Like the prices of any products that compete largely in the consumer market, those of wireless devices such as LANs and cell phones are being squeezed constantly. That, in turn, means relentless downward price pressure on RF ICs. One of the most important ways to keep their costs down is to achieve the lowest possible test cost.
One major challenge associated with testing wireless-LAN devices is the broad-frequency spectrum of each channel. In 802.11a and 802.11b WLAN devices, the bandwidth of each channel can be as high as 22 MHz, which includes the transmit spectrum mask dBr, which represents dB relative to the sin(x)/x peak. For the 802.11a spectrum, the occupied bandwidth is 20 MHz, and the transmit mask is extended to 30 MHz. To be cost-effective, it is desirable to reduce the test time-that is, to make this measurement in one acquisition.
Another test challenge is measuring error vector magnitude (EVM), which is a measure of the transmitter's modulation accuracy representing any deviation from the ideal bit location. In the real world of RF IC test, the measured EVM is a composite measurement of the actual device performance, the performance of the equipment used to make the measurement and all of the interconnections between the two.
There are two primary steps to testing a WLAN receiver: test the individual elements of the receiver chain or perform a bit error rate (BER) test that covers all the elements of the receiver chain in one measurement.
To test the individual elements, a suite of tests must be developed to cover all the major blocks in the receiver chain: low-noise amplifier (LNA), mixer, synthesizer and demodulator. Some of these tests include noise figure, phase noise and jitter, amplitude jitter and quadrature phase error.
BER testing is the second option.
Using either of the two options, we must be able to quantify the receiver performance. For the slowest data rate (6 Mbits/s), 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, requiring a dynamic range greater than 80 dB and as high as 114 dB.
As with the EVM measurement, it is essential to control the electrical characteristics of the contactor and the design of the load board for the BER measurement. For example, too much lead inductance could affect the noise figure of the receiver and yield a distorted test result.Test solutions
Unfortunately, there is no single test system that can economically test all the RF ICs designed into today's wireless applications. For testing low-complexity devices (such as power amplifiers or low-noise amplifiers), available ATE generally exceeds requirements. Consequently, the lowest-cost test approach is to use a custom rack of instruments. Note that testing these simple RF ICs requires little or no digital test capability, so the low-cost RF test system does not need a dedicated digital test subsystem or a traditional test head.
Instruments are readily available from Agilent, Anritsu, Keithley and Rhode & Schwarz. However, integrating them to achieve low test times requires specially designed and optimized software that fully utilizes the instrument's capability. Test integrators like Amkor Technology develop test systems and user-optimized software to effectively integrate multiple instruments into one RF IC test stand.
Midcost testers originally were required to test a single tone stimulus and measure the distortion of the receiver so the digital pins did not require speeds greater than 10 MHz. However, digitally based modulation schemes such as those used by WLANs have generated new test requirements. ATE must now test parameters such as EVM and BER.
With signal bandwidths around 20 MHz and nominal and dynamic range requirements in excess of 80 dB, the test for 3G wireless devices requires a difficult combination of high sample rates, fine resolution and high dynamic range.
The tester must be able to digitize data with at least 14 bits' resolution and a sample rate greater than 65 Msamples/s. To increase dynamic range, the testers are typically designed with a two-stage down-conversion architecture. For GSM and W-CDMA, RF testers must go up to 6 GHz, with baseband and digital capability up to 50 MHz.
The proper design of the load board, contactors and software are key factors in successfully providing the lowest cost-of-test. Contactors should be able to test RF packages up to 0.45-millimeter pitch and should have a minimum downtime caused by package debris. Ideally, a spring-type contactor is desired so that repairs can be made on site.
It has been found that spring contacts work well up to 2 GHz; above that we have experimented with metal over conductive elastomer with excellent electrical and mechanical integrity.
Raimondo P. Sessego is director of the RF Test Product Group at Amkor Technology Inc. (West Chester, Pa.).<.i>
See related chart