alitalf | Hair tearing (Reply)

Since the middle of last week, the main task I have been working on is to stabilise the voltage control loop of a switched mode battery charger. The design is slightly more complex than some, because the output voltage can be both above and below the input - so it uses a sepic converter. Not a big deal, but slightly more difficult to control. The control loop oscillates over a range of 0.2Hz to 500Hz, depending on the components fitted to attempt to stabilise the loop. The one thing it won't do is work in a stable and accurate manner.

The switching voltage regulator would work fine of itself. The current limiting control manages to be stable and accurate, which is a minor victory. However, because the output voltage must be very accurate to charge lithium batteries correctly, an external voltage reference is needed, and this has been incorporated into the control loop by the simple expedient of comparing the output voltage with the reference, using an opamp. The opamp output is used to modify the output voltage that the voltage regulator is trying to supply.

As it stands, this approach would result in the regulator switching on and off rapidly. To avoid that unwanted result, the speed of response of the comparison is limited. Now it acts as an integrator over a wide range of frequencies. The snag is that the battery also acts much like an integrator, in the sense that the voltage rises as it becomes more charged. Integrating sin(x) gives cos(x), and doing this twice gives 180° phase shift. Since this is negative feedback, there will always be a frequency at which the circuit will oscillate.

In effect, about 30dB of gain, and a 90° phase shift has been added to a control loop that would otherwise be stable. If you imagine that, when steering a car, instead of turning the wheel 5° to the right to compensate for a slight drift to the left, instead you turn it as far as it will go, then turn it back the other way to compensate for the suddenly much larger error, you begin to get the picture.

This control loop revision has worked with different voltage regulators "Simple Switchers" but that range does not include one that covers the right range of voltages for this application. It has always surprised me how the approach ever worked, but "Simple Switchers" have a subtle and effective internal control loop which compensates for most external problems. My customer managed to use a technique that, on the face of it, has serious problems. In this application, the problems are impossible to ignore, and so far I have not found a solution that is reliable. It appears impossible to solve using the normal control loop design principles, so the answer is inherently difficult to find.