[rescue] Power supply load balancing

George Adkins george at webbastard.org
Thu Feb 28 19:26:55 CST 2002


On Thursday 28 February 2002 05:05 pm, you wrote:
> That's what I've been designing & working on.  What I need right now is a
> sure fire way to load share across three ATX power supplies, with the
> ability to loose one.  I've got a few ideas, but most of them are UGLY.
> 	Nick
>

The best way to do this is to build a network of comparators and wire them to 
a feedback type control and adjust the incoming voltages from each supply, 
this will insure an even current draw and proper load balancing.  

If you just hook them up in parallel, variations in output voltages between 
different supplies mean the highest one will get maxed out on current (until 
the extreme load causes it's voltage to fall below the next highest one and 
then it begins to pull a portion of the load, etc...)  With switching 
supplies, this results in one of the supplies running at max load all the 
time (unless the load is less than one supply, in which case it works for 
hot-failover, but you need steering diodes and a dummy load before they come 
together, to keep the un-loaded supply "live").
Running at 'Maximum Load' is unhealthy for switchers.  They burn out.

So, working back from the common bus, you have steering diodes to prevent 
voltage from other supplies from affecting the comparator, then the voltage 
comparator (output voltage of the supply is compared to the main bus 
voltage), then the output of this drives a voltage regulator which adjusts 
the output voltage of the individual supply to match the main bus voltage, 
then you have the power supply plugged in and supplying power.

Multiply this times the number of supplies you need to handle the load, and 
add one for redundancy.  Voilla! you have N+1 reduntant load-balancing power 
supplies.



More information about the rescue mailing list