While power supplies are quite reliable, there are many applications where zero downtime is demanded. This requirement necessitates fault tolerant power systems.
Uninterruptible power sources which address the power integrity question of the utility mains are not a whole solution. Consideration has to be given to the reliability of the power converters themselves. It is possible to employ ultra-conservative design philosophies to minimize the likelihood of failure, but it is not possible to eliminate failure entirely. A more effective approach is to design systems that are fault tolerant in that a single failure does not disable the function. In power supplies, multi-unit assemblies are frequently specified to ensure that internal failure of one power supply does not cause loss of output.
Power supplies may be paralleled with "OR-ing" diodes used to isolate one from the other and to prevent current from being fed back into a failed unit. Thus if one fails, the other may continue to supply the common load.
PARALLELING
It is the nature of voltage sources that only ONE unit can establish the voltage level in a system here several are paralleled. If low source impedance voltage stabilizers are paralleled ith no special effort to form a master- slave relationship, the power supplies themselves will sort out who is to be "boss." Only ONE unit will end up controlling the voltage while the others "overload" into their current mode and merely add their current to the total. The combination of such power supplies always ends up with just one voltage stabilizer. All of the other units operate in current mode.
FIGURE 12 - Power supplies paralleled without forced sharing. The voltage is set by the voltage of unit 3 (E3). Units 1 and 2 merely add current I1 and I2 to the total and operate in their current limit mode
The principal disadvantage of this method is that the power supplies will be loaded unequally. The one with the highest voltage setting will initially deliver ALL of the current. Only after it goes into current mode, will the next unit assume part of the load and so on. The LAST unit, the one with the smallest voltage setting, ends up controlling the voltage and providing the least current. A further disadvantage is that the voltage seen by the load will change by an amount equal to the difference between the voltage setting of the individual models as each one goes into its current limit, passing voltage control to the next lowest unit.
A complementary situation will prevail if current sources are connected in series. Only ONE current stabilizer can control the current in a single loop. All of the others will overload to their voltage mode, merely supplying voltage to the load, but not controlling the current.
PARALLELED POWER SUPPLIES WITHOUT FORCED CURRENT SHARING
If the voltage stabilizers being used in parallel are poor regulators, with significant non-zero source impedance, then it may be possible to rely on the voltage "droop" afforded by the internal resistance of each unit as a means of causing the current to share. Adjustment is tricky however. All units must be set to the same voltage and readjusted carefully to share the current. This reliance on the internal resistance of poor stabilizers to equalize the current of paralleled power supplies was a "feature" of early power supply designs.
In modern fault tolerant power supply systems, a means is provided for forced current sharing. This ensures that all units carry an equal part of the burden.
The current sharing circuit is a form of master-slave operation in which the current supplied by the master is measured and the other units are controlled to match the current, ensuring that they are equally loaded. Only the master actually controls the voltage to the load.
N PLUS ONE
The basic idea of fault tolerance through redundancy is to size the system so that there is at least one more unit than the minimum required to carry the load. Thus, if a load is 10 amperes, a fault tolerant redundant system might have three (3) 5 ampere units in parallel;one more than is needed, hence:N+1 redundancy. The failure of any one power module leaves sufficient power available to support the whole load.
Currently Kepco supports N+1 hot-swap parallel redundancy operation with the 200W MST models, the 50, 100, 150, 300, 600, and 1200/1500 HSF 3U models, the 50 and 100 HSF 1U models and the 1000 and 1500W HSP series.
In a "hot-swap" enabled power supply, the idea is to replace bad modules without shutting down the system. This is accomplished by a combination of pre-set adjustments that allow modules to be set off-line and a mechanical arrangement that allows units to be inserted and removed without upsetting the output. When the currents are large, as they are in our 1000W and 1500W HSP hot-swappable modules, the connector to allow the removal and substitution of a new module while the power is live is quite specialized. It is found in the rack adapter, RA 60. A simpler rack adapter without the special hot-swap d-c connectors is available at lower cost. It is called RA 58. The hot-swap rack adapter for the HSF 3U 50W, 100W and 150W models (-PFC) is called RA 19-(X)B, for HSF 1U 50W, 100W and 150W models (-1UR) the rack adapter is RA 19-1U, for the HSF 3U 300W, 600W and 1500W models the rack adapter is RA 19-4C, and for the MST series, the rack adapter is RA 55.
FORCED CURRENT SHARING
Several other Kepco power supplies offer the kind of built-in forced current sharing that is needed for N+1 redundancy without the plug-in hot-swap capabilities. The RCW 350 and 1500W sizes, the RKW and the HSM are examples.
FIGURE 13 - Current sharing by means of a current share bus (marked CB)
