Optimize the control loop for power-converter designs

Article By : Magnus Karlsson

Software allows designers to perform investigations themselves rather than relying on power module suppliers.

The days when converter manufacturers would have little option but to rely on the expertise of power-module specialists to design filters, optimize the control loop, and provide the result, have given way to a new era in which system designers can use free software to achieve fast and easy outcomes. Loop compensation tools embedded in power system design software have evolved to optimize voltage at the concept stage, a convenience that allows engineers to try different configurations and re-run the simulation until the best result is accomplished. So, what are the engineering principles underpinning this advance?

In the first instance, it is worth pointing out that the addition of inductor-capacitor filters to the inputs and outputs of power converters serves a number of purposes. These filters not only reduce reflected ripple current and output noise, but are designed to meet EMC radiation and susceptibility regulations.

Checking the design and simulating the result of such metrics can now be achieved in a simple format using loop compensation software routines, which means engineers are better informed about what power modules would suit the system by evaluating any issues and challenges beforehand. A well-specified loop compensator can design, simulate, analyze and configure a power converter directly in the digital Z-domain within minutes.

Assessing control loop stability

To achieve voltage/filter optimization, a loop compensator tool can be used in a number of ways. Priority number one is to look for any voltage deviations that are outside of requirements, not just in terms of peak values, but the whole area around nominal voltage. As a point of note, loop optimization routines function best for well-designed external decoupling filters, namely those with a short distance between the power module and the load (the closer the better, but certainly within 10 cm). Low-ESR (equivalent series resistance) capacitors of 15 mΩ or less are also beneficial in terms of achieving optimized control loops.

An alternative option is to use smaller capacitors (of the same type and time constant) in parallel, which will serve to lower ESR. In control theory terms, nothing is better than the accuracy of the model. As a consequence, using the correct ESR values is necessary, as well as using a proper estimate of parasitic inductance, caused by the traces between the power module and the load, which in turn will provide a better projection of voltage deviation during load transients and the robustness of the design.

The impact of digital-control technology

The introduction of digital-control technology in power converters has caused many users to struggle with designing appropriate control loop compensation. A common approach is to use traditional analog tools to determine a solution, then transform that solution into the digital Z-domain. However, this strategy can be time-consuming and often generates a non-optimal solution.

This is one reason fueling demand for effective, reliable voltage optimization technology. Further factors hinge around the inherent way that design projects progress. Unfortunately, power is often an afterthought, with many design engineers adding this function to the corners of boards or any areas left over as the project draws to its close. The upshot is that power modules are not close to the load, risking the potential for key measures to demonstrate poor behavior. Furthermore, as a result of budgetary constraints, many designers will select a 45 mΩ capacitor, for example, rather than a 15 mΩ capacitor, which can cost around twice as much.

The simplest way to determine which capacitors to include in the output filter section of a loop compensation tool is to calculate and sort them by total capacitance per type and longest time constant (Figure 1).

diagram of output filter configurationFigure 1 The initial output filter configuration enables detailed modeling of filter capacitors and trace inductance values, including associated ESR values. Source: Flex Power Modules

This process will reveal the two capacitor types for inclusion, namely those with the largest total capacitance and longest time constants. The other capacitors’ time constants are usually so small that they will not interfere with the control loop, or can be handled using margins for additional component/model uncertainty.

Achieving effective simulation

In all instances, the need for voltage optimization through effective simulation is paramount. Take the case of a filter design with an ESR of 45 mΩ close to the load, and a significant inductance (between module and load) of about 20 nH. Here, loop compensation can be applied to indicate the areas of concern. For instance, Figure 2 shows that a huge deviation would be evident at load release, after just 1 ms in this instance.

graphs and tool screenshot show deviation at load releaseFigure 2 These graphs show significant deviation at load release with 20 nH inductance and large ESR close to load due to an aggressive control loop. Source: Flex Power Modules

Clearly, having 20 nH inductance and large ESR close to the load would create an aggressive control loop, while another issue would be the frequency bandwidth range, as it would extend from the nominal of 14 kHz to the maximum of 128 kHz. The goal of this particular simulation was set to 20 kHz, so it was apparent that component variations were making the control loop highly-sensitive. Looking at the output impedance curve (Figure 3) also highlighted the problem – the high impedance at high frequencies was causing large voltage deviations and, in turn, large currents.

output impedance curveFigure 3 The output impedance curve shows corresponding variations of the cross-over frequency (0 dB) in the open loop bode plot. Source: Flex Power Modules

Eliminating current spikes

Using a loop compensation tool, one solution would be to decrease the gain until the output impedance appeared constant, or until the current spikes had gone (Figure 4). In this case example, with control loop optimization, the gain was reduced from 26 to 12.5 dB. As a result, the voltage deviations appeared as more or less symmetrical – the current spikes disappeared in both the load and load-release graphs, while the frequency bandwidth range narrowed considerably (10 to 14 kHz). The variance in output impedance was also very small. In short, applying optimization process meant that the robustness and performance of the loop design could be increased significantly.

graphs and loop compensation tool screenshot of decreasing gainFigure 4 Reduce the loop gain until the output impedance looks constant and the current spikes are gone. Source: Flex Power Modules

As an alternative solution, the optimization tool could be re-run with different goals (Figure 5). For instance, using the same case example, the frequency bandwidth goal was reduced from its original value of 20 to 15, 10, and 5 kHz.

optimization tool screenshot with different goalsFigure 5 Decrease the bandwidth goal and run optimization. Source: Flex Power Modules

Subsequent analysis revealed this to be a robust alternative, as the parameter spread proved small (Figure 6). Notably, the positive and negative voltage deviation appeared almost identical in shape, with a load current absent of spikes.

graphs and tool screenshot show load current and voltage deviations after optimizationFigure 6 The load current and the voltage deviations are better behaved after optimization. Source: Flex Power Modules

The re-design option

Of course, a third option is to re-design the filter using the same real-estate area, by moving components around or, preferably, using better components, as in Figure 7. In the case example, the high inductance pointed to trying a different bulk capacitor strategy – 25% at the power module and 75% close to the load.

tool screenshots show redesigned filterFigure 7 Another solution was to re-design the filter by moving bulk capacitors close to the load. Source: Flex Power Modules

Observations from the simulation of the re-design (Figure 8) indicated that the output impedance peaks were almost equal and, while the negative voltage deviations did not alter much, the positive voltage deviation improved significantly. In addition, there was no extreme bandwidth and the results looked a lot more robust.

simulation results and tool screenshot after alternative placementFigure 8 The alternative placement with the same BOM produced these results. Source: Flex Power Modules

Adopting the latest power system design software has many benefits beyond mere converter configuration. For instance, such software enables designers and system architects to track or simulate the efficiency of their entire power system, taking full advantage of the latest digital power technology. Certain suites allow users to define relationships across power rails, including phase spreading, sequencing, and fault spreading, for instance. As a result, it is easier to understand behavior at a system level, which helps reduce time-to-market.

Ultimately, software of this type permits designers to perform investigations themselves rather than having to rely on the expertise of power module suppliers.

Images in this article are taken from the Flex Power Designer Tool.

This article was originally published on EDN.

Magnus Karlsson is a senior engineer for Flex Power Modules.

Related articles:


Leave a comment