POWER MEASUREMENTS
Typical scalar power meters have good 50 Ω matches. If used in well-matched systems, they provide good power references or power transfer standards. When measuring devices directly with a scalar power meter, however, the devices typically do not have 50 Ω matches, therefore frequency dependent mismatch uncertainty can become quite large. This can be problematic even when making Pout and P1dB compression measurements at 2 GHz. At 50 GHz with the added overall socket and connection mismatch uncertainty approaching 1 dB, this approach is untenable.
Only with a vector corrected power measurement technique can a P1dB measurement in this environment have credibility. Traditional vector network analyzers (VNAs) measure only ratios such as gain and RL, but recently many of the new breed of VNAs now incorporate power accuracy calibrations to address this need.
MULTI-SIGNAL GENERATION
Since RF source power at 50 GHz is harder to come by and is degraded rapidly by transmission losses, amplifying and then combining signals close to the device under test is the only way to effectively source clean two-tone signals for testing. This is not typically available in commercial VNAs; therefore, a custom network analyzer test set must be assembled and the calibration necessary to characterize and compensate for inherent system losses, such as directivity and crosstalk, must be created. Some vendors are starting to offer more customizable network analyzer tools sets to address this need.
ONE-PORT S-PARAMETER ERROR CORRECTION
Figure 1 Simplified one-port error model.
National Institute of Standards and Technology (NIST) articles have been published with the rigors of six-port power and implied phase reference approaches of network analysis for addressing the challenges of S-parameter measurements of frequency translation devices with multiple ports and frequencies. These are especially relevant when testing at higher frequencies up to 50 GHz. Employing these methods (see Figure 1) greatly simplifies the measurement of S-parameters needed to error-correct the mismatch and loss effects more prevalent at 50 GHz.
The trade-off is the loss of absolute phase measurement capability through the device. Relative phase can still be measured accurately; however, providing the ability to make all the ratio S-parameter measurements needed for error correction. Some new VNAs incorporate frequency transitional device testing methods using these approaches.
VECTOR VOLTAGE AVERAGING
Measuring phase and amplitude of signals (vectors) enables signal processing techniques to be used in measurements at 50 GHz that extends the dynamic range and sensitivity in high frequency and high bandwidth systems. This is effectively a spectrum analyzer but with a vector receiver.
Figure 2 Signals below the noise floor (a) while the vector average of the coherent signals is above the noise floor (b).
Figure 3 Cassini ATE (a) and block diagram (b).
As frequency becomes higher, like bandwidth increasing, the signal-to-noise ratio begins to decline because more noise is present along with the desired signals. Figure 2 shows how vector measurements with random noise (shown in blue) average to a much smaller magnitude since opposite phases effectively cancel. A coherent signal (shown in red), however, averages to a stable version of itself that may be smaller in magnitude than the original noise floor, but much larger than the level of the noise vector average.
MODEL-BASED TESTING
Modulating signals such as WCDMA at 2 to 6 GHz is challenging with significant errors in both the modulation and demodulation measurement systems, meaning that effectively testing any device with less than 6 to 8 percent EVM is an exercise in futility to maintain any type of accuracy.
NIST typically recommends measurement equipment be 10x more capable than what is measured for best accuracy and anything under 2× more capable is effectively useless. This leaves many RF components unmeasurable when their individual specs are as little as 0.7 to as much as 6 percent EVM. When moving to 50 GHz, the ability to modulate these signals is further reduced, in part because the signals must be up-converted, effectively increasing errors dependent on the additive effect of linearity and phase noise characteristics in the upconversion process.
For high volume production testing, an error model approach like the way RF simulation systems predict EVM based on intrinsic device performance is much better, especially at 50 GHz. For power amplifiers, it is the AM-to-AM and AM-to-PM effects of the devices, and how they perturb the symbol constellation from the ideal, that leads to the EVM and ACPR calculation.
Error model EVM can provide measurement times approaching 30 msec with repeatability less than 0.1 percent. For I/Q modulators and demodulators, the DC offsets as well as the magnitude and phase errors of the devices are easily measured and characterized with traditional swept measurement methodologies.
Seven different measurements, four different RF carrier measurements at different DC offsets and three RF image measurements at different magnitude and phase balances, solve for the DC and AC intrinsic errors. This method minimizes baseband stimulus error, eliminates test system phase noise errors, minimizes waveform capture errors, is very fast at less than 100 msec and improves standard deviations.
Contributions by Roos Instruments engineers in the field of RF/microwave and mmWave device production test techniques regarding these novel production-proven approaches have been documented.1-5
COMMERCIALLY AVAILABLE IMPLEMENTATION
One of the currently available RF/microwave automated test equipment (ATE) systems that addresses these issues at 50 GHz is Cassini by Roos Instruments. Cassini is a modular architecture where test instrument modules (TIMs) provide the building blocks for a configurable microwave system from DC to 110 GHz (see Figure 3). This enables an enhanced measurement capability and frequency extension as well as instant multi-site expansion. TIMs are air-cooled, shielded instruments that provide all the source, receive, measurement and signal processing capability for a broad range of DC, digital, mixed-signal, RF and mmWave applications. Fixtures carry the modular architecture of Cassini into the device interface environment providing seamless integration with Cassini’s software, extending and enhancing the capabilities of the test instruments with an integrated calibration layer that guarantees signal accuracy to the device pin.
Designed to address the previously described challenges is the newly released System RF Core (SyRF Core) 25 GHz TIMs with support test set TIMs in contiguous bands to 50, 70, 86 and 110 GHz.
The focus of this article is the 50 GHz test set TIM paired with the new 25 GHz source and receiver TIMs. This synergistic design is one that optimizes the latest available technology to build the best performing measurement architecture on the market.
The key features of the new TIMs are:
- Fundamental mixing up to 25 GHz and complimentary low harmonic double balanced mixers for all frequencies up to 110 GHz
- Broadband compact high speed frequency synthesized source
- Excellent impedance match up through 50 GHz in coax using integrated 60 GHz blind mate connections to the device interface board
- Direct frequency domain high speed vector measurement receiver with a built-in fundamental LO
- Production-proven modulated measurements techniques.
The 50 GHz test set TIM can produce three simultaneous RF signals with independent frequency multipliers to provide the device with two-tone stimulus and mixer LO sources at the same time, which are required to test any up- or down-converting device at mmWave frequencies. It can down-convert and route both incident and reflected signals for each of these signal paths to the high speed vector receiver. Power measurements at mmWave frequencies are performed through the vector receiver and take advantage of vector error correction to compensate for mismatch issues.
CONCLUSION
The Roos Instruments Cassini ATE is the premier microwave/mmWave test system available on the market with the lowest risk and highest throughput for handling high volume millimeter frequency devices.
References
- M. S. Heutmake, “The Error Vector and Amplifier Distortion,” Proceedings of Wireless Communications Conference, August 1997.
- A. Nassery, S. Ozev, M. Verhelst and M. Slamani, “Extraction of EVM from Transmitter System Parameters,” Sixteenth IEEE European Test Symposium, May 2011.
- S. Yamanouchi, K. Kunihiro and H. Hid, “An Efficient Algorithm for Simulating Error Vector Magnitude in Nonlinear OFDM Amplifiers,” Proceedings of the IEEE Custom Integrated Circuits Conference, October 2004.
- D. Morris, W. R. Eisenstadt, A. Paganini, M. Slamani, T. Platt and J. Ferrario “Synthetic DSP Approach for Novel FPGA-Based Measurement of Error Vector Magnitude,” International Test Conference, November 2010.
- A. Nassery, S. Ozev and M. Slamani, “Analytical Modeling for EVM in OFDM Transmitters Including the Effects of IIP3, I/Q Imbalance, Noise, AM/AM and AM/PM Distortion,” 18th IEEE European Test Symposium, May 2013.
