February 2, 1998

• Choosing test equipment for characterizing wireless-communication products in R&D is no simple task.

• General-purpose instruments offer flexibility and accuracy, but only a few support testing in accordance with the specialized standards that govern the products under test.

• One-box test sets provide support for the specialized standards but, traditionally, have sacrificed flexibility and accuracy to achieve ease of use and short test times.

• A new generation of flexible, high-accuracy, protocol-specific signal generators is emerging to satisfy R&D engineers' needs.

• Even with these signal sources, R&D engineers often need even more highly specialized test-equipment.

Fading—an acid test of wireless-system robustness

Unlike old soldiers, digital wireless signals do die, or "drop out," when they fade away. And although the signals sometimes rise from the dead, resynchronizing the receiver with the transmitter after a dropout often takes an appreciable time. During that time, data is lost, making dropouts at best undesirable, and often unacceptable.

Immunity from the effects of signal impairments is hardly the only measure of a system's robustness. But fading characteristics are important, especially for mobile and handheld devices. For Elektrobit, Noise/Com, and Telecom Analysis Systems, a major line of business is supplying test equipment that design engineers use to characterize their systems' fading performance. Rohde and Schwarz also offers some fading-simulation capabilities in its SMIQ signal generators. With this option, engineers can perform certain fading tests while avoiding the cost of a full-fledged fading simulator. (Full-featured fading simulators can cost more than $50,000. Most R&D labs own no more than one or two such units.)

Most digital wireless-communication systems operate at 800 MHz to 2 GHz. Reception can be tricky at these frequencies, especially in cities, where buildings reflect signals and shield receiving antennas. Another problem area is in fast-moving vehicles, where transmission-path characteristics change rapidly. It should surprise no one that wireless communication is nowhere more popular than in moving vehicles in cities.

Radio propagation is a subject of much study and analysis. Researchers have characterized and named several types of fading. The types that most concern developers of wireless systems are Rayleigh, Rice, and log-normal.

Rayleigh, or short-term, and Rice, or Rician, fading are consequences of multipath reception, in which signals reflect from different objects and arrive at the receiving antenna at slightly different times. In Rayleigh fading, the direct, unreflected path is obstructed, and all arriving signals are reflections. A Rayleigh probability-density function describes the distribution of arriving-signal amplitudes. In Rician fading, there are both reflections and a line-of-sight path to the transmitting antenna. Because of the line-of-sight path, the amplitude varies less than in Rayleigh fading.

When the receiving antenna is moving, two other multipath effects come into play. Both effects broaden the received signal's spectrum. Typically, multiple reflected signals arrive from different angles, resulting in multiple Doppler frequency shifts. Each Doppler shift depends on the moving antenna's velocity and the angle between the direction of motion and the path from the reflecting object to the antenna. The result is an effect called Doppler spread.

Doppler spread is a frequency-domain characterization of multipath effects on a signal that a moving antenna receives. Delay spread is a time-domain characterization of the same phenomenon. In the time domain, the effects are easier to understand; they can cause intersymbol interference. The signals representing data from multiple symbol periods can simultaneously arrive at the antenna. The result is unpredictable outputs from the receiver's demodulator.

Digital modulation—a matter of I-Q

I-Q or vector modulation does not directly relate to intelligence quotients; I-Q refers to in-phase and quadrature. Nevertheless, many engineers think that developing the I-Q-modulation concept and technology involved a stroke of genius.

Vector modulation (of which quadrature-amplitude modulation, or QAM, is a popular type) is at the heart of most digital wireless-communication systems. QAM packs multiple data bits into single symbols, each of which modulates the carrier's amplitude and phase. Phase is measured with respect to that of an unmodulated carrier. In systems currently in production, one symbol can represent as many as 2¹⁰ (1024) values, or 10 bits, but systems that pack smaller numbers of values (usually 16 or 64, equivalent to 4 or 6 bits) into one symbol are more common.

The easiest way to visualize vector modulation is in the I-Q plane (Figure A). The length of a vector drawn from the origin represents the carrier amplitude during transmission of each symbol. The angle between the vector and the horizontal (I) axis represents the phase of the carrier when the symbol is transmitted. The time allowed for transmitting one symbol is called the symbol time, or symbol period. The data rate in bits per second is N times the symbol rate, where the symbol rate is the reciprocal of the symbol time and N is the number of bits per symbol.

Thinking of signals in the I-Q plane

Although you can think of amplitude and phase as independent quantities, it is both conceptually and practically easier to deal with the I and Q components (the projections of the vector onto the I and Q axes). In 16-level QAM (16QAM), I and Q can each have four possible values or states. Representing four states requires 2 bits. So when you pack 4 bits into one 16QAM symbol, the I component might represent the 2 most significant bits and the Q component might represent the 2 least significant bits.

If, at the end of a symbol period, the signal could jump in zero time from one pair of I and Q values to a different pair, the signal would be discontinuous and would occupy infinite bandwidth. Clearly, a system that allowed such discontinuities would make inefficient use of bandwidth and, indeed, would defeat the purpose of packing many bits into one symbol.

So the I-Q-modulated signal must be filtered to limit its bandwidth. Generally, you accomplish this filtering by developing baseband I and Q signals and passing them through lowpass filters.

Viewed as functions of time, the I and Q signals are a pair of multivalued analog waveforms. In 16QAM, each of the waveforms has four possible values, corresponding to the states 0/0, 0/1, 1/1, and 1/0 of the bit pair that the signal represents. Before filtering, the waveforms have vertical rising and falling edges and perfectly square corners. After filtering, the rising and falling edges are more gradual, the corners are more rounded, and, depending on the filter characteristics, overshoot or ringing may exist at the corners.

Raised-cosine filters

To limit the signal bandwidth, many vector-modulation systems use so-called raised-cosine filters. These filters' characteristics are flat in both the passband and the stopband. The transfer ratio is 1 in the passband and 0 in the stopband. On a linear frequency scale, the characteristics in the transition band resemble 180° of a 0.5-amplitude cosine function raised by (added to) a fixed value of 0.5.

The amplitude response passes through 0.5 (that is, ­6 dB) at the cutoff frequency, which is usually ˆ2 times the symbol rate. (Remember that the fundamental-frequency component of the baseband I and Q signals is at one-half the symbol rate.) The parameter r measures the sharpness of the filter cutoff. An r of zero indicates that the transition band has zero width in the frequency domain (a so-called brick-wall filter). Practical filters have r values as low as 0.11, but typical values are usually above 0.3. The maximum value of r is 1. With this value, the amplitude response begins rolling off gradually at dc and reaches zero at twice the cutoff frequency.

The bandlimited waveforms go to the inputs of an I-Q modulator. The I signal then modulates a carrier whose phase has not been shifted. The Q signal modulates an equal-amplitude carrier whose phase has been shifted within the modulator by 90° (pi/4 radians) with respect to the first carrier. Each modulation operation can affect the associated carrier's amplitude and either reverse or leave unchanged the carrier's phase. Vectorially adding the two modulated carriers produces an I-Q-modulated carrier.

You can use a stand-alone, two-channel arbitrary-waveform generator (ARB) to generate baseband I and Q signals. In fact,  Tektronix offers a PC-based software package, IQSim, which synthesizes data streams that you can send to a two-channel ARB to produce the baseband signals. You can apply these signals to the modulation inputs of an RF-signal generator that includes an I-Q modulator. You should be sure, however, that the ARB is properly specified for this application. The ARB's two channels must operate from a common clock. The ARB must also include lowpass output filters whose characteristics closely match each other and are properly specified for producing I and Q inputs for a modulator.

You can reach Senior Technical Editor Dan Strassberg at 1-617-558-4205, fax 1-617-928-4205, e-mail
ednstrassberg@cahners.com.