Optimal digital power control using LLC resonant converters

Bilal Akin PhD and Daniel Chang, Texas Instruments Incorporated (TI)

April 24, 2012

Bilal Akin PhD and Daniel Chang, Texas Instruments Incorporated (TI)

SR PWM timing considerations
Synchronous rectifier (SR) current has a positive half-wave sinusoidal shape. Ideal SR timing would have the MOSFET on during non-zero positive current and off at all other times, the same way a diode conducts. This means the SR would turn on at zero current just as current begins and turns off at zero current just as current ends, achieving Zero Current Switching (ZCS).

SR turn on timing can be easily obtained based on primary side switch timing. This is because SR current starts flowing at the beginning of the half period, when the primary side switch turns on. By setting the SR PWM to turn on at the same time or slightly after its corresponding primary side half-bridge PWM, ZCS can be achieved during SR turn on. SR turn off timing is more difficult to obtain. This is because the SR turn off current zero crossing point is variable with frequency. Above the resonant frequency, SR current actually never reaches zero before the end of the half period. In this situation, SR turn off timing is simply at the end of the half period. Even though ZCS is not achieved, this provides the minimum power loss. At the resonant frequency, SR current reaches zero at the end of the half period. In this situation, SR turn off timing is also at the end of the half period, but ZCS can be achieved. Below the resonant frequency, SR current reaches zero before the end of the half period.

This results in three possible scenarios. First, if SR turn off occurs too late, negative current can flow backwards through the SR MOSFET which is undesirable and can lead to component damage. Second, if SR turn off occurs too early, ZCS is not achieved and additional power loss occurs. Third, if SR turn off occurs at the zero crossing point, ZCS is achieved. The third scenario with ZCS is the one desired.

There are many ways to set SR turn off timing. One simple method is to choose a fixed timing (relative to either the beginning or end of the half period) that ensures SR turn off at the ZCS point or earlier for all frequencies, providing some benefits of SR without risk of component damage. A second more advanced method is to adjust the SR turn off timing based on the frequency. This would allow ZCS for all frequencies but, unless the SR turn off timing is updated sufficiently fast, either of the first two scenarios for operation below resonant frequency can occur after sudden shifts in frequency. Both of these methods would also require experimentation to determine the SR turn off timings required for each implementation, which can be time consuming or impractical. A third method is to adjust the SR turn off timing based on the SR current level directly. This would require additional sense circuitry but could simplify development and reduce computational requirements.

Transient state tuning
To keep loop tuning simple and avoid the need for complex mathematics or analysis tools, the number of degrees of freedom have to be considered by remapping them to a more intuitive set of coefficients. For example, working with the five 2P2Z regulator coefficient terms (B0, B1, B2, A1, and A2) can be simplified by remapping these terms to the P, I, and D coefficient gains, each of which can be independently adjusted. This method requires a periodic transient or disturbance to be present and a means to observe the output transient while interactively making adjustments while the built-in active load on the converter board can provide the periodic disturbance (see Figure 3).


Click on image to enlarge.
The compensator block has two poles and two zeros and is based on the general infinite impulse response (IIR) filter structure. The transfer function is given by:

The recursive form of the PID controller is given by the difference equation:

where:
And the z-domain transfer function form of this is:

Comparing this with the general form, we can see that the PID is nothing but a special case of CNTL_2P2Z control where A1 = -1 and A2 = 0.


Click on image to enlarge.
Figure 3: Active load test, from full-load to no-load transient response tuning with various regulator coefficients

Burst mode operation
When the resonant converter is lightly loaded or not loaded, there will be significant primary current flowing through the transformer's magnetizing inductance to maintain soft switching, thereby introducing losses and significantly reducing light-load efficiency. To overcome this problem, the converter can be run in burst-mode to keep the converter's input consumption to a minimum; when the load falls below a certain value, the program will enter burst mode. Burst mode is a series of switching cycles at a nearly fixed frequency and a duty cycle spaced out by long idle periods where either switches are in OFF-state or duty cycles are set to zero as shown in Figure 4. In this way, the average value of the resonant tank current can be reduced to an almost negligible value. Furthermore, the average switching frequency will be considerably lower, thereby reducing switching losses.













(a)



(b)



(c)



Fig. 4 Various burst mode implementations



In this implementation, the burst mode on/off decision is based on output ripple. Since the amount of ripple is not critical at no-load, a bandwidth less than 5% of the output voltage can be defined to turn on and off burst mode. In addition, a software subroutine can be added to adjust the on/off period according to system ripple limitations. When Figure 4a is compared to Figure 4b, ON time can significantly be reduced to improve light-load efficiency. The flexible control capabilities of the microcontroller will allow developers to implement burst mode operation in a hybrid manner and adjust the duty cycle as well.

Figure 4c shows a duty cycle limited to 10%. This allows the system to obtain smoother transients, reduce inrush current and lower stress on components. Depending upon the system specifications, developers can select an optimal combination of all these alternatives to obtain the highest light- or no-load efficiency.

In addition to the burst mode, the hybrid approach enables soft-starting of the converter. LLC converters initially tend to draw huge currents that can be controlled by increasing the switching frequency up to three times higher values. By means of a hybrid approach, inrush current can be efficiently suppressed at relatively low switching frequencies.

Conclusion
Many OEMs are turning to digital power control technology to improve system performance and efficiency. Advanced topologies, such as those based on LLC resonant converters, bring many benefits to OEMs and end-users, including lower system cost, better responsiveness, higher reliability, and optimal power efficiency. By using the flexibility of a programmable approach with integrated hardware components, OEMs can quickly and easily customize operation and maximize efficiency across a wider range of operation than is possible with analog-based implementation. The high level of integration of the Piccolo MCU architecture also optimizes overall performance while lowering system cost by integrating complete system functionality on a single chip. OEMs will experience fast return on investment through system-cost optimization, long-term software and tool compatibility, and the ability to leverage the expansive portfolio across all their power control applications.


[1] Chun-An Cheng, et al. "Efficiency Study for a 150W LLC Resonant Converter" IEEE Power Electronics and Drive Systems, 2009, Taipei.

[2] B. Yang: Topology Investigation for Front End DC/DC Power Conversion for Distributed Power System, PhD dissertation, Virginia Polytechnic Institute and State University, 2003

[3] H. Choi: Analysis and Design of LLC Resonant Converter with Integrated Transformer, Applied Power Electronics Conference, APEC 2007 - Twenty Second Annual IEEE Feb. 2007

[4] Ya Lui, “High Efficiency Optimization of LLC Resonant Converter for Wide Load Range”, Master Thesis, Virginia Polytechnic Institute and State University 2007

Bilal Akin is an Application Engineer with TI's C2000 Embedded Control Group, Texas Instruments Incorporated, Dallas. He received B.S. and M.S. degrees in electrical engineering from Middle East Technical University, Ankara, Turkey, in 2000 and 2003, respectively, and the Ph.D. degree in electrical engineering from Texas A&M University in 2007. From 2005 to 2007, he was an R&D Engineer with Toshiba Industrial Division, Houston, TX. From 2007 to 2008, he was a Postdoctoral Research Associate with Texas A&M University. Bilal's research interests are advanced control methods in motor drives, real-time fault diagnosis of industrial systems, digital power management, and various DSP based industrial applications.

Daniel Chang is an Application Engineer with TI's C2000 Embedded Control Group, Texas Instruments Incorporated, Dallas. He received B.S. and M.S. degrees in electrical engineering from the University of Illinois at Urbana-Champaign, in 2004 and 2008, respectively. Daniel's interests include advanced digital control, real-time control systems, and digital power supply design.