CRPS and Rack Power Supply Testing: The Six Tests That Matter

A server power supply is tested less for what it does when everything is normal and more for what it does when something is not. The grid is unbalanced, the feeder is long, the load steps from nothing to everything, a sibling module is pulled out of the chassis, and all of it has to keep running for years.

This note sets out the six test areas GW Instek’s application material covers for CRPS modules and high-density rack-mount power supplies, and what the bench has to be able to do for each.

The six areas

CRPS validation divides into input behaviour, dynamic response, redundancy switching, efficiency and harmonics, immunity to line disturbance, and long-duration aging. They are listed separately because they fail separately, and a bench built for only one of them will pass a module that fails in service.

Input behaviour: imbalance and the long feeder

Be precise about which DUT takes which feed, because the two in this article differ. A rack power shelf takes a three-phase supply. A CRPS module does not: the form factor is single-phase, with a C14 or C20 inlet and a nominal 100 to 127 V or 200 to 240 V input, and the shelf is what distributes to it.

The ASR-6000 Series covers both from one instrument. In 3P4W output it feeds a shelf directly and programs voltage imbalance between phases, which is how you establish whether the shelf and the modules behind it regulate and protect under an input the field will eventually present. For module-level work the same three-phase output feeds individual CRPS modules one phase each, so an imbalance programmed across the phases exercises several modules at different supply conditions at once.

The more interesting capability is programmable output impedance. A data-hall feeder is long, and the supply voltage at the rack sags as the power supply draws current. A laboratory source with very low output impedance does not do that, so a module tested on it is tested on a stiffer supply than it will ever see. Programming the source impedance reproduces the line loss and the voltage drop, which changes the module’s input current waveform, its hold-up behaviour and its efficiency figure.

This is the capability most worth asking about when specifying the bench, because programming the impedance lets you sweep feeder conditions from the front panel instead of rebuilding a passive network between source and DUT for every case.

Dynamic response and the repeating disturbance

Dynamic testing uses an electronic load on the DUT output, programmed for a 0 to 100 percent load step. GW Instek’s material adds a 20 ms periodic pulse disturbance superimposed at 50 percent load, which is the more revealing of the two: a single step shows the control loop recovering once, while a repeating disturbance shows whether it recovers the same way every time or accumulates error.

Take the step magnitude, slew and repetition rate from the DUT specification. The figures above are the ones in the source material, and they are a starting shape rather than a requirement.

Surge immunity on the input line

The ASR-6000’s arbitrary waveform function imports a surge waveform and applies it on the input to check anti-interference behaviour, which shows whether the module rides through, shuts down cleanly, or fails.

Be precise about what this does and does not amount to. Applying a surge waveform from a programmable source is a design-stage immunity check. Formal surge-immunity testing to IEC 61000-4-5 is done with a dedicated combination-wave generator into a coupling and decoupling network, and a programmable AC source is not that instrument. Other parts of the same series are a different matter: voltage dips and short interruptions to IEC 61000-4-11 are performed with exactly this kind of programmable source. Use the arbitrary waveform to find problems early and to screen design changes, and an accredited setup when you need a surge compliance result.

Redundancy, efficiency and harmonics

Redundancy is tested by running the modules in their redundant configuration, removing one, and watching the output rail while the electronic load holds current. The number that matters is the depth and duration of the output disturbance during the handover, because downstream converters have their own undervoltage thresholds.

Efficiency and harmonic performance come off the same bench. The ASR-6000 carries harmonic analysis up to the 100th order, so the input current distortion measurement does not need a separate instrument.

Aging, and scaling the bench to the DUT

Burn-in is where bench capacity usually runs out. A single ASR-6000 stand-alone unit is rated at 4.5 to 6.6 kVA depending on model, and units can be combined in a rack-parallel system up to 39.6 kVA. GW Instek’s material uses that expansion for high-power burn-in of dual-path server supplies operating in parallel.

For the load side of a long burn-in, dissipating the energy as heat is the constraint rather than the electronics. Where the test plan allows, a regenerative load returns a large part of the absorbed energy instead of heating the test floor, which is what the RBS Series is for.

Automation is the part that makes it a test, not a demonstration

All of it runs over LAN from a host PC: sequencing, parameter configuration, measurement control and data logging. This is not a convenience. A six-area validation run by hand produces results that cannot be compared between units, between revisions or between sites, and a redundancy or aging result that cannot be compared is not evidence of anything.

Script the sequence, log the instrument configuration alongside the measurements, and keep the firmware and calibration state in the record.

Instruments and support from GSAS

GSAS Micro Systems is an authorized GW Instek engineering partner for Karnataka, Andhra Pradesh and Telangana, supporting customers across India. We supply the ASR-6000 Series, PEL Series electronic loads, the RBS Series and the DAQ-9600 with application engineering on bench architecture and automation, and calibration support, from Bengaluru, Chennai, Pune, Hyderabad, Mumbai and Delhi NCR.

For the wider rack and busbar picture, see AI data center testing: power shelves, busbars, burn-in and optics. To scope a CRPS bench against your own module specification, talk to us.

Source: GW Instek application material for the ASR-6000 Series. Load-step and pulse parameters quoted are the ones in that material; your own test parameters, DUT ratings and parallel configuration come from the applicable product specification.