Four-port calibrated structures
Engineers at M/A-COM developed a new calibration theory that allowed them to compensate for earlier errors in calibration.
Martin Rowe, Senior Technical Editor -- Test & Measurement World, 8/1/2005
READ OTHER AUGUST ARTICLES:
Table of contents, August 2005
AUGUST FEATURES:
Testing for life's sake
Fears temper career satisfaction
Continuity testing
Quick EMC screening
Four-port calibrated structures 
Microwave amplifiers, filters, transmitters, receivers, and radar chip sets that operate at frequencies to 40 GHz. These balanced differential devices are made on silicon or gallium arsenide substrates and are on wafer during tests.
THE CHALLENGEDevelop a test method and the equipment to calculate S-parameters for four-port differential microwave devices based on multimode techniques. Calibrate a test station that consists of a vector network analyzer at frequencies to 40 GHz. Overcome problems found in commercial calibration standards.
THE TOOLS- Agilent Technologies: vector network analyzer, attenuator/switch driver. www.tm.agilent.com.
- Cascade Microtech: automatic wafer prober and ground-signal-signal-ground (GSSG) RF wafer probes. www.cascademicrotech.com.
- Electroglas: automatic wafer prober.www.electroglas.com.
- M/A-Com: custom RF calibration structures. www.macom.com.
- The Mathworks: technical computing software. www.mathworks.com.
- Pico Probe: GSSG RF wafer probes. www.picoprobe.com.
Before engineers at M/A-Com (Lowell, MA) can characterize balanced, 24-GHz ICs with a vector network analyzer (VNA), they need to calibrate their test system. To compensate for errors introduced by cables, switches, connectors, and wafer probes, the engineers measure the responses of calibration standards and compensate for errors in software. After making the calibration using custom calibration standards and compensating for measurement errors, engineers repeat the measurements on the calibration standards to verify accuracy in the process.
"When we used commercially available calibration standards at frequencies above 10 GHz, we found that the magnitude of S21 would be greater than unity when we measured the calibration standards again," said principal engineer Alan Jenkins. "That's impossible, because the calibration standards are passive structures." To solve the problem, M/A-Com developed its own on-wafer calibration standards.
| |
| Switches let a modified two-port VNA make four-port measurements. |
The standards are a uniform set of microwave transmission lines. One is a planar "through" standard with effectively no delay. Two delay standards introduce known, fixed propagation delays to signals passing through them. A fourth, "reflect" standard is a specific open circuit between the structure's test pads. All structures use two signal traces surrounded by two ground traces, thus making four ports. This configuration lets engineers perform a four-port calibration with results consistent with two-port calibrations for balanced circuits. With four-port measurements, the common- and differential-mode S-parameters are available. (See diagrams of the standards.)
To make the VNA into a four-port instrument, M/A-Com engineers added an attenuator/switch driver. They drilled through the case and connected coax cables close to the instrument's input circuits. The coax cables connect to the attenuator/switch driver, which then connects to a patch panel. The VNA, attenuator/switch driver, and patch panel reside in an instrument rack located next to a wafer-probing station.
Jenkins, along with principal engineers Tekamul Buber, Noyan Kinayman, and others, developed a "multimode through-reflect-line" calibration theory, dubbed MTRL. After making measurements on the calibration standards, they use custom software to calculate eigenvalues that they apply to DUT measurements. The software uses the eigenvalues to correct the DUT measurements to produce the final results.
LESSONS LEARNED"We learned more about calibration methods than we thought we would," said Jenkins. "Finding a gain above 10 GHz for a through line was an eye opener." By using their own calibration standards, M/A-Com engineers were able to prove that the errors were caused by the commercially available calibration standard, not by the compensation algorithms they had developed. "The technique lets us separate common-mode gain from differential gain," added Kinayman.
 |
| Diagram of the standards |
