I have long held to the belief that electrons are smarter than people. Even the best engineers can fall prey to subtleties that electrons will readily act upon, especially when it comes to finding sneak feedback paths that can really screw up an otherwise stable feedback loop. We look here at a case study.
There was this multiple output DC power supply whose design was occasionally loop unstable and I was looking for the reason why and seeking to find a remedy. I set up an injected signal, call that “E-test”, as shown in Figure 1 so that by examining E2 with respect to E1, I could look at the gain and phase properties of the feedback loop.
Figure 1 Basic loop gain test plan for E-test.
There was a galvanic isolation barrier in the design, so the test setup was placed as follows (Figure 2):
Figure 2 Loop gain test plan in more detail.
We now look at how the Isolation Barrier Circuit was configured (Figure 3):
Figure 3 The alternating action clamping isolation barrier circuit.
The signal input voltage, E (on the left), gets transferred to the signal output voltage, E (on the right) by having a DC current source that drives the transformer secondary center tap to induce alternate clamping action via the two diodes on the transformer’s primary. We lose some level due to the Vcesat of the two NPN transistors and the forward voltage drops of the four diodes, but a linear transfer function from input to output is very closely achieved. A more detailed circuit is shown in Figure 4.
Figure 4 The alternative action clamping isolation barrier circuit in a bit more detail than Figure 3.
Please take mental note of the 8-volt power supply at the 1N4623 zener diode. We will return to consider the nature of those two parts a little later.
This pair of curves shown in Figure 5 shows the output of the isolation barrier circuit and the subsequent output of a PWM control signal versus input to the isolation barrier circuit. For the sake of feedback loop control, that is all we need.
Figure 5 The isolation barrier circuit linearity.
Although now superseded, the Hewlett-Packard 4395A Network Analyzer was used for loop testing (Figure 6).
Figure 6 To the left, the HP 4395A network analyzer used for our E-test. To the right, its attachment to the unit under test (UUT).
The 4395A was attached to the UUT via the 1:1 interface transformer shown in Figure 7. The braid of the coaxial cable served as the test transformer’s primary while the center conductor of the cable served as the test transformer’s secondary. The two 100 Ω resistors provide a nearly 50 Ω load for the analyzer’s RF output while the 100 Ω and 10 Ω resistors create a very small E-test in order to keep the power supply’s operating status as close to normal as possible while we do our measurements.
Figure 7 The test transformer and its attachment to the HP 4395A network analyzer.
We ran our loop gain tests at various levels of excitation for E-test and got a big surprise.
As the test signal level from the analyzer was taken from 0 dBm downward to -12 dBm, we had different test results (Figure 8).
Figure 8 Loop gain Seen for (a) 0 dBm, (b) -3 dBm, (c) -6 dBm, (d) -9 dBm, (e) -10 dBm and (f) -12 dBm excitation from the network analyzer.
While the loop gain roll-off characteristic looked good at first when the network analyzer was set to an output level of 0 dBm, the roll-off characteristic changed dramatically as the excitation level was changed.
The culprit was discovered as follows (Figure 9):
Figure 9 The sneak feedback path.
The 8 volts of power was being derived from the very same inverter that the PWM action was controlling, which led to a sneak feedback path as shown above. The on resistance of the zener diode facilitated that path and, in my suspicion, Rzener varied versus test excitation which led to the weird test results.
The zener was replaced with an active IC as follows (Figure 10):
Figure 10 Eliminating the sneak feedback path.
By using the LM136 with its extremely low dynamic resistance and changing one resistor to restore the PNP transistor’s Q-point, the sneak feedback path was eliminated.
Test results became the following (Figure 11):
Figure 11 Loop gain and loop phase with sneak path removed.
With the sneak feedback path broken, the gain-phase results were good and alike to each other at all levels of test drive.
We had incorrectly assumed the power supply was a linear system. Because of the zener’s behavior, the power supply was really a non-linear system.
Starting from scratch as it were, loop gain and loop phase tests should always be run at varying levels of excitation to see if the test results match each other at every excitation level. If they don’t, you have a non-linearity somewhere that may cause trouble for you and/or for your end user.
John Dunn is an electronics consultant, and a graduate of The Polytechnic Institute of Brooklyn (BSEE) and of New York University (MSEE).
Related articles:
