click here for article

To meet the needs of emerging wireless applications such as 3G handsets, hardware designers are creating a new generation of RF integrated circuits (RFICs). These devices offer increased bandwidth, more complex digital modulation and increased dynamic range.

Test equipment originally created to support continuous wave (CW)-narrowband designs, however, cannot meet the performance and accuracy requirements of this new generation of RFICs. This drives RFIC designers to increase guardbanding and, in turn, limits these new devices from realizing their full performance potential. Now, however, RF automatic test equipment (ATE) receiver designers are turning to new approaches that match increased RFIC bandwidth, dynamic range, linearity and noise requirements. As a result, a new generation of RF ATE receivers promises to maximize test accuracy and throughput — enabling designers to maximize performance of new RFICs needed for more complex wireless markets.

3G demands

Mobile handsets is the largest market segment for wireless RFIC applications. Research firm Dataquest estimates that as many as 500 million units will be sold in 2001 with compound annual growth rates (CAGR) remaining in double digits beyond 2005.

The needs of current RFICs are being met primarily with classic CW-narrowband techniques. Classic tests for linearity have been extended to measure the linear demands of these non-linear wideband signals. Although good in some cases, correlations have started to break down as the bandwidth and complexity of the modulations have grown. This breakdown has forced test guard bands to increase, causing the RFIC designer to provide even greater performance margins in the design.

3G cell phone standards are increasing the data rates, modulation complexity, channel bandwidths and linearity requirements well beyond today's 2G designs. These changes demand that ATE measurement receivers have dynamic ranges of more than 70 dBc and information bandwidths of 15 MHz or greater to maximize measurement accuracy and throughput. In contrast, today's RF ATE receivers typically have dynamic ranges of 60 dBc and information bandwidths of about 5 MHz or less.

In addition to performing well in a wideband mode, the receiver must be capable of making measurements that are inherently narrowband, such as noise figure and phase noise, as well as other classic linearity figures of merit. In turn, to handle the phase noise and noise figure, the ATE receiver needs to add a narrowband path to address these inherently narrowband measurements, which further complicates the receiver.

Receiver boundary conditions

The instantaneous dynamic range of next-generation wideband receivers is being driven by adjacent (or alternate) channel power ratio (ACPR). This modulated power amplifier figure of merit also characterizes the wideband requirements in a next-generation RF ATE design. Figure 1 shows the wideband spectrum at the output of a cellular power amplifier. ACPR is the ratio of the integrated powers of the occupied channel to either the adjacent or alternate channels.

CDMA (IS-95) has the widest bandwidth, 1.25 MHz, and the greatest ACPR requirements (45 dB on the alternate channel) of today's 2G cell phone standards. By comparison, CDMA 2000 (one of the more demanding 3G standards) has three times the channel bandwidth and an adjacent-channel ratio requirement of 70 dB. Using CDMA 2000 as a reference for the RF ATE receiver design, the information bandwidth (final IF) should be at least three times this.

Having this much bandwidth allows the occupied channel and its two adjacent channels to be captured in a single acquisition. This strategy has two primary advantages: faster measurements and increased accuracy as the ratios are calculated from the same time stamp and the same device under test (DUT) conditions.

In the case of dynamic range, old CW tricks, such as lots of averaging to pull signals out of the noise, can't be used on the pseudo-noise like signals used in these standards. Averaging will eliminate, or at least minimize, the magnitude of random signals (which is what noise is) while leaving the stable signal of interest standing unchanged. Applying averaging to these noise-like signals will have the same effect; only this time it is reducing the magnitude of the signal that is being measured. Therefore, the linearity requirements for this kind of measurement receiver must exceed the dynamic range of ACPR and give additional signal to noise margin so that averaging is not required, or is at least minimized.

Measurements in the lab have shown that this goal is achieved when there is at least an additional 10 dB signal-to-noise margin to the worst-case measurement. In the case of CDMA 2000, this means the spurious-free dynamic range (SFDR) must be at least 80 dB across the full information bandwidth (2^nd IF).

Linearity and thermal noise

To achieve these kinds of numbers, the trade-off between the limits of linearity (linear floor) and thermal noise must be optimized. The thermal noise is one form of noise over which the receiver designer has direct control as opposed to 1/f noise or others determined by the underlying technology. Thermal noise is proportional to bandwidth and, in the case of the measurement receiver, to its resolution bandwidth (RBW).

Once a “working” RBW (trade-off in resolution and measurement) is chosen, the minimum “working” thermal noise floor is established (the thermal noise floor can be further reduced by a narrower RBW at the cost of measurement time). The upper bound on the thermal floor for an optimized system is the linear floor. The difference determines the signal amplification budget. Therefore, the signal path amplifiers must be chosen to maximize the measured signal delivered to the digitizer, but not so large as to increase the noise levels beyond the linear floor.

The primary non-linear device in any receiver architecture is the mixer. The chosen mixers and their local oscillator (LO) drive levels must achieve or exceed the linearity requirements (in this case the ACPR of the next generation of cellular phone devices).

In Figure 2, an example of the trade-offs in thermal noise (RX sensitivity) and linearity (TOI) are shown. Obviously, the optimum point is the bottom of the “V” at the intersection of these two lines. In today's sampler-based RF ATE systems, this point must also coincide with the choice of digitizer (the analog-to-digital converter, or A/D converter). The greater number of effective bits, the lower the theoretical linear floor (about 6 dB of dynamic range per bit). However, these sample-based receivers must also obey Nyquist to prevent aliasing and also live within the confines of current technology.

Frequency plan

Like all RF receivers, the RF signals must be down-converted to the desired final IF, or information bandwidth. Limitations in filter linearity and roll-off must be one of the considerations in a good frequency plan. Contrary to today's trend toward zero-IF receiver designs, an RF ATE measurement receiver must also have the flexibility to deal with a variety of signal types and does not enjoy the a priori knowledge of the incoming signal.

For example, in the case of a CDMA phone, the characteristics of the signal are known. This allows the use of pre-selection filters and the coding that is applied to the signal itself to extract the desired signal. The RF ATE designer does not have this advantage and must create a general-purpose wideband receiver. For this reason, although it complicates the design, a super-heterodyne architecture offers maximum flexibility. Figures 3 and 4 show an example of what can happen in a wideband system.

In the presence of a strong second harmonic of the modulated RF signal being measured, as well as the second harmonic of the mixer's LO, another spectrum of signals with twice the bandwidth is produced and partially overlaps the desired signal. This overlap in bandwidths is unavoidable in a general-purpose, single down conversion wideband receiver. In a dual down conversion receiver, however, this overlap is easily avoided by the naturally occurring frequency separation at the higher-frequency first IF and its associated bandpass filter.

Once the receiver architecture is known, the frequency plan can be established. Starting at the back-end (see Figure 5), a digitizer must be picked that has the required number of bits for linearity and a sample rate that will allow the fixed information bandwidth (2^nd IF) to be properly processed according to Nyquist criteria.

Now that this is established, anti-aliasing filters can be selected and the rest of the signal processing can be chosen such that they do not alias into the information bandwidth. To keep unwanted spurs from the 2^nd LO from reappearing as an alias signal within the information bandwidth, an LO that is a harmonic multiple of the sample clock will be used and converted to DC in a sample-based receiver.

This is not the only consideration in determining the 2^nd LO. Because of the wideband nature of the RF ATE front end, one must also beware of the mixing products coming from the harmonics and spurs of the 1^st LO and the RF signal being measured. These products can be avoided by allowing the system software to vary both the first and second LOs such that the 2^nd IF remains fixed and 1^st IF moves so that unwanted products are filtered out.

Scalability and upgradeability

A final consideration for the user of RF ATE is the need to further increase measurement throughput with the addition of parallel receivers. If the receiver and associated signal mixing are modular in design, then it is just a matter of adding the required number of receiver modules to maintain or improve the throughput when the measurement requirements increase because of situations such as multiband RFICs or multisite test. This modularity also helps prevent obsolescence. With the super-heterodyne receiver, a well-designed front end can remain unchanged for years. The back-end signal processing technology, on the other hand, is advancing everyday. For instance, a digitizer that may have 12 bits today with the desired sample rate may be 14 or 16 bits tomorrow. Once again, if the receiver is modular, the customer's equipment can be upgraded with improved performance at minimum expense.

As with any receiver, the RF ATE receiver needs to match the application for which it is intended. In the case of next-generation RF ATE, the driving force is the next generation of cell phones. These phones are extending the bandwidth and linearity requirements in their own application space, thus forcing similar changes in test equipment. 3G ACPR requirements drive RF ATE receiver bandwidth and linearity requirements to values greater than 12 MHz and 80 dB. In addition, the nature of 3G signals prevents the use of traditional ATE averaging measurement techniques, requiring linear performance and higher signal-to-noise ratios (spurious-free dynamic range) across the full measurement bandwidth. Combined with a careful frequency plan to avoid unwanted spurs and alias signals, a super-heterodyne architecture provides maximum flexibility needed to meet these requirements.

By combining all of these attributes into the design of the receiver along with a scalable architecture, the next-generation RF ATE customer can meet traditional needs of fast throughput, high accuracy and low cost of test. With reduction in guard bands afforded by these new designs, the additional performance that was being designed in to accommodate older ATE technologies can now turn into an improved yield and bottom line.

About the author

Terry Wilson is the RF product marketing manager at Credence Systems. He has more than 20 years of experience in the field of RFIC design and test and holds an MSEE from Oregon State University.