Click here for the enhanced PDF version of this article

Broadband connections are difficult to come by in remote or sparsely populated locations because the customer base is too small to justify the expense of installing wired networks. Satellite service may be available in these areas, but it has the significant disadvantage of requiring “line-of-site” access for reliable transmission. Trees, buildings or even the weather, may cause interference. Worldwide interoperability for microwave access (WiMAX) is a wireless technology that solves this problem by providing “last-mile” broadband connections using radios instead of cables or telephone lines. Unlike satellite connections, WiMAX does not need direct line-of-sight access to provide wide-area broadband access. A single WiMAX base station can provide broadband speed service to thousands of customers within a 3 km distance or backhaul functions at distances to 30 km (Figure 1).

There is a variety of IEEE 802.x standards that include Wi-Fi (802.11), ZigBee (802.15.4) and WiMAX (802.16), among others. These standards define wireless connectivity within certain RF bands. The most well-known example of this is the 802.11x standard for Wi-Fi, used in wireless home networks and cyber cafés. However, setting a standard for radio transmission and reception does not mean that the equipment made by competing vendors will work together. Multiple factors pose obstacles to interoperability, including physical layer (PHY) performance, media access controller (MAC) protocols, layer messaging, and encryption methodologies, to name a few.

A broadband network must be accessible to all notebooks, PDAs and other equipment, from every vendor, so interoperability becomes a big issue. This is where the WiMAX Forum — an industry-led, non-profit corporation formed to promote and certify compatibility and interoperability of broadband wireless products that operate on the 802.16 standard — steps in. Worldwide member companies support the industry-wide acceptance of IEEE 802.16 and European Telecommunications Standards Institute (ETSI) HiperMAN wireless metropolitan area network (MAN) standards. WiMAX-labeled products must complete a certification process, specified by the WiMAX Forum, to demonstrate their ability to interoperate. To facilitate this process, the forum holds periodic sessions called interops, during which vendors test the interoperability of their products with competing vendors.

WiMAX, based on IEEE standard 802.16, is intended for use in stationary equipment, such as desktop PCs and mobile equipment, which includes notebook computers, mobile phones, personal media players (PMPs) and PDAs. Because there are substantial differences in the characteristics of radio signals being transmitted and received by stationary vs. mobile devices, WiMAX profiles are based on multiple IEEE 802.16 standards: one for stationary equipment and another for mobile equipment that may be in motion while receiving or transmitting signals. To get the standard deployed as quickly as possible and support various degrees of mobility, the WiMAX Forum has described five stages for the implementation of the standard: fixed, nomadic, portable, simple mobility, and full mobility.

An inherent challenge to deploying any broadband network is getting enough people to subscribe to it to make it useful to consumers and economically feasible for service providers. The best way to foster adoption of the standard is to make it affordable. Toward this end, WiMAX CPE vendors have set a target bill of materials (BOM) cost of $100.

A major factor affecting the cost of any wireless system is the architecture of the radio. Radios that are not designed specifically for WiMAX applications may need hundreds of external components in order to transmit and receive signals. IEEE standard 802.16-2004 radios use orthogonal frequency duplex modulation (OFDM) to modulate the data.

The OFDM technique splits raw data into different frequencies called channels. The number of available channels is dependent on the frequency band of the standard and the channel bandwidth being used in the application. This approach reduces the processing effort required to compensate for multipath interference. Since each subcarrier operates at a relatively low bit rate, the duration of each symbol is relatively long. Synchronization of the signal timing is much easier due to the low bit rate and long duration.

The 802.16-2004 WiMAX band is split into three different radio frequency bands, 2.5 GHz and 3.5 GHz for licensed bands and 5.8 GHz for unlicensed, each of which has unique processing requirements that are incompatible with the other frequency bands. The channel bandwidths in the licensed bands are 1.75 MHz, 3.5 MHz, 7 MHz, 14 MHz, and 28 MHz. For the unlicensed band, the channel bandwidths are 5 MHz, 10 MHz and 20 MHz. The bandwidth for each channel is determined by the number of channels required for a given application.

For example, in the 3.5 GHz band, 3.5 MHz bandwidth allows 1024 channels. The huge number of possible combinations of frequency band and channel bandwidth could lead to an equally huge number of WiMAX profiles, significantly complicating the specification and certification process within the WiMAX Forum. It could also lead to higher-cost equipment, since vendors might be forced to a provide solutions for every possibility.

To avoid this unnecessary complexity, the WiMAX Forum considers only a small number of profiles for inclusion in the standards as they are finalized. For example, the 802.16-2004 standard included only five profiles when it was first certified. Two more were added later. The forum is in the process of determining which profiles will be included for certification in the newly ratified 802.16e standard. However, even a small subset of profiles poses a problem in terms of selecting a proper radio for WiMAX applications.

To address the issue of multiple 802.16 bands, the engineer can select a radio that is reconfigurable across a large range of frequencies and bandwidths, or a particular frequency band and bandwidth and use a radio that works just in that band. Radios with multiple frequency bands and multiple bandwidths provide the most flexibility. They are usually implemented in a “double-conversion” architecture that requires expensive SAW filters to define each different bandwidth. Supporting three bands immediately increases the BOM by about $30. Each frequency band also requires its own voltage-controlled oscillator (VCO) to set the frequency band. Each VCO requires hundreds of additional external components to get a clean signal in all bands. Flexibility notwithstanding, the high $200+ system cost associated with a multiple bandwidth radio may make systems prohibitively expensive and severely hamper market adoption.

Performance may be another issue with multiple bandwidth radios. The WiMAX Forum vision for long-distance communication and high throughput can be demanding for transceiver output power and receive sensitivity. Changing the frequency and/or the bandwidth alters transmit and receive performance of the radio. At the higher spectrum, it may cause transmit power or receive sensitivity to fall below what is needed for interoperability. Recommended transmit power at the antenna for a WiMAX CPE device is +30 dB, while receive sensitivity -80 dB. This problem can be overcome by adding high-performance low-noise amplifiers (LNAs) and power amplifiers (Pas) to get the system into the desired range for full inter-operability. However, it will increase the system cost.

A second option is to use a radio architecture that operates in a single band of the 802.16 bandwidth. Using a single band limits the radio's flexibility. However, the bandwidth limitation may not be as confining as it appears. Initially, the business model for WiMAX is expected to be similar to that of the mobile phone industry. End customers will subscribe to a carrier service that provides the WiMAX equipment and connection, just as mobile phone companies provide the mobile phone and the connection today. Thus, any WiMAX consumer will need to communicate only within the network to which he or she has subscribed, at whatever bandwidth the carrier selects. Service providers and consumers will not need multiband radios, in the same way that subscribers to AT&T's GSM phone service do not need CDMA radios. The service provider can select a frequency band and allocate the bandwidth of that channel, as required, to meet end-user demand. For example, in the 3.5 GHz band, carriers may operate at 1.75 MHz, 3.5 MHz, and 7 MHz. A 7 MHz channel bandwidth allows more data to be transmitted in each packet, but limits the subscriber's distance from any base station. A 1.75 MHz channel bandwidth allows less data to be transmitted in each packet, but users can be farther away from the base station. This mixture of different channel bandwidths provides more efficient coverage for all subscribers. The IEEE 802.16 standard maintains data throughput by adapting the modulation technique to the total area of coverage.

The use of multiple band and channel bandwidth combinations is unique to the 802.16 standards and mandates careful evaluation of the variety of single-band radio architectures available. There are three basic radio architectures including double conversion, direct-conversion zero-IF (intermediate frequency) and direct-conversion low-IF. The different radio architectures will affect total system cost and performance in different ways.

Double conversion, also called superheterodyne, architectures use two intermediate frequencies to filter and amplify the incoming weak RF signal. This method results in two image frequencies, which are filtered to eliminate interference. The advantage of dual conversion is that because the first intermediate frequency is typically fixed, it is easier to compensate for the local oscillator (LO) phase noise. This is a good solution for applications in which high performance and good receive sensitivity are important. The disadvantage of this architecture is that the required additional filters and external components may result in a system cost that puts WiMAX out of reach of the mass market. In fact, to meet the 802.16 specification, a double-conversion radio will require about 600 external components that result in a BOM of more than $150; thereby making it less than ideal for highly integrated systems (Figure 2).

Direct-conversion radios virtually eliminate sensitivity to image interference by offsetting the signal from the zero sub-band and then using a direct current (dc) offset correction to compensate for the offset effect from the radio. The channel filtering and amplification are done at the baseband frequency, allowing a large number of components to be integrated into the RF silicon. This feature makes direct-conversion radio architectures ideal for 802.11a and 802.11g Wi-Fi and WLAN applications because modulation techniques required for ODFM fit easily with the architecture.

In the case of 802.16, however, direct conversion may actually create interference because the initial frequency difference between a base station and a subscriber could be equivalent to one or more of the subcarriers of the OFDM channel frequencies. This dc offset could disturb some subcarriers in the OFDM symbol around the zero subcarrier. This deficiency can be resolved by adding a high-resolution temperature-controlled, voltage-controlled crystal oscillator (TCVCXCO) or a high-resolution synthesizer that tunes the radio frequency to within 1% to 2% of the subcarrier frequency spacing (85 parts per billion at the 3.7 GHz and 3.5 MHz bandwidths). The ac coupling frequency of the offset correction must be less than a few kilohertz during this operation. The drawback to this approach is that the small frequency spacing may require settling time of as long as 100 us when switching the transceiver from transmit (Tx) to receive (Rx) mode.

A solution to this issue is to use a frequency dynamic offset correction, which operates like an offset sample and hold. However, the zero-IF receive path requires coordinated control of both the frequency and offset correction that is extremely difficult to integrate into the radio and equally difficult to manage between the radio and the baseband. As a result, the zero-IF radio subscriber takes more time to get into synchronization with the base station, which limits mobility whenever the subcarrier spacing is tight. Handover to another base station requires fast switching (Figure 3).

The third single-channel radio option is a direct-conversion low-IF radio architecture with a bandwidth-programmable integrated channel filter for receive and transmit paths, and an offset cancellation circuit that rejects the dc offsets inherent in the receive gain path in a mobile radio. The settling time of this circuit is much faster because the lowest-signal subcarriers are far away from the dc offset frequency. Low-IF radios are easier to integrate with other components than other radios.

It's also easier to integrate RF components on the same silicon with low IF radios, than other radio architectures.

They can include, on a single piece of silicon, a single completely integrated synthesizer, digital gain settings for the receive path that improve sensitivity and digital transmit power control within a large control range, integrated image rejection, LO leakage digital control settings, and calibration detectors. This solution minimizes the number of external components to about 250 or less, while still allowing the implementation of programmable channel bandwidths for the different WiMAX profiles. The total BOM with a highly integrated low IF radio is less than $100 — a 33% reduction when compared to other options (Figure 4).

The synthesizer of a WiMAX radio is the other demanding component. The -30 dB transmit error vector magnitude (EVM) certification limit for subscriber stations must be split between the transmit components and the synthesizer. A -37 dB EVM target for the synthesizer means that it contributes 20% of the total EVM, allowing more headroom for the power amplifier (PA) distortions and production margins. Because the PA is a critical component and the largest consumer of power, it is extremely important to consider the efficiency of this block when designing battery-powered mobile terminals. A higher EVM budget improves the total power efficiency of the system. The best way to get a better EVM is to implement the frequency correction in an integrated programmable synthesizer with a frequency resolution up to the required subcarrier accuracy. A synthesizer with a fast (10 ¼s to 50 ¼s) settling time can support Rx/Tx frequency switching in hybrid frequency-division duplex (HFDD) systems.

The sub-channelization option of WiMAX requires a power control range of more than 50 dB. In a low-IF radio this can be implemented with full digital control and a resolution of less than 1 dB. Instead of contributing to the Tx EVM budget, transmit path imperfections in a low-IF radio contributes to the Tx emission mask. These masks, defined in Europe by ETSI for licensed frequency bands, are prone to leakage and image imperfection. However, this problem can be corrected easily using a calibration algorithm. The analog detectors that support the calibration can be integrated into the low-IF radio, but must be controlled by the baseband firmware.

The only real drawback of low-IF radios with integrated programmable synthesizers is that they tend to be more expensive than other radio architectures. However, the added cost is typically more than offset by the fact that they can significantly reduce the external component count and the BOM cost for consumer applications by $50 or more, while still allowing the implementation of programmable channel bandwidths for the different WiMAX profiles.

The unique characteristics of Wi-MAX IEEE 802.16 standards complicate the task of choosing the appropriate radio architecture. Selecting a multiband or single-frequency band; or selecting one of the dual-conversion, zero-IF, or low-IF radios will affect the cost and influence performance of the application being developed. Because of the need to achieve early adoption by a large number of end-users, the external component count and total BOM cost are critical. It is equally important that performance not be sacrificed to cost considerations In most cases, a low-IF radio with integrated synthesizer will be the best option.

Michael Livingston is a product manager at Atmel in Colorado Springs, Colo.

Reiner Franke is a principal senior RF design engineer at Atmel in Duisburg, Germany.