Powering Single-corded Equipment in a Dual Path Environment

Overview

The three options are a single feed, a point-of-use transfer switch in the rack, or a large centralised static transfer switch (STS) that creates a new bus. Single-corded equipment still makes up about 10-20% of IT equipment in mission-critical facilities; per WP 48, a 100% dual-corded data centre with independent paths can have 10,000 times less downtime than a single-path design. The paper explains how static (SCR-based) and electromechanical (relay-based) switches work and how they fail.

The failure modes are the paper's most distinctive part: SCRs fail short 98% of the time, and a shorted SCR differs from a good one by typically less than 0.5 volts, which makes detection hard. Timing comes from the ITIC (CBEMA) curve: IT equipment keeps running for 20 milliseconds at zero volts, so any switch must transfer in 20 ms or less; a heavily loaded power supply tested held its output for 18 ms. An STS transfers in about 4 ms, an electromechanical ATS in 8 to 16 ms.

The reliability argument is built on numbers. Relays are typically rated for 100,000 operations, while data-centre transfer switches see on average four transfers a year. STS units range from 5 kVA to 35 MVA, most in the 100-300 kVA range with a footprint of about two IT racks. Indicative TCO is $200-300/kW for a large STS, $550-700/kW for a rack STS and $100-150/kW for a rack ATS; MTBF estimates are 400,000-1,000,000 h for STS against 700,000-1,500,000 h for rack ATS. The conclusion is rack ATS below 10 kVA.

Key points

1. Single-corded gear is about 10-20% of IT equipment in mission-critical sites; a 100% dual-corded site with independent paths can have 10,000 times less downtime than single-path (per WP 48).
2. STS units range from 5 kVA to 35 MVA, most 100-300 kVA with a footprint of about two IT racks; rack-mount STS are 5-10 kVA.
3. STS switching takes about 4 ms and electromechanical ATS 8 to 16 ms; rack ATS rarely switch beyond 10 kVA.
4. SCRs fail short 98% of the time, and a shorted SCR differs from a good one by typically under 0.5 V, so detection is hard.
5. Per the ITIC (CBEMA) curve IT equipment runs 20 ms at zero volts, so a switch must transfer in 20 ms or less; a heavily loaded PSU held output for 18 ms.
6. Relays are typically rated for 100,000 operations while data-centre transfer switches average four transfers a year.
7. Indicative TCO: large STS $200-300/kW, rack STS $550-700, rack ATS $100-150; MTBF 400,000-1,000,000 h for STS vs 700,000-1,500,000 h for rack ATS.

Using it in design

• Record the cord count (single or dual) of each IT device in the rack model and compute the single-corded share.
• In a 2N design, warn when a single-corded device is connected to only one path and suggest a rack ATS for 10 kVA or less rather than a central STS.
• Give transfer switch catalogue entries type (STS, ATS), kVA, transfer time in ms, MTBF and mounting (rack, floor), and require a transfer time of 20 ms or less for IT loads.
• In availability calculations treat a large STS as a common-mode element feeding many racks and show its blast radius.

Who should read it

Electrical designers and availability analysts feeding single-corded devices in 2N power architectures.

Related standards

• ITIC (CBEMA) curve
• IEC 61000-4-11

Publication details

Authors

Victor Avelar

Publisher

Schneider Electric

Year

2011

Document number

WP 62

Document language

English

Source type

Manufacturer

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation