Many tests, many setups

READ OTHER FEBRUARY ARTICLES:  Contents, February 2006

Broadband differential variable-gain IC amplifiers. The amplifiers are designed to operate at frequencies from DC to 10 GHz with a maximum gain of 27 dB. The amplifiers are used in medical instrumentation and test and measurement equipment.

Fully characterize the amplifiers under operating conditions such as temperature, power-supply voltage, and input signals. Tests on the amplifier include gain, common-mode rejection, linearity, differential output voltage, total harmonic distortion, and power dissipation. Develop unique setups for each test. Conduct all tests at temperatures of 25°C, 55°C, and 85°C and with supply voltages 4.75 V_CC to 5.25 V_CC and –4.94 V_EE to –5.46 V_EE.

• Agilent Technologies: Vector network analyzer, network/spectrum/impedance analyzer, transmission/reflection test set, sweep generator. www.tm.agilent.com.
• Anritsu: Microwave power divider. www.us.anritsu.com.
• Picosecond Pulse Labs: Bias insertion tees. www.picosecond.com.

Inphi (Westlake Village, CA; www.inphi-corp.com) manufactures broadband amplifiers that require dozens of measurements during design verification. When engineers developed a DC-to-10-GHz differential amplifier, they knew that customers would need these measurements to specify the amplifier. For each of the measurements, principal design engineer Mike Case configured a circuit.

A 100-mV AC signal from the VNA feeds a power splitter that produces two outputs, each of which is 50 mV. The outputs connect to the DUT through phase-matched cables. Case measured the amplifier's output at three gain settings (high, middle, and low). Because the amplifier's common-mode gain is 6 db higher than the value measured by the VNA, Case adds the single-ended gain (6 dB less than the common-mode gain) to the measured CMR. That produces the common-mode rejection ratio measurement.

Inphi engineers used this setup to measure common-mode rejection in a DC-to-10-GHz differential amplifier IC.

When Case measured CMR with signals above 10 kHz, the amplifier performed within specifications. But he needed to make CMR measurements with AC signals down to DC, and he still needed to produce a common-mode voltage to measure CMR. The manufacturer of the bias tee, however, specifies the device's low-frequency limit at 10 kHz. Below 10 kHz, Case encountered CMR measurements that didn't make sense. "We expected the bias tee to function like a linear high-pass filter below 10 kHz," he said, "and we assumed that we could compensate for those effects."

The DC common-mode voltages changed unexpectedly when the AC signals went below 10 kHz. The nonlinearity appeared to produce gain errors, which Case also observed while measuring the amplifier's single-ended gain. The bias tees altered the amplitude of the AC signal, which produced the illusion of a nonlinear gain-versus-frequency measurement in the amplifier.

"We attributed the unexpected differences in gain and CMR to a nonlinear inductance in the bias tees below 10 kHz," said Case. As a result, he hasn't yet made CMR measurements below that frequency, except with the common-mode voltage set to 0 V.

If you use an electronic component outside its specified range, you can't always predict how it will perform. When I interviewed Case, he was looking into other methods to make the CMR and single-ended gain measurements with input signals below 10 kHz. He also learned that hysteresis in the bias tees made characterizing them difficult below that frequency because the hysteresis makes them perform nonlinearly.