Keeping power factor high at light loads
Techniques for increasing power factor at light loads - from desktop PCs to servers - are explained by Edward Ong, product marketing manager, Power Integrations
From desktop PCs to servers, the computer industry is looking for ways to meet increasingly-demanding energy efficiency and power quality regulations. While power supply designers have achieved high power factor at high loads, when it comes to low loads, the technology has proven to be challenging.
The main driving force to increase power factor in computer systems comes from the 80 Plus organisation, a voluntary certification program launched in 2004 that certifies computer power supplies. There are different tiers of certification from bronze to titanium, and in the top class of these, the power factor has to be above 0.95 at 20% load. As the technology has developed to serve the tiered standards, power supply makers have been searching for better and more cost-effective power factor correction (PFC). Standards-makers for energy use in applications beyond computer systems have also adopted the higher standards, broadening the requirement and application of the technology.
Hitting such high PF targets is made even more challenging as designs need to allow for some margin of error due to variations in production and EMI filters. This means that designers often aim for PFs of 0.97 or 0.98 to guarantee 0.95 at 20% load in operation. On top of that, some large server companies specify a PF of 0.9 at just 10% load – a significantly more difficult task. This is because multiple-redundant server power supplies may operate for long periods at a very light load. Whether the PF is measured at 10 or 20% load, the solution has proven to be increasingly costly.
Limiting factors
The big technical difficulty in achieving a high power factor at light loads is the presence of X capacitors in the EMI filter. X capacitors are used to reduce the differential mode EMI that is being fed back into the AC line. The problem is that the voltage lags behind the current, which has an adverse effect on the power factor. At high loads, the reactive power is relatively very small compared to the real power so it does not have a serious impact on power factor. However, at light loads the effect is more noticeable and the phase shift causes a severe power factor displacement. At 20% load the effect is dominant - even more so at 10%.
Techniques
To compensate for the X capacitor issue, manufacturers have adopted a number of measures, such as reducing the value of the capacitor or substituting them for more expensive differential chokes. This compromises the power supply’s ability to pass international EMI regulations and it only really works for low-power applications because the EMI is lower to start with; this method is not suitable for use with large servers.
At light loads when the EMI is also lower and there is subsequently less requirement for the X capacitors, some engineers have chosen the approach of designing a circuit that will disconnect the capacitors when the load falls below a certain threshold.
This is quite a crude method and adds more components and testing, and hence cost. Even worse, there are sometimes multiple X capacitors that each need their own circuits that not only switch out the capacitors but also detect the load.
A more elegant method is to use signal processing to compensate for the phase difference, and this is preferred by many power supply companies that address the server industry. Digital signal processors (DSPs) are already used for PFC, so they can also handle a signal processing algorithm to compensate for the phase shift. DSP is expensive, however, which makes this approach suitable only for high-end markets.
Another way
A better way is to have a digital function on the PFC IC that compensates for the input voltage phase lag by phase shifting the input current. This is both low-cost and effective, and is the method used in the HiperPFS-3 family of PFC ICs from Power Integrations. Put simply, these ICs delay the current to compensate for the voltage lag. This function is transparent to the user and involves no additional development work. Since it is already integrated, it is low-cost. This allows the use of large and relatively inexpensive C capacitors for EMI suppression and eliminates the need for more expensive EMI mitigation strategies such as extra differential mode chokes.
These ICs are best suited to applications with continuous power demands - up to 405W for universal input and 900W for high line. Efficiency levels are better than 95% at 10% load and devices consume less than 60mW under no-load conditions. A power factor above 0.92 is easily achievable at 20% load. In tests at various voltages under light-load conditions, the 275W PFS3 produced a higher power factor than equivalent products. It also provided the lowest overall total harmonic distortion, the highest efficiency and the input current waveform was more sinusoid than the competitors.
Figure 1: The HiperPFS-3 IC eliminates the PFC converter’s need for external current sense resistors
The HiperPFS-3 IC eliminates the PFC converter’s need for external current sense resistors and the associated power loss, and uses an innovative control technique that adjusts the switching frequency over output load, input line voltage and even input line cycle. This control technique increases efficiency over the entire load range of the converter, particularly at light loads. Additionally, it significantly reduces the EMI filtering requirements due to its wide bandwidth spread spectrum effect. Advanced digital techniques are used for line monitoring functions, line feed-forward scaling and power factor enhancement, while analogue techniques are used for the core controller to maintain extremely low no-load power consumption.
The HiperPFS-3 also has an integrated non-linear error amplifier for enhanced load transient response, a user programmable power-good signal and user selectable power limit functionality. Standard protection features include integrated under-voltage, over-voltage, brown-in and out, and hysteretic thermal shutdown. Devices also provide cycle-by-cycle current limit and safe-operating-area protection of the power MOSFET, power limiting of the output for overload protection and pin-to-pin short-circuit protection.
At a time when all industries are seeking to achieve better energy efficiency, the computer industry has received a boost with the HiperPFS-3 family of PFC ICs from Power Integrations. These devices will allow power supply manufacturers to achieve a 0.95 power factor at 20% output load and meet the latest 80 Plus Platinum and Titanium standards.
This is also achieved with fewer parts than previous members of the HiperPFS range – 22 compared with 25 for the HiperPFS-2, and 30 for the original HiperPFS. The net result is that designers can increase the size of the X capacitors while reducing or eliminating the use of differential-mode chokes, thereby reducing EMI without degrading light load power factor performance. The result is a smaller, lower cost EMI filter stage, which crucially meets the high efficiency and PF bar set by the regulatory authorities.