Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Power/Alternative Energy
Power/Alternative Energy??

Paralleling power supplies: Benefits and drawbacks

Posted: 23 Mar 2016 ?? ?Print Version ?Bookmark and Share

Keywords:paralleling? power supplies? resistors? amplifier?

For a parallel array of supplies to deliver increased levels of usable current to a load, some type of control-loop strategy that factors in array use is needed. A popular control strategy is to run the supplies with no internal voltage regulation amplifier, but instead group them together with a common control-signal input which is controlled by a single error amplifier. This error amplifier regulates the output of the system, and then its single feedback signal is distributed to all of the power supplies in the system.

A major benefit of this popular control strategy is the regulation of the output voltage is excellent, and sharing errors are dominated by part-to-part variations in modulator gain. On the downside, use of a single error amplifier and single wired control bus represents a single point of failure, which may present a problem for some types of high-reliability systems. Also, parametric errors on modulator gain can be difficult to control, often leading the manufacturer to trade off yield to control sharing errors.

For a single control-loop approach, sharing errors are minimised if the supplies feature tight tolerance on their control-node inputs. If the sharing errors are large, then either the power rating of the array must be reduced to avoid overloading of any single supply in the array due to sharing imbalances, or specific countermeasures need to be used. Techniques for sharing errors which result from part-to-part variations of the control node can include a production-based adjustment to calibrate out errors (an expensive approach), or adding a current-control loop around local to each supply inside the array to cancel such errors (which adds some complexity and parts). Sensing current for these local loops typically involves adding a shunt resistor to the supply.

There's a second obstacle for isolated power supplies that have their control nodes referred to the primary side of the DC-DC: transmitting the output of the error-amplifier across the primary-to-secondary isolation boundary. Isolation techniques often add cost, take valuable board real-estate, and can have adverse effects on reliability, depending on the isolation components used.

A second control-loop strategy which permits separate supplies to be arranged in a parallel array uses a load line to emulate the path resistance of the ballast resistor method. By implementing what is called the "droop-share" method of load sharing, each supply has a separate reference and integrating error amplifier, but the reference is deliberately and linearly reduced by some nominal amount as the load current of the supply increases.

Paralleling supplies may have negative consequences on transient response and load regulation. The droop-share method deliberately uses a negative load-regulation term to distribute the load across modules in the array. Therefore, load regulation tends to be worse for droop-share arrays than for arrays created with a single traditional error amplifier. If this is a problem, an external control loop can be used around the droop-share array, to effectively cancel out the negative-regulation term. The resulting static-regulation error is identical to the traditional error-amplifier case, since the external loop is itself an error integrator.

Supply design can simplify, enhance parallel configurations
Supply vendors can take steps to ease the paralleling challenge. For example, Vicor's DCM DC-DC converters in Converter housed in Package (ChiP) packaging feature a built-in negative-slope load-line; thus, as the load increases, the DCM's internal regulator reduces the output voltage slightly. This effectively acts like the small ballast resistor approach but without any actual resistors (figure 3) and with a few additional key differences.

Figure 3: Vicor's DCMs in ChiP packaging are designed to be paralleled by simply connecting their outputs together; no diodes, ballast resistors, or other load-balancing components are needed.(reference 1)

First, it's a different way to implement a ballast resistor, and one which doesn't involve wasted heat as there is no physical resistor and no VI heat generated. A second difference relates to dynamic response, since for frequencies up to hundreds of kilohertz, a real resistor can be considered as having unlimited "bandwidth" in its I-V transfer-function curve due to lack of high-frequency parasitic concerns. As a result, any instantaneous change in voltage across the resistor results in an immediate corresponding change in current.

In the DCM converters, the load line is implemented through a discrete-time modulation of the digital/analogue converter that creates the reference for the error amplifier. The correct reference value is calculated primarily based on an estimate of the DCM's output current, and involves some averaging to reduce noise. Therefore, the resistor that the DCM load-line emulates is one that acts like it has a significant capacitor in parallel with it, and the resulting RC-time constant is evident when looking at the datasheet figures which show the supply's response to a load step.

?First Page?Previous Page 1???2???3???4?Next Page?Last Page

Article Comments - Paralleling power supplies: Benefits...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top