The Common Redundant Power Supply (CRPS) specification defines a standardized mechanical, electrical, and thermal form factor for enterprise and data center power supply units. Widely adopted across Open Compute Project (OCP) platforms and hyperscale architecture, CRPS allows hardware designers to source hot-swappable power supplies from multiple independent vendors without modifying internal backplane layouts or mechanical housing frames.
Central to CRPS inter-vendor compatibility is its unified card-edge connector interface, which consolidates primary DC power output, system grounding, standby power, and control/monitoring communication channels into a single standardized board edge connection.
The CRPS Physical Card-Edge Interface
The CRPS connector uses a high-reliability card-edge architecture. The power supply unit contains a gold-plated PCB edge finger array that slides directly into a mating backplane receptacle mounted inside the server power distribution assembly.
CRPS POWER SUPPLY UNIT
| | | (PCB Edge Fingers)
========================================================= [Backplane Slot]
[Ground Pins] | [+12V Power Pins] | [Staged Control Signals]
=========================================================The pinout distribution is functionally divided into four key operational zones:
- Main +12V Power Output Zone: High-current copper pads designed to deliver the bulk DC voltage to the baseboard. Multiple redundant contacts run in parallel to spread the total load current and keep temperature rise within specified limits.
- +12VSB (Standby) Rail Zone: Dedicated auxiliary power pins providing low-power current to the BMC, security chips, and management logic while the main system power rails are toggled off.
- PMBus / I2C Communication Lines: Standardized serial lines (SDA, SCL, SMBALERT#) enabling real-time monitoring of input/output voltage, current draw, fan speeds, and internal power supply thermal telemetry.
- Control and Status Signals: Discrete hardware lines including PS_ON# (power switch trigger), PS_KILL (fast hardware safety interlock), and PWOK (power good status flag).
Hot-Swap Engineering and Staged Contact Pinning
In live data center environments, power supply replacement must occur without interrupting active compute workloads. CRPS connectors achieve reliable hot-swapping through a staggered contact length strategy, a first-mate, last-break pin architecture:
Insertion Direction ===>
Long Pins (First Mate) : [Ground / Frame Shield]
Medium Pins (Second Mate) : [Main +12V Output & +12VSB]
Short Pins (Last Mate) : [PS_KILL / Control Logic]- First Mate (Ground): Long pins ensure system chassis ground is connected first, safely discharging any accumulated static charge before power tracks touch.
- Second Mate (Power & Standby): Intermediate-length pins engage main power and standby voltage lines.
- Last Mate / First Break (PS_KILL): The shortest pin on the connector is PS_KILL. During insertion, this pin mates last, signaling internal PWM controllers to enable the primary output stages only after all high-current contact fingers are fully seated. During removal, PS_KILL disconnects first, disabling power conversion instantly to prevent destructive electrical arcing across main output contacts.
Vendor Qualification and Integration Considerations
When qualifying CRPS-compliant power supplies across different vendors, hardware engineers must evaluate specific connector integration variables:
- Gold Plating Thickness: Ensure edge-card contact fingers feature minimum spec-compliant hard gold plating, commonly 30 microinches or higher over a nickel underlayer, to withstand repeated mating cycles without exposing underlying copper to oxidation.
- Thermal De-rating: While CRPS defines standardized pinout currents, internal PCB trace heat sinking varies between manufacturers. Evaluate power supply thermals under full load to ensure localized backplane temperature limits are respected.
Standardizing power interfaces through the CRPS specification simplifies chassis supply chains, reduces tooling overhead, and secures high-availability power redundant topologies for enterprise servers.
