ARM, the ARM Powered logo, Thu m b and StrongARM are registered t rademarks of ARM Limited.
The ARM logo, AMBA, Angel, ARMulator, EmbeddedICE, ModelGen, Multi-ICE, ARM7TDMI,
ARM7TDMI-S, ARM9TDMI, PrimeCell, and STRONG are trademarks of ARM Limited.
All other products or services mentioned herein may be trademarks of their respective owners.
Neither the whole nor any part of the information contained in, or the product described in, this document may
be adapted or reproduced in any material form except with the prior written permission of the copyright
holder.
The product describe d in this document is subject to co nt inuous developments an d im provements. All
particulars of the produc t and its use contained in this document are given by ARM Limit ed in good faith.
However, all warrantie s im plied or expressed, including but not limited to impli ed warranties or
merchantability, or fitness for purpose, are excluded.
This document is intended only to assist the reader in the use of the product. ARM Limited shall not be liable
for any loss or damage arisi ng from the use of any informati on in this document, or any error or om ission in
such information, or any incorrect use of the product.
Document confidentiality status
This document is Open Access. T his do cument has no restriction on distribution.
Product status
The information in this documents is Final (information on a developed product).
This section contains electromagnetic conformity (EMC) notices.
Federal Communications Commission Notice
NOTE: This equipment has been tested and found to comply with the limits for a class
A digital device, pursuant to part 15 of the FCC rules. These limits are designed to
provide reasonable protection against harmful interference when the equipment is
operated in a commercial environment. This equipment generates, uses, and can radiate
radio frequency energy and, if not installed and used in accordance with the instruction
manual, may cause harmful interference to radio communications. Operation of this
equipment in a residential area is likely to cause harmful interference in which case the
user will be required to correct the interference at his own expense.
CE Declaration of Conformity
This equipment has been tested according to ISE/IEC Guide 22 and EN 45014. It
conforms to the following product EMC specifications:
The product herewith complies with the requirements of EMC Directive 89/336/EEC
as amended.
This document describes how to set up and use the ARM Integrator/CM940T core
module.
Intended audience
This document has been written for experienced hardware and software developers to
aid the development of ARM-based products using the ARM Integrator/CM940T as
part of a development system.
Organization
This document is organized into the following chapters:
Chapter 1Introduction
Chapter 2Getting Started
Chapter 3Hardware Description
Read this chapter for an introduction to the core module.
Read this chapter for a description of how to set up and start using the
core module.
Read this chapter for a descr iption of the header’s hardwar e architecture,
including clocks, resets, and debug.
viii
Chapter 4Programmer’s Reference
Read this chapter for a description of the header memory map and
registers.
Appendix A Signal Descript i ons
Refer to this appendix for a description of t he sign al s on the HDRA and
HDRB connectors.
Appendix B Specifica tions
Refer to this appendix for electrical, timing, and mechanical
specifications.
The Integrator/CM940T core module provides you with the b a sis of a flexible
development system which can be used in a number of differ ent ways. Wi th power and
a simple connection to a Multi-ICE debugger, the core module provides a basic
development system. By mounting th e core mod ule onto a mo therboard , you can build
a realistic emulation of the system being developed. Through-board connectors allow
up to four core modules to be stacked on one motherboard.
The core module can be used in the following ways:
•as a standalone development system
•mounted onto an ARM Integrator/SP develo pm ent motherboard
•mounted onto an ARM Integrator/AP development motherboard
•integrated into a third-party development or ASIC prototyping system.
Figure 1-1 on page 1-3 shows the layout of the ARM Integrator/CM940T.
The major components on the core module are as follows:
•ARM940T microprocessor core
•core module FPGA which implements:
—SDRAM controller
—system bus bridge
—reset controller
—interrupt controller
—status, con f iguration, and interrupt registers.
•volatile memory comprising:
—up to 256MB of SDRAM (optional) via DIMM socket
—256KB SSRAM.
•SSRAM controller
•clock generator
•system bus connectors
•Multi-ICE debug connector.
1.2.1System architecture
1-4
Figure 1-2 illustrates the architecture of the core module.
The FPGA provides system control functions for the core module, enabling it to operate
as a standalone development system or attached to a motherboard. These functions are
outlined in this section and descri bed in detail in Chapter 3 Hardware Description.
SDRAM controller
The SDRAM controller is implemented within the FPGA. This provides support for
Dual In-line Memory Modules (DIMMs) with a capacity of between 16 and 256MB.
See SDRAM controller on page 3-6.
Reset controller
The reset controller initializes the core and allows the core module to be reset from five
sources:
•reset button
•motherboard
•other core modules
•Multi-ICE
•software.
Introduction
ARM DUI 0125A
For information about the reset controller, see Reset controller on page 3-8.
System bus bridge
The system bus bridge provides an interface between the memory bus on the core
module and the system bus on a motherboard. It allows the local processor access to
interface resources on the motherboard and to the SDRAM on other core modules. It
also allows access to the local SDRAM from the PCI bridge on the motherboard and
processors on other core modules (see System bus bridge on page 3-11).
The status and configuration space contains status and configuration registers for the
core module. These provide the following information and control:
•type of processor and whether it has a cache, MMU, or protection unit
•the position of the core module in a multi-module stack
•SDRAM size, address configuration, and CAS latency setup
•core module oscillator setup
•interrupt control for the processor debug communications channel.
The status and control registers can only be accessed by the local processor. For more
information about the status and control registers see Chapter 4 Programmer’ s
Reference.
1.2.3Volatile memory
The volatile memory system includes an SSRAM device, and a plug-in SDRAM
memory module (referred to as local SDRAM when it is on the same core module as
the processor). These areas of memory are closely coupled to the processor core to
ensure high performance. Th e cor e mo dul e u ses s eparate memory and system buses to
avoid memory access performance being degraded by bus loading.
The SDRAM controller is implemented within the core module controller FPGA and a
separate SSRAM controller is implemented with a Programmable Logic Device (PLD).
The SDRAM can be accessed by the local processor, by processors on other core
modules, and by other system bus masters.
The SSRAM can only be accessed by the local processor.
1.2.4Clock generator
The core module uses two on-board clocks:
•CPU clock up to 160MHz
•memory bus clock up to 66MHz.
These clocks are supplied by two clock generator chips. Their frequencies are selected
via the oscillator control register (CM_OSC) within the FPGA. A reference clock is
supplied to the two clock generators and to the FPGA (see Clock generators on
page 3-17). The memory bus and system bus are asynchro nous. This allows each bus to
be run at the speed of its slowest device without compromising the performance of othe r
buses in the system.
The Multi-ICE connector enables JTAG hardware debugging equipment, such as
Multi-ICE, to be connected to the core module. It is possible to bo th drive and sense the
system-reset line (nSRST), and to drive JTAG reset (nTRST) to the core from the
Multi-ICE connector. See Multi-ICE su pport on page 3-21.
JTAG test equipment supplied by other vendors may also be used.
The core module provides one link and four surface-mounted LEDs. These are
illustrated in Figure 1-3.
CONFIGMISCFPGA OKPOWERCFGLED
1.3.1CONFIG link
1-8
Figure 1-3 Links and indicators
The core module has only one link, marked CONFIG. This is left open during normal
operation. It is only fitted when downloading new FPGA and PLD configuration
information.
The core module provides two ground and five signal test points as an aid to debug.
These are illustrated in Figure 1-4.
Voltage
regulator
TP1
TP3
TP5
TP4
TP6
TP7
1-10
TP2
Figure 1-4 Test points
The functions of the test points are summarized in Table 1-2.
Table 1-2 Test point functions
T est pointNameFunction
TP1GNDGround
TP2GNDGround
TP3FCLKOUTClock output from the ARM940T microprocessor core
TP4LBCLKLocal memory bus clock
TP5Core clock Clock suppl ied to the ARM940T microprocessor core
TP6VDD940TOutput from voltage regulator
TP7ADJADJ pin of voltage regulator
This section contains safety information and advice on how to avoid damage to the core
module.
1.5.1Ensuring safety
To avoid a safety hazard, only Safety Extra Low Voltage (SELV) equipment should be
connected to the JTAG interface.
1.5.2Preventing damage
The core module is intended for use within a laboratory or engineering development
environment. It is supplied without an encl osure which leaves the board sensitive to
electrostatic discharges and allows electromagnetic emissions.
To avoid damage to the board you should observe the following precautions.
Introduction
Warning
Caution
ARM DUI 0125A
•Never subject the board to high electrostatic potentials.
•Always wear a grounding strap when handling the board.
•Only hold the board by the edges.
•Avoid touching the component pins or any other metallic element.
•Do not use the board near equipment which could be:
—sensitive to electrom a gnetic emissions (such as medical equipment)
—a transmitter of electromagnetic em issions.
To set up the core module as a standalone development system:
1.Optionally, fit an SDRAM DIMM.
2.Supply power.
3.Connect Multi-ICE.
2.1.1Fitting an SDRAM DIMM
You should fit the following type of SDRAM module:
•PC66- or PC100-compliant 168pin DIMM
•unbuffered
•16MB, 32MB, 64MB, 128MB or 256MB.
To install an SDRAM DIMM:
1.Ensure that the core module is powered down.
2.Op en the SDRAM retaining latches outwards.
3.Press the SDRAM module into the edge connector until the retaining latches
click into place.
Note
The DIMM edge connector has polarizing notches to ensure that it is correctly oriented
in the socket.
2.1.2Using the core module without SDRAM
The core module can be operated without SDRAM because it has 256KB of SSRAM
permanently fitted. However, in order to operate the core module with ARM deb uggers,
you must change the internal variable
0x00080000 (= 512KB) to 0x00040000 (= 256KB) before running programs that are
linked with the standard libraries.
For further information about ARM debugger internal variables, refer to the Software Development Toolkit Reference Guide.
When using the core modul e as a s tandal on e de vel opm ent s yst em, y ou sho ul d conn ect
a bench power supply with 3.3V and 5V outputs to the power connector, as illustrated
in Figure 2-1.
Getting Started
5V
3.3V
GND
5V
3V3
GND
Figure 2-1 Power connector
Note
This power connection is not required when the core module is fitted to a motherboard.
When you are using the core module as a standalone system, Multi-ICE debugging
equipment can be used to download programs. The Multi-ICE setup for a standalone
core module is shown in Figure 2-2.
Multi
Multi-ICE
server/debugger
Parallel
cable
ICE
2-4
Multi-ICE unit
Power supply
Core module
Figure 2-2 Multi-ICE connection to a core module
Caution
Because the core module does not provide non-volatile memory, programs are lost
when the power is removed.
Multi-ICE can also be used when a core module is attached to a motherboard. If more
than one core module is attached, then the Multi-ICE unit must b e connected to the
module at the top of the stack. The Multi-ICE server and the debugger can be on one
computer or on two networked co mputers.
2.2Attaching the ARM Integrator/CM940T to a motherboard
Attach the core module onto a motherboard (for example, the ARM Integrator/SP) by
engaging the connectors HDRA and HDRB on the bottom of the core module with the
corresponding connectors on the top of the motherboard. The lower side of the core
module has sockets and the upper side of the core module has plugs to allow core
modules to be mounted on t op of one anot her. A maximum of f our core module s can be
stacked on a motherboard.
Figure 2-3 illustrates an example development system with four core modules attached
to an ARM Integrator/SP motherboard.
The ID of the core module is configured automatically by the connectors (there are no
links to set) and depends on its p osition in the stack:
•core module 0 is installed first
•core module 1 is installed next, and cannot be fitted without core module 0
•core module 2 is installed next, and cannot be fitted without core module 1
•core module 3 is installed next, and cannot be fitted without core module 2.
The ID of the core module also defines the ID of the microprocessor it carries and the
system bus address of its SDRAM. The position of a core module in the stack can be
read from the CM_STAT register. See CM_STAT (0x10000010) on page 4-12.
2.2.2Powering the assembled Integrator development system
Power the assembled Integrator development system by:
•connecting a bench power supply to the motherboard
•installing the motherboard in a card cage or an AT X-type PC case, depending on
type.
For further information, refer to the user guide for the motherboard you are using.
The ARM940T cached processor macrocell is a member of the ARM9 Thumb family
of high-performance 32-bit system-on-a-chip processors. It provides the following:
•ARM9TDMI RISC integer CPU
•4KB instruction and data caches
•write buffer
•protection unit
•AMBA ASB bus interface.
The ARM940T processor employs a Harvard cache architecture, and so has separate
4KB instruction and 4KB data caches. Each cache has a 4-word line length.
The protection unit allows eight regions of memory to be defined, each with individual
cache and write buffer configurations and access permissions.
The cache system is software-configurable to provide highest average performance or
to meet the needs of real-time systems. Software configurable options include:
•random or round-robin cache line replacement algorithm
•write-through or writeback cache operation
•cache locking with granularity
1
/64 th of cache size.
3-2
The caches and write buffers improve CPU performance and minimize accesses to
off-chip memory, thus reducing overall system power consumption. The ARM940T
includes support for coprocessors, allowing a floating point unit or other
application-specific hardware acceleration to be added.
The ARM9TDMI core executes both the 32-bit ARM and 16-bit Thumb instruction
sets, allowing you to trade between high performance and high code density. It is
binary-compatible with ARM7TDMI, ARM10T DMI, and StrongARM pr ocessors, and
is supported by a wide range of tools, operating systems, and application software.
The ARM940T also features a TrackingICE mode which allows an approach similar to
a conventional ICE mode of operation.
For more information about the ARM940T, refer to the ARM940T Technical Refe rence Manual.
The SSRAM controller is implemented in a Xilinx 9572 PLD which enables the
SSRAM to achieve single-cycle operation. In addition to controlling accesses to the
SSRAM, the controller generates the processor response signals (BWAIT, BERROR,
BLAST) for all accesses to:
The core module FPGA contains five main functional blocks:
•SDRAM controller on page 3-6
•Reset controller on page 3-8
•System bus bridge on page 3-11
•Core module registers on page 4-7
•Debug interrupt controller, see Debug communications interrupts on page 3-27.
The FPGA provides sufficient functionality for the core module to operate as a
standalone development system, although with limited capabilities. Sy stem bus
arbitration, system interrupt control, and input/output resources are provided by the
system controller FPGA on the motherboard. See the user guide for your mother boar d
for further information.
Figure 3-1 illustrates the function of the core module FPGA and shows how it connects
to the other devices in the system.
At power-up the FPGA loads its configuration data from a flash memory device.
Parallel data from the flash is serialized by the Programmable Logic Device (PLD) into
the configuration inputs of the FPGA. Figure 3-2 shows the FPGA configuration
mechanism.
A[18:0]
D[7:0]
OE
WE
CS
FPGA
configuration
ROM
(flash)
FPGA
Multi-ICE
DIN
CCLK
DONE
PLD
Figure 3-2 FPGA configuration
Multi-ICE can be used to reprogram the PLD, FPGA, and flash when the core module
is placed in configuration mode. See Multi-ICE support on page 3-21.
The core module provides support for a single 16, 32, 64, 128, or 256MB SDRAM
DIMM.
3.4.1SDRAM operating mode
The operating mode of the SDRAM devices is controlled with the mode set register
within each SDRAM. These registers are set immediately after power-up to specify:
•a burst size of four for both reads and writes
•Column Address Strobe (CAS) latency of 2 cycles.
The CAS latency and memory size can be reprogrammed via the SDRAM control
register (CM_SDRAM) at address 0x10000020. See CM_SDRAM (0x10000020) on
page 4-14.
Note
Before the SDRAM is used it is necessary to read the SPD memory and program the
CM_SDRAM register with the para meters indicated in Table 4-8 on pag e 4-14. If these
values are not correctly set then SDRAM accesses may be slow or unreliable.
3.4.2Access arbitration
The SDRAM controller provides two ports to support reads and writes by the local
processor core and by masters on the motherboard. The SDRAM controller uses an
alternating priority scheme to ensure that the processor core and motherboard have
equal access (see System bus bridge on page 3-11).
3.4.3Serial presence detect
JEDEC-compliant SDRAM DIMMs incorporate a Serial Presence Detect (SPD)
feature. This comprises a 2048-bit serial EEPROM located on the DIMM with the first
128 bytes programmed by the DIMM manufacturer to identify the following:
•module type
•memory organization
•timing parameters.
The EEPROM clock (SCL) operates at 93.75kHz (24MHz divided by 256). The t ransfer
rate for read accesses to the EEPROM is 100kbit/s maximum. The data is read out
serially 8 bits at a time, preceded by a start bit and followed by a stop bit. This makes
reading the EEPROM a very slow process because it takes approximately 27ms to read
all 256 bytes. However, during power-up the content s of the EEPROM are copied into
a 64 x 32-bit area of memory (CM_SPD) within the SDRAM controller. The SPD flag
is set in the SDRAM control register (CM_SDRAM) when the SPD data is available.
This copy can be randomly accessed at 0x10000100 to 0x100001FC (see CM_SPD (0x10000100 to 0x100001FC) on page 4-16).
Write accesses to the SPD EEPROM are not supported.
The core module FPGA incorporates a reset controller which enables the core module
to be reset as a standalone u nit or as part of an Integrator de velopment syst em. The core
module can be reset from five sources:
•reset button
•motherboard
•other core modules
•Multi-ICE
•software.
Figure 3-3 shows the architecture of the reset controller.
BnRES_MProcessor r esetOut putThe BnRES_M signal is used to reset the processor core. It is
generated from nSRST LOW when the core module is used
standalone, or n SYSRS T LOW when the core module is
attached t o a motherboard.
It is asserted as soon as the appropriate input becomes active.
It is deasserted synchronously from the falling edge of the
processor bus clock.
nDONEFPGA config uredInputThe nDONE signal is an inversion of the open collector
signal FPGADONE which is generated by all FPGAs when
they have completed their configuration. The FPGADONE
signal is routed round the system through the HDRB
connectors to the inputs of all other FPGAs in the system. The
signal nSRST is held asserted until nDONE is driven LOW.
nMBDETMotherboard detectInputThe nMBDET signal is pulled LOW when the core module is
attached to a motherboard and HIGH when the core module is
used standalone.
When MBDET is LOW, nSYSRST is used to generate the BnRES_M signal.
When nMBDET is HIGH, nSR ST is used to genera te the
BnRES_M signal.
PBRSTPush-button resetInputThe PBRST signal is generat ed by pressing the reset button.
nSRSTSystem reset BidirectionalThe nSRST open collector output signal is driven LOW by
the core module FPGA when the signal PBRST or software
reset (SWRST) is asserted.
As an input, nSRST can be driven LOW by Multi-ICE.
If there is no motherboard present, the nSRST signal is
synchronized to the processor bus clock to generate the
BnRES-M signal.
nSYSRSTSystem resetInputThe nSYSRST signal is generated by the system controller
FPGA on the motherboard. It is used to generate the
BnRES_M signal when the core module is attach ed to a
motherboard. It is selected by the motherboard detect signal
The core module FPGA provides a software reset which can be trigg ered by writi ng to
the reset bit in the CM_CTRL register. This generates the internal reset signal SWRST
which generates nSRST and resets the whole system (see C M_CTRL ( 0x1000000C) on
page 4-11).
Write transactions from the processor to the system b us normally complete on the local
memory bus in a single cycle. The data, address, and control information associated
with the transfer are posted into FIFO, and the transfer on the system bus occurs some
time later when that bus is available. This means that system bus error responses to
write transfers are not reported back to the pro cessor as data aborts. If the FIFO is full,
the processor receives a wait response until space becomes available.
Processor reads
The data routing for processor reads from the system bus is illu strated in Figure 3-5.
Processor
core
SDRAM
controller
FIFO
FIFO
Motherboard
Figure 3-5 Processor reads from the system bus
SDRAM
For reads from the system bus, the address and control information also pass through
the FIFO. The returned data from the system bus bypasses the FIFO.
The order of processor transactions is preserved on the system bus. Any previously
posted writes are drained from the FIFO (that is, completed on the system bus) before
the read transfer is performed. The processor receives a wait response until the read
transfer has completed on the system bus, when it receives the data an d any associated
bus error response from the system bus. For information about SDRAM addresses, see
SDRAM accesses on page 4-4.
The second FIFO supports read and write accesses by system bus masters on the
motherboard and other core modules to the local core module memory.
System bus writes
The data routing for system bus writes to SDRAM is illustrated in Figu re 3-6.
Hardware Description
Processor
core
SDRAM
controller
FIFO
FIFO
Motherboard
SDRAM
Figure 3-6 System bus writes to SDRAM
Write transactions from the system bus to the SDRAM normally complete in a single
cycle on the system bus. The data, address, and control information associated with the
transfer are posted into the FIFO, and the transfer into the SDRAM completes when th e
SDRAM is available. If the FIFO is full, then the system bus master receives a retract
response indicating that the arbiter may grant the bu s to another master and that this
transaction must be retried later.
The data routing for system bus reads from SDRAM is illustrated in Figure 3-7.
Processor
core
For system bus reads, the address and control information also pass through the FIFO,
but the returned data from the SDRAM bypasses the FIFO.
The order of transactions on the system bus and the memory bus is preserved. Any
previously posted write transactions are dr ained from the FIFO (that is, writes to
SDRAM are completed) before the read transfer is performed.
3.6.3Multiprocessor
The two FIFOs operate independently, as described above, and can be accessed at the
same time. This makes it possible for a local processor to read local SDRAM via the
system bus (through both FIFOs). This feature can be used to support multiprocessor
systems that share data in SDRAM because the processors can all access the same
DRAM locations at the same addresses.
The core module is mounted onto a motherboard via the connectors HDRA and HDRB.
As well as carrying all signal connections between the boards, these provide mech anical
mounting (see Attaching the ARM Integrator/CM920T to a motherboa rd on page 2-5).
HDRA
The signals on the HDRA connectors are tracked between the socket on the underside
and the plug on the top so that pi n 1 conn ects to pi n 1, pin 2 to pi n 2 and so on. That is,
the signals are routed straight through.
HDRB
A number of signals on the HDRB connectors are rotated i n groups of four between the
connectors on the bottom and top of each module. This ensures that each processor (or
other bus master device) on a module connects to the correct signals according to
whether it is bus master 0, 1, 2, or 3. The ID for the bus master on a module is
determined by the position of the module in the stack.
This signal rotation scheme is illu strated in Figure 3-8.
The example in Figure 3-8 illustrates how a group of four signals (labelled A, B, C, and
D) are routed through a group of four con nector pins u p through the s tack. It highligh ts
how signal C is rotated as it passes up through the stack and only utilized on module 2.
All four signals are rotated and utilized in a similar way, as follows :
•signal A on core module 0
•signal B on core module 1
•signal C used on core module 2
•signal D used on core module 3.
For details of the signals on the HDRB connectors, see HDRB on page A-4.
Note
The JTAG signals are discussed in Multi-ICE support on page 3-21.
3.6.5Bus operating modes
The bus operating modes are programmed by writing to coprocessor 15 register 1 within
the ARM940T microprocessor core.
The Integrator system supports:
•asynchronous and FastBus clocking
•little-endian addressing.
3-16
The Integrator system does not support:
•synchronous clocking
•big-endian addressing.
For details of how to set the bus operating parameters, refer to the ARM940T Technical
The core module provides its own clock generators and operates asynchronously with
the motherbo ard. The clock generator provides two programmable cloc ks:
•processor core clock CORECLK
•processor local memory bus clocks LCLK and nLCLK.
In addition, a fixed-frequency reference clock REFCLK is supplied to the FPGA.
These clocks are supplied by two MicroClock ICS525 devices and by the SSRAM
controller PLD, as illustrated in Figure 3-9.
Hardware Description
FPGA
CM_OSC register
24MHz
crystal
ICS525
(U6)
2XCLK
ICS525
(U7)
REFCLK
CORECLK
LCLK
SSRAM controller
(PLD)
nLCLK
Figure 3-9 Core module clock generator
The ICS525s are supplied with a reference clock signal from a 24MHz crystal
oscillator. The 2XCLK output from the first ICS525 (U6) is supplied to the PLD and
divided by two to produce the signals LCLK and nLCLK. The output from U7
provides the CORECLK signal. The reference output from U6 supplies the reference
input to U7 and the reference output from U7 supplies the FPGA reference clock.
The output frequencies from the ICS525s are configured using divider input pins to
produce a wide range of frequencies.
The frequency of CORECLK is controllable in 1MHz steps in the range 12MHz to
160MHz. This is achieved by setting the Voltage Controlled Oscillator (VCO) divider
and output divider for the CORECLK generator via the CM_OSC register. The VCO
divider is controlled b y the C_VDW bi ts and ou tput di vide r is cont rolled by the C _OD
bits. The reference divider value is fixed.
Table 3-2 shows the values placed on the divider input pins and how the clock speeds
are derived. The bits marked:
•C are programmable in the CM_OSC register
•1 are tied HIGH
•0 are tied LOW.
Table 3-2 CORECLK divider values
C_RDW R[6:0]C_VDW V[8:0]C_OD S[2:0]
00101100CCCCCCCCCCC
22 (fixed value)4 – 152 (<4 and >152 not allowed) 2-10
12 – 160MHz in 1MHz steps
3-18
The frequency of CORECLK can be derived from the formula:
freq = 2*((C_VDW+ 8)/C_OD)
where:
C_VDW is the VCO divider word for the core clock.
C_OD is the output divider for the core clock.
For details about programming C_VDW and C_OD, see CM_OSC (0x10000008) on
page 4-9.
Note
Values for C_VDW and C_OD can be calculated using the ICS525 calculator on the
Microclock website.
The CORECLK is buffered with a PI49FCT3805 to convert the Phase-Locked Loop
(PLL) output to 3.3V signal level. The clock is series terminated with a 33Ω resistor and
then drives a single load on the microprocessor core.
The frequency of the processor bus clocks LCLK and nLCLK is determined by the
frequency of 2XCLK. The clock signal 2XCLK is divided by 2 by the SSRAM
controller PLD to produce LCLK and nLCLK.
The frequency of LCLK is controllable in 0.5MHz steps in the range 6MHz to 66MHz.
This is achieved by programming the VCO and output divider bits for the 2XCLK
generator in the CM_OSC regist er. The VCO divi der is co ntroll ed by th e L_VDW bits
and the output divider is controlled by the L_OD bits.The reference divider is fixed.
The maximum speed of 2XCLK is limited by the speed of the SSRAM PLD.
Table 3-3 shows the values placed on the divider input pins and how the clock speeds
are obtained. The bits marked:
•L are programmable in the CM_OSC register
•1 are tied HIGH
•0 are tied LOW
L_RDW R[6:0]L_VDW V[8:0]L_OD S[2:0]
.
Hardware Description
Table 3-3 2XCLK divider values
00101100LLLLLLLLLLL
22 (fixed value)4 – 124 (<4 and >124 not allowed)2-10
12 – 132MHz in 1MHz steps (for 2XCLK)
The frequency of LCLK can be derived from the formula:
freq = (L_VDW+ 8)/L_OD
where:
L_VDW is the VCO divider word for the processor bus clock.
L_OD is the output divider for the processor bus clock.
Note
Values for L_VDW and L_OD can be calculated using the ICS525 calculator on the
Microclock website.
For details about programming L_VDW and L_OD, see CM_OSC (0x10000008) on
page 4-9.
The LCLK clock signal is buffered by a 5-output low-skew buffer PI49FCT3805 to
drive five loads. These are:
•SDRAM_CLK[3:0]
•SSRAM_CLK.
The nLCLK clock signal is a phase-aligned inversion of the LCLK signal. It is
buffered by a 5-output low-skew buffer PI49FCT3805 to four loads. These are:
•ARM_BCLK_M
•PLD_BCLK_M
•FPGA_BCLK_M
•LA_BCLK_M.
All clocks are series terminated with 33Ω resistors placed as close to the source as
possible.
3.7.3FPGA reference clock (REFCLK)
The REFCLK signal is used by the FPGA to generate the SDRAM refresh clock and
SPD EEPROM clock. This is a fi xed- freq uency clock of 24MHz which is output from
the reference pin of the second ICS525 chip U7.
The core module provides support for debug using JTAG. It provides a Multi-ICE
connector and JTAG scan paths aro und the development sys tem. Figure 3-10 sho ws the
Multi-ICE connector and the CONFIG link.
Hardware Description
Multi-ICE connector
CONFIG link
ARM DUI 0125A
CFGLED
Figure 3-10 JTAG connector, CONFIG link, and LED
The CONFIG link is used to enable in-circuit programming of the FPGA and PLDs
using Multi-ICE (see Debugging modes on page 3-23).
The Multi-ICE connector provides a set of JTAG signals allowing third-party JTAG
debugging equipment t o be used (see JTAG si gnals on page 3-24). If you are debugging
a development system with multiple core modules, connect the Multi-ICE hardware to
the top core module.
Figure 3-11 shows a simplified diagram of the scan path.
Core module
TDI
Multi-ICE
HDRB
HDRB
TDO
TDI
Processor
core
TDI
nMBDET
Core module
TDO
TDI
Processor
core
HDRB
HDRB
Motherboard
Figure 3-11 JTAG scan path (simplified)
When the core module is used as a standalone development system, the JTAG scan path
is routed through the processor core and back to the Multi-ICE connector.
When the core module is attached to a motherboard, either directly or with other core
modules, the TDI signal from the top core module is routed down through the HDRB
connectors to the motherboard. The path is routed back up the stack through the
processor core on each core module, before being returned to the Multi-ICE connector.
The motherboard detect signal nMBDET signal is used to control the routing for a
standalone or an attached core module.
3-22
The PLDs and FPGAs are included in the scan chain when the core module is in
configuration mode.
The core module is capable of operating in two modes:
•normal debug mode
•configuration mode.
Normal debug mode
During normal operation and software development, the core module operates in debug
mode. The debug mode is selected by default (when a jumper is not fitted at the
CONFIG link, see Figure 3-10 on page 3-21). In this mode, the processor core and
debuggable devices on other modules are accessible on the scan chain, as shown in
Figure 3-11 on page 3-22.
Configuration mode
In configuration mode the debuggable devices are still accessible and, in addition, all
FPGAs and PLDs in the system are added into the scan chain. This allows the board to
be configured or upgraded in the field using Multi-ICE or other JTAG debugging
equipment.
To select configuration mode, fit a jumper to the CONFIG link on the co re module at
the top of the stack (see Figure 3-10 on page 3-21). This has the effect of pulling the
nCFGEN signal LOW which illuminates the CFG LED (yellow) on each module in the
stack and reroutes the JTAG scan path. The LED provides a warning that the
development system is in the conf iguration mode.
Hardware Description
ARM DUI 0125A
Note
Configuration mode is guaranteed for a single core module attached to a motherboard
but may be unreliable if more than one cor e module is attached. The larger load s on the
TCK and TMS lines may cause unreliable operation.
The configuration mode allows FPGA and PLD code to be updated as follows:
•The FPGAs are volatile, but load their configuration from flash memory. Flash
memory, which itself does not have a JTAG port, can be programmed by loading
designs into the FPGAs and PLDs which hand le the transfer of data to the flash
using JTAG.
•The PLDs are non-volatile devices which can be p rogrammed directly by JTAG.
Figure 3-12 shows the pinout of the Multi-ICE connector and Table 3-4 on page 3-25
provides a description of the JTAG signals.
1192
nTRST
TDI
TMS
TCK
RTCK
TDO
nSRST
DBGRQ
DBGACK
20
3V33V3
GND
GND
GND
GND
GND
GND
GND
GND
GND
Figure 3-12 Multi-ICE connector pinout
Note
In the description in Table 3-4 on page 3-25, the term JTAG equipment refers to any
hardware that can drive the JTAG signals to devices in the scan chain. In most cases this
will be Multi-ICE, although hardware from third-party suppliers can also be used to
debug ARM processors.
DBGRQ is a request for the processor core to enter the debug
state. It is provided for compatibility with third-party JTAG
equipment.
DBGACKDebug acknowledge
(to JTAG equipment)
DBGACK indicates to the deb ugger that the processor core has
entered debug mode. It is provided for compatibility with
third-party JTAG equipment.
DONEFPGA configuredDONE is an open-collector signa l which indicates when FPGA
configuration is complete. Although this signal is not a JTAG
signal, it does effect nSRST. The DONE signal is routed between
all FPGAs in the system through the HDRB connectors. The
master reset controller on the motherboard senses this signal and
holds all the boards in reset (by driving nSRST LOW) until all
FPGAs are configured.
nCFGENConfiguration enable
(from jumper on module at the
top of the stack)
nCFGEN is an active LOW signal used to put the boards into
configuration mode. In configuration mode all FPGAs and PLDs
are connected to the scan chain so that they can be configured by
the JTAG equipment.
nRTCKENReturn TCK enable (from core
module to motherboard)
nRTCKEN is an active LOW signal driven by any core module
that requires RTCK to be routed back to the JTAG equipment. If
nRTCKEN is HIGH, the motherboard drives RTCK LOW. If
nRTCKEN is LOW, th e motherboard drives the TCK signal
back up the stack to the JTAG equipment.
nSRSTSystem reset (bidirectional)nSRST is an active LOW open-collector signal which can b e
driven by the JTAG equipment to reset the target board. Some
JTAG equipment senses this line to determine when a board has
been reset by the user.
The open collector nR S T reset sign al may be driv en LOW by the
reset controller on the core module t o cause the motherboard to
reset the whole system by driving nSYSRST LOW.
This is also used in configuration mode to control the
initialization pin (nINIT) on the FPGAs.
Though not a JTAG signal, nSRST is described because it can be
controlled by JTAG equipment.
This active low open-collector is used to reset the JTAG port and
the associated debug circuitry on the ARM940T processor. It is
asserted at power-up by each module, and can be driven by the
JTAG equipment. This signal is also used in configuration mode
to control the programming pin (nPROG) on FPGAs.
3-25
Hardware Description
NameDescriptionFunction
Table 3-4 JTAG signal description (continued)
RTCKReturn TCK
(to JTAG equipment)
TCKTest clock
(from JTAG equipment)
TDITest data in
(from JTAG equipment)
Some devices sample TCK (for example a synthesizable core
with only one clock), and this has the effect of delaying the time
at which a component act ually captures data. RTCK is a
mechanism for returning the sampled clock to the JTAG
equipment, so that the clock is not advanced until the
synchronizing de vice captured the dat a . In adaptive clocking mode, Multi-ICE is required to detect an edge on RTCK before
changing TCK. In a multiple device JT AG chain, the RTCK
output from a component connects to the TCK input of the
down-stream device. The RTCK si gnal on the module
connectors HDRB returns TCK to the JTAG equipment. If there
are no synchronizing components in the scan chain then it is
unnecessary to use the RTCK signal and it is connected to
ground on the motherboard.
TCK synchronizes all JTAG transactions. TCK connects to all
JTAG components in the scan chain. Series termination resistors
are used to reduce reflections and maintain good signal integrity.
TCK flows down the stack of modules and connects to each
JTAG component. However, if there is a device in the scan chain
that synchronizes TCK to some other clock, then all down-stream
devices are connected to the RTCK signal on that component
(see RTCK).
TDI goes down the stack of modules to the motherboard and then
back up the stack, labelled TDO, connecting to each component
in the scan chain.
TDOTest data out
(to JTAG equipment)
TMSTest mode select
(from JTAG equipment)
3-26
TDO is the return path of the data input signal TDI. The module
connectors HDRB have tw o pin s labelled TDI and TDO. TDI
refers to data flowing down the stack and TDO to data flowing up
the stack. The JTAG components are connected in the retu rn p a th
so that the length of track driven by the last compone nt in the
chain is kept as short as possible.
TMS controls transitions in the tap controller state machine.
TMS connects to all JTAG components in the scan chain as the
The ARM940T processor core incorporates EmbeddedICE hardware and provides a
debug communications data register which is used to pass data between the processor
and JTAG equipment. The processor accesses this register as a normal 32-bit read/write
register and the JTAG equipment reads and writes the register using the scan chain. For
a description of the debug communications channel, see the ARM940T Technical Reference Manual.
Interrupts can be used to signal when data has been written into one side of the register
and is available for reading from the other side. These interrupts are supported by the
interrupt controller within the core module FPGA and can be enabled and cleared by
accessing the interrupt registers (see Interrupt registers on page 4-19).
This section describes the memory map. For a standalone core module, the memory
map is limited to local SSRAM, SDRAM, and core module registers. For the full
memory map of an Integ rator deve lopment system, which in cludes a mot herboard, y ou
should refer to the user guide for the motherboard.
4.1.1Core module memory map
The core module has a fixed memory map which maintains compatibility with other
ARM modules and Integrator systems. Table 4-1 shows the memory map.
Table 4-1 ARM Integrator/CM940T memory map
nMBDETREMAPAddress rangeRegion sizeDescription
000x00000000 to 0x0003FFFF256KBBoot ROM (on motherboard)
010x00000000 to 0x0003FFFF256KBSSRAM
1X0x000 00000 to 0x0003FFFF256KBSSRAM
XX0x00040000 to 0x0FFFFFFF 256MBLocal SDRAM
XX0x10000000 to 0x10FFFF FF16MBCore module registers
0X0x11000000 to 0xFFFFFFFF4GB to 272MBSystem bus address space
1X0x11000000 to 0xFFFFFFFF4GB to 272MBAbort
4.1.2Boot ROM and SSRAM accesses
The boot ROM on the motherboard and the SSRAM on the core module share the same
location within the Integrator memory map. Accesses to either the boot ROM or
SSRAM are controlled by the REMAP bit and the motherboard detect signal
(nMBDET) from the motherboard, as shown in Table 4-1 on page 4-2.
Remap
The REMAP bit only has effect if the core module is attached to a motherboard
(nMBDET=0). It is controlled by bit 2 of the CM_CTRL register at 0x1000000C and
functions as follows:
REMAP=0As it is after a reset. The Boot ROM on the motherboard appears in the
address range 0x0 to 0x3FFFF.
REMAP=1 The SSRAM appears in the 0x0 to 0x3FFFF address range.
The nMBDET signal operates as follows:
nMBDET=0 The core module is attached to a motherboard, and accesses in the
address range 0x0 to 0x3FFFF to the boot ROM or SSRAM are
controlled by the REMAP bit.
nMBDET=1 The core module is not attached, and accesses in the address range 0x0 to
0x3FFFF are routed to the SSRAM.
Note
The SSRAM is local, which means that it can only be read by the processor on the same
core module. It provides fast local memory for the local processor and cannot be
accessed by other system bus masters.
The Integrator memory map provides a 256MB address space for SDRAM. When a
smaller sized SDRAM DIMM is fitted, it is mapped repeatedly to fill the 256MB space.
For example, a 64MB DIMM appears four times, as shown in Figure 4-2.
0x0FFFFFFF
64MB
Local SDRAM
(repeat image)
64MB
Local SDRAM
(repeat image)
64MB
64MB
256KB
Figure 4-2 SDRAM repeat mapping for a 64MB DIMM
Local SDRAM
(repeat image)
Local SDRAM
SSRAM
0x0003FFFF
Local SDRAM
The local processor can access the local SDRAM (that is, the SDRAM on the same core
module) at 0x00000000 to 0x0FFFFFFF in the core module address space. However,
the lowest 256KB (0x00000000 to 0x0003FFFF) is hidden by the SSRAM or boot
ROM, depending upon whether the core mo dule is attached to a motherboard and upon
the state of the REMAP bit.
4-4
The SDRAM cannot be accessed within this address space, although it can be accessed
at one of its repeat images or at its alias location. In the case of a 256MB DIMM which
fills the local SDRAM space, the first 256KB can only be accessed at the alias location
(see System bus accesses to SDRAM below) .
If the core module is mounted on a motherboard, the SDRAM is mapped to appear at
the al iased module memory region of the combine d Integrator system bus memory map .
The SDRAM can be accessed by all bus masters at its alias location, and accessed by
the local processor at both its local and alias locations.
The system bus address for a core module is automatically controlled by its position in
the stack (see Core module ID on page 2-6). Figure 4-3 show s the local and alias
address of the SDRAM on four core modules.
System bus address
All masters
0xBFFFFFFF
0xB0000000
0xAFFFFFFF
0xA0000000
0x9FFFFFFF
0x90000000
0x8FFFFFFF
0x80000000
Local address
0x0FFFFFFF
SDRAM
core module 3
0x00000000
0x0FFFFFFF
SDRAM
core module 2
0x00000000
0x0FFFFFFF
SDRAM
core module 1
0x00000000
0x0FFFFFFF
SDRAM
core module 0
0x00000000
Figure 4-3 Core module local and alias addresses
Module 3
Module 2
Module 1
Module 0
ARM DUI 0125A
By reading the CM_STAT register, a processor can determine which core module it is
on and, therefore, the alias location of its own SDRAM (see CM_STAT (0x10000010)
on page 4-12).
The convention for ARM cores is to map the exception vectors to begin at address 0.
However, the ARM940T core allows the vectors to be moved to 0xFFFF0000 by
writing to the V bit in coprocessor 15 register 1 (CP15c1). The value of the V bit at reset
is determined by the level on an external pin (VINITHI). To mainta in compatibility
across all cores, the default reset value maps the vector to begin at address 0 (see the
ARM940T Technical Reference Manual).
When running applications wi th high vectors (at 0xFFFF0000 to 0xFFFFFFFF),
software must write the correct value to the coprocessor register. However, Integrator
motherboards have no physical memory at this location. This means that the MMU
must be programmed t o m a p an ar ea of physical memory to this virtual addres s . W hen
the core module is being used with a motherboard, an alternative is to implement an area
of physical memory on an expansion card or logic module.
The core module status and control registers allow the processor to determine its
environment and to control some core module operations. The register s, listed in Table
4-2, are located at 0x10000000 and can only be accessed by the local processor.
Table 4-2 Core module status, control, and interrupt registers
Register NameAddressAccessDescription
CM_ID0x10000000ReadCore module identification register
CM_PROC0x10000004ReadCore module processor regist er
CM_OSC0x10000008Read/writeCore module oscillator values
CM_CTRL0x1 000000CRead/writeCore module control
CM_STAT0x10000010ReadC ore module status
CM_LOCK0x10000014Read/writeCore module lock
CM_SDRAM0x10000020Read/writeSDRAM status and control
CM_IRQ_STAT0x10000040ReadCore module IRQ status regi ster
CM_IRQ_RSTAT0x10000044ReadCore module IRQ raw status r e gister
CM_IRQ_ENSET0x10000048Read/writeCore module IRQ enable set regi ster
CM_IRQ_ENCLR0x1000004CWriteCore module IRQ enable clear register
CM_SOFT_INTSET0x10000050Read/writeCore module software interrupt set
CM_SOFT_INTCLR0x10000054WriteCore module software interrupt clear
CM_FIQ_STAT0x10000060ReadCore module FIQ status register
CM_FIQ_RSTAT0x10000064ReadCore module FIQ raw status register
CM_FIQ_ENSET0x10000068Read/writeCore module FIQ enable set re gister
CM_FIQ_ENCLR0x1000006CWriteCore module FIR enable clear register
CM_SPD0x10000100 to 0x100001FCRead SDRAM SPD memory
Note
All registers are 32 -bit s wi de a nd do not support b yt e writes. Write operations must be
wordwide. Bits marked as res erv ed in th e following secti ons should be preserved usin g
read-modify-write operations.
The core module ID register (CM_ID) is a read-only register that identifies the board
manufacturer, board type, and revision.
31242316 1512114 30
MAN
Table 4-3 describes the core module ID register bits.
BitsNameAccessFunction
31:24MANReadManufacturer:
23:16ARCHReadArchitecture:
15:12FPGAReadFPGA type:
11:4BUILDReadBuild value (ARM internal use)
3:0REVReadRevision:
4.3.2CM_PROC (0x10000004)
ARCH
FPGA
BUILD
REV
Table 4-3 CM_ID register bit descriptions
0x41 = ARM
0x00 = Generic ARM7x0T or AR M9x0T,
4 word SDRAM bursts
0x00 = XC4036
0x0 = Rev A
0x1 = Rev B
4-8
The core module processor register (CM_PROC) is a read-only register that contains
the value 0x00000000. This is provided for compatibility with processors that do not
have a system control coprocessor (CP15). For the ARM940T, information about the
processor can be obtained by reading coprocessor 15 register 0 (CP15 c0).
The core module oscillator register (CM_OSC) is a read/write register that controls the
frequency of the clocks generated by the two clock generators (see Clock generators on
page 3-17). In additio n, it provides information about process or bus mode setting.
Programmer’s Reference
31
Reserved
25
24
23 2220
L_ODBMODE
19
L_VDW
12 11 10780
C_OD
C_VDWR
Before writing to the CM_OSC register, you must unlock it by writing the value
0x0000A05F to the CM_LOCK register. Aft er writing t he CM_OSC regi ster, relock it
by writing any value other than 0x0000A05F to the CM_LOCK register.
Table 4-4 describes the core module oscillator register bits.
Table 4-4 CM_OSC register
BitsName AccessFunction
31:25ReservedUse read-modif y-write to preserve value.
24:23BMODReadThis field c ontains 00 whic h indica tes that the
processor bus mode is selected by writing to
CM_CTRL register (see CM_CTRL (0x1000000C) on page 4-11).
22:20L_ODRead/writeMemory clock ou tput divider:
000 = divide by 10
001 = divide by 2 (default)
010 = divide by 8
011 = divide by 4
100 = divide by 5
101 = divide by 7
110 = divide by 9
111 = divide by 6.
ARM DUI 0125A
19:12L_VDWRead/writeProcessor bus clock VCO divider word.
Defines the binary value of the V[7:0] pins of
the clock generator (V[8] is tied low).
00000100 = 6MHz (def ault with OD = 2).
The core module control register (CM_CTRL) is a read/write register that provides
control of a number of user-configurable features of the core module.
3101234
Programmer’s Reference
Reserved
Table 4-5 describes the core module control register bits.
BitsNameAccessFunction
31:8ReservedUse read-modify-write to preserve value.
7:4ReservedUse read-modify-write to preserve value.
3
2
1
RESET
REMAP
nMBDET
WriteThis is used to reset the core module, the
motherboard on which i t is mounte d, and an y core
modules in a stack. A reset is triggered b y writing
a 1. Reading this bit always returns a 0 allowing
you to use read-modi fy-write operations without
masking the RESET bit.
Read/writeThis only has affect when the core module is
mounted onto a motherboard. When this is the
case, and this bit is a 0, accesses to the first
256KB (0x00000000 to 0x0003FFFF) of memory
are redirected into the motherboard.
ReadThis bit indicates whether or not the core module
is mounted on a mother board:
0 = mounted on motherboard
1 = standalone.
LEDnMBDETREMAPRESET
Table 4-5 CM_CTL register
ARM DUI 0125A
0LEDRead/writeThis bit controls the green MISC LED on the core
The core module status register (CM_STAT) is a read -on ly r e gister that can be r ead to
determine where in a multi-core module stack this core module is positioned.
31078
Reserved
Table 4-6 describes the core module status register bits.
Table 4-6 CM_STAT register
BitNameAccessFunction
31:8ReservedUse read-modify-write to preserve value.
7:0IDReadCard number:
The core module lock register (CM_LOCK) is a read/write register that is used to
control access to the CM_OSC register, allowing it to be locked and unlocked. This
mechanism prevents the CM_OSC register from being overwritten accidently.
310151617
Programmer’s Reference
Reserved
LOCKVALLOCKED
Table 4-7 describes the core module lock register bits.
Table 4-7 CM_LOCK register
BitsNameAccessFunction
16LOCKEDReadThis bit indicates if the CM_OSC register is
locked or unlocked:
0 = unlocked
1 = locked.
15:0LOCKVALRead/writeWrite the value 0x0000A05F to this register
to enable write accesses to the CM_OSC
register.
Write any other value to this register to lock
the CM_OSC register.
The SDRAM status and control register (CM_SDRAM) is a read/write r egister used to
set the configuration parameters for the SDRAM DIMM. This control is necessary
because of the variety of module sizes and types available.
Writing a value to this register automatically updates the mode register on the SDRAM
DIMM.
310765 4211181216 151920
Reserved
CASLATMEMSIZESPDOKRNROWSNCOLSNBANKS
Table 4-8 describes the SDRAM status and control register bits.
Table 4-8 CM_SDRAM register
BitsNameAccessFunction
31:20ReservedUse read-modify-write to preserve value.
19:16NBANKSRead/writeNumber of SDRAM banks. Should be set to
the same value as byte 5 of SPD EEPROM .
15:12NCOLSRead/writeNumber of SDRAM columns. Should be set
to the same value as byte 4 o f S PD E EPROM .
11:8NROWSRead/writeNumber of SDRAM rows. Should be set to
the same value as byte 3 of SPD EEPROM .
7:6ReservedUse read-modify-write to preserve value.
5
1:0CASLATRead/writeThese bits specify the CAS laten cy set for the
MEMSIZE
Read/writeThese bits specify the size of the SDRAM
module fitted to the core modu le. Th e bits are
encoded as follows:
000 = 16MB
001 = 32MB
010 = 64MB (default)
011 = 128MB
100 = 256MB
101 = Reserved
110 = Reserved.
core module. The bits are encoded as follows:
00 = Reserved
01 = Reserved
10 = 2 cycles (default)
11 = 3 cycles.
Note
Before the SDRAM is used it is necessary to read the SPD memory and program the
CM_SDRAM register with the parameters indicated in Table 4-8. If these values are not
correctly set then SDRAM accesses may be slow or unreliable. See CM_SPD (0x10000100 to 0x100001FC) on page 4-16.
This area of memory contains a copy of the SPD data from the SPD EEPROM on the
DIMM. Because accesses to the EEPROM are very slow, the data is copied to this
memory during board initialization to allow faster random access to the SPD data (see
Serial presence detect on page 3-6). The SPD memory contains 256 bytes of data, the
most important of which are as shown in Table 4-9.
Table 4-9 SPD memory contents
ByteContents
2Memory type
3Number of row addresses
4Number of column a ddresses
5Number of banks
31Modu le bank density (MB
divided by 4)
18CAS latencies supported
4-16
63Checksum
64:71Manufacturer
73:90Module part numb er
Check for valid SPD data as follows:
1.Add together all bytes 0 to 62.
2.Lo gically AND the result with 0xFF.
3.C o mpare the result with b yte 63.
If the two values match, then the SPD data is valid.
Note
A number of SDRAM DIMMs do not comply with the JEDEC standard and do not
implement the checksum byte. The Integrator is not guaranteed to operate with
non-compliant DIMMs.
The code segment shown in Example 4-1 on page 4-17 can be used to correctly setup
and remap the SDRAM.
CM_BASEEQU0x10000000; base address of Core Module registers
SPD_BASEEQU0x10000100; base address of SPD information
Programmer’s Reference
lightled
readspdbit
setupsdram
LDRr0, =CM_BASE; load register base address
; turn on header LED and remap memory
MOVr1,#5; set remap and led bits
STRr1,[r0,#0xc]; write the register
; setup SDRAM
; check SPD bit is set
LDRr1,[r0,#0x20]; read the status register
ANDr1,r1,#0x20; mask SPD bit (5)
CMPr1,#0x20; test if set
BNEreadspdbit; branch until the SPD memory has been read
; work out the SDRAM size
LDRr0, =SPD_BASE ; point at SPD memory
LDRBr1,[r0,#3]; number of row address lines
LDRBr2,[r0,#4]; number of column address lines
LDRBr3,[r0,#5]; number of banks
LDRBr4,[r0,#31]; module bank density
MULr5,r4,r3; calculate size of SDRAM (MB divided by 4)
MOVr5,r5,ASL#2; size in MB
CMPr5,#0x10; is it 16MB?
BNEnot16; if no, move on
MOVr6,#0x2; store size and CAS latency of 2
Bwritesize
not16
not32
ARM DUI 0125A
CMPr5,#0x20; is it 32MB?
BNEnot32; if no, move on
MOVr6,#0x6 ; store size and CAS latency of 2
B writesize
CMPr5,#0x40; is it 64MB?
BNEnot64; if no, move on
MOVr6,#0xa; store size and CAS latency of 2
B writesize
CMPr5,#0x80; is it 128MB?
BNEnot128; if no, move on
MOVr6,#0xe; store size and CAS latency of 2
Bwritesize
; if it is none of these sizes then it is either 256MB, or
; there is no SDRAM fitted so default to 256MB.
MOVr6,#0x12; store size and CAS latency of 2
MOVr1,r1,ASL#8; get row address lines for SDRAM register
ORRr2,r1,r2,ASL#12 ; OR in column address lines
ORRr3,r2,r3,ASL#16 ; OR in number of banks
ORRr6,r6,r3; OR in size and CAS latency
LDRr0, =CM_BASE; point at module registers
STRr6,[r0,#0x20]; store SDRAM parameters
The core module provides a 3-bit IRQ controller and 3-bit FIQ controller to support the
debug communications channel used for passing information between applications
software and the debugger. The interrupt control registers are listed in Table 4-10.
Table 4-10 Interrupt controller registers
Register NameAddressAccessSizeDescription
CM_IRQ_STAT0x10000040Read3 bitsCore module IRQ status register
CM_IRQ_RSTAT0x10000044Read3 bitsCore module IRQ raw status register
CM_IRQ_ENSET0x10000048Read/write3 bitsCore module IRQ enable set register
CM_IRQ_ENCLR0x1000004CWrite3 bitsCore module IRQ enable clear register
CM_SOFT_INTSET0x10000050Read/write1 bitCore module software interrupt set
CM_SOFT_INTCLR0x10000054Write1 bitCore module software interrupt clear
CM_FIQ_STAT0x10000060Read3 bitsCore module FIQ status register
CM_FIQ_RSTAT0x10000064Read3 bitsCore module FIQ raw status register
CM_FIQ_ENSET0x10000068Read/write3 bitsCore module FIQ enable set register
CM_FIQ_ENCLR0x1000006CWrite3 bitsCore module FIR enable clear r e gister
The IRQ and FIQ controllers each provide three registers for controlling and handling
interrupts. These are:
•status register
•raw status register
•enable register, which is accesse d u sing th e en able set an d enable clear locations.
The way that the interrupt enable, clear, and status bits function for each interrupt is
illustrated in Figure 4-4 on page 4-20 and described in the following subsections. The
illustration shows the control for one IRQ bit. The remaining IRQ bits and FIQ bits are
controlled in a similar way.
The raw status register indicates the signal levels on the interrupt requ est inputs. A bit
set to 1 indicates that the corresponding interrupt request is active.
The core module interrupt controller provides a register for controlling and clearing
software interrupts. This register is accessed using the software interrupt set and
software interrupt clear locations. The set and clear locations are used as follows:
•Set the software interrupt by writing to the CM_SOFT_INTSET location:
1 = SET the software interrupt
0 = leave the software interrupt unchanged.
•Read the current state of the of the software interrupt register from the
CM_SOFT_INTSET location. A bit set to 1 indicates that the correspondi ng
interrupt request is active.
•Clear the software interrupt by writing to the CM_SOFT_INTCLR location:
1 = CLEAR the software interrupt.
0 = leave the software interrupt unchanged.
The bit assignment for the software interrupt register is shown in Table 4-12.
Table 4-12 IRQ register bit assignment
4-22
Bit NameFunction
0SOF TSoftware interrupt
Note
The software interrupt described in this section is used by software to generate IRQs or
FIQs. It should not be confused with the ARM SWI software interrupt instruction. See
the ARM Architecture Reference Manual.
The signals present on the pins labeled A[31:0], B[31:0], and C[31:0] are described in
Table A-1.
Table A-1 Bus bit assignment (for an AMBA ASB bus )
Pin labelName (ASB)Description
A[31:0]System address busSystem address bus
B[31:0]Not usedC[31:0]System control bus See remainder of table.
C[31:16]Not usedC15BLOKLocked transaction
C14BLASTLast response
C13BERRORError response
C12BWAITWait response
C11BWRITEWrite transaction
C10Not use d-
ARM DUI 0125A
C[9:8]BPROT[1:0]Transaction protection type
C7Not usedC[6:5]BURST[1:0]Transaction burst size
C4Not usedC[3:2]BSIZE[1:0]Transaction width
C[1:0]BTRAN[1:0]Transaction type
D[31:0]System data busSystem data bus
Note
Table A-1 shows signal descriptions for an AMBA ASB bus implementation.
The signals on the pins labeled E[31:0] are cross-connected between the plug and socket
so that the signals are rotated through the stack in groups of four. For example, the first
block of four are connected as shown in Table A-2.
For details about the signal rotation scheme, see System bus signal routing on
page 3-15.
The signals on the pins labeled F[31:0] are connected so that pins on the socket are
connected to the corresponding pins on the plug.
Table A-2 Signal cross-connections (example)
Plug Socket
E0connects to E1
E1connects to E2
E2connects to E3
E3connects to E0
A-6
The signals on G[15:8] and G[5:0] are connected so that pins on the socket are
connected to the corresponding pins on the plug.
Pins G[7:6] carry the JTAG TDI and TDO signals. The signal TDO is routed through
devices on each baord as it passes up through the stack (see JTAG si gnals on page 3-24).
Table B-1 shows the core module electrical characteristics for the system bus interface.
The core module uses 3.3V and 5V source. The 12V inputs are supplied by the
motherboard but not used by the core module.
SymbolDescriptionMinMaxUnit
3V3Supply voltage (int erface signals)3.13.5V
5VSupply voltage4.755.25V
The core module is designed to be stackable on a num ber of different mothe rboards. Its
size allows it to be mounted onto a CompactPC I mo therboard while allowing the
motherboard to be installed in a card cage.
Figure B-1 shows the mechanical outline of the core module.
10.0
200-way connector
(4 col x 50 row)
Plug on top and
socket on underside
Detail A
148.0
128.0
Detail A
HDRAHDRB
130-way connector
(4 col x 30 row)
Plug on top and
socket on underside
Memory DIMM
Memory DIMM connector is 25º type
which overhangs the edge of the board
10.0
10.0
Pin numbers for 200-way plug,
viewed from above board
1
2
3
Detail B
101
102
103
10.0
100.0
81.0
Pin numbers for 200-way socket,
viewed from below board
101
102
103
Detail B
Pin numbers for 168-way
SDRAM DIMM connector
shown
SDRAM operating mo de3-6
SDRAM repeat mapp ing4-4
SDRAM status and control register 4-14
SDRAM, SPD memory 4-16
Serial presence detec t3-6
Setting CAS latency 4-15
Setting SDRAM size 4-15
Setup
power connections2-3
standalone 2-2
Software interrupt registers 4-22
Software reset 4-11
SPD memory 4-16
SPDOK bit 4-14
SSRAM accesses 4-2
SSRAM controller 3-3
Standalone core module 2-2
Status and configurat ion registers 1-6
Status register 4-12
Status register, interrupt 4-20
Supplying power 2-3
System architectu r e 1-4
System bu s
operating modes 3-16
System bus bridge 1-5, 3-11
System bus signal routing 3-15