Home>Manufacturing Test Center>Test How-To

ATE challenges for mission-mode testing of wireless transceivers

Steven Fields, Scott Therrien, and Dr. Steve Lyons, Teradyne- October 31, 2012

Wireless devices have become critical tools in everyday life.  They allow us to transfer voice, data, images and video with ease.  The evolving quest for increasing quantities of data to be transferred to the furthest reaches has driven modulation standards to allow higher date rates.  In order to produce devices that utilize new high-data-rate modulation standards, the baseband and RF modulator chipsets which comprise these devices must deliver increasing levels of performance and precision. This makes the performance of the automatic test equipment used to test the semiconductor devices in production more demanding. Combined, all of this requires new testing strategies.

Transmitting and receiving data over the air is challenging because wireless devices need to be able to decode symbols accurately in the presence of numerous uncontrolled sources of noise and interference.  This is especially critical when using increased data rates for WLAN and cellular devices, as well as when trying to maintain very high data rates at the edges of the wireless range.  Due to the environment in which these devices operate, it is critical that the wireless transmitter and receiver elements produce little or no impairment to the modulated / demodulated signal. For heterodyne systems that use an intermediate frequency, controlling impairments can be made easier by the use of better components and multiple down conversions that can maintain image frequency rejection.

Impairments
Heterodyne systems also can have the ability to process the in-phase (I) and the quadrature-phase (Q) signals in DSP and avoid exposing the baseband signals to analog circuitry.  This generally reduces the amount of IQ impairments that are introduced into the system at the cost of increased design complexity and power consumption.  Homodyne (Zero-IF) conversion systems on the other hand are less expensive, less complex and require less power. They perform a direct conversion from baseband signals in the analog domain to RF or vise versa.  It is this exposure to the analog domain and less control of the image frequency that can introduce IQ impairments.

No matter what transceiver architecture is used, it can be difficult to produce a mixed-signal IC with no IQ impairments. For transceivers with less than perfect native performance attributes, it may be necessary to adjust baseband signals to overcome impairments in the transceiver, in the baseband to transceiver link or in the system itself. This has the effect of improving the modulation accuracy along with improving the spectral properties of the modulated signal.  

Impairments that may be acceptable for low bandwidth standards could be unacceptable for latest generation higher bandwidth standards.  Take, for example, the evolution of the cellular standards GSM, WCDMA, and LTE.  3G GSM and WCDMA employ GMSK and QPSK modulation schemes respectively, providing two bits per symbol each.  4G LTE, however, employs a 16QAM or 64QAM scheme with the capability of tripling the bit density of each symbol relative to its 3G counterpart.  In wireless LAN systems, the demand for higher data rates drives 256QAM.  This requires the modulator to produce higher quality modulated signals in order to maintain higher data rates at the furthest extent of the required coverage area in the presence of broadband noise.  An example of this higher quality modulation requirement is the 802.11ac standard [1], where the error vector magnitude (EVM) system requirement for QPSK is 22.4% and for 256QAM is 3.1%.  As future standards drive increased data rates over increased distances, the importance of compensation for IQ impairments will become even more important.

Correct for impairments
There are two fundamental ways to design systems with improved modulation accuracy.  One is to improve the architecture of the devices to remove the sources of I and Q impairments. This approach comes at cost of complexity and price. Another approach is to measure the inaccuracies and to perform a one-time or real-time correction to the I and Q signals to correct for these impairments that result in improved modulation accuracy.  Figure 1A shows an example of such a device. A feedback path is used to evaluate the IQ impairments.  The detection of IQ impairments is typically done by using either side band suppression measurements or digital demodulation to evaluate these impairments.  Once impairments have been determined, depending on the system architecture, corrections can be made by affecting the analog circuits or pre-distorting either the I or Q waveforms digitally to achieve the I and Q balance required.



ATE typically only tests a subsection of the end application at a time. This means that the feedback path and correction mechanism may not be present during test.  Therefore, in order to understand how the device under test (DUT) will perform in the entire system, the ATE needs to emulate the feedback path along with the correction methodology to simulate the end use of the system.  We call testing of the DUT as would be used in its end-use application vs. a simplified alternative test mode a ‘mission mode’ test strategy.  Mission mode testing is a more robust test strategy because it tests the device in the same mode of operation that the device will actually be used. To implement mission mode testing there are numerous challenges that must be overcome when using ATE instruments and fixtures. One challenge is overcoming additional sources of IQ impairment that could be introduced by the analog baseband instrument as well as the quality of the test fixture (see Figure 1b).

These impairments, whether from the DUT, ATE or test fixture, can be grouped into two categories; frequency dependent and frequency independent error.  Frequency dependent errors, such as filter group delay, are not uniform across bandwidth and are beyond the scope of this article. Frequency independent errors have the same effect regardless of bandwidth.  Quadrature error (quad error), IQ amplitude imbalance and IQ timing skew are all considered frequency independent errors. Quad error is the orthogonal error between the I and Q signals. Ideally, I and Q should be orthogonal (90 degrees apart) at the output of the mixers. However due to quad err they are not 90 degrees apart.  This error typically can only be introduced by the quadrature modulator / demodulator of the DUT.  

Another type of error is I/Q gain imbalance, this is the ratio of the gain of the I signal path vs. Q signal path either before or after the mixers. The ideal gain ratio should be one. Gain imbalance can be introduced by the quadrature modulator / demodulator as well as the ATE and test fixturing.  The third type of error is baseband IQ timing skew error. This is the amount of time based skew is between the I and Q data streams of the baseband signal. In ATE systems, time based skew is often due to an unequal start of I relative to Q or differences in path lengths of the test fixture [2].  

Often the automated test system’s load board design can be a critical element which introduces undesired I/Q amplitude and timing skew impairments.  In an end application, such as a cellular handset, the baseband IC and wireless transceiver IC may be physically located mere centimeters or less from one another.  In the test environment however, the baseband AC instrumentation acting as proxy for the baseband IC is much farther away than a couple of centimeters.  It may be a foot or more away from the wireless transceiver DUT.  This opens the door for a multitude of error sources to be introduced that would not normally exist within the handset.  Trace length mismatches and parasitic errors from non uniform trace routing, potentially on multiple layers can lead to unwanted skew impairments.  Long trace lengths can lead to the addition of active components in the path such as amplifiers which can add further impairments.

Figure 2 shows the effects of these impairments on the constellation.  The effects vary depending on the type of modulation used.  For example, single carrier modulations, such as 802.11B (CCK), are demodulated in the time domain and the amplitude of I and Q components at points in time determines the constellation location.  Multi-carrier modulations on the other hand, such as LTE and 802.11ac (OFDM), are demodulated in the frequency domain. A complex fast Fourier transform is used to determine each constellation point for each carrier frequency individually.  For this reason the effect of quad err, gain imbalance and time skew look different for multi-carrier modulated signals.  Notice that gain imbalance in single carrier modulation takes on a rectangular shape whereas a multi-carrier modulation the constellation produces a miniature constellation diagram at each constellation point.  Quad err also takes on different shapes depending on the modulation type, trapezoidal for single carrier, and mini-constellation for multi-carrier.