PM5310-BI - PMC-Sierra, Inc.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE

CHESS

TECHNICAL OVERVIEW

PROPRIETA R Y A ND CONFIDENTIAL

PRELIMINARY

ISSUE 2: MAY 2000

PM7390-BI

CB924102A

M9845

UNI-

MACH48

PM5316-BI

CB924102A

M9845

SPECTRA-

155-QUAD

PM5372-BI

CB924102A

M9845

TSE

PM5310-BI

CB924102A

M9845

TBS

PM5315-BI

CB924102A

M9845

SPECTRA-

2488

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE

PUBLIC REVISION HISTORY

Issue

No. Issue

Date Details of Change

1 Feb.

2000 Document created.

2May

2000 Update document (power, thermal information,

synchronization methodology). Added TSI mappings for

CHESS devices. .

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE i

CONTENTS

1 INTRODUCTION...............................................................................................................1

1.1 T ARGET AUDIENCE............................................................................................1

1.2 NUMBERING CONVENTIONS............................................................................1

2 REGISTER DESCRIPTION...............................................................................................2

3 THE CHESS ARCHITECTURE.........................................................................................3

3.1 THE CHESS SET DEVICES................................................................................3

3.1.1 SPECTRA-2488 ......................................................................................3

3.1.2 SPECTRA-622 ........................................................................................4

3.1.3 SPECTRA-4X155....................................................................................4

3.1.4 TBS..........................................................................................................4

3.1.5 TSE..........................................................................................................4

3.1.6 S/UNI-MACH48.......................................................................................4

3.1.7 S/UNI-ATLAS 3200..................................................................................5

3.2 CHESS SET EXAMPLE ARCHITECTURES........................................................6

3.2.1 THE BASIC ARCHITECTURE ................................................................6

3.2.2 ONE PORT - ONE SERVICE..................................................................7

3.2.3 COMPLETE CROSS-CONNECT SWITCH ............................................8

3.2.4 BUILDING A SMALL ADM.......................................................................8

3.2.5 1+1 PROTECTION ON LINE CARDS.....................................................9

3.2.6 BUILDING REDUNDANT SWITCH FABRICS......................................10

3.2.7 IMPLEMENTING DS-1/0 CARDS WITH PROTECTION......................12

4 F ABRIC ARCHITECTURES............................................................................................14

4.1 SINGLE STAGE FABRIC: UP TO 160 GBPS ....................................................14

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE ii

4.2 THREE STAGE FABRIC: UP TO 2560 GBPS ...................................................15

4.3 DEVICE COUNT REQUIRED FOR FULLY POPULATED FABRICS.................16

5 DEVICE INTERFACES....................................................................................................18

5.1 LVDS INTERFACE.............................................................................................18

5.2 HOT SWAPPING AND FLOATING LINKS ON THE LVDS INTERFACE...........18

5.3 CALCULATING THE MAXIMUM DIFFERENCE IN TRACE LENGTHS............19

5.4 INTERFACING A 19.44 MHZ TELECOMBUS TO A 77.76 MHZ TELECOMBUS20

6 SYSTEM SYNCRONIZATION.........................................................................................22

6.1 PURPOSE OF SYNCHRONIZATION ................................................................22

6.2 IMPLEMENTING SYNCHRONIZATION IN CHESS SET DEVICES..................22

6.3 IMPLEMENTING SYNCHRONIZATION IN CHESS SET DEVICES..................25

7 PROGRAMMING CONNECTIONS.................................................................................26

7.1 INTRODUCTION................................................................................................26

7.2 PROBLEM DESCRIPTION ................................................................................26

7.2.1 THE ROUTING ALGORITHM ...............................................................27

7.2.2 MULTICAST CONSIDERATIONS FOR PROTECTION AND PORT-

MIRRORING..........................................................................................27

7.2.3 DEVICE PROGRAMMING CONNECTIONS OVERVIEW....................27

7.2.4 REARRANGEMENT COORDINATION.................................................28

7.2.5 IMPLICATIONS OF TOPOLOGIES.......................................................29

7.3 TIME-SLOT-INTERCHANGE.............................................................................30

7.3.1 OPERATION OF THE TSI.....................................................................30

7.3.2 PROGRAMMING THE TBS ..................................................................32

7.3.3 PROGRAMMING THE TSE ..................................................................34

7.3.4 PROGRAMMING THE S/UNI-MACH48................................................34

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE iii

7.3.5 PROGRAMMING OPTIMIZATIONS......................................................34

7.4 TSE SPACE SWITCH........................................................................................35

7.4.1 OPERATION OF THE TSE SPACE SWITCH.......................................35

7.4.2 PROGRAMMING THE TSE SPACE SWITCH......................................36

7.4.3 STS-1 TIMESLOT MAPPINGS IN THE CHESS SET...........................37

8 SYSTEM POWER AND PACKAGE CONSIDERATIONS...............................................39

8.1 CHESS SET DESIGN REFERENCE.................................................................39

8.2 ANALOG AND DIGITAL POWER BY RAIL........................................................40

8.3 POWER SEQUENCING RULES FOR 1.8V AND 3.3V SUPPLIES:..................41

8.4 ESTIMATING SYSTEM POWER.......................................................................41

8.5 MAXIMUM CHESS SET SYSTEM POWER ......................................................43

8.6 CHESS THERMAL INFORMATION...................................................................43

8.6.1 HEAT SINKING: ....................................................................................43

8.6.2 EXAMPLE, USING THE S/UNI-MACH48:.............................................45

8.6.3 CHART OF θJA VERSUS AIR FLOW FOR CHESS CHIPSET ............46

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE iv

LIST OF FIGURES

FIGURE 1: REPRESENTATIVE SONET/SDH DATA STREAM OF THE CHESS DEVICES.........3

FIGURE 2: ANY SERVICE (AT ANY RATE) ON ANY CHANNEL..................................................6

FIGURE 3: ONE PORT ONE SERVICE (OC-N TO DS3)..............................................................7

FIGURE 4: ANY PORT ANY SERVICE (OC-N TO DS-1)..............................................................8

FIGURE 5: METRO/ACCESS RING ADM .....................................................................................8

FIGURE 6: 1+1 PROTECTION SWITCHING ................................................................................9

FIGURE 7A: IMPLEMENTING REDUNDANT SWITCH FABRICS..............................................10

FIGURE 7B: REDUNDANT SWITCH FABRICS WITH DECOUPLED LINE & SERVICE CARDS11

FIGURE 8: OC-48 TO DS-1/0 ON MULTIPLE CARDS................................................................12

FIGURE 9: OC-12 TO DS-1/0 WITH 1:N PROTECTIONS..........................................................13

FIGURE 10: SINGLE STAGE 160 GB FABRIC............................................................................15

FIGURE 11: THREE STAGE 320 GB FABRIC.............................................................................16

FIGURE 12: INTERFACING THE TEMUX TO THE TBS.............................................................21

FIGURE 13: J0 SYNCHRONIZATION CONTROL......................................................................24

FIGURE 14: TIME-SPACE-TIME SWITCHING...........................................................................29

FIGURE 15: OPERATION OF THE TIME SLOT INTERCHANGE .............................................30

FIGURE 16: THE TBS TIME SLOT INTERCHANGE OPERATION ...........................................31

FIGURE 17: SELECTING AN OUTPUT STS-12 DATA-STREAM...............................................32

FIGURE 18: THE TSE SPACE SWITCH OPERATION ...............................................................36

FIGURE 19: THERMAL SCHEMATIC OF HEAT DISSIPATION ..................................................45

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE v

LIST OF TABLES

TABLE 1 DEVICE COUNT FOR VARIOUS BANDWIDTHS.................................................17

TABLE 2 SYNCHRONIZATION PINS & REGISTERS..........................................................25

TABLE 3 SPECTRA-2488 TIMESLOT MAPPINGS..............................................................37

TABLE 4 SPECTRA-4X155 TIMESLOT MAPPINGS...........................................................38

TABLE 5 S/UNI-MACH48 TIMESLOT MAPPINGS ..............................................................38

TABLE 6 DEVICE SUMMARY..............................................................................................39

TABLE 7 ANALOG AND DIGITAL POWER BY RAIL..................................................................40

TABLE 8 POWER SEQUENCE RULES......................................................................................41

TABLE 9 SYSTEM POWER VERSUS LINK UTILIZATION........................................................43

TABLE 10 θJA VERSUS FORCED AIR........................................................................................46

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 1

1 INTRODUCTION

This document introduces the reader to the overall system considerations for

implementing components of the CHESS chipset. It also provides design

architectures available when using the CHESS chipset. This document is a

supplement to the individual CHESS Set device datasheets. Due to the vast

number of configurations available with the CHESS chipset, this document may

not cover all possible configurations. Please contact a PMC-Sierra Field

Applications or Applications Engineer for specific uses not covered in this

document.

Although every effort has been taken to ensure that this document is consistent

with the datasheet, some errors may occur. Where there are discrepancies with

this document, the datasheet and datasheet errata (if any) will take precedence.

1.1 Target Audience

This document has been prepared for Design Architects that are implementing

the CHESS set devices and require a quick reference on how to implement the

devices in a design. It is assumed the reader is familiar with SONET/SDH and

other common telecom/datacom terminology.

1.2 Numbering Conventions

Binary 0001b, 1110b

Decimal 198, 234, 2

Hexadecimal 2H, 200H

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 2

2 REGISTER DESCRIPTION

The CHESS Set Device Registers have the following characteristics:

• All values written into unused register bits should be written with logic 0

unless otherwise stated. This action will ensure software compatibility with

future versions of the product. Reading back unused bits can produce either

logic one or logic zero; hence, unused register bits should be masked off by

software during a register read access.

• Certain register bits are reserved. To ensure that the CHESS Set Devices

operate as intended reserved register bits must only be written with their

default values unless otherwise stated in the datasheet.

The CHESS Set Devices have 2 types of register spaces -- the “normal”

registers, which are accessed directly by the microprocessor bus interface of the

CHESS Set Device, and the “Indirect” registers which are internal to the CHESS

Set Device. Indirect registers are accessed through multiple writes to normal

registers and polling of a BUSY bit, ensuring data is visible at the microprocessor

bus or written into a normal register stored internal to the CHESS Set Device.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 3

3 THE CHESS ARCHITECTURE

The CHESS (CHannelizer Engine for SONET/SDH) chipset enables the salient

benefits of SONET/SDH to be leveraged in new multi-service switches and

routers. Specifically, multi-service switches and routers can be architected to

implement SONET ring and protection functions including UPSR, BLSR, or 1+1

protection. The chipset is designed to process channelized (down to a STS-

1/STM-0/AU3 granularity) SONET/SDH traffic carrying a variety of services

including ATM, POS and TDM, as shown in Figure 1. The CHESS architecture

enables like traffic to be groomed to the appropriate processing cards. That is,

ATM traffic from the STS-48 data stream can be groomed down to an STS-1

granularity to the ATM processing card and POS traffic to a packet-processing

card. Not only is grooming available with the CHESS set, but other features

inherent to CHESS make it easy to implement ADM’s, provide line and

equipment card protection and cross-connect switch fabrics, and even

channelize to a further granularity (sub DS3) once the initial grooming is

performed.

Figure 1: Representa tive SONET/SDH data stream of the CHESS Devices

TDM

ATM

POS OC-48

The CHESS set is composed of modular components that allow the customer to

pick and choose the system architecture. They are not limited to a single type of

design but rather support a variety of flexible Layer 1/ Layer 2 architectures.

Some examples of possible system architecture are presented in the coming

sections.

3.1 The CHESS Set Devices

The CHESS set consists of six different devices:

3.1.1 SPECTRA-2488

The SPECTRA-2488 is an STS-48 (2.488 Gbps) or 4 x STS-12 (622.08 Mbps)

channelized SONET/SDH framer. The chip provides section, line, and optional

path termination. The devices provides 48 pointer processors for

multiplexing/demultiplexing traffic down to STS-1 granularities and includes a

Time Slot Interchange for switching STS-1 level to one of 48 slots. The device

can seamlessly interconnect to the TBS for back-plane serialization or protection

purposes. Alternatively, the SPECTRA-2488 can be mated directly to any device

using the TeleCombus, including the S/UNI-MACH48.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 4

3.1.2 SPECTRA-622

The SPECTRA-622 is an STS-12 (622.08 Mbps) channelized SONET/SDH

framer with section, line and optional path termination. The chip is very similar in

functionality to the SPECTRA-2488, and includes STS-12 clock/data recovery as

well as 12 x DS3 mappers.

3.1.3 SPECTRA-4x155

The SPECTRA-4x155 is a 4xSTS-3 (155 Mbps) channelized SONET/SDH

framer with section, line and path termination. The chip is very similar in

functionality to the SPECTRA-2488, and includes 4 STS-3 clock/data recovery

units.

3.1.4 TBS

The TBS (TeleCombus Serializer) serializes and deserializes each of four 77.76

MHz TeleCombus data-paths to an 8B/10B encoded STS-12 serial link (777.6

MHz), thereby supporting an STS-48 bandwidth equivalent of traffic. There are

twelve 777.6 MHz LVDS full duplex ports available (3 sets of 4 links called

working, protect and auxiliary). The TBS provides the ability to drive 1.6 metres

of back-plane and board trace to enable line, equipment or ring protection and

has its own STS-1 Time Slot Interchange (TSI) for grooming traffic. The TSI

provides the ability to groom traffic at an STS-1 level to diff e re n t devices or

boards. Most users will use one of the three ports as the main port for grooming

to the TSE switch fabric or connecting to a layer 2 processing card. The second

port may be used for connecting to a redundant fabric or layer 2 processing card.

The third port might be used to pass non-drop traffic to another TBS/SPECTRA-

2488 pair. This specific example is shown later in Figure 5 for a Metro/Access

Ring ADM.

3.1.5 TSE

The TSE (Transmission Switch Element) is a scaleable 40 Gbps non-blocking

STS-1 cross-connect fabric. A single stage fabric can be built to 160 Gbps using

only 4 TSE devices. A three-stage fabric can be built to provide fabrics of 320

Gbps to 10.24 Tbps, using 24 to 768 TSE devices respectively.

3.1.6 S/UNI-MACH48

The S/UNI-MACH48 provides ATM mapping and packet mapping for both

SONET/SDH and DS3 frames for channelized STS-48/STM-16 streams. The

device contains:

• 48 DS3 framers for ATM and packet traffic

• 48 PLCP framers (option for mapping ATM cells into DS3 frames)

• 48 DS3 bit HDLC mappers for packet over DS3

• 48 AT M cell delineation blocks (ATM)

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 5

• 48 byte SONET/SDH HDLC processing blocks for Packet Over

SONET/SDH

Any mix of the above traffic is workable, with the constraints that traffic must be

DS3, STS-1, STS-3c, STS-12c, and STS-48c and must total no more than an

STS-48 worth of bandwidth. For example, the S/UNI-MACH48 can

simultaneously support 12 x DS3 packet, 12 x DS3 ATM, 4 x STS-3c ATM, 1 x

twelve STS-12c POS. The S/UNI-MACH48 has both a standard parallel

TeleCombus to mate to SPECTRA devices and a 1+1 serial LVDS TeleCombus

interfaces to mate directly to the TSE without the need of serialization by the

TBS.

3.1.7 S/UNI-ATLAS 3200

Although the S/UNI-ATLAS-3200 is not a member of the CHESS set, it does

interface to the S/UNI-MACH48. The S/UNI-ATL AS-3200 is a monolithic single

chip device which handles ATM Layer functions for one direction including

VPI/VCI address translation, cell appending, cell rate policing, per-connection

counting and I.610 compliant OAM requirements for 64K VCs (virtual

connections). It supports an instantaneous transfer rate of 3200Mbit/s equivalent

to a cell transfer rate of 5.68x106 cells/s (one STS-48c or four STS-12c). Two or

more S/UNI-ATLAS-3200 devices can be cascaded to support additional VC’s.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 6

3.2 CHESS SET Example Architectures

There are many possible ways to interconnect this chipset. The following section

illustrates some common examples.

3.2.1 The Basic Architecture

The example below demonstrates the power of the CHESS chipset to implement

Any Service, at Any Rate on Any Channel. While the following example in Figure

2 does not give exact chip placement, it does demonstrate how a system can be

partitioned.

Figure 2: Any Service (at any Rate) on any Channel

Packet/Cell

Switch

Fabric

Sub Rate

DS-3/1/0

VT/TU

Mapped

T1/E1

TSE

SPECTRA

2488

TBS

S/UNI

MACH48

TBS

Telecombus LVDS LVDS

LVDS

Telecombus

UL3/

PL3

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 7

3.2.2 One Port - One Service

The following example (Figure 3) illustrates a system in which e ach port card has

its own service. There is no grooming of traffic between cards. One card may

be strictly ATM, another TDM, and yet another POS, etc.

Figure 3: One Port One Service (OC-N to DS3)

PM7390

S/UNI-MACH48

Packet & Cell

Switch Fabric

S/UNI-

ATLAS-3200

IP Procesor

CHOC-48/STM-16 Port Card

Opt PM5315

SPECTRA-2488

SPECTRA-155

QUAD

16xOC-3/STM-1 Port Card

Opt

4xOC-12/STM-4 Port Card

Opt

PM5315

SPECTRA-2488

( 4 x 622 Mode)

PM5316

SPECTRA-155

QUAD

x16

4xOC-12/STM-4 Port Card

PM5316

SPECTRA-155

QUAD

PM5316

SPECTRA-

4x155

Opt

SPECTRA-155

QUAD

PM5316

SPECTRA-155

QUAD

PM5316

SPECTRA-155

QUAD

PM5313

SPECTRA-622

Processor

PM7390

S/UNI-MACH48 TM

Processor

PM7390

S/UNI-MACH48 TM

Processor

PM7390

S/UNI-MACH48 TM

Processor

S/UNI-

ATLAS-3200

IP Procesor

S/UNI-

ATLAS-3200

IP Procesor

S/UNI-

ATLAS-3200

IP Procesor

Layer 1 -

SONET/SDH Framers Layer 2 -

ATM & POS Framing

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 8

3.2.3 Complete Cross-Connect Switch

Below is an example (Figure 4) utilizing the TSE fabric to provide switching of

STS-1s from any port card to any processing card. In this fashion, any type of

traffic can be groomed to its correct processing card regardless of the source.

The Layer 2 rates supported include DS-1 through to STS-48. The TSE fabric

can be scaled from 40 Gbps – 160 Gbps in a single stage design. 4 TSE

devices are used in the 160 Gbps fabric. Note that the TBS devices are used to

drive the backplane in this system and can participate in switching (TSI).

Figure 4: Any Port Any Service (OC-N to DS-1)

3.2.4 Building a Small ADM

The following example (Figure 5) demonstrates an implementation of a small

North-South (STS-48) Add/Drop Multiplexor. Aside from the serializing and a

backplane driving functions, the TBS can provide STS-1 switching and with use

of its auxiliary port, drop/ add traffic to the S/UNI-MACH48. For example, in a

North-South ADM one SPECTRA 2488 device receives from (for example) the

counter-clockwise ring and transmits on the clockwise, while the other SPECTRA

receives and transmits the opposite. In this case, pass-through traffic can be

passed from one card to the other via the TBS so that it can continue around the

ring. In an East-West ADM, one SPECTRA-2488 would both receive and

transmit on the same ring. This would require that traffic be passed form the

Drop bus to Add bus. This approach is less popular.

Figure 5: Metro/Access Ring ADM

Other Tributary card alternatives:

Frame Rel ay Pr oc essi ng

T1/E1 Tri butary Card

OC- 3 Tributary Card

Etc.

East Lin e/ A ggr egat e Card

Optics SPECTRA

2488 TBS

West Line/Aggre gate Card

Optics SPECTRA

2488 TBS

Layer 1 or Layer 2 Tribut ary Card

S/UNI

MACH48

ATM/ IP Processi n g

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 9

3.2.5 1+1 Protection on Line Cards

The CHESS chipset provides the ability to perform 1 + 1 line card protection by

using the TBS as shown in Figure 6. The secondary (protection) and auxiliary

LVDS ports of the TBS give the flexibility to perform many types of protection and

redundancy configurations. In this example, the LVDS links would drive the

system back plane. Each set of LVDS ports can handle a full 2.488 Gbps, and

can be configured independently. In the event that the optics of card 1 failed

then traffic can be routed from the S/UNI-MACH48 device of card 1 through TBS,

SPECTRA-2488 and optics of card 2. This is illustrated more clearly by the

dashed line below.

Figure 6: 1+1 Protection Switching

SPECTRA

2488 S/UNI

MACH48

TBS

Telecombus LVDS UL3/PL3

UL3/PL3

Optics SPECTRA

2488 S/UNI

MACH48

TBS

Telecombus LVDS

Optics

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 10

3.2.6 Building Redundant Sw itch Fabrics

Again the TBS chips give the ability to create a completely redundant switch

fabric without the use of muxes or other external circuitry. Figure 7a illustrates a

very simple redundant fabric, but in the same manner, much larger redundant

fabrics can be built. This architecture is ideal to implement a switch fabric for slot

limited chassis or to leverage on an existing line interface card (SPECTRA-2488,

TBS and S/UNI-MACH-48 card) to build a switch fabric. This configuration can

be used if the design is slot limited and separate line and port cards cannot be

accommodated in the frame. The one limitation is that the top SPECTRA-2488

cannot groom traffic to the bottom S/UNI MACH-48. Figure 7b illustrates a more

popular approach that decouples the Layer 1 line card from the Layer 2 service

card to provide for full grooming.

Figure 7a: Implementing Redundant Switch Fabrics

PM5315

SPECTRA-

2488

PM5310

TBS

PM7390

S/UNI-

MACH48

PM5372

TSE

PM5315

SPECTRA-

2488

PM5310

TBS

PM7390

S/UNI-

MACH48

LVDS x4

PM5372

TSE

working fabric protect f abri c

LVDS x4

LVDS x4 LVDS x4

LVDS x4

TeleCombus

UL3/PL3

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 11

Figure 7b: Redundant Switch Fabrics with Decoupled Line & Service Cards

Protect

40 Gb Fabric

Working

40 Gb Fabric

OC-48/STM-16 Port Card

Opt PM5315

SPECTRA-2488

SPECTRA-155

QUAD

16xOC-3/STM-1 Port Card

Opt

4xOC-12/STM-4 Port Card

Opt

PM5315

SPECTRA-2488

( 4 x 622 Mode)

PM5316

SPECTRA-155

QUAD

x16

Layer 1 -

SONET/SDH Framers Layer 1 -

Cross-Connect Layer 2 -

Processing

4xOC-12/STM-4 Port Card

PM5316

SPECTRA-155

QUAD

PM5316

SPECTRA-

4x155

Opt

SPECTRA-155

QUAD

PM5316

SPECTRA-155

QUAD

PM5316

SPECTRA-155

QUAD

PM5313

SPECTRA-622

PM5310

TBS

PM5310

TBS

PM5310

TBS

PM5310

TBS

PM5372

TSE TEMUX FREEDM

Frame Relay Processing Card

TEMAPTUPP-622 QDSX

PM5310

TBS

QDSX

T1/E1 Tributary Card

PM7390

S/UNI-MACH48 S/UNI-

ATLAS-3200

ATM Processing Card

Packet Over SONET Processing Card

PM5310

TBS

PM7390

S/UNI-MACH48 Layer 2/3

Processor

Working LVDS Links

Protect LVDS Links

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 12

3.2.7 Implementing DS-1/0 Cards with Protection

The first system (Figure 8 ) illu strates an im plemen tation for an OC-48 to DS1/0

card. Each TBS feeds one STS-12 link to a service card. Twelve TEMUX

devices can terminate an STS-12 worth of traffic. Four such cards can terminate

the entire traffic load (STS-48). Note that in this example “a” can have a value of

0 to 3 (up to 4 cards). Service card protection is available by using the protection

links. This is better illustrated in the next example. The second example system

illustrates an OC-1 2 to DS1/0 system with equipme nt protection (Figure 9). Each

interface card uses only one LVDS link to interface to the service card providing

an STS-12 worth of traffic. A spare link is connected to the protection service

card. The protection card TBS has 12 links (4 working, 4 prote ct and 4 a uxiliary),

and can thus protect up to 12 such cards. Therefore n can have a value of 12

(12 working cards for one protection card).

Figure 8: OC-48 to DS-1/0 on multiple cards

PM5315

SPECTRA-2488

PM5310

TBS

PM5310

TBS

TEMUX #1

TEMUX #12

Card 1

(STS-12)

OPTICS

HDLC

C.E.

IMA

PM5310

TBS

TEMUX #1

TEMUX #12

Card a

(STS-12)

HDLC

C.E.

IMA

LVDS Link #1

(STS-12)

LVDS Link #a

(STS-12)

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 13

Figure 9: OC-12 to DS-1/0 with 1:n protections

PM5310

TBS

PM5313

SPECTRA-

622

PM5310

TBS

PM5310

TBS

PM5313

SPECTRA-

622

PM5310

TBS

PM5310

TBS

PM5313

SPECTRA-

622

PM5310

TBS

1:n SPARE

TEMUX84

TEMUX

(1 to 12)

TEMUX84

TEMUX

(1 to 12)

TEMUX84

TEMUX

(1 to 12)

PM5310

TBS

TEMUX84

TEMUX

(1 to 12)

Optics

LVDS Link #1

(STS-12)

LVDS Link #1-n

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 14

4 FABRIC ARCHITECTURES

A few common multi-service switch architectures will be shown to assist in

understanding how to build the fabric and connect the TBS with the TSE chips.

The fabric can also provide for sub-lamda grooming without SONET path

termination by replacing the S/UNI-MACH48 with the TBS and SPECTRA-2488.

4.1 Single Stage Fabric: Up to 160 Gbps

Since each TSE can support 40 Gbps of bandwidth, a very large single stage

fabric can be built without the requirement for multi-stage architectures. Recall

each TBS has four STS-12 LVDS ports (actually 3 sets of 4 – working, protect,

auxiliary) which can connect to the TSE. Therefore, a non-blocking single stage

fabric can be built with up to 4 TSE’s by connecting a single LVDS from the TBS

to each of the TSE devices. This gives a 160 Gbps fabric as shown in Figure 10.

32 TBS and 32 S/UNI-MACH48 chips will be connected in this fashion to fully

utilize the 160 Gbps fabric (assuming both the TX and RX links of the devices

are used).

A single TSE can be used to build a 40 Gbps fabric and two-chips yield a 80

Gbps fabric. . In a 40 Gbps fabric all four ports of the TBS and S/UNI-MACH48

feed a single TSE. Thus 8 total TBS and 8 S/UNI-MACH48 devices would

interface to a single TSE. In an 80 Gbps fabric two of the TBS STS-12 LVDS

links mate to each of the two TSE devices. A total of 16 TBS and 16 S/UNI-

MACH48 devices would interface to the two TSE cross connects. In addition, an

equivalent number of SPECTRA-2488 devices could interface to the line side of

the TBS. This example is bi-directional in which both the TX and RX links of the

respective devices are connected to the TSE.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 15

Figure 10: Single Stage 160 Gb Fabric

TSE

TBS

#1 x1

TBS

#32

SPECTRA

-2488 #1

MACH48

SPECTRA -

2488 #32

TSE

MACH48

#32

4.2 Three Stage Fabric: Up to 2560 Gbps

Moving beyond 160 Gbps requires a three-stage architecture to guarantee all

possible port mappings and a non-blocking architecture. The example in Figure

11 shows a 320 Gb fabric. Not all connections are shown for simplicity and clarity

in this example. The 64 ports on the TSE are sent in groups of eight between

TSE chips in each stage. All four links from a TBS are connected to a single

TSE. This architecture requires 64 TBS and S/UNI-MACH48 devices and 24

TSE chips. A 2560 Gbps fabric is constructed similarly, with each TSE feeding a

single STS-12 LVDS link to one of the 64 other TSE devices in the next

stage.Note that in Figure 11 the TBS and S/UNI-MACH48 device sit on both

sides of the first and third fabric.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 16

Figure 11: Three Stage 320 Gb Fabric

TSE

#10

TSE

#11

TSE

#12

TSE

#13

TSE

#14

TSE

#15

TSE

#16

TSE

#18

TSE

#17

TSE

#19

TSE

#20

TSE

#24

TSE

#23

TSE

#22

TSE

#21

TBS

#64

x8 MACH48

MACH48

#64

4.3 Device Count Required For Fully Populated Fabrics

The follo wing table illu strates the n umber of devices required for a fully

populated TSE fabric as shown in Figure 7b. Note that the TSE device count

doubles if implementing fabric protection on the TSE (single point of failure). In

general, the device count scales directly with the bandwidth (as the bandwidth

doubles, the chip count doubles). Note that a three stage TSE fabric is required

for bandwidths beyond 160 Gbps. This produces a tripling of the TSE chip count

on top of the normal doubling with bandwidth at the 320 Gbps fabric size.

Beyond 320 Gbps all devices scale with bandwidth until 20480 Gbps where a 5

stage TSE fabric is required to produce a non-blocking architecture. Planes refer

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 17

to the number of different TSEs that a TBS is connected to. If all four TBS links

are connected to a single TSE, then this represents a plane of 1. If half (two) of

the TBS links are connected to the first TSE, then the other two are mated to a

second TSE then this is a plane of two. Finally if each of the f our TBS link

services an individual TSE, then this represents a plane of 4. Note that for a

10240 Gbps fabric, that each TBS link feeds 1 of four 2560 Gbps three stage

TSE fabrics.

Table 1 Device Count For Various Bandwidths

Bandwidth Fabric Layout # of Devices Required

Stages Planes SPECTRA-2488, TBS

and S/UNI-MA CH48 SPECTRA-2488, TBS

and S/UNI-MA CH48 TSE (x2 for fabric

protection)

Unidirectional Bi-directional

40 Gb 1 1 16 8 1

80 Gb 1 2 32 16 2

160 Gb 1 4 64 32 4

320 Gb 3 1 128 64 24

640 Gb 3 1 256 128 48

1280 Gb 3 1 512 256 96

2560 Gb 3 1 1024 512 192

5120 Gb 3 2 2048 1024 384

10240 Gb 3 4 4096 2048 768

20480 Gb 5 1 8192 4096 2560

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 18

5 DEVICE INTERFACES

5.1 LVDS Interface

The LVDS interface implemented on the TBS and TSE follows the IEEE 1596.3-

1996 specification with some minor differences. The changes are implemented

to customize and optimize the LVDS interface for the system and are described

in detail below. Even with these differences the LVDS interface should function

with the physical layer of other LVDS interfaces. The differences include:

1. Faster rise/fall times (200 – 400) ps versus (300 - 500) ps. Faster edge rates

are commonly used with higher speed LVDS interfaces in the industry to ease

the interfacing. The IEEE 1596.3-1996 edge rates are optimized for data

rates of 400 Mbps and below.

2. Hysteresis is not implemented on the receive LVDS interface. Hysteresis is

used in many implementations to negate the effect of noise that may exist on

any of the unused LVDS links. Hysteresis was not implemented in the

CHESS set devices to minimize circuit complexity, power and cost. The RX

interface and the DRU can be both disabled (powered down) by the

appropriate register to prevent any sensitivity to noise.

3. The LVDS transmitter contains an on-chip 100-ohm termination. Most

implementations have single 100-ohm termination on the receiver. By

implementing a double termination (on both the LVDS receiver and

transmitter) a higher signal integrity and matching is ensured.

4. Although not a difference with the layer 1 IEEE 1596.3-1996 specification, the

layer 2 8B/10B encoding is presented here for completeness. 8B/10B

encoding guarantees transition density as compared to scrambled encoding,

which provides only a certain probability of transition density. This

guaranteed transition density allows a simpler and more power-effective data

recovery unit, provides a more robust serial interface (greater trace or back-

plane distance achievable). It also negates the need for complete SONET

framing since the A1A2 and J0 bytes can be encoded into special escape

characters of the LVDS data stream.

5.2 Hot Sw apping and Floating Links on the LVDS Interface

The interface is a LVDS, not a CMOS interface; therefore there is no electrical

problem in leaving the LVDS interfaces floating. Note that the LVDS receiver

consists of a differential amplifier with a wide common-mode range. The power

dissipation is independent of the data transitions (that is, if the input is

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 19

connected). There is an internal 100 Ω termination across the positive and

negative input. Floating inputs will settle to an arbitrary voltage (between VDD

and VSS) determined by leakage paths. Regardless of this arbitrary voltage, the

input structure of the receiver will operate in its proper range and the receiver

output will be logic 1 or 0 depending on internal offsets. Noise events (power

supply noise, crosstalk) may induce the receiver to toggle randomly from 1 to 0

generating "ambiguous" data. This ambiguous data will not result in any

problems but is not a desirable condition. Recall that it is the responsibility of the

user to configure the appropriate links for calls.

Unused links should be disabled in software. The consumed power for that link

will be nearly 0mW. There is no requirement for how quickly the link should be

disabled. Disabling the link, simply results in lower power dissipation since the

circuitry will be shut down. This action is not mandatory, but is good practice to

improve margins. There are no problems with hot-swapping. The “hot-swap”

channel can be left enabled and the device will sync up once the far end

transmitter is connected. There are no affects on other channels. Hot swapping

of ca rd s is still allowed by reprogramm in g of the links in software.

5.3 No device damage will occur when driving an unpowered input board. (For

example, if a board is inserted and power has not come up yet, or a fuse has

blow n on the board). If a LVDS driver is connected to an unpowered

receiver, the LVDS transmitter will force a maximum of 7mA to one of the

input pins. This produces a forw ard bias condition on the ESD protection

scheme but not sufficient to elevate the power supply of the device at the

receiving end and produce device damage (nor to result in a latch-up

condition). Also note that the LVDS interface should not be damaged by

interfacing to a 3.3V TTL slot. Calculating the Maximum Difference in Trace

Lengths

The TSE utilizes 64 different input and output differential LVDS pairs. It is critical

to match the lengths of the positive and negative traces of each differential pair

to minimize skew and maximize the eye opening. However, matching one

differential pair to another pair is not as important. To accommodate this, the

high-speed serial LVDS links have a 24 word (10 bit byte) FIFO. Of this 24 word

FIFO, 8 words should be allocated for clock skew and wander between cards or

within devices. The remaining 16 words are then available to accommodate

differences in trace lengths between LVDS pairs.

The 16 word FIFO yields an allowable delay of 205.8 ns or 41.2m.

➩ 16 words x 10 bits/word = 160 bits of margin in FIFO

➩ 1/777.6 Mb/s = 1.29 ns/bit on the serial link

➩ 160 bits x 1.29 ns/bit = 205.8 ns of margin or 16 clock cycles (at 77.76 MHz)

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 20

A transmission speed of 2/3 the speed of light, this corresponds to a trace length

difference of 41.2m.

➩ 205.8 x 10-9 s x 2/3 x 3 x 108 m/s = 41.2 m

It is important to note that the LVDS interface itself is designed to drive 1m of

backplane plus 30cm of trace length on either side. Therefore this will be the

limiting factor. This 1.6m of trace length corresponds to 8ns of delay or 0.6

clock cycles (77.76 MHz) between any LVDS link. Low loss cable or an optical

interface can be used to connect to the LVDS interface to realize greater back-

plane distances.

5.4 Interfacing a 19.44 MHz Te leCombus to a 77.76 MHz TeleCombus

The task of interfacing the 19.44 MHz TeleCombus to the 77.76 MHz

TeleCombus will be illustrated by means of the TEMUX devices (19.44 MHz) and

TBS (77.76 MHz) as shown in Figure 12. Three TEMUXs together drive a

19.44MHz TeleCombus. Four such 19.44MHz buses are muxed together into a

77.76MHz bus, so twelve TEMUXs are required to interface to each of the four

77.76 MHz TBS buses.

Each 19.44 MHz stream is muxed in sequentially- first, second, third, fourth, first,

… The exact order doesn’t matter, as long as it’s consistent and the timing

signals are muxed in the same order. The J1 pulses must be muxed together in

the same order as the data streams. The same criterion applies for demuxing.

Synchronizing the clock will have to be done carefully to avoid timing problems.

The only criterion is that set-up and hold times are met throughout the system.

Note the IJ0J1 signal into TBS expects only one J0 pulse, marking the first J0

byte in the incoming stream. But there will be one J0 (C1) pulse coming from

each of the four sets of TEMUXs. Only the first J0 (C1) must be sent to TBS; the

others are discarded. The converse is true in the opposite direction; one J0 pulse

from TBS needs to be replicated into four, one fo r each of the TEMUX sets.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 21

Figure 12: Interfacing the TEMUX to the TBS

FPGA

Add Direction

SYSCLK

TBS

ID1[7:0]

IDP1

IPL1

IJ0J11

IPAIS1

ITV51

ITAIS

OD1[7:0]

ODP1

OPL1

OJ0J11

OPAIS1

OTV51

OTPL1

OTAIS

ID[2:4][7:0]

IDP[2:4]

IPL[2:4]

IJ0J1[2:4]

IPAIS[2:4]

ITV5[2:4]

ITAIS[2:4]

OD[2:4][7:0]

ODP[2:4]

OPL[2:4]

OJ0J1[2:4]

OPAIS[2:4]

OTV5[2:4]

OTPL[2:4]

OTAIS[2:4]

3*TEMUX

LREFCLK

LADATA[7:0]

LADP

LAPL

LAC1J1V1

LAOE

LAC1

LDDATA[7:0]

LDDP

LDPL

LDTPL

LDV5

LDAIS

77.76MHz

3*TEMUX

LREFCLK

LADATA[7:0]

LADP

LAPL

LAC1J1V1

LAOE

LAC1

LDDATA[7:0]

LDDP

LDPL

LDTPL

LDV5

LDAIS

3*TEMUX

LREFCLK

LADATA[7:0]

LADP

LAPL

LAC1J1V1

LAOE

LAC1

LDDATA[7:0]

LDDP

LDPL

LDTPL

LDV5

LDAIS

3*TEMUX

LREFCLK

LADATA[7:0]

LADP

LAPL

LAC1J1V1

LAOE

LAC1

LDDATA[7:0]

LDDP

LDPL

LDTPL

LDV5

LDAIS

SYSTEM

FRAME PULSE

Mux Selection

Circuitry

19.44 MHz

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 22

6 SYSTEM SYNCRONIZATION

6.1 Purpose of Synchronization

Synchronization is required to align the serial (STS-12) LVDS links (such as the

64 serial links of the TSE). A global frame pulse (8 kHz, 125µs) is generated and

fed to each CHESS device and the J0 byte of the STS-12 data stream is used to

synchronize each STS-12 link. Synchronization is accomplished on each link by

programming the clock cycle delay from the global frame pulse to the time when

the J0 byte is received on that link. Thus the device is programmed to expect

the J0 byte a set number of clock cycles following the frame pulse.

There are two primary sources of delay. First, the delay can be introduced by

clock skew or trace length differences between different links. However these

forms of delay can be ignored in practice because the input receive links

compensate for them. A 24-word input FIFO is available to accommodate for

any mismatches. As discussed earlier, 1 m differential plus clock skew is easily

tolerable from link to link. The second and most significant source of delay is

introduced by the device latency.

6.2 Implementing Synchronization in CHESS SET Devices

Consider the implementation shown in Figure 13. All devices receive the global

frame pulse simultaneously at time t0 (ignoring any trace length differentials).

The SPECTRA-2488 emits the J0 byte onto the TeleCombus upon receiving the

global frame pulse on the DJ0REF input. This action is entirely independent of

receiving a J0 byte from the optical line. SPECTRA-2488 pointer adjustments

will define the start of the payload envelope (the J1 byte indicates start of

payload) and this payload will be outputted over the TeleCombus. The

SPECTRA-2488 can be viewed as the master by which the synchronization of

the other CHESS devices is determined. The TBS expects the four incoming

eight bit 77.76 MHz TeleCombus data paths to be synchronized and upon

processing emits the serialized data with J0 character 33 clock cycles after

receiving the J0 on the parallel TeleCombus. The J0 byte on each of the twelve

independent 777.6 MHz LVDS links are not exactly simultaneous and may have

a slight amount of skew relative to each other (because of presence of an 8 word

FIFO on the LVDS transmitter output). The LVDS links are then mated to the

TSE through a back-plane. The TSE is programmed (via indirect register access

of the RJ0DLY[13:0] word) to expect the J0 byte a certain number of clock cycles

after it receives the global frame pulse.

The ingress FIFOs permit a variable latency in J0 arrival of up to 16 clock cycles.

That is, the largest tolerable delay between the slowest and fastest LVDS link of

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 23

the 64 TSE LVDS links is 16 bytes. Consequently, the external system must

ensure that the relative delays between all the 64 receive LVDS links be less

than 16 bytes. The minimum value for the internal programmable delay

(RJ0DLY[13:0]) is the delay to the last (slowest) J0 character plus 15 bytes. The

maximum value is the delay to the first (fastest) J0 character plus 31 bytes. The

actual programmed delay should be based on the delay of the “slowest” of the

64 links – the link in which J0 arrives last plus a small safety margin of 1 or 2

words. The magnitude of the clock cycle delay is bounded by two parameters.

First, the programmed delay register RJ0DLY is 14 bits. This implies that a clock

cycle delay of 214 –1 or 16,383 clock cycles can be programmed. However, the

second parameter, the frame rate (125 µs), bounds the delay to one STS-12

frame or 9719 (9720 unique values but 0 is the value for no delay) clock cycles

(125 µs x 77.76 MHz), after which the next SONET frame begins. The TSE,

upon receiving the global frame pulse, will wait the programmed amount of time

(56 clock cycles + cable length delays) before searching each of the 64 links for

the location of the J0 pulse to initialize synchronization.

The number of clock cycles can be determined by simply adding the relevant

device and cable length latencies. In practice, the programmed delay can be

obtained by measuring the clock difference between the global 8kHz frame pulse

and the presence of the J0 on the TJ0FP pin. The programmed delay is this

clock cycle difference plus a few clock cycles for margin.

This synchronization mechanism is flexible to accommodate differing path

lengths. Consider a TSE that is mated to a S/UNI-MACH48 on one link and a

SPECTRA-2488 feeding a TBS on the other link. The alternate data paths have

different delays; the SPECTRA-2488/TBS link has a greater delay than the S/UNI-

MACH48 link delay. In this case, the S/UNI-MACH-48 is programmed to emit the

J0 pulse later than SPECTRA-2488 (but aligned with the TBS serial output) such

that the J0 from both sources arrive at the TSE within the allowed 16-clock cycle

wi ndow. The S/UNI-MACH48 programmed delay is 24 clock cycles after the

receipt of the frame pulse.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 24

Figure 13: J0 Synchronization Control

SPECTRA-

2488

DFP

AJ0J1

DJ0J1

TBS

RJ0FP

OJ0J1

IJ0J1

TSE

RJ0FP

S/UNI-MACH48

RJ0FP

OJ0REF

Parallel

TelecomBus Serial LVDS

TelecomBus

8 kHz reference frame pulse

distributed to all devices at t

RJ0DLY

TSE

23 23

RJ0DLY

MACH

RJ0DLY

TBS

TJ0DLY

MACH

In the line side RX direction:

1. An 8kHz frame pulse is received by all devices at time t

2. Upon receipt of the 8kHz frame pulse, the SPECTRA-2488 outputs data (and J0) onto the

TeleCombus at time t

3. The TBS emits the serialized J0 byte and data 33 clock cycles later at time t

4. The J0 byte arrives at the input of the TSE A clock cycles later at time t

. A and B represents the

clock cycle delay of the link

5. 23 clock cycles after the receipt of the J0 on the input link, the various J0 bytes are centered in

the TSE's 24 word FIFO. This cumulative time from t

is the programmed RJ0DLY of the TSE.

6. The TSE emits the J0 66 clock cycles after receipt of the J0 at time t

7. The J0 byte is present at the MACH48 device B clock cycles later at time t

8. 23 clock cyles after the receipt of the J0 on the serial link, the four incoming J0 bytes of each

LVDS link are centered in the 24 word FIFOs. This cumulative time from t

is the programmed

RJ0DLY of the TSE.

In the line side TX direction:

1. An 8kHz frame pulse is received by all devices at time t

2. Nine clock cycles after receiving the 8kHz frame pulse, the S/UNI-MACH48 outputs data (and J0)

to the TSE device. Since t

and t

must be equal (synchronous) for the TSE to function, a 24 clock

cycle delay must be programmed as the TJ0DLY of the S/UNI-MACH48. This assumes the link

delay A = the link delay B or is negligible. Actually, t

- t

is the propogation time of the signal

between the devices measured in 77.76 MHz clock cycles.

3. The TSE outputs the J0 byte at time t

4. The TBS receives the J0 byte at time t

and 23 clock cycles later, each of the four J0 bytes are

centered in the 24 word FIFOs. This cumulative time from t

is the programmed RJ0DLY of the TBS.

Direction

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 25

6.3 Implementing Synchronization in CHESS SET Devices

Table 2 Synchronization Pins & Registers

Device Pin Name Register Bit(s)

Name

Transmit J 0

Frame Pul se Receive J0

Frame Pul se Transmit J 0

Delay Receive J0

Delay

TBS TJ0FP RJ0FP RJ0DLY[13:0]

TSE TJ0FP RJ0FP

S/UNI-

MACH48 TJ0FP RJ0FP/

OJ0REF

OJ0DLY[13:0] RJ0DLY[13:0]

SPECTRA-

2488 DFP

SPECTRA-

622 DFP

SPECTRA-

4x155 TBD

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 26

7 PROGRAMMING CONNECTIONS

7.1 Introduction

This section describes the process of physically mapping a call onto a fabric

comprised of CHESS devices. The operation of the various blocks within the

devices is described, and the programming of the devices in a coordinated

fashion to provide hitless rearrangement of the fabric is addressed.

7.2 Problem description

Connections within a switch fabric made from PMC-Sierra’s CHESS devices are

made at the STS-1 level. Programming multiple parallel STS-1 connections

supports higher levels of concatenation. Any level of concatenation can be

supported in this way through a TSE fabric. As the devices have aggregate

capacity of an STS-48 or higher and physical connections between the devices

are made with STS-12 links, a method is needed to calculate how to route a call

through the fabric.

A call is defined as a single STS-1 connection from a timeslot on an input port of

an input device to a timeslot on an output port of an output device. Depending

on the topology of the fabric, the call may traverse one or more core stages

comprised of TSEs.

The CHESS architecture is a re-arrangeably non-blocking Time-Space-Time

STS-1 switch fabric. The “Space” portion of the fabric takes data from a set of

available inputs and moves it to a set of outputs. There are multiple settings for

the space switch each corresponding to an STS-1 timeslot. The first “Time”

portion of the fabric is responsible for moving the data from the timeslot it is

received on to the timeslot that matches the space switch’s opportunity to move

the data to the desired output. The second “Time” portion of the fabric is

responsible for moving the data from the output of the space switch to the

desired timeslot on the output port. Depending on existing connections within

the fabric, there may not be a time where the space stage has both the desired

input and the desired output free. In this case existing connections may need to

be rearranged to add the desired connection.

Setting up a call therefore requires the following steps:

1) Determining a set of STS-1 level connections needed for the aggregate

bandwidth of the desired call,

2) Determining a path for each of these connections through the fabric,

3) Programming the connections into each of the devices in the path,

4) Committing the new connection definitions in all devices simultaneously.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 27

7.2.1 The Routing Algorithm

A call can be described as a collection of Device-Port-Timeslot (DPS) pair. A

single DPS pair describes a single STS-1 call. N DPS pairs therefore describe

an STS-N call. Reference software is available from PMC-Sierra to perform the

routing of each STS-1 call through a fabric. The output of the algorithm identifies

which timeslot through the space portion of the switch (called a wave) can be

used to make the desired connection. Based on the topology of the fabric, this

information is used to derive the actual device settings required to set up the call.

7.2.2 Multicast considerations for protection and port-mirroring

The routing algorithm provided by PMC handles multicast as well as unicast

connections. In general, a mix of multicast connections from any port to any port

will result in an indeterminate route calculation time and in some cases the

multicast routing may fail to find a path. The documentation for the algorithm

provides a set of conditions for which multicast routing is guaranteed.

7.2.3 Device Programming Connections Overview

There are currently three devices in the PMC CHESS Set which together form a

complete switch fabric. These devices are the TBS, TSE, and S/UNI-MACH48.

The S/UNI-MACH48 and TBS implement the “time” portion of the fabric while the

TSE implements both the “space” portion and optionally the ingress and egress

“time” portion in some configurations.

The TBS has a parallel STS-48 TeleCombus interface for connection to a

SONET framer and twelve STS-12 serial interfaces for connecting to the fabric.

Only four of the twelve are required for connecting to a fabric. The other eight

are available for redundant configurations and other unique topologies. Any

STS-1 can be mapped from the STS-48 TeleCombus to any one or more of the

STS-1s on the outgoing serial links. In the reverse direction, any of the STS-1s

on any of the twelve serial links can be mapped into any STS-1 on the parallel

TeleCombus.

The S/UNI-MACH48 has the same STS-1 interchange functionality as the TBS

with the exception that it has eight serial STS-12s instead of twelve. The S/UNI-

MACH48 will thus be tre ated as a subset of the TBS in the following discussions

and not treated directly except when discussing individual register addresses.

The TSE implements the “Space” portion of the fabric and optionally part of the

“time” portion. It has 64 serial STS-12 interfaces. The time-based portion of the

TSE can rearrange the order in which the STS-1s on an incoming STS-12 link

are presented to the space fabric. STS-1s from the 64 incoming STS-12s are

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 28

presented to the input of the space switch in twelve time periods. Connections

from the inputs to the outputs of the space switch within each of these time

periods are described as a wave. Each wave can have a unique mapping of

inputs to outputs.

7.2.4 Rearrangement coordination

All internal switching blocks of the CHESS Set devices have two configuration

tables. One of the tables describes the active switch settings. The second table

is used to modify the switch settings off-line. The devices can be instructed by a

global signal to switch the definition of the active and inactive tables on the

second 125µs STS-12 frame boundary after receiving the global switch over

signal. (As STS-1’s are transported over STS-12 links, the switch is coordinated

on the first STS-1 within an STS-12). This allows a coordinated hitless

reconfiguration of a fabric containing multiple devices. Alternatively the user can

change which page is active via software control on a per-STS-48 block basis

within each device. This is useful mainly in situations where the modification

being made does not affect other existing calls or coordinated switchover with

other devices is not required. Example configurations of this type could include

devices in an inactive redundant switch plane or small fabrics comprised solely

of TBS and/or S/UNI-MACH48 devices. In all cases, the selected page bit within

each functional block is exclusive-OR’d with the external page-select pin allowing

for a mixture of the coordinated switch models.

The user may choose one of several methods for updating the switch table

settings when adding or deleting calls. One method is to write the entire new

switch configuration for each device in all of the inactive tables, then force the

switch between active and inactive settings. This is inefficient in terms of

managing the hardware when only incremental changes are required for adding

a new connection but it may be useful for implementing network connection

preplans. The second method is to write only the necessary changes to the

inactive tables before forcing the switch. This requires that the software keep

track of the existing state of the inactive tables both for configuration purposes

and when routing new calls. To ease this burden, it is assumed that the user will

perform the following steps when reconfiguring the devices:

1) perform a call routing calculation,

2) write the necessary changes to the inactive switch setting tables,

3) force the switch between active and inactive tables,

4) update the now inactive tables with the same changes made to the now active

tables.

This enables the user to keep track of only one set of switch settings, which

simplifies the call setup and route calculation steps.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 29

7.2.5 Implications of Topologies

There are many topologies available for interconnecting the devices in the

CHESS chipset as discussed earlier. Topologies using the TSE device have

multiple time-stages – those in the TBS/S/UNI-MACH48 devices and those in the

TSE. The question arises as to whether all of the time stages are necessary,

and if not which of these time stages should be used. For a topology other than

a 40 Gbps or smaller fabric using one TSE (or two TSE’s for redundant fabrics),

the time stages in the TBS/S/UNI-MACH48 devices should be used and the TSE

time stages left in the “identity mapping” or pass-through mode. If there is only

one active TSE in the switch fabric, then the TSE’s time stages can be used and

the TBS/S/UNI-MACH48 time stages can be left in the identity-mapping mode.

The re ason for this is illustrate d in Figure 14. In this configuration there are two

TSE’s. Two TBS’s with four STS-12 ports each are shown. For simplicity of the

diagram, the other 30 TBS’s supported by the TSE’s are not shown. In order to

make the desired connections in this example shown by the dashed lines, the

time stages within the input TBSs must be used to direct the STS-12’s to the

proper TSE for final connection to the desired output TBS. The time stage in the

output TBS is used to direct the STS-12’s to the desired output ports. This can

always be done without making any timeslot rearrangements within the TSE time

stages.

Figure 14: Time-Space-Time Switching

TBS

TSE

TBS

STS-12 #1

STS-12 #2

STS-12 #3

STS-12 #4

STS-12 #1

STS-12 #2

STS-12 #3

STS-12 #4

STS-12 #1

STS-12 #2

STS-12 #3

STS-12 #4

STS-12 #1

STS-12 #2

STS-12 #3

STS-12 #4

In summary, the guidelines fo r which TSI blocks need to be used in which

devices can be broken down by topology as follows:

Single TSE fabrics (<=40Gbps): The time stages in the TSE can be used and

the time stages in the TBS/MACH devices can be bypassed.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 30

Single-stage, multi-TSE fabrics (80Gbps to 160Gbps typically): The time stages

in the TBS/MACH must be used and the time stages in the TSE are bypassed.

Multi-stage TSE fabrics (320Gbps to 1.2Tbps typically): The time stages in the

TSE can be used and the time stages in the TBS/MACH devices can be

bypassed.

7.3 Time-Slot-Interchange

7.3.1 Operation of the TSI

The TBS, TSE, and S/UNI-MACH48 all have STS-1 time-slot interchange

capabilities. To describe how these work, first consider the basic case of a blo ck

with a byte-serial STS-12 stream input and byte-serial STS-12 output. Timeslot

interchange capability in such a device is achieved by storing the incoming data

in a shift register. Once the entire STS-12 stream has been received in this

byte can be written to any output register byte. Shifting out the contents of the

output register forms the outgoing STS-12 stream. This operation is shown in

Figure 15. Note that multicasting in this case is trivial. A single byte in the input

basic operation of the Ingress Time-Switch Element (ITSE) and Egress Time-

Switch Element (ETSE) blocks in the TSE device.

Figure 15: Operation of the Time Slot Interchange

Byte Serial

STS-12 TSIN

12 TSIN

11 TSIN

10 TSIN

09 TSIN

08 TSIN

07 TSIN

06 TSIN

TSIN

03 TSIN

TSIN

08 TSIN

12 TSIN

11 TSIN

10 TSIN

TSIN

08 TSIN

07 TSIN

06 TSIN

01 TSIN

02 TSIN

03 Byte Serial

STS-12

Extend Figure 15 by allowing multiple aligned STS-12 streams at the input and

output and extend the switching such that bytes from any of the input registers

can be switched to any one of the output registers in any of the multiple

locations. For example, the TBS has three copies (working, protect and

auxiliary) of a configuration with four inputs and four outputs. In the ingress

direction (STS-48 parallel to 4xSTS-12 serial), the STS-48 stream is

demultiplexed into four STS-12s and each STS-12 is shifted into three of the

input shift registers (working, protect and auxiliary). This operation is illustrated

in Figure 16. There are three identical blocks in the TBS that implement the

function in this diagram, all sharing the same four Byte Serial input streams

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 31

originating from the TeleCombus. These are the Transmit W o rking Timeslot

Interchange (TWTI), the Transmit Protection Timeslot Interchange (TPTI), and

the T ransmit Auxiliary T i meslot Interchange (TATI).

Figure 16: The TBS Time Slot Interchange Operation

Byte Serial

STS-12 #1 TS-1

12 TS-1

11 TSI-1

10 TS-1

09 TS-1

08 TS-1

07 TS-1

06 TS-1

TS-2

TS-1

03 TS-1

TS-1

08 TS-1

12 TS-2

11 TS-2

09 TS-2

TS-1

09 TS-1

07 TS-1

08 TS-4

03 TS-4

02 TS-4

01 Byte Serial

STS-12 #1

TS-4

09 TS-4

08 TS-4

07 TS-4

06 TS-4

05 TS-1

06 TS-1

05 TS-1

04 TS-2

06 TS-2

05 TS-2

04 Byte Serial

STS-12 #2

TS-2

TS-4

10 TS-2

10 TS-4

12 TS-2

03 TS-2

02 TS-1

03 TS-1

02 TS-1

01 TS-1

11 TS-1

10 TS-4

01 Byte Serial

STS-12 #3

TS-3

12 TS-3

11 TS-3

10 TS-3

09 TS-3

08 TS-3

06 TS-3

05 TS-3

04 TS-3

03 TS-3

02 TS-3

01 Byte Serial

STS-12 #4

Byte Serial

STS-12 #2 TS-2

12 TS-2

11 TSI-2

10 TS-2

09 TS-2

08 TS-2

07 TS-2

06 TS-2

05 TS-2

03 TS-2

TS-2

Byte Serial

STS-12 #3 TS-3

12 TS-3

11 TS-3

10 TS-3

09 TS-3

08 TS-3

07 TS-3

06 TS-3

05 TS-3

03 TS-3

TS-3

Byte Serial

STS-12 #4 TS-4

12 TS-4

11 TS-4

10 TS-4

09 TS-4

08 TS-4

07 TS-4

06 TS-4

05 TS-4

03 TS-4

TS-4

In the egress direction, the TBS must choose from one of three output registers

on a time-slot by time-slot basis to form each STS-12 byte-serial output. This

selection is achieved by programming an enable into the block (working, protect

or auxiliary block) that is to be chosen for th e output (see Figure 17).

Alternatively, an external pin can be used to force the simultaneous selection of

either all of the “working” STS-12s’ timeslots or all of the “protection” STS-12s’

timeslots (see Figure 17). The blocks in the TBS that implement this function are

the Receive Working Timeslot Interchange (RWTI), Receive Protection Timeslot

Interchange (RPTI), and Receive Auxiliary Time slot Interchange (RATI). The

operation of selecting a single output from several candidate STS-12 streams is

illustrated in Figure 17.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 32

Figure 17: Selecting an Output STS-12 Data-stream

12 TS

11 TS

10 TS

09 TS

08 TS

07 TS

06 TS

05 TS

03 TS

Byte Serial

STS-12

111010000000

000100111000

12 TS

11 TS

10 TS

09 TS

08 TS

07 TS

06 TS

05 TS

03 TS

000001000111

12 TS

11 TS

10 TS

09 TS

08 TS

07 TS

06 TS

05 TS

03 TS

Enable

Working

Output

Protection

Output

Auxiliar

Output

Working/Protect ion Select

The S/ UNI-MACH48 has the same capability as the TBS with the exception that

it has eight byte-serial STS-12 links as opposed to twelve. The blocks are

named Input W o rking Timeslot Interchange (IWT I) and Input Protection Timeslot

Interchange (IPTI) in the ingress direction toward the system side of the S/UNI-

MACH48 and Output Working Timeslot Interchange (OWT I) and Output

Protection Timeslot Interchange (OPTI) in the opposite direction.

7.3.2 Programming the TBS

The TBS contains three Timeslot Interchange (TSI) blocks in each direction. The

transmit direction is defined as from the parallel TeleCombus to the serial STS-

12 interfaces. The receive direction is defined as from the serial STS-12

interfaces to the parallel TeleCombus. For programming purposes, the 32-bit

TeleCombus is identified as four individual 8-bit bus segments numbered 1 to 4.

The three TSI blocks are the W o rking, Protection, and Auxiliary respectively

labeled TWTI, TPTI, and TATI in the transmit direction and RWTI, RPTI, and

RATI in the receive direction. Each block controls four outgoing STS-12 serial

streams numbered 1 to 4. Timeslots are numbered 1 to 12. Accesses to these

blocks are via indirect reads and writes.

Any time-slot can be over-written with a programmable idle pattern in the

transmit direction. This is normally done when a connection to the output

timeslot is torn down. The pattern is programmed as a 10-bit code as it is

inserted after 8B/10B encoding of data has been performed internal to the

device. Care must be taken not to insert 8B/10B control characters or invalid

8B/10B codes. The following steps perform the programming an idle code

insertion function:

1) Poll the indirect address register until the BUSY bit is low.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 33

2) W rite ‘11’ to the IDLE[1:0] bits and the desired 8B/10B encoded idle pattern in

the IC[9:0] bits of the indirect data register (1010110001b for encoded all 1’s).

3) W rite the desired output stream number in IWDSEL[1:0], the desired timeslot

number in IWDTSEL[3:0], and the inactive page number in the PAGE bit in

the indirect address register.

4) Perform other system configurations in other devices if desired.

5) Force switchover to the inactive page using the TCMP pin or CMPSEL

6) Update the newly inactive page with the same configuration information.

The change will take effect on the second STS-12 frame boundary after the

inactive page is selected as the active page. Consider this example, to send an

8B/10B encoded idle all ‘1s’ pattern in timeslot 3 of the protection LVDS link #2

with page 1 being active. The following writes would be made after polling BUSY

inactive in the TPTI Indirect Address register: TPTI Indirect Data register =

0x1xxx, TP TI Indi rect A ddress register = 0x4032. After the active page switch

has been performed, the following two writes would occur (simply inverting the

polarity of the PAGE bit in the indirect address register): TPTI Indirect Data

A connection can be set up by writing the source timeslot and input TeleCombus

segment to the indirect data register, f ollowed by writing the target serial STS-12

stream number and timeslot in the indirect address register. As with the idle

code assertion, this should first be written to the inactive configuration page, then

a page switch should be performed, followed by the updating of the newly

inactive page. Consider this example, to set up a connection from timeslot #10

of STS-12 #1 on the parallel TeleCombus to timeslot #5 of working serial STS-12

link #0 with page 0 active. The following writes would be made after polling

BUSY low in the TWTI indirect address register: TWTI Indirect Data register =

0x00a1, TWTI Indirect Address register = 0x4450. After the active page switch

has been performed, the following two writes would occur (simply inverting the

polarity of the PAGE bit in the indirect address register): TWTI Indirect Data

Programming the receive direction of the TBS (from the serial STS-12 streams

toward the parallel TeleCombus) follows the same procedure as for the transmit

direction. The only differences are that

1) there is no idle code insertion available in the receive direction, and

2) the enable bit for one and only one of the three streams (RWTSEN, RPTSEN,

RATSEN for working, prote ction, or au xiliary respe ctively) must be set for any

given timeslot when setting up a connection or cleared to tear down a

connection.

Note that per-timeslot selection between W, P, A links can only be done when

RWSEL_EN register bit = 0. Care must be taken that two timeslots are not

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 34

programmed to the same destination with both having the enable bit set.

Besides programming the enable bits, there is an alternative way to choose

between the working and protection serial links in the receive direction. This

scheme is available when mixing of timeslots between the three links is not

necessary. In this scheme, RWSEL_EN register bit must be set to logic 1 to

enable the external RWSEL pin. The external RWSEL pin is used to select

between the working and protection links. Switchover based on the RWSEL pin

occurs on the J0 boundary.

7.3.3 Programming the TSE

There are 16 Ingress Time Slot Interchange Element (ITSE) and sixteen Egress

Time Slot Interchange Element (ETSE) blocks within the TSE device. Each

controls four serial STS-12 inputs and four STS-12 outputs. Programming these

blocks follows the same procedure as the TSI blocks in the TBS device with the

restriction that timeslots may only be interchanged within an STS-12 stream and

not between the four STS-12 streams controlled by a given block. This

restriction requires that the DINSEL[1:0] and DOUTSEL[1:0] bits in the ITSE

blocks be programmed with the same value when setting up a TSI connection.

Both the ITSE and ETSE functional blocks support idle code insertion via the

same method used in the TBS TWTI, TPTI, and TAT I blocks.

7.3.4 Programming the S/UNI-MACH48

The S/UNI-MACH48 contains two Timeslot Interchange (TSI) blocks in each

direction. The ingress direction is defined as from the serial TeleCombus

towards the UL3/PL3 interface. The egress direction is defined as from the

UL3/PL3 interface toward the serial TeleCombus. The two TSI blocks are the

Working and Protection blocks labeled IWTI and IPTI in the ingress direction and

OWTI and OPTI in the egress direction. These blocks are programmed and

operated similarly to the working and protection TSI blocks in the TBS device.

Exceptions are that the idle code insertion feature is not available. Also note that

the selection between the working and protection timeslots in the ingress

direction is defined by the state of the SER_PRT _SEL bit in the IWT I block when

RWSEL_EN register bit is logic 0. The external RWSEL pin can be used to

select between the working and protection ingress serial links when RWSEL_EN

operation as on the TBS.

7.3.5 Programming optimizations

Each of the programming sequences described above includes the polling of a

“BUSY” bit that indicates when the device is in the process of implementing a

previously requested indirect read or write operation. The setup of a connection

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 35

normally involves the programming of several different functional blocks within a

device or even multiple devices. , Therefore the time spent waiting for BUSY to

clear can be reduced by performing reads and writes in an order such that

successive operations are not performed to a given functional block.

7.4 TSE Space Switch

7.4.1 Operation of the TSE Space Switch

The TSE device has what is called a space switch stage as well as the time

switch stages. A space switch is able to connect any of the inputs to any of the

outputs. Connecting multiple inputs to the same output is not allowed, though

multicasting a single input to multiple outputs is allowed. A space switch can be

conceptualized as a multiplexer at each output that is able to choose from all the

inputs. The TSE has 64 inputs and 64 outputs. It supports twelve settings for

the 64 output multiplexers. These settings are used to allow a unique setting of

the space switch for each of the twelve time periods during the shift-out of the

ingress timeslot output registers. The operation of a simple 2x2-space switch

with three ingress tim eslots is illustrated in the following figure. During each of

the three timeslots there are unique settings programmable for each of the

output multiplexers. Figure 18 shows the intermediate and final results after

each step of the process.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 36

Figure 18: The TSE Space Switch Operation

TS-1

03 TS-1

TS-1

TS-2

03 TS-2

TS-2

TS-1

TS-2

Sp ace Configuration d uring T imeslot #1

TS-1

TS-2

02 TS-1

TS-1

02 TS-2

Sp ace Configuration d uring T imeslot #2

TS-1

TS-2

03 TS-1

TS-2

03 TS-1

TS-1

Sp ace Configuration d uring T imeslot #3

7.4.2 Programming the TSE Space Switch

The space-switch configurations are made using pair of registers to perform

indirect write accesses. As with other CHESS device functional blocks, there are

two pages of configuration memory allowing changes to be made to the inactive

page and then committed on the second STS-1 frame boundary after selection

of the inactive page in a hitless manner. Writing the source port address (from 1

to 64) to the indirect data register, followed by writing the destination port

address and the timeslot in which the connection is to be made in the indirect

address register can set up a connection. This connection information should

first be written to the inactive configuration page, then a page switch should be

performed followed by the updating of the newly inactive page. Consider this

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 37

example, to set up a connection during timeslot #10 from the STS-12 input port

#28 to the STS-12 output port #5 with page 0 active. The following writes would

be made after polling BUSY low in the SSWT Indirect Control Address register:

SSWT Indirect Data register = 0x001c, SSWT Indirect Control Address register =

0x2a05. After the active page switch has been performed, the following two

writes would occur (simply inverting the polarity of the PAGE bit in the indirect

address register): SSWT Indirect Data Register = 0x001c, SSWT Indirect

Control Address register = 0x0a05.

7.4.3 STS-1 Timeslot Mappings in the CHESS set

The following tables provide the STS-1 Timeslot mappings from the line to

TeleCombus Interface for various CHESS devices assuming the TSI is

bypassed. This mapping information is important to properly execute the time

portion of the switching before the data gets to the TSE for instance.

Table 3 SPECTRA-2488 Timeslot Mappings

Spectra-2488 (1x2488 mode)

Line (Time Slots) #1:12 123456789101112

x,y: (Byte x of ST S-12 # y) 1,1 1,2 1,3 1,4 2,1 2,2 2,3 2,4 3,1 3,2 3,3 3,4

Line (Time Slots) #13:24 13 14 15 16 17 18 19 20 21 22 23 24

x,y: (Byte x of ST S-12 # y) 4,1 4,2 4,3 4,4 1,5 1,6 1,7 1,8 2,5 2,6 2,7 2,8

Line (Time Slots) #25:36 25 26 27 28 29 30 31 32 33 34 35 36

x,y: (Byte x of ST S-12 # y) 3,5 3,6 3,7 3,8 4,5 4,6 4,7 4,8 1,9 1,10 1,11 1,12

Line (Time Slots) #37:48 37 38 39 40 41 42 43 44 45 46 47 48

x,y: (Byte x of ST S-12 # y) 2,9 2,10 2,11 2,12 3,9 3,10 3,11 3,12 4,9 4,10 4,11 4,12

Spectra-2488 (1X2488 or 4 x 622 mode) TeleCombus Mappings

TeleCombus #1 12341718192033343536

x,y: (Byte x of ST S-12 # y) 1,1 1,2 1,3 1,4 1,5 1,6 1,7 1,8 1,9 1,10 1,11 1,12

a, b: (Byte a of STS-3 # b) 1,1 2,1 3,1 4,1 1,2 2,2 3,2 4,2 1,3 2,3 3,3 4,3

TeleCombus #2 56782122232437383940

x,y: (Byte x of ST S-12 # y) 2,1 2,2 2,3 2,4 2,5 2,6 2,7 2,8 2,9 2,10 2,11 2,12

TeleCombus #3 91011122526272841424344

x,y: (Byte x of ST S-12 # y) 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 3,10 3,11 3,12

TeleCombus #4 13 14 15 16 29 30 31 32 45 46 47 48

x,y: (Byte x of ST S-12 # y) 4,1 4,2 4,3 4,4 4,5 4,6 4,7 4,8 4,9 4,10 4,11 4,12

Spectra-2488 (4x622 mode)

Line 1 123456789101112

Line 2 131415161718192021222324

Line 3 252627282930313233343536

Line 4 373839404142434445464748

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 38

Table 4 SPECTRA-4x155 Timeslot Mappings

Spectra-4x155

Line1 1 2 3

Line2 4 5 6

Line3 7 8 9

Line4 10 11 12

TeleCombus M apping (SPECTRA 4 x 155)

TeleCombus 147102581136912

a, b: (Byte a of STS-3 # b) 1,1 2,1 3,1 4,1 1,2 2,2 3,2 4,2 1,3 2,3 3,3 4,3

Table 5 S/UNI-MACH48 Timeslot Mappings

S/UNI-MACH-48 TeleCombus Mapping (External Mapping)

TeleCombus #1 123456789101112

x,y: (Byte x of ST S -12 # y) 1,1 1,2 1,3 1,4 1, 5 1,6 1,7 1 , 8 1,9 1,10 1,11 1,12

a, b: (Byte a of STS-3 # b) 1,1 2, 1 3,1 4,1 1,2 2,2 3, 2 4,2 1,3 2, 3 3,3 4,3

TeleCombus #2 13 14 15 16 17 18 19 20 21 22 23 24

x,y: (Byte x of ST S -12 # y) 2,1 2,2 2,3 2,4 2, 5 2,6 2,7 2 , 8 2,9 2,10 2,11 2,12

TeleCombus #3 25 26 27 28 29 30 31 32 33 34 35 36

x,y: (Byte x of ST S -12 # y) 3,1 3,2 3,3 3,4 3, 5 3,6 3,7 3 , 8 3,9 3,10 3,11 3,12

TeleCombus #4 37 38 39 40 41 42 43 44 45 46 47 48

x,y: (Byte x of ST S -12 # y) 4,1 4,2 4,3 4,4 4, 5 4,6 4,7 4 , 8 4,9 4,10 4,11 4,12

Internal Mapping of the S/UNI-MACH-48 for OC48c*

TeleCombus #1 1132537 5172941 9213345

TeleCombus #2 2 14 26 38 6 18 30 42 10 22 34 46

TeleCombus #3 3 15 27 39 7 19 31 43 11 23 35 47

TeleCombus #4 4 16 28 40 8 20 32 44 12 24 36 48

* The S/UNI-MACH-48 uses a di ff erent set of mappings i nternal to the devi ce in OC-48c mode.

For modes other than OC-48c, the external mapping i s used

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 39

8 SYSTEM POWER AND PACKAGE CONSIDERATIONS

8.1 CHESS SET Design Reference

Table 6 DEVICE SUMMARY

Device Power

(max.)*!

Supply Package

Informatio

Interfaces**

Line Side

System Side

TBS PM5310

(Tel eCombus S eri al i zer) 4.2W 1.8 V

3.3 V 352 UBGA

27x27mm

1mm pitch

θjc = 1.8

4x8x77 MHz TeleCombus (TTL)

3 sets of 4 x 777 MHz (LVDS )

TSE PM5372

(40Gb STS-1 Cros s

Connect)

13 W 1.8 V

3.3 V 560 UBGA

40x40mm

1mm pitch

θjc = 1.2

64 x 777 MHz (LV DS )

SPECTRA-2488

PM5315

(SONET/SDH terminating

device for 1xOC-48 or

4xOC-12)

6.6 W 1.8 V

3.3 V 520 SBGA

40x40mm

1.27m m pitch

θjc = 1.0

16 bit x 155 MHz (PECL) in STS -48 mode

or 4x8x77 MHz (TTL) in 4 x STS-12 mode

32 bit x 77 MHz TeleCombus (TTL) or

4x8x77 MHz TeleCombus (TTL)

SPECTRA-622

PM5313

(1xOC-12 SONET/S DH

term i nating device)

4 W 3.3 V 520 SBGA

40x40mm

1.27m m pitch

θjc =1.0

8 bit x 77MHz (TTL) or serial 622 MHz

(PECL)

8 x77 MHz TeleCombus (TTL) or

32 bit x 19.44 MHz TeleCom bus (TTL)

SPECTRA-4x155

PM5316

(4xOC-3 SONET/SDH

term i nating Device )

3.9 W 3.3 V 520 SBGA

40x40mm

1.27m m pitch

θjc =1.0

4x155 MHz (PECL)

8x77 MHz TeleCombus (TTL)

S/UNI-MACH48

PM7390

(OC-48 A TM/POS

term i nating device)

6.5 W 1.8 V

3.3 V 560 UBGA

40x40mm

1mm pitch

θjc = 1.2

2 sets of 4x777 MHz (LVDS ) or

4x8x 77M Hz TeleCombus (TTL)

32 bit x 104 MHz (UL3/PL3)

S/UNI-

ATLAS3200

PM7325

(OC-48 S/UNI AT M Layer

Solution)

4.4W 1.5V

2.5V

3.3V

576 TBGA

40x40mm

1.27mm

θjc =TBD

32 bit x 104 MHz UL3 or 32 bit x 104 MHz

PL3

32 bit x 104 MHz UL3 or 32 bit x 104 MHz

PL3

* Maximum power (high voltage, full functionality). Typical Power ~ 90% of max.

! Consult datasheet for most up to date power numbers

**Exact rates are: 77 = 77.76, 777=777.6, 155=155.52, 622=622.08

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 40

8.2 Analog and Digital Power By Rail

Table 7 Analog and Digital Power By Rail*

Device TSE2TBS S/UNI-

MACH484SPECTRA-

24885SPECTRA-

4x155

Current I(AVDL@1.8V) 965 225 210 505.8 -

Typ. I(AVDH@3.3V) 1180 230 160 307.2 124

(mA) I(VDDITotal@1.8

V) 3 3655 1310 2450 1657.7 -

I(VDDO@3.3V) 0.001 135 240 366.9 938

Power P(AVDL@1.8V) 1737 405 378 910.5 -

Typ. P(AVDH@3.3V) 3894 759 528 1013.6 409

(mW) P(VDDITotal@1.8

V)6574 2538 4410 2983.1 -

P(VDDO@3.3V) 0.005 446 792 1210.9 3096

Total Power 12210.0 3968 6108 6188.8 3505

Power P(AVDL@1.89V) 1824 425 397 956 -

Max. P(AVDH@3.63V) 4283 835 581 1115 450

(mW) P(VDDITotal@1.8

9V)6908 2476 4631 3133 -

P(VDDO@3.63V

)0.005 490 871.2 1332 3405

Total Power 13015 4226 6479 6537 3855

* Consult datas heet for most up to dat e power num bers

1. 1.8V supplies have a ± 5% toleranc e; 3.3 V s uppl i es have a ± 10% tol erance

2. TSE I(VDDO) current extremel y l ow - singl e (non-CBI/non-JTAG ) di gi tal output w/ low switching f requency

3. I(VDDI Total@1. 8V) = I(VDDI A nal og@1.8V) + I(V DDIDigital@1.8V)

4. Assuming both s eri al and parallel TeleCombus is active

5. Assuming only 21 RX PECLs, and 18 TX PECLs

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 41

8.3 Powe r Sequencing Rules for 1.8V and 3.3V Supplies:

This section describes the power-sequencing for the device power pins.

Table 8 Power Sequence Rules

Technology Pow e r Sequencing Relevant PMC-Sierra -

Part

First VDDO, QAVDPO

Second VDD, AVD

0.35

µm

Third Data to I/Os

SPECTRA-622

SPECTRA-4x155

First VDDO no later than AVDH

and CSU_AVDH

Second*VDDI, A VDL

0.18

µm

Third Data to I/Os

SPECTRA-2488,

S/UNI-MACH48,

TBS, TSE

First VDD33

Second#VDD25, VDDQ25

0.18

µmThird VDD15, VDDQ15

S/UNI-ATLAS3200

* The 3.3 V and 1.8 V s uppl i es can come up at t he same t i me as l ong as there is never more than 0. 3V di ff erence between

the 1.8V supply and the 3. 3V supply. There are no power up ramp rate restric tions.

# During power-up, VDD33 must be brought up bef ore or at the same time as VDD25 and VDDQ25, which must be brought

up before or at the same tim e as VDD15 and VDDQ15. The VDD33 and VDD25 power m ust be applied bef ore i nput pins are

driven or the input current per pin be li mited to less than the maximum DC i nput current specif i cation (20mA). P ower down

the device i n the reverse sequence.

8.4 Estimating System Power

The following formulas are useful for estimating device power when powering

down certain modes of operation.

TBS Power (W) = Core Power + nRX x Receive Power + nTX x Transmit

Power = 3.0 + 0.036nRX + 0.064nTX

Definitions & Assumptions:

• Core power utilization is independent of the number of link used and hence is

constant

• The TBS has 3 groups of 4 RX/TX LVDS 777 MHz links (working, protect,

auxiliary)

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 42

• n

RX and nTX can have values of 1 through 12 dependent on the number of

links

• The values of nRX and nTX can be chosen independently

TSE Power (W) = Core Power + α x TX CSU Power + nRX x Receive Power +

nTX x Transmit Power

= 4.8 + 0.45α + 0.036nRX + 0.064nTX

Definitions & Assumptions:

• Core power utilization is independent of the number of link used and hence is

constant

• The TSE has 64 transmit and receive LVDS 777 MHz links

• n

RX and nTX can have values of 1 through 64 dependent on the number of

links

• The values of nRX and nTX can be chosen independently

• There is one transmitter CSU (clock synthesis and reference) block for each

16 transmit interfaces

• The TSE has 4 CSU blocks; therefore α is integer multiple between 1 and

• Note that it is possible, by using only 4 TX LVDS links, that all CSUs will be

powered if each of the four TX link’s clock source is from each of these 4

different CSUs. That is if the LVDS TX links number 1, 17, 33, 49 are used

then all 4 CSUs will need to be powered. However if TX links 1 through 16

are used, then only one CSU is powered.

S/UNI-MA CH48 Po wer (W) = Core Power + θ x Parallel Bus + β x DS3 + φ

(Serial B us Power1)

= 4.7 + 0.35θ + 0.2β + φ (0.45 + 0.036nRX + 0.064nTX)

Definitions & Assumptions:

• Core power utilization is independent of the number of link used and hence is

constant

• ATM and packet mode are always powered

• The DS3 block can be powered down; thus β can have a value of 1 or 0

• The parallel bus can be powered down; thus θ can have a value of 1 or 0

• The serial LVDS can be powered down; thus φ can have a value of 1 or 0

• Even with the serial bus powered, individuals links can be powered down

• The serial LVDS bus has 8 TX/RX links (4 working, 4 protect)

• n

RX and nTX can have values of 1 through 8 dependent on the number

of lin ks

1 Seria l Bus Power (W) = TX CSU Power + nRX x Receiver Power + nTX x Transmitter Power

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 43

• The values of nRX and nTX can be chosen independently

8.5 Maximum CHESS SET System Power

The following power calculations are based on the TSE Fabric Cross Connect as

in Figure 7b, using SPECTRA-2488, TBS with Auxiliary Links Powered Down,

Spare TSE and S/UNI-MACH48 devices for Layer 2. All devices connected to

the TSE are operated bi-directionally. These powers represents maximum power

consumption and are based on sections 8.2 and 8.3 above. The shaded portion

of the table represents systems that could be more power efficiently produced

with smaller fabrics. For example, a 160 Gb fabric with 25% link utilization is

better constructed by using a 40 Gb fabric with 100% link utilization. In this case,

the 40 Gb fabric power consumption is 275.5 W versus 313.9 W for the larger

fabric. However this 14% power tradeoff may be acceptable since the system is

future proofed for a four times seamless fabric expansion.

Table 9 System Power versus Link Utilization

System Power (W) Pow er

(100% Link

Utilization)

Powe r (75%

Link

Utilization)

Powe r (50%

Link

Utilization)

Powe r (25%

Link

Utilization)

40 Gbps Cross Connect 291.6 221.9 152.2 82.5

80 Gbps Cross Connect 583.2 443.8 304.4 165.0

160 Gbps Cross Connect 1166.4 887.5 608.7 330.0

320 Gbps Cross Connect 2749.3 2138.2 1527.9 917.5

8.6 CHESS Thermal Information

The following section provides information on the PMC-Sierra CHESS devices to

assist in thermal calculations and choosing appropriate heat sinks. The CHESS

devices include the SPECTRA-2488, SPECTRA-4x155, S/UNI-MACH48, TBS

and TSE. Please refer to the specific datasheets on each of these devices for

further information. This section is strictly informative.

8.6.1 Heat Sinking:

The amount of heat that can be removed with a specific temperature difference

between the heat sink and the air determines the heat sinks performance. It is

most often characterized by thermal resistance, i.e., the lower the thermal

resistance, the better the performance. Heat-sink performance can be improved

by increasing the physical size of the heat sink (i.e., changing the surface area)

or by moving more air across the sink (i.e., changing from natural convection to

forced convection). As a general rule, to reduce the thermal resistance by 50

percent, the heat-sink volume must be approximately quadrupled.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 44

The function of a heat-sink is to protect the semiconductor from the heat it

produces as a by-product of its normal operation. Depending on environmental

conditions, failure to remove this heat may cause the semiconductor device to

exceed its safe operating temperature. In this circumstance, the

semiconductor's performance, life, and reliability may be reduced. The objective,

then, is to hold the junction temperature below the temperature allowed by the

device reliability criteria.

The CHESS devices are guaranteed to work at 120°C junction temperature

(transient condition). For long-term reliability, it is recommended to keep the

junction temperature below 105°C.

The selection of a heat sink requires knowledge of:

(1) The device configuration (package size, orientation);

(2) The power dissipated by the device;

(3) The maximum allowable device junction temperature

(4) The available volume of space to be occupied

(5) Ambient conditions (temperature, air velocity).

A heat-sink mounted on a semiconductor has three major thermal resistances.

The total resistance θtot, is the sum of these individual resistances:

θtot = θjc + θcs + θsa

Thermal resistance from junction to case. θjc. Expressed in °C/W, this

resistance is a function of design and manufacturing methods and is specified in

the device's datasheet. Thermal resistance from case to sink. θcs. Expressed in

°C/W, this is a variable that can be reduced by applying a thermal grease or

paste that decreases the high thermal impedance of the air gap between the

case and sink.

Thermal resistance from sink to ambient. θsa. Expressed in °C/W, this

resistance is used in selecting a heat-sink. The smaller the value is, the more

power the device can dissipate without exceeding its junction temperature limit.

Heat sinks specifications should provide θsa as a function of airflow. The

thermal schematic is represented in Figure 19.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 45

Figure 19: Thermal Schematic of Heat Dissipation

Maximum semi conductor junction t emp. (Tj)

θjc

Sink Te m p . (Ts)

θcs

Case Temp. (Tc)

θsa

mbient air temp. (Ta)

In most applications, values for all these parameters are known, and can be

used as the selection criteria for the heat sink (i.e. to find a heat sink, with

corresponding air flow, that provides an acceptable θsa). The follow equation

can be used to calculate the required θsa, and is applicable for both natural and

forced convection cooling.

Q * θsa = Tsa = (Tj - Ta) - Q (θjc + θ cs)

where

Q = power dissipated, watts

Tj = junction temperature, °C

Ta = ambient air temperature, °C

θjc = thermal resistance from junction to semiconductor case, °C/W

θcs = thermal resistance from case to heat sink, °C/W

θsa = thermal resistance from surface of heat sink to ambient air, °C/W

Tsa = temperature difference between sink and air

8.6.2 Example, using the S/UNI-MACH48:

Total Power Consumption = 6.5W.

The power consumption assume 3.6V (high voltage on 3.3V rail) and 1.89V (high

voltage on 1.8V rail), fully loaded and worst case process. A tighter power supply

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

Proprietary and Confidential to PMC-Sierra Inc., and for its Customers’ Internal Use 46

(i.e. VDD max = 3.4 5, 1.9V ) will allow you to de-ra te the power consumptio n by

the following ratio: V2(new) / V2(VDD max).

Q = 6.5W

Tj = 105°C (fixed; for long term reliability)

Ta = 70°C (variable)

θjc = 1°C/W (fixed, see Thermal Information in Datasheet)

θcs = 0.09°C/W (variable, depending of thermal joint compound)

Q * θsa = Tsa = (Tj - Ta) - Q (θjc + θ cs)

θsa = [(105°C - 70°C) – 6.5W (1°C/W + .09°C/W)] / 6.5 W

= 4.3 °C/W

This is the largest value of θsa that can be used in this example. Heat sinks

providing smaller values of θsa are also acceptable. With this value, you can

then proceed to the data presented for various heat sinks and locate one that will

provide this performance.

8.6.3 Chart of θ

θθ

θja versus Air Flow for CHESS Chipset

Table 10 θ

θθ

θja versus Forced Air

Package Part (deg C/Watt) Conv 100.0 200.0 300.0 400.0 500.0

PM5310 TBS Dense Board 24.9 22.9 21.5 20.7 20.3 20.2

JEDEC Board 16.0 15.2 14.7 14.4 14.2 14.0

PM5372 TSE Dense Board 11.1 9 .1 7.7 6.9 6.5 6.4

JEDEC Board 7.2 6.4 6.0 5.7 5.5 5.2

PM7390 S/UNI-MACH48 Dense Board 12.2 10.2 8.8 7.9 7.5 7.5

JEDEC Board 7.7 6.9 6.5 6.2 6.0 5.7

PM5315 SPECTRA-2488 Dense Board 14.4 12.4 11.0 10.2 9.8 9.7

JEDEC Board 8.1 7.3 6.8 6.5 6.3 6.1

PM5316 SPECTRA-4x155 Dense Board 14.2 12.2 10.8 9.9 9.5 9.5

JEDEC Board 8.3 7.5 7.0 6.7 6.5 6.3

520 SBG

Forced Air

(

Linear Feet

er Minute

)

352 UBGA

560 UBGA

560 UBGA package is preliminary. θja values subject to change.

PRELIMINARY

CHESS CHIPSET CHESS

APPLICATION NOTE

PMC-1991797 ISSUE 2 PM5372 CHESS USER’S GUIDE

None of the information contained in this document constitutes an express or implied warranty by PMC-Sierra, Inc. as to the sufficiency, fitness or

suitability for a particular purpose of any such information or the fitness, or suitability for a particular purpose, merchantability , performan c e, compatib ility

with other parts or systems, of any of the products of PMC-Sierra, Inc., or any portion thereof, referred to in this document. PMC-Sierra, Inc. expressly

disclaims all representations and warranties of any kind regarding the contents or use of the information, including, but not limited to, express and

implied warranties of accuracy , completeness, merchantability, fitness for a particular use, or non-infringement.

In no event will PMC-Sierra, In c. be liable for any direct, indirect, special, incidental or consequential damages, including, but not limited to, lost profits,

lost business or lost data resulting from any use of or reliance upon the information, whether or not PMC-Sierra, Inc. has been advised o f th e possibility

of such damage.

PMC-1991797 (P2)

PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE

CONTACTING PMC-SIERRA, INC.

PMC-Sierra, Inc.

105-8555 Baxter Place Burnaby, BC

Canada V5A 4V7

Tel: (604) 415-6000

Fax: (604) 415-6200

Document Information: document@pmc-sierra.com

Corporate Information: info@pmc-sierra.com

Application Information: apps@pmc-sierra.com

(604) 415-4533

W eb Site: http://www.pmc-sierra.com