 
|
  
   |
| About
CE-Mag Free Subscriptions Current Issue Article Archives ESD Help Mr. Static Web Gallery Staff Info Contact us
|
|
|
||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||
Unlicensed Wireless Data Communications, Part I: Defining RequirementsTim Cutler Choosing the right frequency band and protocol is critical to designing successful wireless data products. The need for fast and reliable wireless data transfer is exploding. As wireless products have gained mainstream acceptance, developers are hard at work applying wireless technologies, particularly those using license-exempt frequency bands, to free such applications as inventory, process, and remote control from the physical tether of wiring. One key to these efforts is the optimization of operational parameters for the specific application; a design approach that works best in one setting is not necessarily portable to another. A successful design begins with a determination of operational criteria and an analysis of such performance characteristics as network topology, interoperability, power, and range. Defining Wireless Data Requirements Some of the questions that must be addressed in the design stages of a wireless application lead to trade-offs—like speed versus range. Others have mutually exclusive answers—like proprietary versus standards-based protocols. Topology. The nature of wireless applications generally leads to a star configuration: a centrally located access point communicating with one (point-to-point) or more (point-to-multipoint) remote devices. Even when multiple access points are used, each is the center of a star arrangement. In determining the right topology for a wireless application, the following must be answered: Is the best configuration point-to-point or point-to-multipoint? Are peer-to-peer communications required? Will a polling scheme be used, or will remotes transmit data as needed? Will remotes have to forward data from other units? How important is seamless roaming between access points? Speed and Latency. Speed is important in straight-line racing, but road-course drivers know that many other factors, such as handling and braking, are just as critical. The same is true for wireless data: Speed is not the only issue, and faster is not always better. In fact, speed has tradeoffs with two other important criteria: range and latency. The best designs limit connection speed according to data quantity and time requirements. Closely related to speed, latency measures the time required to transmit a data payload between devices. Latency and throughput are generally counterproductive; with decreased transmission overhead, longer packets increase throughput but increase latency when other devices have to delay transmission as a result. This makes latency tolerance (and variance) an important application design factor. Range. Any wireless application must account for both how devices will be used and over what distances. How far apart will the devices be? Are they in line of sight or around obstacles? If needed, range and operability can be extended with multiple access points and repeaters. Are these options acceptable? Interoperability. Does the product need to communicate with other vendors' products over the wireless link? Multivendor interoperability is often perceived as a benefit that lowers product cost through market competition, but does this make sense in the application? Because they must encompass a wide range of uses, standards are accompanied by performance tradeoffs and security issues. All aspects of a radio-frequency (RF) standard must be thoroughly evaluated for compliance with product requirements. Power Consumption. Power consumption is particularly important in battery-powered devices such as wireless remotes. The design must determine the run time required by the application, the battery type used (e.g., rechargeable versus disposable), the total amount of power consumed, and voltage levels available to the device. Size. Consider the desired form factor for the wireless device and the space available to it. Wireless functionality is easily accommodated in a new design, but retrofitting may limit size and placement. Host Intelligence. How much host intelligence is present? It is likely to be unequal at the ends of the RF link. Will the link require more intelligence? If sufficient intelligence exists at both ends of the RF link, issues such as the handling of protocol stacks are nonexistent. But if one end offers only limited intelligence, a transparent communications scheme may be required. Is intelligence sufficient at both ends to support a networking or addressing scheme at the application layer, or should that be part of the transceiver? What interface types does the host device have? Are the ports parallel or serial? Is there a PC card interface? Environment. What is the device's intended operating environment—office, home, or factory? What operating temperature range is needed? How rugged does the device need to be? Will it be subjected to wash downs? Selecting a Frequency Band Because the selected frequency band will drive many other design factors, it's a good place to begin the design process. (If interoperability with a specific standard is a requirement, that interoperability will define the frequency band.) There are three main unlicensed-spectrum frequency bands suitable for sophisticated data transmission. The industrial, scientific, and medical (ISM) bands in the United States include 900 MHz, 2.4 GHz, and 5.8 GHz. The unlicensed national information structure (U-NII) and high-performance radio local-area networks (HiperLAN2) bands are in the 5 GHz band. Unfortunately, the 5.8 GHz ISM band, the U-NII band, and the HiperLAN2 band use slightly different portions of the 5 GHz band. Table I shows unlicensed bands and the requirements for operating in each.
One decision involves whether a particular band will meet technical requirements. Is there sufficient bandwidth for data throughput and collocation? The 900 MHz band has 28 MHz of spectrum available, and the 2.4 GHz band has 83.5 MHz of spectrum. The ISM 5.8 GHz band has 125 MHz; the U-NII band has two bands of 100 MHz and a third with 75 MHz. The HiperLAN2 band provides one band of 100 MHz and a second band of 275 MHz. Although a lack of bandwidth can be countered by using a higher symbol-per-bit modulation technique, this technique reduces a radio's receive sensitivity and thus the range. If multiple independent networks must be collocated, is sufficient bandwidth available to support them? For example, 802.11b uses 22 MHz per network. The 83.5-MHz-wide 2.4 GHz band makes possible three nonoverlapping networks. One important consideration in frequency-band selection is the geographic market. The 900 MHz band is unlicensed in the United States and Canada, but is unavailable in Europe. The 2.4 GHz band is unlicensed throughout most of the world, but some countries restrict it. And the 5 GHz bands are slightly different in the United States and Europe. Another consideration is the number and types of existing products using the band. Many designers are concerned about interference from 900 MHz phones, but many 900 MHz data products are used in plants where there aren't many phones. The 2.4 GHz band is also used by microwave ovens. The pulsed operation of these ovens usually has no impact on a well-designed radio. A more important consideration is the proliferation of 2.4 GHz wireless-LAN products. But these products obey spread-spectrum rules, allowing products to coexist. At this time, the 5 GHz bands are less utilized. It is important to note that the U-NII and HiperLAN2 bands do not require spread-spectrum technology, and systems collocation could be problematic. Selecting Multipath Indoor and typical short-range RF communications in unlicensed bands must allow for multipath fading and the presence of other unlicensed radios. The three main techniques for this are frequency hopping, direct sequence (spreading techniques), and orthogonal frequency-division multiplexing (OFDM). Although OFDM is similar to frequency hopping and direct sequence, it is not technically a spreading technique. Frequency bands operating under FCC Part 15.247 have specific spreading requirements, and FCC recently allowed OFDM to operate in the 2.4 GHz band. Direct Sequence. Direct sequence multiplies a data signal with a pseudorandom noise signal to spread it. Because of the number of chipsets available for implementing 802.11b products, implementing a direct-sequence radio in the 2.4 GHz band is straightforward. Direct sequence as implemented in 802.11b occupies 22 MHz of spectrum while providing a maximum over-the-air data rate of 11 Mb/sec. The over-the-air data rate is the most appealing feature of direct sequence. Occupying only 22 MHz of spectrum, however, this option is limited in power density when compared with frequency hopping. This limitation makes direct sequence more vulnerable to interference from frequency hopping. Direct sequence, when occupying 22 MHz of spectrum, can have just three nonoverlapping collocated systems. Direct-sequence radios can meet FCC requirements without being 802.11b compatible. FCC requires at least a 10 to 1 spreading ratio, so more-narrowly occupied bandwidths with higher power densities are possible. The trade-off is lower over-the-air data rates, or less receive sensitivity. Implementing this product requires a custom digital-signal processing (DSP) design for direct-sequence spreading. Frequency Hopping. Frequency-hopping radios transmit for a short period on a single frequency and hop to other frequencies in pseudorandom fashion, which is generally regarded as having better interference immunity than direct sequence. The trade-offs are the narrower spectrum occupied by each hopping channel and the reduced over-the-air data rate. FCC rules require a frequency hopper in the 2.4 GHz band to hop over 75 MHz of spectrum at least once every 30 seconds, with each portion of spectrum used equally. Previously, hopping channels were limited to 1 MHz; new rules allow hopping channels up to 5 MHz at reduced transmit power levels. Frequency hopping allows a number of collocated systems because of the relatively large number of frequencies used. Frequency hopping avoids interference by hopping over a wide range of frequencies, theoretically with no wideband interferer. Multipath fading is handled by hopping over different frequencies, each of which has a different multipath effect. If multipath fades one channel, other channels are not faded. Although frequency hopping does not have the DSP requirements of direct sequence, it has its own challenges, including rapidly changing frequencies and synchronizing radios to the hopping. OFDM. OFDM is similar to frequency hopping except that it transmits over multiple channels, or subcarriers, simultaneously. All subcarriers are orthogonal, each with a null at the center frequency of all other subcarriers to avoid intercarrier interference. The RF spectrum of an OFDM signal looks similar to a direct-sequence spectrum, but the similarity ends there. The theory is that although one or more subcarriers may be affected by multipath fading or an interferer, others will get through. OFDM also makes use of training channels to characterize the RF paths, and it interleaves the data over multiple channels to improve its reconstruction when subcarriers are impacted. For 802.11a, the use of forward error correction is required. OFDM has been promoted as a superior non-line-of-sight technology, particularly for last-mile applications. OFDM is similar to 802.11b in the use of collocated networks because they occupy similar spectra. Because OFDM has less field experience than direct sequence and frequency hopping, however, the jury is still out. Chipsets for 802.11a are beginning to appear on the market. Similar to direct sequence, OFDM radios will be implemented using chipsets because of the DSP complexity involved. Which technique is best? It depends on the application and other design considerations. Frequency hopping generally offers superior reliability in a rising noise-floor environment. Direct sequence can provide higher over-the-air data rates. OFDM is promising, but is just now coming to market. Selecting a Protocol Interoperability with other manufacturers' products requires a standard protocol. The two dominant standards are IEEE 802.11b and Bluetooth. 802.11b was developed primarily for office LANs. Bluetooth was developed originally for personal area networks (PANs) and was optimized for very-short-range applications. If interoperability among multiple manufacturers' products is not a requirement, protocol choice is open to standard or proprietary options, including custom protocols. The protocol defines both the transmission of data over the air and the RF device interface, and different applications have different needs. For example, is the data flow primarily between remote units and a central location, or is the data flow primarily between peers? Because the protocol affects the overhead associated with the data transmission, the protocol must fit the application. Error Detection and Correction. Error detection and correction is a critical protocol consideration. Although some systems make the host device responsible for this, most designs focus on the radio. The main approach to error detection is cyclic redundancy code (CRC) checks. Some basic designs rely on checksums, but a CRC is necessary for reliable data. Shorter CRCs allow too many errors, and longer CRCs don't significantly add protection for the increased overhead. A 24-bit CRC is usually the best choice; for a radio with a bit-error rate of 1 x 10–5, a 24-bit CRC should theoretically allow one error to pass undetected in 24 hours of continuous transmission. Forward error correction (FEC) and automatic retransmit request (ARQ) are the main error-correction techniques. ARQ is error correction by detection and transmission, whereas FEC corrects receive errors. FEC includes additional information that allows data to be reconstructed at the receiver upon error detection. The drawback is the overhead of additional data that are always sent, even when no error occurs. Depending on the level of FEC implemented, the overhead can approach 50%. In typical applications, FEC adds 3–4 dB of coding gain to the link budget. ARQ uses an acknowledgment to indicate that data were received without error. If an acknowledgment is not received, data are retransmitted. The approach offers significantly lower overhead—when data are received without error, the only overhead is the acknowledgment—typically a few bytes. Multipoint Operation. The protocol also performs the important function of addressing intended receivers, with some mechanism to accommodate multiple remotes transmitting at about the same time. The two schemes used in unlicensed applications are carrier sense multiple access/collision detect (CSMA/CD) and time-division multiple access (TDMA). CSMA/CD provides good bandwidth utilization in lightly loaded networks, but with wide latency variations and collisions. TDMA does not provide the same bandwidth utilization, but provides fixed bandwidth per remote, smaller variations in latency, and no collisions. When multiple remotes frequently transmit data of similar amounts, TDMA can provide optimum bandwidth utilization. Because every remote is assigned a time slot and every remote transmits frequently, there are no collisions and few unused time slots. CSMA/CD in such an application might require more bandwidth to allow for frequent collisions or back-offs. When remotes transmit infrequently (or some remotes transmit more data than others), CSMA/CD can provide better bandwidth utilization. Collisions and back-offs are infrequent, but there would be fewer unused time slots than with TDMA. Flexibility. Flexibility is a very important protocol consideration. Can the protocol be configured to meet device needs? How easy is it to configure? Standard protocols have inherent flexibility limits, and proprietary protocols typically offer greater flexibility. Standards-Based Protocols. Four primary standards-based protocols are viable: 802.11b, 802.11a, 802.11g, and Bluetooth. 802.11b. The most widely used standard protocol, 802.11b, requires direct-sequence spread-spectrum (DSSS) technology, specifying a maximum over-the-air data rate of 11 Mb/sec and a scheme to reduce the data rate when higher data rates cannot be sustained. This protocol supports 5.5 Mb/sec, 2 Mb/sec, and 1 Mb/sec over-the-air data rates in addition to 11 Mb/sec. It uses CSMA/CD for multipoint operation and allows peer-to-peer communication as well as access points. To support the 802.11b protocol stack, sufficient intelligence is required at all nodes. Devices autoauthenticate, meaning they can always join an 802.11b network. The 802.11b standard has developed good interoperability for basic functions. (Some functions, such as seamless roaming, are manufacturer dependent.) PC cards are available at low cost but tend to be power hungry, consuming more than a watt during transmit. Designed as wireless Ethernet, 802.11b requires the most host intelligence and is complex to integrate. 802.11a. The recently approved 802.11a standard moves 802.11b functionality to the U-NII band, and chipsets are just now coming to market. The 802.11a standard specifies OFDM using 52 subcarriers for interference and multipath avoidance, supports a maximum data rate of 54 Mb/sec using 64QAM, and mandates support of 6, 12, and 24 Mb/sec data rates. FEC is also required, with coding rates of 1/2, 2/3, and 3/4. This protocol specifies minimum receive sensitivities ranging from –65 dBm for the 54-Mb/sec rate to –82 dBm for 6 Mb/sec. Also designed as wireless Ethernet, 802.11a requires the same host intelligence and integration effort as 802.11b. 802.11g. The 802.11g standard, which is under development, provides 802.11b functionality, data rates up to 54 Mb/sec, and operation in the 2.4 GHz band. Bluetooth. The Bluetooth protocol specifies frequency-hopping spread-spectrum (FHSS) technology. With a fixed hop time of 625 microseconds and fixed over-the-air data rate of 1 Mb/sec, Bluetooth delivers 768 Kb/sec maximum throughput. Bluetooth supports access points and peer-to-peer communications, and it organizes devices into eight-member piconets that can then form ad hoc connections among piconets to build larger networks. Bluetooth supports multiple data types, including voice, by providing variable-bandwidth channels to different devices. Although Bluetooth is fairly complex, it makes use of different profiles to support such connections as serial or Ethernet ports. There are currently few suppliers and products, but Bluetooth is expected to gain popularity due to the potentially low cost of chipsets. Bluetooth was designed for short-range communications (about 10 m), but also has specifications for higher-power and longer-range applications. Bluetooth is similar in complexity and host intelligence requirements to 802.11b. Proprietary Protocols. Choosing an appropriate proprietary protocol requires ensuring that it meets application requirements. In many cases, a proprietary protocol may fit better than a standard; when interoperability among wireless manufacturers' devices is not a consideration, proprietary protocols often provide more flexibility. The key is to choose products from a supplier with a proven track record and sound financial standing. It is also important to make sure the company's product roadmap aligns with your future application needs. Custom Protocols. The remaining option—if a radio is to be designed in-house— is to develop a custom protocol. This option may yield a highly customized solution, but is also the most difficult route. A custom design requires substantial RF protocol engineering experience, and ensuring its reliability is an extremely time-consuming and iterative process. RF Requirements RF parameters for wireless devices can be specified once the appropriate frequency band and protocol have been identified. The topic is fully covered in the related article, "Unlicensed Wireless Data Communications, Part II: Specifying RF Parameters," on page 145. Regulatory Requirements Regulatory requirements for the United States and the European Union are straightforward. In the United States, the radio or the device containing it must be submitted to an approved laboratory for testing to appropriate FCC rules. The test data are submitted to FCC for review and type approval. If a radio is submitted by itself and subsequently approved, the radio can be integrated into another device without having to submit the combined device for type approval. However, FCC may require other tests on the combined device (e.g., computing devices), and the FCC ID must appear on the outside of the combined device. The European Union harmonized the majority of nations' requirements for unlicensed devices in the 2.4 GHz band under the Radio and Telecommunications Terminal Equipment (R&TTE) Directive of 2000. Beginning in April 2001, all radio devices operating in the 2.4 GHz band must be approved under a CE marking process. Meeting the essential requirements enables declaration of CE conformity. Manufacturers must notify each EU country of intentions to market a device in that country. Each country has four weeks to deny the right to market, and lack of denial notice means acceptance. As in the United States, in Europe an unlicensed device can be CE marked at the radio or device level. If a CE marked radio module is to be integrated into a device, the device manufacturer or marketer (although not needing to have the device retested for the radio) must submit the device for the required testing to meet the CE criteria for that device. The manufacturer or marketer is also required to notify each country of its intention to market the product in that country. The rest of the world is somewhat problematic. Many countries accept either FCC or CE test results, but also have their own testing requirements. Additional testing is often required because only part of the 2.4 GHz band may be unlicensed in a given country. This homologation process can be a simple submission of test reports, with approval within weeks, or a complete retesting and approval process that can take several months. Make or Buy As usual, the make or buy decision is driven by two factors: cost and time to market. Cirronet (Norcross, GA) conducted a study of the costs associated with implementing a reference design versus buying a third-party radio. The study shows that, assuming appropriate RF ability and related test equipment in-house, an annual volume of between 5000 and 10,000 units is required to reach cost parity with a quality third-party radio. In terms of time-to-market, an existing modular third-party solution with FCC and CE certification can save anywhere from three to six months, based on the certification requirements. Designing an unlicensed-band radio requires consideration of many details and the mastering of many skills. The availability of chipsets and reference designs can mitigate the amount of required RF design. However, modifying a reference design can mean that much of the reference design benefit is lost. Designing a radio without a reference design requires care to completely understand the time and effort involved. RF design is as much art as science. Even if a reference design is used unmodified, unless the reference layout remains unchanged, the printed circuit board layout can provide a daunting challenge to those without RF engineering experience. Expenses that are often overlooked include those for the RF test equipment needed to implement even a reference design. The price of this equipment can reach $100,000. Another cost involves testing for FCC and European Telecommunication Standards Institute (ETSI) certification and the associated certification and notification fees, which can exceed $10,000. Finally, and possibly the source of highest cost, is delayed market entry. A more qualitative consideration is the role of the wireless feature in the product. Does the device perform a core function and use wireless technology as an adjunct to that function, or is the wireless function the core function? This addresses the question of the core competency of the integrating company and the strategic nature of the wireless capability. Conclusion Unlicensed radio offers end-users the convenience and power of radio communications without the need to obtain frequency licenses. But for the equipment provider, it offers a vast array of choices and decisions that are not easy to make without RF experience. Fortunately, third-party radio solutions allow wireless capability to be added without adding an RF design team, when those products are appropriate. The key is to completely understand the wireless requirements of the device, and to fit the RF solution to those needs as closely as possible. Over-designing the wireless capability can lead to operational problems, but choosing the cheapest solution can cause more problems. Take the time in the beginning to understand the needs of the wireless application so that the solution chosen is the right one. Tim Cutler is vice president of sales and marketing for Cirronet Inc. (Norcross, GA). He can be reached at tcutler@cirronet.com. |