For the best way to measure the loop gain of a switching power supply while the circuit is running closed-loop with a regulated output, see the page on Loop Gain Measurement Setup. You will need a Frequency Response Analyzer to do this, as it cannot be done with a signal generator.

 

Crossover Frequency

With Voltage-Mode Control, the crossover frequency must be at least 3 times the equivalent LC filter resonant frequency if the converter is in CCM mode-- in order to attenuate the LC filter characteristics.

For flyback and boost converters, the crossover frequency should not exceed about 1/4 of the power stage Right-half Plane (RHP) zero frequency. If this is too low to be able to control the supply tightly enough, the power stage should be redesigned to move the RHP zero to a higher frequency.

The crossover frequency should be below 1/8 of the switching frequency.
Filtering, noise, and higher-order parasitics in the power stage restrict the practical attainable crossover frequency.

If you need a very high crossover frequency above 15 kHz or so in order to reduce component size on a high frequency, high density converter, you will probably need some additional high-performance components for the feedback amplifier.

Remember, crossover frequency is not an objective or specification of the power supply. The important quantities are output impedance and line rejection which determine the step load response of the converter, and the input noise rejection.

 

Loop Gain Measurement

Releasing a power supply to production without a comprehensive set of loop gain measurements is poor engineering practice. This measurement is as important as checking the peak voltage stress on the power switches with an oscilloscope.

Many companies continue to design and release power supplies without a loop gain measurement. And many power supplies in the field oscillate and fail. Don't fall into this trap, modern equipment makes it inexpensive, easy and quick to verify loop gains. A couple of days extra work in characterizing and optimizing the control loop is much better than stopping the production line or losing a valued customer.

Loop gain measurements will often differ significantly from prediction-- that's one of the reasons you should always make a measurement. The existing models are theoretically sound, but they often do not take into account circuit parasitics, especially in multi-output converters.

And predictions are frequently inaccurate at the extremes of power supply operation-- particularly at high line and light load for many converters.

© copyright Ridley Engineering, Inc. 2007