Monday, July 21, 2008

Paralleling power supply outputs for redundancy

The highest reliability attainable in a single power supply is not nearly as good as that of a redundant power system, in which the outputs of two (or more) power supplies are connected so that - even if one were to fail - the other(s) would continue to provide uninterrupted power to the load But implementing redundancy is not as straightforward as it may appear to be

To have a redundant power system that will function as intended requires careful consideration not only of the power supplies to be used and the electrical configuration, but also of the physical arrangement.

As every power supply will fail at some point, it's necessary to allow for quickly and easily replacing a supply that's failed or is in need of refurbishment.

For example, if the supplies are within an assembly mounted in an equipment rack, equip the assembly with slides so that it can be extended out of the rack for access - and don't forget to include handles on the front panel to pull it out.

Alternatively, some power supply manufacturers make supplies that can be plugged into the front panel of an enclosure or rack adapter, permitting a supply to be, quite literally, changed in a matter of seconds.

Another approach is to have the system's controls and indicators located on a main control panel, but to mount the power supplies in a more accessible location a few feet away.

And the supplies should be mounted in such a way that they can be easily and quickly removed and replaced - for example, by using thumbscrews.

Similarly, it must be possible to do the actual connecting and disconnecting of the power supplies quickly and easily.

If the supplies have screw terminals or lugs rather than connectors, then use insulated connectors that can be easily pulled apart in the wiring cable to each supply.

An isolation diode must be used in series with the output of each power supply, for two reasons - to avoid the possibility of the combined output being shorted if the output of one supply should short, and to prevent current from one supply flowing back into the other and reverse biasing it (which could cause it to malfunction and possibly damage it).

Obviously, the use of diodes introduces a voltage drop in the output voltage from the supply as seen by the load.

This is especially significant at low voltages; for example, a 5V output might drop to only 4V.

Using Schottky diodes can minimise the drop, but doesn't eliminate the need to allow for it.

Keep in mind that the supply must be capable of providing a voltage equal to the sum of the voltage required across the load, the diode drop and the drops (round trip.) in the wiring.

Particularly at low voltages, the lower drop of a larger gauge wire can be a big help.

A typical power supply can compensate up to a volt or so of drops in the wiring, but may not be capable of compensating both the wiring and the diode drops of a redundant system.

And if you're using remote sensing to regulate the voltage across the load, you might not be able to solve this problem by simply stepping up to a supply with a higher nominal output voltage (for example, going from a 5V supply to a 6V supply), because then the sense lines of that supply would try to maintain 6V across the load rather than 5V.

Therefore, be sure to use a supply that is capable of putting out a voltage high enough to compensate both the diode and wiring drops under worst-case conditions (usually, at low line voltage and with maximum rated load current being drawn), and also has the desired load voltage within its adjustment range.

A supply's maximum output voltage is usually considered to be the high end of its adjustment range; for example, a supply with an output specified as 24+/-1V could be relied on to provide a maximum of 25V, so if the load requires 24V and if the combined drops will be no more than 1V, you're in good shape.

Sometimes an easy solution to this potential problem is to use a wide range power supply; in the above example, a 0-30V supply adjusted to 24V would be capable of compensating "round-trip" drops up to 6V.

If two sources of AC power are available, providing separate AC wiring for each power supply permits using a different source of input power for each supply, resulting in the additional advantage of input power redundancy.

Even using two different branches of the same building power source will offer improved input redundancy.

A battery-backup UPS may also be used in series with one of the inputs, further improving overall reliability by permitting continued normal operation of the load even if both of the AC sources should fail simultaneously.

Although meters and/or indicator lights are helpful for monitoring, they don't command attention and may not be checked regularly.

However, an audible alarm can't be easily ignored.

Include an undervoltage alarm circuit on the output of each supply to detect if its output is lower than normal (or a relay can be used if you simply wish to know if an output is there or not), and use it to control an audible alarm, either built into the assembly containing the power supplies or remotely located where it will be heard.

The contact wiring for two or more relays can be cascaded so that only one audible alarm is required.

Checking the meters or indicator lights will then disclose which of the power supplies is low.

Power supply outputs don't always go low when they fail; with linearly regulated supplies, the series pass transistors can short and the voltage can instead go high - by 50% or more in some cases - and quickly fry the load.

Therefore, it's vitally important that power supplies used in redundant applications be equipped with overvoltage protection to assure that the output voltage can't go much higher than the nominal under any circumstances Don't use output fuses.

Virtually all power supplies today have output current limiting circuits that will drop the output faster than the time required for a fuse to blow, so including a fuse won't accomplish anything.

And with most supplies the current limiting automatically resets after a surge, while a blown fuse is counterproductive to the intent of a redundant power system - always having the output present.

Space the power supplies away from sources of heat.

If convection air flow is restricted, use a fan.

Overheating dries out capacitors, which is probably the single greatest cause of power supply failure.

And speaking of capacitor dryout, schedule testing of both supplies at least annually to be certain that each is capable of functioning properly.

If the capacitors are drying out (reducing the output current capability of the supplies) and the supplies are sharing the load, it's possible that working together they can support the load, but if one should fail the other won't be able to support it by itself.

Or, if one supply is set slightly higher than the other, the first will provide all of the current because the isolation diode of the other won't be forward biased; the voltmeter of the other may show that it's maintaining its output voltage, but that doesn't necessarily mean that it can support the entire load.

Slightly increase the output voltage of each power supply occasionally so that it will assume the entire load and verify that it can support the load by itself.

In summary, configuring a redundant power system isn't as simple as connecting two power supplies in parallel.

It requires careful consideration not only of how the power supply outputs may be affected by the manner in which they are connected, but also of factors that may affect both short and long term performance, and a physical arrangement that permits safe and fast maintenance while the system remains on line.

comments: 0