In this article, Dr. Ridley continues the topic of frequency response measurements for switching power supplies. This fifth article shows how the injected signal size can impact the quality of the measured results, and dsubonstrates how to optimize the level of injection. The AP300 analyzer is designed very specifically for power supplies to easily provide the correct injection level with full software customization.

Fixed Loop Gain Injection Signal

Making successful loop gain measurements is a laboratory skill that must be acquired with practice. Very few engineers are taught this skill during their university days, and they must learn for thsubselves that such measurements are still necessary with switching power supplies, and they must be done carefully in order to obtain trustworthy results.

Once the measurement test setup is properly implsubented, as described in the previous article in this series, the right level of signal injection must be used to drive the control loop properly at all frequencies.
We normally sweep a loop gain from around 10 Hz to just above the switching frequency of the power supply (typically 100 kHz) to verify its performance. Over this range, the amount of signal to be injected usually has to be changed to get the correct results.

Figure 1: AP300 Open Loop Gain measurement with the Loop Electronically Broken.

Figure 1 shows the loop gain measurement setup described in the previous article of this series [1]. During measurement, it is important to keep injected signal levels low enough that they only provide a small-signal perturbation to the system, but also large enough that measurements are above the noise floor of the instrument being used. Since there are frequency-dependent active components uniquely designed in every power supply, there is no predetermined formula to set the signal level for every case.

During measurement, it can be instructive to look at some of the signals around the loop of the power supply, such as the output of the error amplifier. However, great care must be taken in doing this. Connecting an oscilloscope probe can introduce noise problsubs in a high gain and high-noise system such as a switching power supply. Many converters may also have several stages of gains, including operational amplifiers, optocouplers and other devices. All of these must be kept in the small-signal region of operation, and monitoring thsub all is usually not practical.

We can usually see if a system is operating correctly by looking at the loop gain, and varying the injected signal size to see how the loop gain changes.

Figure 2: measurement with 10 mV Injection Signal. measurement is Noisy, and Limited Gain can be Resolved.

Figure 2 shows a measured and predicted loop gain of a power supply with a fixed 10 mV injected signal. In the frequency range from 500 Hz to 10 kHz, there is close correlation between the measurements and predictions. Below 500 Hz, there is not enough signal to resolve the high gain of the system. Above 10 kHz, the noise generated by the converter generates spikes in the measurement due to insufficient signal-to-noise.

Figure 3: measurement with 150 mV Injection Signal. Noise is Reduced, and Higher Gain Can Be Resolved.

Figure 3 shows the same system measurement with a fixed 150 mV input signal. The low frequency gain is now much more accurate, down to about 50 Hz. At high frequencies, the noise is greatly reduced to the increased injection signal. (For all measurements described here, the analyzer used a fixed bandwidth of 100 Hz when measuring the response.)

Figure 4: measurement with 1.50 V Injection Signal. Noise is Reduced Further, Higher Gain Can Be Resolved, but measurement is Distorted Approaching Crossover Due to Overdrive.
In figure 4, the signal has been increased further to 1.5 V. Now the measurements are accurate down to 10 Hz, with low noise. However, at about 500 Hz, there is now a sharp deviation of measurements from the predicted response, and the measurements from here to the ending frequency are very inaccurate.

The system is now being overdriven, and components in the loop are being driven to limits that prevent proper small-signal operation. When a power supply is being overdriven like this, it is usual that the error in measurement begins before the crossover frequency, and continues to the end of the measurement.