This summer, I have had the privilege of mentoring an intern. Helping “John” troubleshoot his system design projects has reminded me about some of the important lessons I've learned throughout my career. Lessons I often take for granted.
The other day, John was taking transient stability data on a design and learned one of the bigger lessons: The step size of the output signal is extremely important with regard to achieving accurate results.
The situation came up while John was working on a design where an op amp had to buffer a 1μF load. Most op amps can't directly drive large capacitive loads, so he had to design an appropriate compensation scheme. Since the output didn't need to deliver much current, he inserted a series isolation resistor (RISO ) between the op amp output and the capacitor to compensate the circuit, as shown in Figure 1.
I helped John verify the stability of the circuit in TINA-TI, TI's SPICE-based analog simulation program. After a little manipulation, he selected the RISO value that would achieve 45 degrees of phase margin. Now he was ready to test the results on the bench.
If you find yourself in a similar situation, beware. Most advanced methods used to test for stability in simulation are generally not practical on the bench. To compensate for this, input a step to the system that causes the output of the op amp to change by a small amount.
The results for a typical op amp system should appear with a standard overshoot and ringing response, as shown in Figure 2. By measuring the percent overshoot, you can determine the phase margin using the table also shown in Figure 2. The results in the figure show 20 percent overshoot on the 10mV step, corresponding to 45 degrees of phase margin.
After explaining the theory, I asked John to conduct the percent overshoot test in the lab.
He came back and told me he was getting “strange” results that were distorting the percent overshoot measurement. He noted that the function generator had a minimum step size of 100mVpp and showed me the results in Figure 3.
His results clearly show a different response when compared to the response shown in Figure 2, and will not correlate to phase margin. Since the amplitude of the step was the only difference, I showed him a trick that maintains the 50Ω termination for the generator, while also dividing down the input signal using R1 and R2, as shown in Figure 4.
Figure 5 shows the results of the test with a 10mV input step. Sure enough, the odd behavior was gone. These results allowed us to measure overshoot and verify that our design achieved a phase margin of roughly 45 degrees.
The linear part of the waveform, and odd spike in Figure 3, suggests the op amp was actually displaying a large-signal limitation, such as slew rate or output current limit. When performing a percent-overshoot-based stability test on an op amp system, the output change must be small, or the large signal limitations of the op amp will obscure the results.
John learned an important lesson. If an output limitation alters the step during the measurement, the results may appear strange and will not correlate to the phase margin.
Next time you're testing the stability of an op amp, limit the output change to between 10 to 20mV when performing percent overshoot tests on the bench. This will ensure the output remains small-signal. If the results are not consistent or usable, consider performing ac gain/phase sweeps.
That might be a good topic for my next blog post…
For weekly tips, tricks, and techniques from TI precision analog experts, visit our companion blog, the TI Precision Designs Hub.
About the author: Collin Wells is an analog applications engineer in the Precision Linear group at Texas Instruments. He supports industrial products and applications with a focus on current-loop and voltage transmitters for process control. Collin received his BSEE from the University of Texas, Dallas before joining Texas Instruments in 2007. Related posts: