Power supplies operated in parallel are frequently organised so that there are sufficient individual units to handle the full load, plus one extra to cater for the failure of any single unit – this is N+1 operation.
There are two established techniques used for improving the load sharing accuracy between power supplies connected in parallel, explains Andy Skinner, Chief Technology Officer at TDK-Lambda EMEA. One method is to use a share bus but this is not suitable for high reliability applications because of the failure of the share bus can cause the whole system to fail. Droop sharing is the second technique, where the load regulation of the power supply – ie the variation of output voltage with current – is deliberately made higher; this forces the output voltage to reduce as the power supply is loaded. Unfortunately in many implementations of droop share, the load regulation (the total excursion of the output voltage from nominal) needed to provide good current sharing is unacceptably high.
The new Load Sharing Apparatus developed by TDK-Lambda overcomes this issue by applying droop over the full load range of each power supply but with a two-stage profile, which is less steep at lighter loads than at heavier loads. “As a result, high sharing accuracy can be achieved at high loads, whilst still allowing for some load sharing at lighter loads,” concludes Skinner. “Thus, by improving the sharing accuracy at high loads, the number of power supplies required to achieve N+1 redundancy is reduced, which in turn reduces both the overall cost and size of the power supply system. In addition, by allowing for load sharing at lighter loads, a reduced spread in MTTF of the power supplies is achieved.”
[picture shows a recently-introduced TDK power supply, story here]
TDK Lambda, www.uk.tdk-lambda.com