[024] Frequency Response Measurement Part 2: Single-Frequency Measurement
Single-frequency injection and measurement is used to analyze and test power supplies.
Introduction
In this article, Dr. Ridley continues the topic of frequency response of switching power supplies. Last month’s article focused on the broadband noise generation of power supplies. This article shows how we can extract single frequencies one-by-one from the noise with the AP300 frequency response analyzer. This allows us to perform control measurements on our switching power supplies.
Single Frequency Measurements for Switching Power Supplies
The previous article in the Design Tips series of this magazine [1] discussed the wide range of frequencies generated by a switching power supply. High noise is unavoidable when working with switchers, and the converter must also be tightly controlled in the presence of this noise.
This presents a challenge in two areas: firstly, the control chip must run reliably with the noise and predictably set the duty cycle of the switch from one cycle to the next. Secondly, measurements must be taken on the power supply once a control loop is closed to ensure that it is always stable.
Analog controllers do an excellent job of managing the first problem through the simple sawtooth ramp and reference technique of a standard PWM controller. Once the controller is working properly, we use traditional Bode plot measurements to assess performance and stability.
Power supplies are one of the few areas of analog electronics that still make conventional measurements rather than depending on prediction and modeling alone. The loop gain and stability margin can vary tremendously for a converter operating over its full range. Figure 1 shows the range of variation that can exist for a simple buck converter when variations of line, load, temperature, and EMI filter are considered.
Figure 1: Variability of Bode plots for a “simple” buck converter with variations of line, load, temperature, and EMI input filter.
These curves show extreme changes in the gain and phase of the converter. The curves, however, assume linearized operation, with small-signal circuit models used to generate the plots. In the real world, the switching power supply may have regions of operation that are not well modeled, and even more variation is possible. Clearly, in the face of such extreme behavior, measurement of the power supply is an essential step in making sure the design of the control loop is rugged.