View Single Post
Old May 5th 05, 12:21 PM
Michael Brown
external usenet poster
Posts: n/a

Rick wrote:
Michael Brown wrote:

From a purely
theoretical perspective, a single large rail handles load changes

That claim is downright silly.

Perhaps I need to be more precise. Start with two identical devices, A and
B, each consuming a constant 8 amps. Consider two PSUs: one with two rails
of 12A each and one with a single rail of 24A. Given two identical devices,
A and B, the sensible configuration for the split-rail PSU is one device per
rail, and the only configuration for the single-rail PSU is both on one
rail. Assume that all three rails have the same hold time curve (the usual
theoretical one for buck regulators) with respect to the maximum rated
current with respect to their maximum rated current. If device A instantly
increases it's current draw by 2A then in the average case:
{} On the split-rail setup, device A will see a voltage drop of a certain
amount (since current usage has increased by 25%), device B will experience
no voltage drop.
{} On the single-rail setup, device A will see a smaller voltage drop, and
device B will see an identical voltage drop. This voltage drop will be
slightly more than half the voltage drop experienced by device A in the
split-rail setup.

Sum-of-squares-wise, the single-rail setup will have a smaller deviation
from ideal than the split-rail setup. Additionally, the maximum deviation
from the ideal voltage will be smaller, which is a good thing. The benefit
of a split-rail configuration is that one of the devices (device B in this
case) experiences no (or in most cases, minimal) deviation from the ideal
voltage. IMO the single-rail situation is preferable in most situations, as
it's usually the maximum deviation that gets you stability wise, as opposed
to noise on the power line. OTOH, if you are pushing your PSU right to the
limit, such that out-of-tolerance voltage sags would occur on a single-rail
setup, then a dual-rail setup would possibly be preferred as it would
isolate a problem spike to a single rail (though the size of the deviation
would be much larger than a single rail setup, which could cause additional


Michael Brown : OOS/RSI software and more
Add [email protected] to ---+--- My inbox is always open