Programmable and Bench Power Supplies for Test and Measurement

In most applications a power supply is infrastructure. In test work it is part of the instrument, and its behaviour is inseparable from the result. A supply whose output noise exceeds the signal being measured has set the noise floor of the measurement, and no amount of care downstream recovers it.

This changes which specifications matter. Cost per watt, the figure that drives most industrial selection, is close to irrelevant. What matters is whether the supply can be commanded, whether it does the same thing twice, and whether it survives a unit under test that fails short — because sooner or later one will.

Programmability is the requirement everything else follows

An automated rig needs to set voltage and current from a script, read back what the supply is actually delivering, and sequence outputs in a defined order with defined timing. A supply with front-panel controls and no interface cannot participate in that, whatever its electrical performance.

Three capabilities are worth separating when reading a datasheet, because vendors describe them loosely:

• Remote programming — setting output voltage and current limit from a host, whether by analogue control signal or a digital interface.
• Readback — the supply measuring and reporting its own actual output, which is what lets a test record what the device under test was really given rather than what it was asked for.
• Sequencing — bringing multiple rails up and down in a specified order with controlled timing, which many devices require and some are damaged by not receiving.

The programmable power category covers this ground across a wide power range: the RSP family from around 750 W to 3 kW, the RST and SHP families at several kilowatts and above, SPV at lower powers, and specialised units including the water-cooled PHP family for installations where air cooling is impractical.

Where a rig is built around an existing framework, the interface it speaks is a hard constraint rather than a preference. Confirm it against the specific model before designing the fixture around it.

Ripple and noise set the measurement floor

Every switching supply puts some residual on its output. Two components are usually quoted together and behave quite differently: low-frequency ripple related to the line and the switching cycle, and higher-frequency noise from the switching edges themselves.

Which one matters depends on the measurement. A high-frequency noise spike that is irrelevant when characterising a motor will dominate a low-level analogue measurement, and ripple at the switching frequency will appear directly in a sensitive measurement made on the same rail.

How the figure is specified deserves attention as much as the figure itself. A peak-to-peak value measured over a stated bandwidth and an RMS value are not comparable numbers, and comparing two datasheets that use different conventions is meaningless. Where the measurement is genuinely sensitive, the useful comparison is made on the bench with the rig's own wiring in place — the loop area of the connection between supply and device often contributes more than the supply does.

Regulation, and where it is measured

Line regulation describes how far the output moves when the input supply varies. Load regulation describes how far it moves when the load current changes. Both are quoted at the supply's own output terminals.

The device under test is not at the supply's terminals. It is at the end of the fixture wiring, and the voltage there is lower by the drop across that wiring — a drop that varies with current, which means the regulation the device actually experiences is worse than the datasheet figure, by an amount the datasheet cannot know.

Remote sensing is the answer where the supply offers it. A separate pair of sense wires lets the supply regulate against the voltage at the load rather than at its own terminals, and because the sense pair carries almost no current it has almost no drop of its own. On a high-current fixture with any appreciable cable length, this is the difference between a specification met and a specification merely quoted.

Transient response is the specification that testing exposes and steady-state work does not. When a load switches abruptly, the output deviates and then recovers, and the size and duration of that excursion matter when the device under test is doing exactly that — as anything with a duty-cycled radio or a switching load will.

Multi-rail fixtures and configurable supplies

A test fixture for a real product usually needs several voltages at once, often with different current requirements and sometimes with a required power-up order. Assembling that from individual supplies works but consumes rack space, and every additional unit is another mains connection and another thing to control.

The configurable category takes a different approach: a chassis fitted with output modules chosen to suit the application, so one unit provides the rail set a fixture needs. The MP, NMP and UMP families cover this. Where a laboratory tests several product variants, reconfiguring a chassis is considerably faster than rebuilding a wall of separate supplies.

Isolation between outputs is the specification to check when rails must be referenced to different points, or when one rail needs to sit on top of another — a requirement that appears quickly in any fixture testing a device with its own internal isolation barrier.

Protection, on the assumption that the DUT will fail

In production test the unit under test is expected to fail sometimes; that is the point of testing it. The supply's protection behaviour therefore determines whether a failure costs a few seconds or a morning.

Overcurrent behaviour differs in ways that matter here. A supply that limits current and keeps running holds the fault current at the set value, which allows a measurement of what the fault is doing. A supply that shuts down and latches off requires an operator to reset it, which stops an unattended rig until someone notices. Neither is wrong, but a rig running overnight needs the behaviour chosen deliberately.

Overvoltage protection is worth setting tightly in test work. The supply is connected to something whose purpose is to be proven, and a control fault that puts 24 V onto a 5 V rail destroys the device and, sometimes, the fixture with it. A protection threshold set just above the intended rail is cheap insurance.

For fixed rails in the same rig — fans, lighting, the fixture's own logic — the enclosed category covers the requirement without paying for programmability that nothing will use.