Major integrated-device manufacturers and foundries are moving from 200-mm fabs to 300-mm fabs to reduce manufacturing costs—until recently, the single most expensive component of the end product. While automated test equipment (ATE) design continues to address the needs of ever-increasing device complexity with more complex and flexible systems, the cost of test has not been following the ever-falling average selling price curve. Obviously, this mismatch must be reconciled if these next-generation products are going to meet the demands of the consumer.
On the technology side of the equation, these next-generation RF devices are more complex, wideband, and linear than ever. Third-generation (3G) cell-phone ICs will be multimode to satisfy consumer demand that phone appliances work anywhere at any time.
Not only will devices need to work in the cellular and personal communications service (PCS) bands, but they also must transmit and receive the standards of first-generation (1G) advanced mobile phone service (AMPS); second-generation (2G) such as code division multiple access (CDMA), global system for mobile communications (GSM), and North American digital cellular (NADC); and 3G such as CDMA2000, wideband CDMA (W-CDMA) devices. 1G will remain in rural areas of the United States.
Market Demand for RF ATE
Cellular handset production continues to be the growth engine for RF devices. Production has grown 60% to 70% annually, and Dataquest projects 430 million handsets for 2000 and more than a billion annually by 2004.
To capitalize on this growth, component manufacturers are driving demand for RF ATE to new heights. This demand is expected to continue its rapid growth through 2005. 3G devices could make their debut in early 2001 with predictions that they will follow the trends of their predecessors by increasing both the demand for and requirements of RF ATE.
Technology Requirements for 3G-Ready ATE
Better connectivity, higher data rates, and wider information bandwidths are core consumer requirements of next-generation cell phones. In addition, service providers need greater system capacity, resulting in increased linearity and power-control specifications.
These requirements are prompting RF ATE vendors to rethink their solutions for 3G. Since the first commercialization of cellular technology in the early 1980s, the bandwidths associated with each generation of cellular phone have followed a logarithmic trend.
Earlier standards yielded channel bandwidths from 30 kHz (NADC) to >1.25 MHz (CDMA). 3G standards, such as CDMA2000, have bandwidths more than three times larger than the widest current standards.
Today’s ATE systems use a traditional narrowband measurement receiver design to test current 2G devices. This has resulted in traditional approaches in specifying and testing the parts.
The most obvious issue of applying a narrowband receiver to a wideband component is acquisition time. Despite the improved technology in data acquisition and digital signal processing (DSP), without the bandwidth to capture all of the information at once, throughput and accuracy are compromised.
Reduction in throughput is easy to understand, but loss of accuracy is less intuitive. The time-varying nature of these modulated signals requires that the same snapshot of data be used when calculating parameters as simple as gain or the more complex adjacent channel power ratio (ACPR) to achieve the highest accuracy (Figure 1).
However, with a narrowband receiver, the stimulus waveform must be replayed many times to capture the entire information bandwidth, which reduces accuracy and throughput. This approach also has resulted in new figures of merit to correlate the modulated behavior of 2G devices.
Faulty evaluation of the data has allowed good devices to be rejected, which drives up costs. It also has permitted defective devices to pass tests, then fail in the end application, which decreases quality. These traditional approaches are failing both the device supplier and the handset manufacturer.
The simple solution is to increase the measurement receiver’s bandwidth and provide compatible modulation capability. As with most things in life, it’s not that simple. Without improvements in other parts of the receiver’s design, nonlinearity and noise also will increase with wider bandwidths, contrary to the demands of 3G devices.
ACPR, a key figure of merit for modulated power-amplifier linearity, is increasing from -55 dBc to -80 dBc (decibels down from the occupied channel to the alternate channel). Modest device performance improvements in intermodulation distortion, -8 dBm vs. -10 dBm, and noise also are being implemented on the downconverter’s data sheets.
An 80-dB, spurious-free dynamic range now is a common request from customers evaluating next-generation RF ATE. Phase-noise requirements for 3G devices, due to more demanding bit error rate (BER) specs, also are forcing the ATE vendor to design corresponding improvements in the measurement receiver.
Beyond the hard-core technical requirements of RF ATE is the issue of reducing the cost of test with the use of multisite ATE configurations that allow parallel testing. Trade-offs in complexity and cost vs. parallel test must be investigated and optimized. Many of today’s RF ATE platforms claim multisite capability but actually operate serially to some degree due to the lack of parallelism in RF ATE design.
Modulated vector network analysis (MVNA) is a new technology for making classic RF measurements on wideband-modulated signals. It has the capability to make measurements in a user’s environment, reducing the issue of correlating test to the end application.
At the core of this technology is modulated vector-corrected stimulus and measure. This capability allows you to make measurements with just one instrument that once required the use of network analyzers, spectrum analyzers, vector signal analyzers, and power meters.
In addition to the obvious cost savings, throughput and accuracy are enhanced. The same data captured in a single acquisition can be used to measure S-parameters, ACPR, error-vector-magnitude (EVM), and power (scalars).
Fourier proved why all of this is possible. Every signal is a composite of an infinite sum of sinusoids. Consequently, the use of modulated signals necessary for testing 3G devices offers the capability to gather more information from the same data set. The amount of information gathered is a function of the stimulus applied and the information bandwidth of the measurement receiver.
S-parameters traditionally have been used by RF engineers to describe the device performance and offer the most direct way of correlating performance to the models used to design them. For the device vendor, they also provide the best feedback to design, product, and modeling engineers on how the ongoing process variations are affecting performance and yield.
Unfortunately, traditional sinusoidal narrowband network analyzers do a poor job of predicting modulated behavior of S-parameters. Figure 2 illustrates the effect of applying a modulated signal vs. a single sinusoid to an amplifier.
A single sinusoid allows measurements at a single power level and frequency. Depending on the power level, the amplifier may or may not be operating in its linear region.
Applying a modulated signal, like the CDMA signal in Figure 2, causes the amplifier to undergo linear and nonlinear operation continuously due to its inherent gain compression and the large peak-to-average power. Capturing and analyzing this kind of signal allow you not only to observe this real-world power response as illustrated in Figure 2, but also to study all of the embedded transient behavior as well. In contrast, a constant power sinusoid corresponds to only a single point on the transfer curve.
Following this line of thought, you would correctly suspect the S-parameter response would be quite different for a modulated signal. Indeed, measurements have shown decibels of difference in S21 (gain), even with an applied signal of small dynamic range.
Like a vector network analyzer (VNA), MVNA technology also can make swept S-parameter measurements of the modulated device, but with the actual power as opposed to an average power at a given frequency. Unlike a VNA, however, this swept response can be captured in a single acquisition, saving time and increasing accuracy.
Since the effects of cable loss, mismatch, and other interface issues are removed, the actual performance of the DUT can be observed. Finding correlated figures of merit will become unnecessary and allow the next step in RF device test—functional (as opposed to parametric) test—to become the norm in guaranteeing performance to the customer.
Testing next-generation devices requires new techniques for device manufacturers to achieve their cost and quality goals. The narrowband sinusoidal techniques used today will increase costs due to decreased yield and more customer returns.
Wideband, modulated vector-corrected stimulus and measurement techniques offered by the MVNA technology will allow vendors to test parts in actual end-use environments. Throughput and accuracy will improve due to the benefits of collecting all of the data in a single acquisition and performing multiple tests on this same set of data. It also will give the customer the assurance that the vendor understands the application and can be a partner in driving down costs and time to market.
A Next-Generation RF Test Solution
RFIQ, the next-generation subsystem for RF applications from Credence, meets the challenges of 3G cellular devices as well as the demands of current RF devices. The core of this system is MVNA technology, the capability to force and measure RF devices in their end-use modulated environment. Using Fourier concepts, MVNA technology decomposes these modulated time-domain waveforms into their multitone frequency components, then applies 12-term vector error correction.
Perhaps the simplest way to understand the application of MVNA technology is to compare S-parameter measurements on a classic VNA and a system using MVNA technology. S-parameters are a ratio of the incident and reflected or transmitted signals. However, to accurately measure them, imperfections in the test system and the DUT interface must be eliminated.
Using two-port network theory, these imperfections, as seen by the DUT’s input and output, are mathematically eliminated by the 12-term model. A VNA uses a single frequency stimulus, combines the raw measurements with the mod-el, and extracts the actual S-parameter response.
In contrast, MVNA technology uses the multifrequency content of a modulated signal to perform these same measurements simultaneously. The number of measurements is limited only by the measurement receiver’s information bandwidth (15 MHz for the RFIQ) and the frequency complexity of the stimulus signal.
When using a modulated signal, the actual power level at a given frequency is being applied and measured, so the real end-use response of the DUT is captured, not just the swept average power-frequency response. The measurements are a true reflection of actual device performance with increased throughput and accuracy.
Architecturally, starting from the right side of Figure 3, the RFIQ subsystem has eight 6-GHz RF ports. Each port has the full modulated vector-corrected stimulus and measurement capability of MVNA technology. Backward fixture capability with the company’s 1G RF solution, RFSS, was maintained to ease transition to RFIQ. The eight ports have been partitioned into two banks of four, each with a dedicated RF synthesizer.
Also, an auxiliary synthesizer provides a third stimulus to any of the eight ports. This gives you the ability to drive a third port, perform third-order intercept point (IP3) tests, or create an interfering signal to simulate real-world transmissions.
On the measurement side, each bank of four ports has two receiver channels. These channels function independently to make four simultaneous scalar measurements or in concert to produce differential, reflection (one-port), or S-parameter (two-port) measurements. This flexibility is suitable for true parallel multisite measurements.
In addition, each bank’s pair of receiver channels has a dedicated measurement local oscillator (LO), allowing two different frequencies to be measured simultaneously for DUTs that have multiband functionality. All high-frequency hardware is fully integrated into the test head.
In the instrument rack are the RF synthesizers and the MVNA technology PCI card-based instruments, which provide the stimulus and measurement capability. The cards contain two integral 30-MS/s arbitrary waveform generators for I and Q signal generation along with four 60-MS/s measurement receivers.
The software is backward compatible, providing access to the company’s DSP library used by Quartet and RFSS users.
About the Author
Terry Wilson is a senior engineer at Credence Systems. He has 20 years of experience in the field of RF IC design and test and holds an M.S.E.E. from Oregon State University. Credence Systems, 5975 NW Pinefarm Place, Hillsboro, OR 97124, 503-466-7275, e-mail: [email protected].
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2000 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.