Towards full cognitive radio

Wireless Spectrum is a valuable resource. Any doubt about that statement was dispelled by the AWS-3 spectrum auction, which is reported to have netted a total of $44.9B in bids. This huge amount of money licenses the cellular providers to use specific bands of spectrum in fixed geographical locations so that they can support their customers.

The concept of fixed and predefined frequencies and modulation techniques dates back to the earliest days of radio. It is enshrined in our frequency allocations and has worked well for the last century.

Unfortunately, today the demand for spectrum is outstripping the finite supply. Different techniques are being used to alleviate the crunch. As an example, more efficient modulation such as orthogonal frequency-division multiple (OFDM) is being adopted. With OFDM and efficient digital compression schemes, a degree of rationalization has been applied to terrestrial television channels in many countries. The analog transmissions have been replaced by digital services with greatly reduced bandwidths and guard band requirements. The result is that some spectrum has been freed up for other applications.

Even with these efforts, there is insufficient spectrum available to meet the ballooning needs of services such as cellular. Consequently, the planning and research for the next generation cellular system is driving towards frequencies up to 60 GHz and beyond. Operation at these millimeter wavelengths will be far from easy and the range will be measured in feet (or meters) rather than the current cells where the wireless footprint covers whole districts.

This technical legacy has resulted in a significant underutilization of many parts of the radio spectrum. Some reports suggest that at any one time only 10% of the available spectrum is being used.

Technical Advances
Cognitive Radio (CR) is a concept that promises to offer much higher utilization of the available spectrum. Briefly, a CR-enabled wireless would adapt to the current environment and use resources that are not being used. This simplified definition provides an insight into why CR has not already been adopted.

The ideal CR system would be able to monitor a very wide range of frequencies to identify which are being used and which are free at that moment in time.  

It is obviously impractical to have a bank of receivers individually tuned to scan an entire frequency band. However, the latest semiconductor technology makes it possible to use a programmable receiver that can be swept across a range of frequencies. The LMS7002M from Lime Microsystems, for example, can be tuned anywhere from 100 kHz to 3800 MHz. It features dual transceivers that are fully programmable on the fly. The gain and bandwidth are also programmable, and each channel includes Received Signal Strength Indicators (RSSI) in both the RF and base band domains. This allows the CR to scan RF frequencies of interest and then “home in” on any potential free channel. The CR can boost the gain and narrow the bandwidth to obviate the possibility of mistakenly jumping onto an occupied frequency.

For two or more CR-enabled radios to establish a communication channel, they must have a method of agreeing what frequency, power levels, and modulation scheme to adopt (Figure 1). One technique is to use a base station to pass messages between the participants. The base station need only be used to exchange connection data, and then the CR radios switch to the agreed settings.

Figure 1. CR sets can continuously scan the spectrum


During the call, the CR equipment needs to continuously monitor thechannel and other frequencies, because other (non-CR) users may start upa call. These are often termed Licensed Users, and they have priorityover the CR equipment and require protection from harmful interference.In this eventuality, the CR role will be to sense the Licensed User,establish and agree on another free frequency and to disconnect from thecurrent settings. To facilitate this, the cognitive radiospectrum-sensing methodology needs to provide time slots when it is nottransmitting, This is to enable the system to detect other signals andcan be accommodated within the frame format for the overall CR system.The base station can collate reports on spectral usage from several CRsets to update a table of free and busy channels. The base station canuse this table to allocate free channels against requests for service.The algorithm can protect against excessive hunting by a CR when a newnetwork configuration is required. This cooperative scheme alsomitigates the “hidden node” problem where a transmission from a remoteor shaded user is unnoticed by a particular CR. The direct signal may betoo weak to be detected, but it would cause interference to theexisting user if the CR started to transmit at the same frequency.  

Consider the operation of a CR in a typical wireless environment. Theinitial scenario in Figure 2 is that Channels A, B and D are found tobe currently unused with just wide-band noise detected. The CR setscommunicate via the base station and agree to use Channel A. During theCR call, one of the CR sets detects that a Licensed User has connected,and so the CR callers evaluate other frequency options and switch toChannel B that is still only registering noise. Again during the CR calla Licensed User is detected, and so the CR call switches to Channel Cwhich is now unused.

Figure 2. CR communications in the presence of Licensed Users

Field Programmable Radio Frequency (FPRF) Semiconductor Devices
Thetechnical challenges of being able to scan across a wide range offrequencies, detect activity levels and rapidly switch frequencies areall addressed by the latest FPRF devices. In addition, the CMOStechnology provides a low power and highly cost-effective single chipsolution.

To detect and utilize free channels requires the creation of spectrumsensing algorithms that detect energy levels to allow access in anopportunistic and non-interfering manner. The LMS7002M features threeindependent Low Noise Amplifiers (LNAs), which are respectively used aslow, high and wide frequency band inputs. In initial search mode, thehigh and low LNAs might be depowered, while the RF signal is processedvia the wide band input. This input features an RF received signalstrength indicator (RSSI) on this LNA that detects inputs from -70 dBmto -20 dBm. The signal is digitized and the strength calculated bycomputing the “square root of two”. This allows the automatic gaincontrols (AGC) to set the gain over a 70dB range to ensure that clippingis avoided.

If the signal band of interest is in the range 100 kHz to 2 GHz, thenthe low-band LNA can be used. Alternatively, the high band LNA acceptssignals above 1.5 GHz.

The receive frequency can be rapidly scanned to allow a coarsedetection of signals in the spectrum of interest. The receive bandwidthcan be adjusted on the fly. A wide bandwidth might be selected for theinitial scan, and narrower filters used to “home in” on anytransmissions. Each of the dual transceivers is equipped with a digitalbaseband RSSI that can be used to derive the spectral usage. The CRneeds to be capable of ignoring spurious transmissions and interferenceas well as transmissions made by the cognitive radio system itself.


Transmissions using OFDM will be more difficult for the CR to detect.This is because OFDM uses a large number of subcarriers where each ismodulated with part of the message. The carrier frequencies areorganized so that they do not interfere with each other, because of theorthogonality designed into the system. However, the individual signalsfrom each carrier are detected and aggregated at the receiver and thereconstituted digital data fragments provide a higher data rate. As aresult, OFDM can mimic the characteristics of noise. However, OFDM isdesigned to operate in the presence of noise and so the algorithms usedto recover the data include error correction capabilities. It is thesecharacteristics that make OFDM one of the least intrusive transmissiontypes for CR to employ.

The most significant new scheme is called Multiple-InputMultiple-Output (MIMO). This is a complex configuration that uses two ormore antennas which are separated by a physical distance. MIMOtechniques improve the spectral efficiency and achieve a diversity gainthat improves the link reliability. A configuration using two antennasat the transmitter and two at the receiver is termed a 2 x 2 MIMOsystem. The FPRF is designed with a dual transceiver architecture tosupport 2 x 2 MIMO in a single chip. This can be extended to, say 4 x 4by the addition of more antennas and transceivers.

The receive and transmit bandwidths can be programmed from 1.4 MHz to56 MHz bandwidth. This allows the CR to select a frugal bandwidth forlow data requirements like speech or exploit high throughput forstreaming video or other more demanding uses.

The FPRF (Figure 3) is controlled by data in a simple memory maploaded into the device via an SPI interface. This allows total controlof each parameter, as well as the ability to bypass or depowerindividual blocks. So, during scan mode, all the transmit blocks can bedepowered.

Figure 3. FPRF block diagram

During a duplex call, transmit and receive paths in both transceiverscan be fully programmed on the fly. This allows the device to adjustthe transmit signal levels so that the signal to noise ratio at the farend is adequate. Similarly, the RSSI in the receiver allows the CR toadvise the far end to adjust its transmit power. This allows optimumcommunication with a minimum of potential interference and alsominimizes the power consumption.

The digital baseband data streams take the form of in-phase andquadrature (I&Q) components. The TX carrier is determined by thedata loaded into the SPI memories at that instant and the frequency isgenerated in the TX PLL. The RF and data are combined in the mixer toproduce the complex signal that will be transmitted. Similarly, the RXPLL generates the instantaneous frequency for the direct conversionmixer (sometimes called Zero-IF) to convert the RF to baseband andrecover the I&Q data.

Size, weight, power, and cost, commonly abbreviated to SWaP-C, is ofgrowing importance in all hand held equipment. The low cost LMS7002M hasa frugal power requirement of 550 mW in single input single output(SISO) mode or 880 mW in MIMO, and can be powered by a single supplyrail. The size and weight requirements are addressed by the devicerequiring only a minimum of external components.


Baseband Companion Chip
There are a few optionsfor the baseband chip, but the choice usually boils down to a dedicatedprocessor or an FPGA. The processor is optimized for intelligentdecision making, while the FPGA excels at hardware acceleration ofalgorithmic tasks. However, recent product introductions have seen theintegration of high performance processors embedded into FPGAs. TheArria10 SoCs from Altera offer dual 32-bit ARM cores that clock at 1.2GHz alongside up to 660k logic elements. The combination of theprocessor and FPGA logic fabric on the same device (Figure 4) providesmuch higher performance, without the power drain normally associatedwith running signals off-chip.

Figure 4. FPRF and FPGA configuration

FPGAs are configured on power-up from an external memory such asFlash. This memory can also contain the executable code for the ARMcores, as well as tables of frequency information and SPI memory maps toload into the LMS7002M transceiver. The Flash also provides thecapability of loading mid-life firmware updates to add new capabilities,fresh channel allocation data or to correct bugs.

The benefits of using an FPGA with embedded processors becomes clearwhen the overall functionality is partitioned between software andhardware tasks. Typically the software will control the LMS7002M byprogramming it to scan for a free channel. This is achieved by loadingsuccessive data patterns into the SPI memory to tune the PLL to thedesired frequency. The programmable fabric would accept the data fedback from the RSSI on the transceiver to identify what RF activity ispresent.

The ARM cores would also handle the protocol of negotiating with thebase station for both outgoing and incoming call set up routines.Included in this data would be the frequency, bandwidth, power level,and modulation scheme to be used for the connection. The processor wouldalso control the programmable fabric core to ensure that the datastreams produced the agreed modulation scheme. The hardware accelerationof the fabric is ideal for the encoding and decoding of I&Q dataand for error detection/correction. The abundance of dedicatedvariable-precision digital signal processing (DSP) blocks that canhandle floating point calculations is perfect for high performancefiltering.

If it is required to implement encryption of the data, then this taskcan be performed in the logic fabric. For example, data confidentialityis provided by using the Advanced Encryption Standard (AES) or the morecomplex AES-GCM variant that also ensures that the data has not beenmaliciously altered or corrupted in transmission.

The Arria10 family provide the opportunity for high performancebaseband functions. In applications that do not require the highestspeeds, the cost-effective Cyclone V SE can be used that also featureembedded ARM cores. In both options, designers can match the tasks tothe most appropriate resource and also achieve the lowest powerconsumption.


Applications for Cognitive Radio
The primeapplication area today is for military communications. CR is a logicalextension to Software Defined Radios (SDR). SDR sets have been deployedto military users around the globe and support a range of differentwaveforms.

CR sets will see early adoption for military use for both technicaland logistical reasons. OFDM transmissions are by nature stealthy, sodesigners can exploit the ability of the LMS7002M for fast frequencyhopping with very low lock time to evade attempts to track and interceptsignals. Inherently CR sets also have capabilities suited to thedetection and avoidance of jamming signals or other extraneous signalsthat can be encountered on the battlefield. Logistically, the sets canbe made interoperable between different branches of the military andwith coalition and NATO forces. The bandwidth allocations for militaryuse have been divided between different services and for specified usessuch as radar. This resource is under pressure from a rapidly expandingdemand for bandwidth for Communications, Intelligence, Surveillance, andReconnaissance (CISR) on the one hand, and regulators like the FederalCommunications Commission on the other.

The penetration of CR into other applications from railwaytransportation to public safety applications will be more measured, asprogress will depend on movements on political and regulatory issues.

Political and Regulatory Issues
Overcoming thetechnical obstacles is only part of the problem, because before CR canbecome a mainstream wireless technology there are significant politicaland regulatory barriers that must be overcome.

The concept of unlicensed users utilizing spectrum that is ostensiblyreserved for other purposes is alien to the regulatory regime in placetoday. This is understandable when companies such as cellular providershave invested large sums to acquire spectrum which has a highutilization. That is not necessarily the case for other bands, such asthose reserved for fixed, mobile, or military use. Large swaths ofvaluable spectrum could be opened up to CR users given sufficientassurances on interference mitigation. An extreme analogy of the currentsituation would be the building of a freeway that was reserved for thesole use of delivery trucks.

There are proposals to introduce an addition to cellular systems thathave some parallels to the CR concept. Current planning for LTEnetworks includes a feature called LTE Assisted Access (LAA) technology,or an alternative known as LTE-Unlicensed (LTE-U). LAA is currently inthe standards-setting process with the 3GPP. The idea is that a mobileuser sets up a call or session using the conventional cellular network,and if WiFi is available to the receiver, the bulk of the traffic isthen transferred to the WiFi network. WiFi is unlicensed and so thecellular operators argue that it is a legitimate use of that resource.This is being hotly debated in some quarters to ensure that allunlicensed devices operating in the band have fair and reasonable accessto 3.5 GHz spectrum.

Clearly any CR should avoid the public safety networks for firstresponder organizations or the emergency services. These are bandsreserved for Project 25 (P25) and Terrestrial Trunked Radio (TETRA)services. To help ensure that CR users do not unintentionally infringeon these bands, the equipment can be supplied with frequencies thatshould be avoided at different geographic points. This requires that theCR use geolocation via GPS or equivalent and combine it with “no-go”look up tables.

The spectrum crunch is putting greater pressure on regulators tomaximize the use of this precious resource. Cognitive radio offers onepossible solution for addressing our insatiable appetite for morebandwidth.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.