[Click for a PDF copy of WLAN RFICs TEST Challenges]

Consumers' insatiable thirst for high data rates coupled with mobility is creating rapid growth in wireless local area network (WLAN) market requirements and device volume. In turn, this demand has created unprecedented challenges for radio frequency integrated circuit (RFIC) testing.

Modulation schemes for 802.11, a family of WLAN standards, have been devised to accommodate data rates from 1 Mbps to 54 Mbps in channel bandwidths up to 22 MHz. Such channel bandwidths pose difficulties for traditional automatic test equipment (ATE).

Such systems were originally designed for the requirements of narrowband communication schemes that traded bandwidth for high dynamic range. The complexity of WLAN systems continues in the direction first set by Qualcomm's code-division multiple access (CDMA) IS-95 standard created in the mid 1990s. These systems require not only high dynamic range, but also wide bandwidths — two often mutually exclusive constraints.

Benefiting from the lessons learned from cellular phone technology, WLAN devices are starting out with the latest in device architectures. The convergence of cellular and WLAN device architectures and modulation schemes are pushing wireless ATE manufacturers to design equally new advanced wideband systems to meet these challenging standards. Many 802.11 (WLAN) chipsets combine high integration, with zero-IF architectures, making traditional parametric test, almost useless.

System designers are now demanding that device manufacturers specify these devices with end-to-end (functional) metrics such as error vector magnitude (EVM). This, in turn, places a much greater emphasis on digital signal processing. A distributed approach to this increased processing burden in the ATE system design will relieve the host controller of this task and maintain or possibly improve the throughput. The distributed solution is typically not found in today's wireless ATE systems.

In addition to the wideband requirements found in the WLAN modulation schemes, high linearity demands to reduce inter-symbol-interference (ISI) are imposed on both the RFICs and the wireless ATE. Other WLAN standards, such as Bluetooth, are modest in their linearity and spurious free dynamic range (SFDR) test requirements. On the other hand, 802.11 requires test equipment to exceed 70 db of SFDR.

The 802.11 physical layer (PHY) family — 802.11a, b, and g — have one common ancestry (IEEE 802.11 at www.ieee.org) but utilize a variety of modulation schemes to achieve the desired data rates, such as orthogonal frequency-division multiplexing (OFDM), complimentary code-keying (CCK), and packet binary convolutional coding (PBCC). What do these acronyms mean and how does this affect the next generation ATE systems?

With 802.11, some background information is beneficial. In 1997, the Institute of Electrical and Electronics Engineers (IEEE) developed a standard that dictates the method for transmitting information over the airways for WLANs. IEEE 802.11, or more simply 802.11, is this standard. This standard is comprised of two layers, the medium access control (MAC) and the physical (PHY) layer. The MAC provides the asynchronous data service, security, and controls the order of the MAC service data unit (MSDU).

The MAC is common for all PHY extensions of IEEE 802.11. For 802.11, there are three different methods of data transmission — diffuse infrared (IR), frequency hopping spread spectrum (FHSS), and the most widely used method, direct sequence spread spectrum (DSSS). Each has its own place in the wireless world, but higher data rates are the driving force. Since DSSS allows for both 1 Mbps and 2 Mbps data rates, it is clear why DSSS is the most common PHY layer on the market today.

To transmit at these data rates, 802.11 utilizes differential binary phase shift keying (DBPSK) and differential quadrature phase shift keying (DQPSK) for 1 Mbps and 2 Mbps data rates respectively.

The 802.11b PHY extension of IEEE 802.11 was introduced in 1999. To meet the requirements of the MAC layer, 802.11b must be backwards compatible to the 1 Mbps and 2 Mbps data rates of 802.11. However, to meet consumers' thirst for more and faster data, the use of DQPSK, along with the mandatory CCK or optional PBCC, is what makes the 5.5 Mbps and 11 Mbps rates achievable.

Also ratified in 1999, IEEE 802.11a PHY adds yet another wrinkle to the modulation schemes achieving data rates as high as 54 Mbps. It incorporates OFDM.

Orthogonal frequency-division multiplexing has been around since the 1950s, but until recently, it was not economically feasible to implement. This changed with the incorporation of high-speed fast Fourier transform (FFT) hardware into the integrated circuit. OFDM is a form of multi-carrier modulation (MCM) that allows a single carrier to be split into numerous sub-carriers (52 for 802.11a) that are spaced 312.5 kHz apart and transmitted in parallel. Figure 1 shows three sub-carriers (spaced 300 kHz apart) that illustrate the orthogonality aspect of each¹.

To achieve higher data rates, 802.11a incorporates binary phase shift keying (BPSK) for 6 Mbps and 9 Mbps data, quadrature phase shift keying (QPSK) for 12 Mbps and 18 Mbps data, 16-quadrature amplitude modulation (QAM) for 24 Mbps and 36 Mbps data, and 64-QAM for 48 Mbps and 54 Mbps data. This standard operates in the relatively “clean” unlicensed national information infrastructure (U-NII) band.

There are three 100 MHz frequency bands available for 802.11a, the “lower” band (5.15 GHz to 5.25 GHz), the “middle” band (5.25 GHz to 5.35 GHz), and the “upper” band (5.725 GHz to 5.825 GHz). In each of the bands the maximum power has been increased to 16 dBm, 23 dBm and 29 dBm, respectively. This was done to improve propagation distances at these higher frequencies.

IEEE 802.11g PHY is currently the last member of the 802.11 family, and although not yet ratified, it is the current hot topic among developers. As mentioned above, the higher frequencies, high peak-to-average power ratio (PAPR), and strict power requirements in the U-NII band, mean that 802.11a propagation distances and coverage are less than 802.11b.

However, the consumer has the desire for a robust network with data rates in excess of 20 Mbps. This has led to the formation of the IEEE 802.11 Task Group g (TGg). TGg is now in the process of resolving the numerous issues derived from the approved September 2002 draft. Once TGg passes the five-step IEEE 802 standards ratification process, 802.11g will become a standard. At that point 802.11g will be capable of the 802.11a high data rates, be backward compatible with 802.11b and above all, compete for spectrum within the crowded 2.4 GHz industrial, scientific, and medical (ISM) band. Since this standard is currently in flux, the remainder of this discussion will be limited to the existing standards.

One major test challenge associated with WLAN is the broad spectrum of each channel. The force behind the wide bandwidth of 802.11 and 802.11b is the Barker sequence used to spread the carrier with a chip rate of 11 mega chips per second (Mcps). OFDM and its associated 52 sub-carriers determine 802.11a's bandwidth, each spaced 312.5 kHz apart, within a channel. Each adjacent sub-carrier is orthogonal to the other, which theoretically eliminates intersymbol interference (ISI).

Many wireless testers on the market today do not have the required information bandwidth or digitizer capabilities to measure such wide, high data rate signals. In 802.11 and 802.11b, the bandwidth of each channel is 22 MHz, which includes the transmit spectrum mask as seen in figure 2 (dBr represents dB relative to the sin (x)/x peak). For the 802.11a spectrum, the occupied bandwidth is 16.6 MHz and the transmit mask is extended to 18 MHz.

Assuming the digitizer is capable of meeting Nyquist, a multi-acquisition approach can overcome the bandwidth issue at the expense of test time. However, this still poses a major problem in a very cost sensitive market as WLAN, since longer test times translate directly to lower profit margins.

Error vector magnitude (EVM) is a measure of the transmitter's modulation accuracy, where any deviation from the ideal bit location generates an EVM.

Phase noise from an oscillator in the frequency domain is related to phase jitter in the time domain. Phase jitter effects the phase location of each point on a constellation diagram. For visualization, figure 3 compares an ideal 16-QAM constellation to one with the effects of phase jitter. Figure 3(a) represents ideal conditions where each point could represent one or thousands of points. Figure 3(b) represents the effects of phase jitter where the green dashed line is the ideal vector, the bold line is the actual and the red line is the EVM. As the points cross the “decision boundary”, the bit location will be misinterpreted and will result in an error. As the order of QAM is increased, the decision boundaries are more stringent and could lead to higher bit errors.

For 802.11a, EVM is coined “relative constellation error,” where the allowable errors associated with data rates, is outlined in the IEEE Standard. As mentioned above, the most stringent criteria are placed on the highest data rates (-25 dB at 54 Mbps), which is necessary due to the reduced area of the decision boundary of 64-QAM.

For 802.11 and 802.11b, the decision boundaries of DQPSK are basically the in-phase (I) and quadrature (Q) axis. However, the standard states that the EVM shall not deviate from the ideal ±0.707, normalized sampled chip data point by more than 0.35.

In the real world of test, the measured EVM is a composite of the DUT's performance as well as the ATE's arbitrary waveform generator (AWG) and measurement receiver. Calibration can minimize or eliminate the effects of the AWG, but the measurement receiver's phase linearity and low phase jitter performance must be designed in.

The other side of WLAN device is the receiver, which has its own method of measuring performance. In 802.11 and 802.11b, the method is called frame error ratio (FER) and in the case of 802.11a, it defined as packet error rate (PER). There are some devices on the market that allow for both continuous and packetized data transmission, but to be in compliance with the standards, the ATE needs to be able to analyze the data in packet form. In the case of continuous data transmission bit error rate (BER) is used as a measure of receiver performance, while FER and PER are used as figures of merit for packetized data. PER and FER are not tested with ideal stimulus to the DUT's receiver, as is the case for the transmission tests. They are instead tested with signals that simulate “normal” environmental conditions. This means noise and interferer signals need to be introduced to test a DUT receiver's ability to receive weak or corrupted signals.

However, just like the reference (measurement) receiver, the reference transmitter (ATE) characteristics need to be known, as well. Prior to 802.11 and other modern digitally-based modulation schemes, simple single tone stimulus were combined together to measure the distortion of the receiver, greatly reducing the demands placed on the ATE's stimuli calibration.

Another important test challenge to address is the ability to quantify the receiver performance of the more demanding 802.11a. For the slowest data rate (6 Mbps), the specified minimum sensitivity of the channel is -82 dBm, the adjacent channel rejection is 16 dB, and the alternate channel rejection is 32 dB, so the minimum value which needs to be measured accurately is at a -114 dBm level. These tests demands require today's wireless ATE to have a low noise figure and narrow resolution bandwidth (RBW) capabilities. However, 802.11 and 802.11b do not require such extreme levels, since the receiver minimum input level sensitivity is a less stringent -80 dBm.

Although not explicitly mentioned so far, to have the ability to create (modulate) and break down (demodulate) these signals, the test system must be capable of creating the other half of the communication link. This means that when testing WLANs, an abundance of modulation schemes must be created to simulate the transmitter when testing the DUT's receive characteristics and the tester must have an equal amount of agility to demodulate the DUT's transmitted signals. To speed up test development, ATE vendors will need to supply extensive modulation/demodulation libraries.

To meet the propagation specifications of the WLAN standards, particularly 802.11a, power amplifiers must be able to handle the wide bandwidth and large peak-to-average signals presented them.

Traditional narrowband, sinusoidal vector network analysis may work on some applications, but more often than not require a statistical correlation to predict the behavior of the wideband, modulated signal found in the end-use environment. These correlations lead to larger test guardbands to reduce the number of test escapes due to the measurement uncertainty associated with them. This approach runs contrary to the high device yields required of these low cost components.

Modulated vector network analysis (MVNATM) offers a different approach to this test challenge. Borrowing from Fourier the fact that all signals are composed of an infinite sum of sinusoids, vector network analysis techniques can be extended to the high dynamic range, wideband signals associated with 802.11. MVNA allows the use of the same wideband, high dynamic range input stimulus that the DUT will undergo in its end-use application³. The resulting wideband measurements will in turn more accurately predict real world performance. As seen in figure 4, the difference in wideband and narrowband gain can be quite significant, especially when one considers that power amplifiers are optimized for the maximum power out at the highest level of efficiency. This latter trade-off is of vital importance to mobile applications, where battery life often overshadows all other technology. Giving up one to two dB due to the measurement uncertainty of a correlated narrowband measurement is an opposing strategy to maximizing yield. Using MVNATM technology, measurement uncertainty and the associated test guardbands can be reduced, resulting in greater returns to the bottom line without reducing the quality delivered to the customer.

Consumer demand for more and faster data in a mobile environment has lead to the IEEE 802.11 family of standards, and, in turn, created new challenges for wireless ATE. To meet these challenges, ATE vendors must design in the flexibility to satisfy the various modulations schemes. This flexibility goes beyond the modulation itself; it also extends into other tester design elements such as wide information bandwidth, high dynamic range, linear phase, and low phase jitter. All of this must be done without losing sight of the fact that these WLAN devices should also satisfy the needs of a cost-conscious market. This means that other factors such as throughput must be considered, as well.

The measurement receiver's phase jitter and phase linearity have a direct impact on the ability to measure the EVM performance. The standards require the device and ATE distortion not to exceed -3 dB relative to the normalized chip data point. The thermal noise performance must also allow noise power measurements of -114 dBm in 18 MHz of information bandwidth.

Lastly and of no less importance, the receiver's information bandwidth ideally is at least as wide as the standard's requirement in order to maximize test throughput.

On the stimulus side, the test system's AWG must have the flexibility to create a wide variety of modulation schemes, everything from a relatively simple DBPSK to a highly complex 64-QAM OFDM signal. Test libraries which contain the various modulations are needed to speed develop time. These libraries will also allow the test engineer to focus on testing the part, rather than becoming an expert on the nuances of the various standards.

WLAN, or more succinctly, the 802.11 family, has continued the trend in wireless communications. Bandwidths are getting larger, data rates are getting faster, signal dynamic range is increasing and distortion performance is more stringent than ever. To deliver these devices to the end user, wireless ATE vendors must exceed the device performance requirements so these devices not only perform as advertised, but also can be tested at costs that meet the economics as well.

Warren Strand is a RF marketing applications engineer at Credence Systems Corp. He has more than 12 years of experience in the field of RF and Microwave test and holds a BSEE from Portland State University. He can be contacted at warren_strand@credence.com

1. A. Zamanian, “Orthogonal Frequency Division Multiplex Overview”, Microwave Journal, October 2001, Volume 44, Number 10, page 122.

2. From the WLAN Inc. technical white paper “Wide-band orthogonal frequency multiplexing (W-OFDM) at WLAN, Inc. http://www.wilan.com/technology/index.html.

3. K. Harvey, J. Lukez, D. Metzger, “Vector Analyzer makes Measurements with Modulated Signals, Wireless Systems Design, Number 1, January 2001, page 21.