MCUs Heat Up Digital Loop Control

Aug 1, 2008 12:00 PM
By Tom Stamm, Applications Engineer, and Vipin Bothra, Power Electronics Applications Group Manager, STMicroelectronics, Schaumburg, Illinois

Conventional MCUs meet the digital-control requirements for ac-dc power supplies, including sufficient processing time to do the usual monitoring and communication tasks after the control-loop calculations.

Today's microcontroller units (MCUs) can implement the control functions and feedback loops of off-line ac-dc power converters, which do not require extremely fast control loops. As an example, a relatively simple off-line converter can employ a conventional microcontroller instead of specialized DSP, FPGA or chips containing complex programmable delay lines. Hardware engineers with little or no MCU experience will be surprised at how easy it is to develop such a system.

Response Time is Hardware Limited

Off-line power converters typically have response times limited by their output-filter components. Extremely fast response in the control circuit is not required, so a conventional MCU can operate the feedback loop.

Typically, the output filter (Fig. 1) is the limiting factor in the control-loop design. The inductor is selected to give reasonable ripple current in the output capacitor. In our example, load voltage is 12 V and load current is 4 A. The choke value would be set to give a peak-to-peak ripple current of one-fourth the rated load current, allowing continuous-mode operation down to one-eighth of the rated load current. The choke value is selected based on the ripple frequency. For our 150-kHz converter, the value is 50 µH.

A low-cost electrolytic typically would be selected for the output filter. The electrolytic's equivalent series resistance (ESR) sets the peak-to-peak ripple voltage. A typical ripple specification is 1% of the output voltage, or 0.12 V. ESR, therefore, must be less than 0.12 Ω. A low-cost capacitor of 470 µF and 63 V was used in the demo. It has an ESR of about 0.06 Ω, giving a reasonable safety factor. The high-voltage rating gave some margin during debugging.

Consider the physical limits of the filter's input. Assume a maximum secondary voltage of 48 V set by the transformer turns ratio and source voltage. Since the forward converter has a maximum duty cycle of 50%, the average forcing voltage range for the converter is 0 V to 24 V.

Fig. 2 shows the best damped response time, 425 µs, given the range of forcing voltages. The damping resistor closely matches the impedance of the L-C circuit. In practice, the damping would be applied by the system feedback loop, with the input forcing voltage (duty cycle) reduced as the output voltage reaches its setpoint. The damping resistor serves the same function, in that it removes the excess current stored in the inductor as it approaches the setpoint. The response time cannot be improved with the components and forcing voltages selected. And, since design margins are important, the response time will be even slower.

Demo Verification

To prove that a conventional MCU could do the control job, we constructed a demo. The demo contains an STM32F103 MCU (Table 1), which is largely under-employed. The demo (Fig. 3) consists of two pc boards, a MCU board and the power-converter board. The power converter is a two-switch forward converter, operating from the rectified ac line, feeding an isolated synchronous rectifier and L-C output filter. The system output is 12 V at 4 A. For the benefit of the software side of the development team, we referenced all control to the secondary side, isolated from the line.

The MCU demo board has many unused features. Only the MCU chip itself, the debug interface, the reset button and the 5-V to 3.3-V regulator are used for the control loop. All inputs and outputs are taken directly from the MCU pins, with level shifters added to allow interfacing with gate drivers.

Fig. 4 shows the system architecture and the general interfaces to the MCU. The system requires an isolated housekeeping supply, because the MCU cannot bootstrap its own power supply. Q1 and Q2 implement a 150-kHz two-switch forward converter driven in phase by a common transformer fed by the drive chip on the secondary side. (A 1-A fuse in series with the power-transformer primary helped the debugging phase by saving a lot of replacement parts.)

A current transformer isolates the primary current signal. The burden resistor directly feeds an analog comparator (not among the peripherals on the selected MCU die).

Schottky diodes shunt the synchronous rectifiers, which simplifies debugging. You can remove them once the dead-time is set.

The output ORing diode (D10) was used later to aid in debugging a current-sharing scheme using digital communication. Our planned PMBus sharing scheme will not be discussed here.

A voltage divider consisting of R13 and a resistor added to the MCU board senses the output voltage. There is no phase-lead network. Control-loop compensation is handled digitally.

Acceptable Use Policy blog comments powered by Disqus