary process. Typically, a digital modulated signal carries information over fixed symbol periods, such that the signal exhibits the features of periodic statistics and spectral correlation. CFD makes extensive use of fast Fourier transforms (FFT) to identify the spectral correlation features (see Figure 2).6 The important thing to note is that CFD is robust to noise uncertainties and performs better than energy detection in low noise conditions. This is because noise is uncorrelated, while the information bearing signal has spectral correlation features that show up after the CFD analysis.

In a shared spectrum environment, many cognitive radio researchers use CFD because it allows the detection of far away (hence low received power) incumbent transmissions. A spectrum monitor armed with CFD capability is better able to detect low power signals compared to a threshold based simple energy detector. In a scenario where assessments need to be constantly made of whether a radio frequency is presently occupied by a user, such as in a shared spectrum environment, the measurement from a CFD enabled spectrum monitor is more reliable and gives confidence to the occupancy assessment results.

Figure 3

Figure 3 With a pulsed jamming signal, the identification of subsequent jamming pulses is difficult without continuous acquisition.

Inline CFD computation is a processor-intensive operation that requires access to the real-time time domain samples (I/Q data) captured by the spectrum monitor. Software-defined radio implementations have used host-side processing to perform CFD calculations on I/Q data at limited bandwidths,7 but FPGA implementations of the CFD can provide improved performance at real-time speeds. As the monitoring bandwidth for CFD increases, so does processing time. In such cases, single or multi-channel sub-span extraction through digital down conversion (DDC) is applicable to reduce the spectrum bandwidth subjected to the CFD calculation. For example, in a scenario where the spectrum monitor measures a 100 MHz wide bandwidth, yet where weak signals exist below the noise floor for only 20 percent of that region, DDC can extract that 20 MHz section with low received signal power. CFD is then only performed on that 20 MHz sub-span, greatly reducing overall processing requirements.

Frequency Based Event Detection

Particularly in military scenarios, it is often necessary to identify interference from systems attempting to obstruct a communications channel. Frequently called “jamming” signals, this type of interference signal is able to jam a communications signal by producing unwanted power within the band of interest. Common types of jamming signals include single tones, random white noise, pulsed, frequency hopped and modulated “fake” communications signals. From a jamming perspective, they differ in terms of effectiveness, power requirements, generation complexity and difficulty of detection. For example, the generation of a single carrier in an existing communications channel is relatively simple, but jamming wise, the signal is often ineffective and easily identifiable. Alternatively, the generation of broadband white noise can be extremely effective at obstructing a communications link.

Some of the more interesting types of jamming signals are pulsed or frequency hopped signals. These types of jamming signals are generally effective and can be difficult to detect using a traditional spectrum analyzer. The difficulty lies in the need to capture both time and frequency information regarding the signal of interest. As a result, stream-to-disk systems are commonly used to capture a dedicated portion of RF bandwidth over several hours. Once the signal is recorded, it is possible to use two methods to analyze the power, frequency and timing characteristics of jamming signals: FFT-based analysis and joint time-frequency analysis (JTFA).

Figure 4

Figure 4 Post-processing a pulsed jamming signal using FFT analysis.

When performing an FFT-based analysis of a jamming signal, either inline or post processing can be used. While inline processing provides immediate results, post processing offers the richest data set. Figure 3 illustrates a pulsed jamming signal, showing that the identification of subsequent jamming pulses is difficult in the absence of continuous acquisition. The solution is to record the RF data for a period of time and analyze it after the acquisition is complete. In this scenario, a chunk of RF spectrum is acquired over a long period of time and then analyzed in blocks (see Figure 4). The FFT size can be customized to give the most accurate characterization of the pulse’s spectral information.

The resolution bandwidth (RBW) is inversely proportional to the signal’s acquisition time. In the frequency domain, this affects the displayed power level of a transient signal. The burst might last just a few microseconds, and if a narrow RBW (long acquisition time) is used, the detected power spreads out over frequency. Figure 5 compares the spectrum of a jamming burst using two FFTs: a larger acquisition window (smaller RBW) vs. a smaller acquisition window (larger RBW). A longer acquisition time narrows the RBW of the measurement and reduces the amplitude of the jamming pulse – which may cause the jamming signal power to fall below the noise floor and escape detection. Thus, for frequency-based detection, the FFT parameter size should be properly selected.

Figure 5

Figure 5 Pulsed jamming signal power spectrum with a small (a) and large (b) RBW.

While FFT-based analysis provides useful frequency domain information, to obtain timing statistics about the jamming pulse, joint time and frequency techniques are needed, such as a spectrogram. The spectrogram exposes the timing dimension necessary to identify additional characteristics such as pulse inter-arrival gaps, pulse duration, bandwidth and amplitude. The drawbacks of post-processing are the large storage space and non-real-time identification of the jammer. Such an application can benefit from multi-core and FPGA-based spectrum monitoring hardware. With these technology enablers, it is straightforward to do continuous acquisition, i.e., the FFT and JTFA processing of data in real-time: for example, an existing real-time signal analysis implementation that outputs both the FFT power spectrum and the spectrogram.

Intelligent Signal Identification – “Packet Sniffing”

A second type of interference is a pirating or piggybacking communication signal. Here the interferer attempts to use the existing telecommunications infrastructure to illegally transmit by having the repeater rebroadcast an unauthorized signal. Since the repeater simply amplifies a specified spectral band, the interferer can use it to amplify the unauthorized channel communication with the intended signals.

The “packet sniffing” of such an interference signal can be accomplished by either processing the signal inline or recording a specified bandwidth and post processing the data. Once captured, this data can be post-processed with a variety of methods. Just as with jamming signals, analysis via FFT and JTFA is applicable to identify frequency, power and amplitude information about the interferer. However, for “packet sniffing” applications, the baseband waveform can be demodulated, although demodulating an unknown carrier is not trivial. To accurately demodulate a digital signal, it is important to know the carrier’s symbol rate. This can be estimated by observing the channel bandwidth, but often the symbol rate must be experimentally determined by using the characteristic knowledge of known communications standards.

By demodulating the interfering radio signal, the bit stream being transmitted over the communications channel can be obtained. In some cases, this information is decodable by matching it with known preamble information. However, the greatest challenge occurs in decoding meaningful information from a bit stream, especially if the data is encrypted. Nonetheless, through demodulation and decoding, it is easy to identify the interference signal as a rogue transmitter operating outside authorized broadcasts.

Figure 6

Figure 6 The integrated, hardware-aware design environment of the LabVIEW Communications System Design Suite can speed the development of spectrum monitoring and analysis systems.


This article highlighted several advanced analysis techniques that can enhance the spectrum monitor, transforming it to a more powerful RF measurement tool that doubles as a highly capable signal detector. Researchers are constantly coming up with more powerful and efficient signal processing techniques, while manufacturers continue to leverage newer computing and processing technologies to allow the hardware to keep up with the requisite bandwidth and computational requirements. As the necessity of spectrum monitoring becomes more widespread, due to recent trends in spectrum management – like high-priced auctions and policy changes favoring spectrum sharing – the methodologies discussed and others like them will see greater adoption.

Fortunately, spectrum monitoring software and hardware technologies are keeping pace with trends in the spectrum field. To achieve such capabilities, the spectrum monitoring platform has to be flexible enough to permit advanced signal processing. The hardware platform must have fast parallel cores, a high-speed bus for data transfers and/or support FPGA processing capabilities.Additionally, the platform should have the capability to perform advanced programming (including demodulation and decoding codecs) to permit intelligent signal identification.

One barrier to introducing new spectrum monitoring and analysis methods is dealing with multiple non-integrated processes while prototyping and deploying novel techniques. These include designing, simulating, prototyping, deploying to real-time in-line processing hardware and testing. An engineer working with the tools in one of these steps may not have the tools or skill set for the other steps. A tightly integrated software and hardware platform that instructively brings together all these discrete processes would greatly facilitate researchers and reduce the time to develop and deploy new spectrum monitoring algorithms. As one example, the LabVIEW Communications System Design Suite (see Figure 6) brings together the discrete research, development and deployment steps within a single tool. This flexible software suite tightly integrates with software-defined radios, including one with a programmable FPGA. The suite is especially suited for designing and implementing spectral monitoring systems that benefit from the power of an FPGA.


  1. Federal Communications Commission (FCC) Auction 97 Advanced Wireless Services (AWS-3) Summary, January 2015. Available Online: http://wireless.fcc.gov/auctions/default.htm?job=auction_summary&id=97.
  2. T. Taher, R. Bacchus, K. Zdunek and D. Roberson, “Long Term Spectral Occupancy Findings in Chicago,” IEEE Dynamic Spectrum Access (Dyspan), Aachen, Germany, May 2011.
  3. M. Matinmikko et al., “Overview and Comparison of Recent Spectrum Sharing Approaches in Regulation and Research,” New Frontiers in Dynamic Spectrum Access Networks, IEEE DySPAN, Mclean, Va., April 2014.
  4. R. Bacchus, A. Fertner, C. Hood and D. Roberson, “Long-Term, Wide-Band Spectral Monitoring in Support of Dynamic Spectrum Access Networks at the IIT Spectrum Observatory,” New Frontiers in Dynamic Spectrum Access Networks, IEEE DySPAN,October2008, pp. 1210.
  5. M. Islam et al., “Spectrum Survey in Singapore: Occupancy Measurements and Analyses,” Cognitive Radio Oriented Wireless Networks and Communications,CrownCom 2008,May 2008, pp. 127.
  6. Jayanta Mishra, Deepak Kumar Barik and Manoj Kumar, “Cyclostationary Based Spectrum Sensing in Cognitive Radio: Windowing Approach,” International Journal of Recent Technology and Engineering,March 2014, No. 3, pp. 227723878.
  7. P.D. Sutton, K.E. Nolan and L.E. Doyle, “Cyclostationary Signatures in Practical Cognitive Radio Applications,” Selected Areas in Communications, IEEE Journal, January 2008,Vol. 26, No. 1, pp. 13, 24.