Put Low-Cost Wireless Systems To Work

Many of today's systems engineers still have difficulty specifying detailed radio requirements and evaluating possible performance tradeoffs. At times, it seems easier to select a popular wireless system, such as 802.11 or Bluetooth, rather than sweat the specifics of a customized wireless implementation. For applications that demand interoperability, this is a logical and necessary choice. But many other wireless applications exist. Among them are cordless phones, industrial control, consumer game controllers, meter reading, wireless audio, and security. For these types of applications, systems designers can usually reduce size, cost, and power by using highly integrated RF transceivers along with a simple radio protocol optimized to the specific application.

For anything other than a high-speed, Ethernet-compatible wireless network, the use of an 802.11 solution will almost certainly result in unnecessary cost, size, and power. Likewise, a Bluetooth solution—while much lower in cost and power than 802.11—carries additional overhead. Such overhead is associated with any complex peer-to-peer personal-area network (PAN) that is designed with interoperability as its primary feature.

In a lot of applications, standards-based solutions have many features that aren't required. The result is unnecessary cost and power (see table). The key part of any wireless design is determining the required system performance. In this article, the term "radio" refers to the RF receiver and transmitter circuitry. A total wireless solution also requires a baseband, microcontroller, and software. Fundamentally, the primary parameters that affect size/power/cost are:

• Frequency (in MHz): Choosing the desired frequency of operation is generally the first thing done by a system designer. In the U.S., there are the unlicensed ISM bands at 900, 2400, and 5800 MHz. In the U.K., the narrowband 868 MHz has proven popular for low-data-rate solutions. Presently, the 2.4-GHz band comes closest to a true worldwide solution. As a result, it has the largest number of component vendors. The system designer also should evaluate other issues, including RF propagation (i.e., range), regional regulations, and possible interference from other users.
• Range (in meters): Generally, range is defined as the maximum distance between transceivers that provides a minimum signal-to-noise ratio (SNR). Range has forever been the most important parameter to the systems engineer. For the radio manufacturer, it has been the most difficult one to specify (due to the wide range of performance in different environments).

  Range is theoretically dependent on Tx power, antenna gain (both Tx and Rx), Rx sensitivity, and signal processing gain (if any). Obviously, longer range can be achieved with higher Tx power, but at the expense of shorter battery life. Unfortunately, antenna gain is directly related to size, so longer range can be achieved if more space is available. Rx sensitivity depends on both receiver noise figure and IF bandwidth. Aside from designing a good low-noise radio front end, the designer should try to use the minimum channel bandwidth needed to support the required throughput.

• Throughput (in kbps): Designing a radio to support the required throughput is one of the most important aspects of system design. Mismatching the radio channel bandwidth to the actual data throughput unnecessarily reduces sensitivity and range. Receiver sensitivity is dependent on 10logBW. Using a 1-Mbps radio for a 250-kbps application will therefore reduce sensitivity by 10log(4) or 6 dB, causing a 50% reduction in range. Getting the range back requires a 6-dB increase in Tx power, which effectively quadruples power consumption.
• Link reliability and interference rejection (in dBc): For proper receiver operation, one must accurately know the intended environment. Are there other users on the same or adjacent channels? What about unintended radiators? Can the system tolerate infrequent packet losses due to interference? Obviously, the most robust radio includes enough filtering to reject out-of-band interferers. It also has enough dynamic range to operate with co-channel users. Additionally, the datalink protocols that are intended for wireless links usually have provisions for errors. Such provisions must be included due to the inherently unreliable nature of the wireless channel. The provisions range from simple CRC checks and ACQ capability to more complex forward-error correction. The system designer must try to estimate the worst-case RF environment in which the radio must work as well as the tolerable level of interference.

Once the basic radio system requirements are determined, the designer must choose a radio IC. Many different transceiver architectures have been used in the past. The same can be said for modulation types. However, choosing which transceiver IC is best for the application can sometimes be a difficult decision.

In the world of wireless, three major types of receiver architectures are widely used. The classic double-conversion superheterodyne architecture is utilized in nearly all high-performance radios. Though it has excellent performance, it needs many off-chip components. Generally, it is not used in low-cost applications. Within the last few years, the more advanced direct-conversion (also called zero-IF) architecture is again garnering a lot of attention. Recent advances seem to be overcoming the problems with DC offset and LO isolation. But for now, the primary users of this architecture remain the more complex I/Q modulation approaches like 802.11.

For low-cost applications, the most popular receiver architecture is the low-IF architecture. This architecture avoids the DC offset problem by downconverting to a low IF frequency instead of all the way to baseband. This approach has proven to be very popular in low-cost, highly integrated transceivers including Bluetooth. It is the recommended receiver architecture for the applications discussed here.

Next, the designer must investigate the many different ways to implement the radio's transmitter side. For complex modulation approaches, such as QAM, most ICs implement some form of I/Q upconversion. The baseband processor generates the complex I/Q signals. Today's low-cost data receivers almost always use direct VCO modulation to implement simple, non-coherent FSK modulation schemes. Open-loop modulation has been the most common implementation, although the Micro Linear ML2724 transceiver includes a closed-loop direct VCO modulator.

So how does a systems engineer go about choosing a transceiver IC? With today's integrated transceivers, most of the difficult decisions have already been made and implemented on chip. Optimal performance is obtained because the IC manufacturer has done the complex system-engineering calculations. The manufacturer has matched the receiver performance to the transmitter capability. The design process is therefore reduced to comparing the desired system requirements to the transceiver datasheet. Lastly, compare the total bill-of-materials costs of the various suppliers.

Aside from the radio transceiver, the systems engineer has to consider the performance and cost tradeoffs that are inherent in the protocol choice. The basic goal of any radio protocol is to synchronize communications between Tx and Rx stations. The protocol also strives to recognize packets that were received incorrectly. Using an overly complex protocol with unnecessary features requires additional processing power and/or silicon gates. These aspects will increase cost and power consumption. The system designer must address a number of issues surrounding protocol implementation, including:

• Synchronization: Wireless protocols must provide for quick and easy synchronization to the incoming data. For asynchronous systems or those with transmissions that can occur at any time, the protocol must include enough preamble bits for proper bit and byte synchronization. It also has to include start-of-frame patterns for packet synchronization. For synchronous systems, or those with reserved Tx/Rx timing, the protocol must provide for small corrections to the receiver's own local timing. As a result, it only operates within its reserved timeslots. More complex systems, like FHSS or DSSS, depend on the protocol to speed up synchronization to the frequency-hopping pattern or spread code correlation offset.
• Multiple access: Three common methods exist for sharing the radio airwaves. Time division multiple access (TDMA) allocates different timeslots to each user, thereby preventing any one user from capturing the channel at the expense of others. In the frequency domain, the equivalent to TDMA is frequency-domain multiple access (FDMA). It uses different frequency channels for each user. FDMA is the best choice for users who need long extended access to the radio channel. Finally, spread-spectrum multiple access uses either frequency-hopping (FHSS) or direct-sequence spread spectrum (DSSS).

  FHSS supports multiple users by assigning different hopping patterns. In contrast, DSSS supports multiple access through multiple spreading codes (CDMA). Both FHSS and DSSS systems are more complicated and difficult to implement. They perform well, however, when interference levels are high. For asynchronous RF datalinks that don't assign the user a timeslot, frequency channel, or spread-spectrum code, it is imperative that the protocol implement a listen-before-transmit mode of operation. Otherwise, there will be an unnecessarily high number of collisions, which result in degraded throughput.

• Network architecture: The simplest wireless architecture is a direct point-to-point link. Such a link would be used in a building-to-building link. As there are only two nodes, there is no need for addressing in the protocol.

  Alternatively, there could be a point-to-multipoint architecture in which a master transceiver sends data to multiple nodes. This approach is similar to an access point in the 802.11 protocol. In such a case, the protocol must allow for source and destination addressing. A variation of the basic point-to-multipoint network is the ad-hoc network defined by Bluetooth. Here, any node can act as a master. Finally, there is a fully meshed network in which any station can communicate with any other station. For it to work, there must be provisions for carrier sensing and collision avoidance.

• Error detection/correction: The protocol must be written to deal with the fact that wireless links are notoriously noisy and unreliable. Many higher-level applications will not operate in this environment. It is therefore necessary to correct for this unreliable performance at the lower protocol layers. The simplest implementation is to include error-detection capability in the protocol, such as a parity check, checksum calculation, or CRC. When errors are detected, the packet must be resent. Good packets are acknowledged automatically, while packets with errors are resent. More complicated systems implement forward error correction at the expense of protocol overhead. This method allows for the correction of small numbers of errors without having to resend the packet. Sometime, FEC is useful in voice applications, in which retries would cause unacceptable audio degradation.

Data communication protocols can be the subject of many textbooks. Yet there are many examples of simple protocols. These protocols are based on publicly available implementations that can be used as a starting point.

So how can one implement a proprietary solution that consumes less power at a lower cost? Well, a few simple axioms hold true in wireless just as they do in wired circuits. All IC designers know that complexity equals gates or MIPS, which then equal cost. To reduce that cost, the engineer should restrict functionality and complexity to the absolute minimum required.

A second axiom is that speed equals power. Therefore, all analog circuitry should run at the lowest data rate possible. All digital circuits should use the lowest possible clock frequency. Figure 1 shows the common functions that must be implemented by any wireless system. It serves as a good baseline to appreciate the high-level solution complexity.

As discussed previously, the RF XCVR serves to downconvert the received high-frequency radio signals to baseband. Conversely, it upconverts the baseband outputs to radio frequencies for transmission. The RF XCVR functionality is required for all types of wireless radio systems. Generally, it is a standalone IC.

IEEE 802.11 and Bluetooth are differentiated from low-end proprietary solutions by the baseband, controller, and host implementations. Early 802.11 solutions required three separate chips to implement the RF, baseband, and controller functions. Typically, they interfaced to the host (e.g., a notebook computer) through a standard 16-b PCMCIA interface. Newer implementations have reduced this to just two chips. Though they combine the baseband and controller functions, the complexity of the solution remains.

Bluetooth also implements these basic blocks. Though it has progressed rapidly to single-chip solutions (RF, baseband, and controller), it interfaces differently to the host depending on application profile (i.e., a wireless modem, headset, etc.).

The system designer's fundamental concern is the interface to the host. These standards were fundamentally written to communicate with a separate host processor, which could be a notebook computer, PDA, or even a cell phone. For low-cost, low-power applications, the goal is to implement the entire MAC, protocol, and application software in a simple, low-cost microcontroller. This microcontroller should yield the best solution for the intended application and be fully customizable due to the use of simple, common microcontrollers.

A TYPICAL DESIGN EXAMPLE An infinite number of wireless applications exists. Just thinking through the steps of a typical design, however, helps to highlight the issues addressed here. Consider, for example, a simple wireless temperature-monitoring application: Multiple sensors are to be placed within a small manufacturing facility. Assume that a room monitor is responsible for wirelessly communicating with each of the sensors. It also passes along temperature and battery voltage to a central controller. Next, assume that the monitors will be placed in sight but with no external power source. Small size and low power consumption is therefore critical. In addition, the monitors will be used in very large volumes, making low cost essential.

The assumption is that this design has a worldwide market. The chosen operating frequency is thus 2.4 GHz. This band is regulated in the U.S. per FCC Part 15.247, which determines the maximum transmit power and power spectral density for either frequency hopping or wideband digital modulation. (Formerly, digital modulation was called spread spectrum.)

The simplest solution is to implement a wideband FSK system. To certify it as a digitally modulated system, send short bursts of data with a data rate and deviation that are high enough to spread across >500-kHz bandwidth. In this application, the required throughput is actually pretty low. By sending data at a higher rate than actually necessary, though, it's possible to power down between transmissions and save power. To combat interference, implement a simple adaptive frequency-hopping algorithm. Continuously monitor the packet error rate. If it becomes too high, coordinate a frequency change to an unused channel. This is much simpler to implement than a continuous frequency-hopping system.

Now, it's time to research the available transceivers in this band. The ML2724 from Micro Linear is chosen for this example. It is a highly integrated, 2.4-GHz transceiver. The ML2724 includes a low-IF receiver and direct VCO (closed-loop) modulated FSK transmitter in a single, 32-pin package. It includes voltage regulators, a frequency synthesizer (including the VCO tank circuit), image rejection mixer, IF selectivity filter, and FSK discriminator (FIG. 2). The only external components are bypass capacitors and a simple RC PLL filter. In transmit, a two-port modulator ensures FSK modulation accuracy by incorporating a sigma-delta fractional-N PLL. Because the PLL loop remains closed during transmission, it eliminates frequency drift during transmit.

Note that the true cost is more than the individual component and assembly cost. As such, it must include production alignment costs. Many of today's integrated transceivers require complex production routines that measure and store calibration parameters for each IC.

In this example, the job of programming the transceiver and implementing the simple protocol falls to the low-cost Microchip PIC microcontroller with on-board Flash program memory. The protocol firmware needs to be written, which demands a microcontroller with simple, readily available development tools. Figure 3 shows a block diagram of this hypothetical wireless design. Range is not accurately known at this time, so a low-cost LNA is included for best Rx sensitivity along with a 100-mW Rx transmitter. The antenna will be implemented with a simple tuned microstrip trace. During field trials, the range performance can be evaluated. Components can then be eliminated if possible.

For the lowest cost and power, use the minimum number of components. Also, run the microcontroller at its lowest clock speed. In this example, the same crystal oscillator is used for both the transceiver and the microcontroller. Additionally, the integrated UART is used to synchronize to the incoming asynchronous data. This approach allows the microcontroller to run at the slowest possible speed. Because synchronization is performed in the UART hardware, it doesn't consume the few precious MIPS that are available.

To reduce DC power, this example implements a simple protocol. That protocol sends the data out at the maximum rate of 1.5 Mbps. It then puts the entire radio into sleep mode until the next required transmission.

This system's overall costs are quite low, thanks to the integrated transceiver with minimal external components and the use of a low-cost microcontroller. To help customers implement a wireless system like this one, for example, Micro Linear released the ML2724SK-01 (FIG. 4).

IEEE 802.11 and Bluetooth have certainly been successful in serving their intended applications. Yet a need exists for simple, low-cost wireless solutions. These solutions should be optimized for the proprietary applications in which standards conformance is not desired. After determining the key radio performance requirements, the designer must choose an integrated transceiver IC. A wireless protocol can then be developed to meet the specific application's needs. The hardware is designed and implemented on a microcontroller platform using the minimal processing power for the intended application. By using this straightforward process, the systems designer is guaranteed a highly flexible wireless system with the lowest possible size, power, and cost.

REFERENCES:

1. Larson, L.E., "RF and Microwave Circuit Design for Wireless Communications," Artech, 1996.
2. Rappaport, T.S., "Wireless Communications Principles and Practice," Prentice Hall, 1996.
3. Sayre, C.W., "Complete Wireless Design," McGraw-Hill, 2001.
4. Sklar, B., "Digital Communications Fundamentals and Applications," Prentice Hall, 1988.
5. Heath, S., "Embedded Systems Design," Newnes, 1998.
6. Torvmark, K.H., "Short-Range Wireless Design," Embedded Systems Programming, October 2002.
7. Bensky, A.,"Short-range Wireless Communication," LLH Technology Publishing, 2000.