Transcede 2xxx Hardware Data Sheet (PN 842xx-DSH-001-A) PDF
Transcede 2xxx Hardware Data Sheet (PN 842xx-DSH-001-A) PDF
842xx-DSH-001-A
September, 2013
Contents
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.1 Transcede 2xxx Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22
1.2 Transcede 2xxx Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
1.3 Transcede 2xxx Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
1.4 Device Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25
1.5 Device Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25
1.6 Transcede 2xxx System Memory Map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
1.6.1 CEVA Group Slaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
1.6.2 CEVA Group Partition Slaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
1.6.3 SYS Partition Slaves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
1.6.4 Radio Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
1.6.5 System Masters/Slaves Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
2 Device Configuration Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1 Boot Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .30
2.2 JTAG Debug Modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .30
2.3 Bootstraps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31
3 Pinout and Signal Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1 Transcede 2xxx Ball Map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33
3.2 Summary of Interface Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .47
3.3 Power and Ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .61
3.4 Unused Interface Termination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .61
3.4.1 Unused TDM Bus Signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .62
3.4.2 Unused SPI Bus Signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .62
3.4.3 Unused Radio Timer Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63
3.4.4 Unused SerDes Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63
3.4.5 Unused RGMII Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63
3.4.6 Unused TSU Signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .64
3.4.7 Unused RPDIF CMOS Bus Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .64
3.4.8 Unused RPDIF LVDS Bus Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65
3.4.9 Unused USB Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .66
3.4.10 Unused USIM Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .67
3.4.11 Unused eFuse Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .67
Related Documentation
A comprehensive library of Transcede 2xxx reference and application documentation is available from Mindspeed.
Developers, applications engineers, and other personnel who are looking to implement, support, or otherwise
become familiar with the Transcede 2xxx, log on and register at: www.mindspeed.com to access the full range of
support documentation. For detailed information on specific devices and available features, always consult the
Mindspeed library of titles. The following table gives a summary of Mindspeed publications related to this
document.
Supported Devices
This following devices are included in the T2xxx family:
• T2100
• T2130
• T2120
• T2150
• T2200
Document Conventions
The following conventions are used throughout Mindspeed documentation:
Document Conventions
Convention Description Example
Bold Text in this font denotes items that are emphasized, menu On the File menu, click Save As.
items, buttons, and dialog box options.
Italics This font is used for the names of disk drives, paths, The patch 001-uboot.patch is available under the
directories, file names, and extensions. package/mspd/u-boot/patches/ directory.
Courier Text in this font denotes code sections, programming Create a folder to unpack the filesystem:
examples, functions, arguments, threads, commands, and ~$ mkdir openwrt_fs
syntax examples.
NOTE: In Figure 1-1, CRP Cluster and CPRI are for Transcede 2120, Transcede 2150, and
Transcede 2200.
MAP4
MAP4
MAP4
10kB
10kB
80kB
80kB
DRAM
DRAM
80kB
N
N
E
O
Cortex A9
T2xxx
Chip Rate MAP4 PRAM
80kB E N
DRAM Processor PRAM DRAM
10kB O
Cortex A9 32kB
10kB DRAM N 32kB
24 Correlators PRAM
PRAM ICache DCache
32kB 32kB
CRP Cluster MAP Cluster ICache DCache
DMA FIFO
BitStream DMA FIFO DMA FIFO Snoop CU
ACP
Engine 128bit Port
Addr Filter
64bit
64bit
Mindspeed Proprietary and Confidential
CevaXC323
CevaXC323 256kB
128bit AXI
128bit AXI 256kB DRAM TSU NTG
96kB PRAM
Cache/
PRAM 2D DMA Dual Security
64bit AXI
UART 2X
Cache-DMA AXI-S 2D DMA
AXI-M Accelerator
FEC CTC/Viterbi Cache-DMA AXI-S AXI-M Units TDM/NTG
32bit APB
DMA FIFO X2
Accelerator Units Program 128bit DSP Data 128bit ML-AXI (IPSEC & 4G/3G-
(ULx1) (DLx1) ML-AXI
SNOW) I2C
64bit
ARM 64bit ML-AXI-Trust Zone
64bit
128bit AXI
128bit AXI
128bit AXI
3G/4G AXI DMA DMA Timers 2X
(LTE-A//HSPA+)
SPI 2X
2x JESD207
64bit AXI
64bit AXI
64bit AXI
GPIO
JEDS 256kB
12bit 207 64bit AXI
IRAM
DDR
JEDS
64bit
12bit 207 CEVA 128bit ML AXI-Trust Zone
DDR
AXI
ARM 64bit ML-AXI-Trust Zone
64/128bit AXI
Bridge
32bit AHB
PCIe x1 or PCIe x4
64bit AXI
64bit AXI
64bit AXI
64bit AXI
Serdes (4)
64bit AXI
PCIe
TBI X1 or x4 64bit AXI
PCS
32bit AHB
Quad-Port Arbitrator
CPRIv4.1 GigE
x1 64bit AXI DDR3-1600 Controller GigE MAC MAC
PCS
TBI USB2
w/1588-2008
PCS w/1588-2008 Expansion/
Mux
PCIe PCS
PHY NAND ECC
64bit AXI
PCS
32bit
x1
Introduction
TBI TBI
rGMII
Exp Bus
USB
Sel Mux 3x1
1.25GHz Serdes (1)
1.5V/1.35V DDR3 x32 sGMII or PCIe(x1) or CPRIv4.1
24
Introduction
16 MB APB Top
FE00_0000h
(reserved) Top
FDE0_0000h
1 MB SPAcc/IPSec 0 FDD0_0000h
(reserved)
FA00_0000h
4 MB External PCIe x1/x4
Memory window F600_0000h
2 MB Radio
Rad Group Config Group
1 MB CPRI/JESD/CPDMA..
2 MB F5B0_0000h
(reserved)
F500_0000h
1 MB ARM Semaphores F4F0_0000h
1 MB ARM AHB F4E0_0000h
1 MB DDR & CVP L2 CFG F404_0000h
256 MB IRAM
F400_0000h
1 MB DSP ROM F3F0_0000h
2 MB Ceva GP APB FDE0_0000h
(reserved) F3B0_0000h
Ceva
8 MB CRP Group
F340_0000h
(reserved) F30C_0000h
768 KB CRAM F300_0000h
(reserved)
F100_0000h
1 MB GP Semaphores F0F0_0000h
1 MB GP AHB F0E0_0000h
(reserved)
F020_0000h
1 MB GP CevaXC Slave 1 F010_0000h
1 MB GP CevaXC Slave 0 F000_0000h
External Memory
3.75 GB ~~ Device Space / ~
~
Expansion Bus
0001_0000
64 KB tmp boot (IBR remap) 0000_0000h
NOTE: CPRI, CRP and CPDMA blocks are not included in Transcede 2100 and Transcede 2130
devices and are for Transcede 2120, Transcede 2150, and Transcede 2200.
Table 1-5 SYS Partition Slaves (0xF4) – IRAM, FEC, DDR cfg, Semaphore Blocks (Continued)
Block Name Start Address End Address Access Bus Memory Size
Clock Reset + NTG 0xF4CF_0000 0xF4CF_FFFF APB3_8bit 64 KB
FEC0 DL cfg 0xF4D0_0000 0xF4D0_FFFF APB3 64 KB
FEC0 UL cfg 0xF4D1_0000 0xF4D1_FFFF APB3 64 KB
ARM CA9 & L2 APB 0xF4E0_0000 0xF4E7_FFFF APB3 512 KB
Semaphore ARM 0xF4F0_0000 0xF4FF_FFFF AHB 1 MB
Table 1-7 System Partition - SPAcc, IPSec, RGMII, TDM, GPIO, UART, SPI, I2C Blocks (Continued)
Block Name Start Address End Address Access Bus Memory Size
TIMER 0xFE05_0000 0xFE05_FFFF APB 64 KB
GPIO 0xFE07_0000 0xFE07_FFFF APB 64 KB
UART S1 0xFE09_0000 0xFE09_1FFF APB 8 KB
SPI S1 0xFE09_8000 0xFE09_9FFF APB 4 KB
I2C 0xFE09_C000 0xFE09_FFFF APB 8 KB
USIM 0xFE0E_0000 0xFE0E_FFFF APB 64 KB
eFuse / OTP ROM 0xFE0F_0000 0xFE0F_FFFF APB 64 KB
EXP BUS 0xFE10_0000 0xFE10_FFFF APB 64 KB
GEM1 0xFE19_0000 0xFE19_FFFF APB3 64 KB
UART0/SPI0 (HS) 0xFE80_0000 0xFE80_FFFF AHB 64 KB
USB 0xFE82_0000 0xFE86_FFFF AHB 256KB+64 KB
IBR (BootRom) 0xFF00_0000 0xFF0F_FFFF APB 1 MB
2.3 Bootstraps
Bootstraps are values latched on the following inputs pins as shown in Table 2-2 during assertion state of the
device hardware reset. Upon de-assertion of device hardware reset, their latched values are in effect and pins
directions are switched to output state according to their normal behavioral functions.
EXP_A_2 BOOT_OP[2]
EXP_A_3 Sync Mode Bootstrap pull-up.
EXP_A_4 RSVD This Bootstrap should be pulled high.
EXP_A_5 EPROM_N When asserted "active low", it will enable boot from external EPROM/FLASH.
EXP_A_6 RSVD This Bootstrap is not used and should be pulled high.
EXP_A_7 SRDS_MODE[0] SerDes (X1) Mode:
EXP_A_8 SRDS_MODE[1] • 00: PCIe Mode
• 01: CPRI Mode (Transcede 2120, Transcede 2150, and Transcede 2200)
• 1x: SGMII Mode (default)
EXP_A_9 CORESIGHT_SEL Bootstrap to select between : RPDIF0 (CMOS) or CoreSight interface.
• 0: RPDIF0(default)
• 1: CoreSight
EXP_A_22
EXP_A_23 RSVD These Bootstraps are not used, should be pulled down (low).
EXP_A_24
EXP_A_25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
A
AA
AB
AC
AD
AE
DSS DQS output falling edge setup time 0.25 — — clk CLK rising
DSH DQS output falling edge hold time 0.25 — — clk CLK rising
DSS DQS input falling edge setup time 0.18 — — clk CLK rising
DSH DQS input falling edge hold time 0.18 — — clk CLK rising
NOTE: SerDes for CPRI is supported for Transcede 2120, Transcede 2150, and Transcede
2200.
See Section 2.3, Bootstraps, on page 31 for the configuration.
NOTE: CPRI is for Transcede 2120, Transcede 2150, and Transcede 2200.
external
clk2x calib clk Serdes TX Serdes RX
Clock
Divider
clkx clk2x
Read
Request
Queue
(8)
AXI RD
AXI_GM_64
AXI WR
AXI_GM_64 Write
Request
Queue
(8)
Template 2 Template 3
Template 0
Template Reg 0
Template 1
Template Reg 1
Templates 2 and 3
4 Desc 4 Desc 4 Desc 4 Desc are common to 8
AxC
Template Mux
TSL entry 0
TSL entry 1 DMA data transfer
TSL entry 2 size
TSL entry 3 Desc 0
TSL entry 4 Descriptor symbol offset (16b)
TSL entry 5 Desc 1
TSL entry 6 Template exec. flow
TSL entry 7 exec. flow Desc 2 DMA data fetch
TSL entry 8 Desc 3 address
TSL entry 9 +
TSL entry 10
TSL entry 11
TSL entry 12 Base Pointer - Word aligned
TSL entry 13 (30b) + 2'b00
TSL entry 14
TSL entry 15 current symbol
Current Symbol Buffer Pointer buffer pointer
(CSBP)
(30b) + 2'b00
symbol
pointer
Symbol
Pre
Pre Symbol
Transmit (Downlink)
One signal group per AxC port
req (in)
ack (out)
ack (in)
clk
cpri_map_tx_ready
cpri_map_tx_data
d0 d1 d2 d3
clk
cpri_map_rx_ready
cpri_map_rx_data
d0 d1 d2 d3
CSBP
R=0
Desc 0 - Symbol 6
Symbol 6- SCP
Desc 1 -
Symbol 6 SCP
Desc 2 -
NULL
Desc 3 -
NULL
R=1
Table 7-1 2.5 and 5.0 Gbps PCIe Gen2 Transmitter Specifications
Symbol Parameter 2.5 Gbps 5.0 Gbps Units Comments
Ul Unit Interval 399.88 (min) 199.94 (min) ps The specified Ul is equivalent to a tolerance of ±300
400.12 (max) 200.06 (max) ppm for each Refclk source. Period does not
account for SSC induced variations.
VTX-DIFF-PP Differential p-p Tx voltage 0.8 (min) 0.8 (min) V As measured with compliance test load. Defined as
swing 1.2 (max) 1.2 (max) 2*|VtXD+-- Vtxd- |.
VTX-DIFF-PP Low power differential p-p 0.4 (min) 0.4 (min) V As measured with compliance test load. Defined as
Tx voltage swing 1.2 (max) 1.2 (max) 2*|VtXD+-- Vtxd- |.
Table 7-1 2.5 and 5.0 Gbps PCIe Gen2 Transmitter Specifications (Continued)
Symbol Parameter 2.5 Gbps 5.0 Gbps Units Comments
VTX-DE-RATIO-6dB Tx de-emphasis level N/A 5.5 (min) dB -
6.5 (max)
TMIN-PULSE Instantaneous lone pulse Not specified 0.9 (min) Ul Measured relative to rising/falling pulse.
width
TTX-EYE Transmitter Eye including 0.75 (min) 0.75 (min) Ul Does not include SSC or Refclk jitter. Includes Rj at
all jitter sources 10-12.
See Notes 2. 3, 4, and 10.
Note that 2.5 Gbps and 5.0 Gbps use different jitter
determination methods.
TTX-EYE-MEDIAN-to- Maximum time between 0 125 (max) Not specified Ul Measured differentially at zero crossing points after
MAX-JITTER the jitter median and max applying the 2.5 Gbps clock recovery function.
deviation from the median
TTX-HF-DJ-DD Tx deterministic jitter > 1.5 Not specified 0.15 (max) Ul Deterministic jitter only.
MHz
TTX-LF-RMS Tx RMS jitter< 1.5 MHz Not specified 3.0 ps Total energy measured over a 10 kHz-1.5 MHz
RMS range.
TTX-RISE-FALL Transmitter rise and fall 0 125 (min) 0.15 (min) Ul Measured differentially from 20% to 80% of swing.
time
TRF-MISMATCH Tx rise/fall mismatch Not specified 0.1 (max) Ul Measured from 20% to 80% differentially.
BWTX-PLL Maximum Tx PLL 22 (max) 16 (max) MHz Second order PLL jitter transfer bounding function.
bandwidth
BWTX-PLL-LO-3DB Minimum Tx PLL BW for 3 1.5 (min) 8 (min) MHz Second order PLL jitter transfer bounding function.
dB peaking
BWTX-PLL-LO-1DB Minimum Tx PLL BW for 1 Not specified 5 (min) MHz Second order PLL jitter transfer bounding function.
dB peaking
PKGTX-PLL1 Tx PLL peaking with 8 MHz Not specified 3.0 (max) dB Second order PLL jitter rransfer bounding function.
min BW
PKGTX-PLL2 Tx PLL peaking with 5 MHz Not specified 1.0 (max) dB -
min BW
RLTX-DIFF Tx package plus Si 10 (min) 10 (min) for dB -
differential return loss 0.05 -1.25 GHz
8 (min) for
1.25-2.5 GHz
RLTX-CM Tx package plus Si 6 (min) 6 (min) dB Measured over 0.05 — 1.25 GHz range for 2.5 Gbps
common mode return loss and 0.05 - 2.5 GHz range for 5.0 Gbps. (S11
parameter)
ZTX-DIFF-DC DC differential Tx 80 (min) 120 (max) (] Low impedance defined during signaling. Parameter
impedance 120 (max) is captured for 5.0 GHz by RLTX-DIFF.
Table 7-1 2.5 and 5.0 Gbps PCIe Gen2 Transmitter Specifications (Continued)
Symbol Parameter 2.5 Gbps 5.0 Gbps Units Comments
VTX-CM-DC-ACTIVE- Absolute Delta of DC 0 (min) 0 (min) mV |VTX-CM-DC-[during l0] - VTX-CM-Idle-DC[during Electrical
IDLE-DELTA Common Mode Voltage 100 (max) 100 (max) Idle]
during LO and Electrical
<= 100 mV
Idle.
VTX-CM-DC=DC(avg) of |VTX-D++VTX-D-|/2
VTX-CM--Idle-DC=DC(avg) of |VTX-D++VTX-D-|/2
VTX-CM-DC-LINE-DELTA Absolute Delta of DC 0 (min) 0 (min) mV |VTX-CM-DC-D+[during l0] - VTX-CM-DC-D[during l0]<= 25
Common Mode Voltage 25 (max) 25 (max) mV
between D+ and D- VTX-CM-DC-D+=DC(avg) of |VTX-D+| [during L0] VTX-CM-
DC-D-=DC(avg) of |VTX-D+| [during L0]
VTX-IDLE-DIFF-AC-p Electrical Idle Differential 0 (min) 0 (min) mV VTX-IDLE-DIFFp = | VTX-IDLE-D+ - VTX-IDLE-D-| <= 20 mV.
Peak Output Voltage 20 (max) 20 (max) Voltage must be high pass filtered to remove any DC
component.
VTX-IDLE-DIFF-DC DC Electrical Idle Not specified 0 (min) mV VTX-IDLE-DIFF-DC = | VTX-IDLE-D+ - VTX-IDLE-D-| <= 5
Differential Output Voltage 5 (max) mV.
Voltage must be low pass filtered to remove any AC
component. Filler characteristics complementary to
above.
VTX-RCV-DETECT The amount of voltage 600 (max) 600 (max) mV The total amount of voltage change in a positive
change allowed during direction that a Transmitter can apply to sense
Receiver Defection whether a low impedance Receiver is present.
Note: Receivers display substantially different
impedance for Vin <0 vs Vin > 0.
TTX-IDLE-MIN Minimum time spent in 20 (min) 20 (min) ns Minimum time a Transmitter must be in Electrical
Electrical Idle Idle.
TTX-IDLE-SET-TO-IDLE Maximum time to 8 (max) 8 (max) ns After sending the required number of ElOSs, the
transition to a valid Transmitter must meet all Electrical Idle
Electrical Idle after sending specifications within this time. This is measured
an EIOS from the end of the last UI of the last EIOS to the
Transmitter in Electrical Idle.
TTX-IDLE-TO-DIFF-DATA Maximum time to 8 (max) 8 (max) ns Maximum time to transition to valid diff signaling
transition to valid diff after leaving Electrical Idle. This is considered a
signaling after leaving debounce time to the Tx.
Electrical Idle
TCROSSLINK Crosslink random timeout 1.0 (max) 1.0 (max) ms This random timeout helps resolve potential
conflicts in the crosslink configuration.
LTX-SKEW Lane-to-Lane Output Skew 500 ps + 2 UI 500 ps + 4 UI PS Between any two Lanes within a single Transmitter.
(max) (max)
Table 7-1 2.5 and 5.0 Gbps PCIe Gen2 Transmitter Specifications (Continued)
Symbol Parameter 2.5 Gbps 5.0 Gbps Units Comments
CTX AC Coupling Capacitor 75 (min) 75 (min) nF All Transmitters should be AC coupled. The AC
200 (max) 200 (max) coupling is required either within the media or
within the transmitting component itself.
Notes:
1. SSC permits a +0. - 5000 ppm modulation of the clock frequency at the modulation rate does not exceed 33 kHz.
2. Measurements at 5.0 Gbps require an oscilloscope with at least 12.5 GHz bandwidth, or equivalent, while measurements made at 2.5 Gbps
require a scope with at least 6.2 GHz bandwidth. Measurement at 5.0 Gbps must deconvolve effects of compliance test board to yield an effective
measurement at Tx pins. 2.5 Gbps may be measured within 200 mils of Tx device's pins, although deconvolution is recommended. At least 106
Ul of data must be acquired.
3. Transmitter jitter is measured by driving the Transmitter under test with a low jitter "ideal" clock and connecting the DUT to a reference load.
4. Transmitter raw jitter data must be convolved with a filtering function that represents the worst case CDR tracking BW. 2.5 Gbps and 5.0 Gbps
use different filter functions. After the convolution process has been applied, the center of the resulting eye must be determined and used as a
reference point for obtaining eye voltage and margins.
5. Measurement is made over at least 106 UI.
6. Tx PLL Bandwidth must lie between the min and max ranges given in the above table. PLL peaking must lie below the value listed above. Note
that PLL B/W extends from zero up to the value(s) specified in the above table.
7. Measurements are made for both common mode and differential return loss. The DUT must be powered up and DC isolated, and Ms data+/data-
outputs must be in the low-Z state at a static value.
8. A single combination of PLL BW and peaking is specified for 2.5 Gbps implementations. For 5.O Gbps, two combinations of PLL BW and peaking
are specified to permit designers to make a tradeoff between the two parameters. If the PLL's min BW is >=8 MHz, the up to 3.0 dB of peaking is
permitted. If the PLL's min BW is relaxed to >= 5.0 MHz, then a tighter peaking value of 1.0 dB must be met. In both cases, the max PLL BW is
16 MHz.
9. Low swing output, defined by VTX-DIFF-PP-LOW must be implemented with no de-emphasis.
10. For 5.0 Gbps, de-emphasis timing jitter must be removed. This parameter is measured by accumulating a record length of 106 Ul while the DUT
outputs a compliance pattern. TMIN-PULSE is defined to be nominally 1 Ul wide and is bordered on both sides by pulses of the opposite polarity.
11. Root complex Tx de-emphasis is configured from Upstream controller. Downstream Tx de-emphasis is set via a command, issued at 2.5 Gbps.
Table 7-2 5.0 Gbps Limits for Common Refclk Rx Architecture (Continued)
Parameter Description Min Max. Units Comments
VRX-CM-CH-BRC Common mode noise from -- 300 mVPP 1
Rx
Notes:
1. Accumulated over 106 UI.
2. Minimum eye is obtained by first injecting maximum Dj and then adjusting Rj until a minimum eye (defined by TRX-EYE) is reached. Rj is filtered
with a BPF having FC-LOW and FC-HIGH of 1.5 MHz and 100 MHz, respectively with step rolloff at 1.5 MHz and a 20 dB/decade rolloff on the high
side. Minimum eye width is defined for a sample size equivalent to a BER of 10-12.
3. Different combinations of TRX-HF-DJ-DD and TRX-HF-RMS are needed to measure TRX-TJ-CC. and TRX-DJ-DD-CC.
Table 7-3 5.0 Gbps Tolerancing limits for Data Clocked Rx Architecture
Parameter Description Min Max Units Comments
Ul Unit interval without 200.06 ppm 199.94 ppm ps Over 106 Ul
including SSC
TRX-HF-RMS 1.5-100 MHz RMS jitter - 4.2 ps Spectrally flat, Note 1
RMS
TRX-HF-DJ-DD Max Dj impinging on Rx - 88 ps 1, 3
under tolerancing
TRX-LF-SSC-FULL Full 33 kHz SSC - 20 ns 1
TRX-LF-RMS 10 kHz to 1.5 MHz RMS - 8.0 ps Spectrally flat
jitter RMS
TRX-MIN-PULSE Minimum single pulse 120 - ps 1
applied at Rx
VRX-MIN-MAX-RATIO Min/max pulse voltage — 5 - 1
ratio seen over a time
interval of 2 Ul.
VRX-EYE Receive eye voltage 100 - mVPP 2
opening diff
VRX-CM-CH-BRC Common mode noise from - 300 mVPP 1
Rx
Notes:
1. Accumulated for 106 UI.
2. Minimum eye is obtained by first injecting maximum Dj and then adjusting Rj until a minimum eye (defined by TRX-EYE is reached. Rj is filtered
with a BPF having FC-LOW and FC-HIGH of 1.5 MHz and 100 MHz, respectively with step rolloff at 1.5 MHz and a 20 dB/decade rolloff on the high
side. Minimum eye width is defined for a sample size equivalent to a BER of 10-12.
3. Different combinations of TRX-HF-DJ-DD and TRX-HF-RMS are needed to measure TRX-TJ-DC and TRX-DJ-DD-DC.
Table 7-4 defines the parameters for 2.5 and 5.0 Gbps Receivers.
ZRX-DIFF-DC DC differential impedance 80 (min) Not specified C. For 5.0 Gbps covered under RLRX-DIFF parameter.
120 (max) See Note 5.
MDC
Tis Tih
MDIO
( input)
MDC
Tod
MDIO
( output)
TXD[4] TXD[9]
TX_CTL TXEN TXERR
TskewR
RXD[4] RXD[9]
RX_CTL RXDV RXERR
TskewR
NOTE: The TXC signal requires a typical delay of 1 to 1.5 ns. Use the PHY or switch
specification to determine the required delay.
Data RX
RXCLK
1.25 GHz 625 MHz DDR Clock
Clock
RXCLK
(single ended)
RXCLK
(differential)
RX
(single ended)
RX
(differential)
tclock2q (min)
tclock2q (max)
Note:
1. Measured at 50% of the transition
A small input buffer receives data through the external FIFO interface from the DMA module which will extract data
in 64-bit form. All subsequent processing prior to the final output is performed in bytes.
Transmit data can be output using the GMII/MII interface or through the TBI. If the TBI is selected (gigabit and
SGMII modes only), then the MAC transmitter passes 8-bitdata to the PCS sub-layer for further processing prior to
output on the TBI.
Frame assembly starts by adding preamble and the start frame delimiter. Data is taken from the transmit FIFO
interface a word at a time.
If necessary, padding is added to take the frame length to 60 bytes. CRC is calculated using an order 32 bit
polynomial. This is inverted and appended to the end of the frame taking the frame length to a minimum of 64
bytes. If the no CRC bit is set in the second word of the last buffer descriptor of a transmit frame neither pad nor
CRC are appended. The no CRC bit can also be set through the FIFO interface.
In full duplex mode, frames are transmitted immediately. Back to back frames are transmitted at least 96 bit times
apart to guarantee the inter frame gap. When operating in gigabit mode, late collisions are treated as an exception
and transmission is aborted, without retry. This condition is reported in the transmit buffer descriptor word 1 (late
collision, bit 26) and also in the transmit status register (late collision, bit 7). An interrupt can also be generated (if
enabled) when this exception occurs, and bit 5 in the interrupt status register will be set.
Several frames may be transmitted up to the burst limit of 65,536 bytes. The transmitter relinquishes control of the
medium when there are no more frames queued for transmission or the burst limit is exceeded.
In gigabit mode any collisions occurring after the minimum slot time for the first frame within a burst are treated as
a late collision. The burst is terminated upon this event.
In all modes of operation, if the transmit DMA under runs, a bad CRC is automatically appended using the same
mechanism as jam insertion and the tx_er signal is asserted. For a properly configured system this should never
happen and also it is impossible if configured to use the DMA with packet buffers, as the complete frame is
buffered in local packet buffer memory. By setting when bit 28 is set in the network configuration register the IPG
may be stretched beyond 96 bits depending on the length of the previously transmitted frame and the value written
to the IPG_STRETCH register. The least significant 8 bits of the IPG_STRETCH register multiply the previous
frame length (including preamble) the next significant 8 bits (+1so as not to get a divide by zero) divide the frame
length to generate the IPG. IPG stretch only works in full duplex mode and when bit 28 is set in the network
configuration register.
The IPG_STRETCH register cannot be used to shrink the IPG below 96 bits.
If the back pressure bit is set in the network control register, the transmit block transmits 64 bits of data, which can
consist of 16 nibbles of 1011 or in bit rate mode 641s, whenever it sees an incoming frame to force a collision.
If one of these events occurs, Wake on LAN detection is indicated. These events can be individually enabled. For
Wake on LAN detection to occur receive enable must be set in the network control register, however a receive
buffer does not have to be available.
The GEM has the ability to be programmed to detect when the PTP event messages (Sync, Delay Request,
PDelay Request and PDelay response) are transmitted and received and to measure the hardware timestamp
when those messages are sent or received.
The time-stamp is taken when the message time-stamp point passes the clock time-stamp point. In the case of the
GEM device this measurement is taken when a transmit or receive PTP Event message is processed at the rGMII
or sGMII interface depending on which interface the GEM is programmed for.
The GEM recognizes four different encapsulations for PTP event messages:
• 1588 version 1 (UDP/IPv4 multicast)
• 1588 version 2 (UDP/IPv4 multicast)
• 1588 version 2 (UDP/IPv6 multicast)
• 1588 version 2 (Ethernet multicast)
Table 8-7 to Table 8-8 are examples of 1588 frames in the 1588 version 1 format.
Table 8-7 Example of a Sync Frame in the 1588 Version 1 Format (Continued)
Table 8-8 Example of a Delay Request Frame in the 1588 Version 1 Format
For 1588 version 1 messages sync and delay request frames are indicated by the GEM if the type field of the frame
indicates TCP/IP, UDP protocol is indicated, the destination IP address is 224.0.1.129/130/131 or 132, the
destination UDP port is 319 and the control field is correct. The control field is 0xX0 for sync frames and 0xX1 for
delay request frames.
For 1588 version 2 messages the type of frame is determined by looking at the low nibble of the messagetype field
of the PTP frame. Whether a frame is version 1 or version 2 can be determined by looking at the version PTP field
of both version 1 and version 2 PTP frames.
In version 2 messages, sync frames have a message type value of 0xX0, delay_req have 0xX1, pdelay_req have
0xX2 and pdelay_resp have 0xX3.
Table 8-9 to Table 8-12 are examples of 1588 frames in the 1588 version 2 formats.
Table 8-9 Example of a Sync Frame in the 1588 Version 2 (UDP/IPv4) Format
Table 8-10 Example of a Sync Frame in the 1588 Version 2 (UDP/IPv6) Format
Table 8-11 Example of a Sync Frame in the 1588 Version 2 (Ethernet Multicast) Format
Table 8-12 Example of a Pdelay_req Frame in the 1588 Version 2 (Ethernet Multicast) Format
The GEM contains an optional time stamp unit (TSU) selectable with a tick define. The TSU consists of a timer and
registers to capture the time at which PTP event frames cross the message time-stamp point. An optional interrupt
can be issued when a capture register is updated.
The timer is implemented as a 64 bit register with the upper 32 bits counting seconds and the lower 32 bits
counting nanoseconds. The lower 32 bits rolls over when they have counted to one second. An optional interrupt
can be generated when the seconds increment. The timer value can be read, written and adjusted by software as
needed to implement timer adjustment algorithms as needed by IEEE 1588 clock synchronization algorithms.
The amount by which the timer increments each input reference clocking source cycle is controlled by the timer
increment register.
• Bits 7:0 are the default increment value in nanoseconds. If the rest of the register is written with zero the timer
increments by the value in 7:0 each clock cycle.
• Bits 15:8 of the increment register are the alternative increment value in nanoseconds
• Bits 23:16 are the number of increments after which the alternative increment value is used. If 23:16 are zero
then the alternative increment value will never be used.
An alternative way of adjusting time in the timer is to use the timer adjustment register. This allows for small
adjustments in time without the need to adjust frequency.
Using all of the above options in addition to any frequency adjustments (example, if the Network Timing Generator
is used to source the 1588 timing reference) allows the 1588 clock synchronization algorithms to chose when and
how to either adjust time, adjust reference frequency or both.
– NAND Flash controls: ALE and CLE will be assigned to upper expansion address bus.
• Supports hardware acceleration of multi-bit ECC for NAND, with the following features:
– BCH IP encodes data into a 16383-bit codeword (214 Galois Field)
– 2-32 bit error correction
– 2-1800 data bytes per block
– Pass through flow (no data FIFO)
– Hamming
Chip
APB Bus
GPIO
APB I/F
CS# 1 Ͳ 3
Generic
Data & Strobes
Peripheral
Device
Addr & Control #1Ͳ 3
Expansion Bus
Controller
AHB Bus
Block
AHB I/F Generic
Peripheral
Device
#0
or
Boot Flash
CS# 0
When the Expansion block is accessed by the system, the Expansion controller will fetch/store data from/to these
memory flash like devices until de-selected. Depending on the memory device type being accessed, the required
timing parameters are configured to comply with the device timing-requirements, before initiating any transactions.
RDY pulse sampling, when supported, will be synchronized to AHB clock (Hclk) before casing the de-assertion of
the CS and the RE or WE control signals. The read data will be latched upon de-assertion of the CS or RE signal,
whichever comes first. A configuration bit per chip-select is provided to enable/disable detection of the RDY signal
and should be disabled for the devices that do not provide the RDY signals. If RDY signal, when enabled, is not
detected, will cause an error response on the AHB interface.
As for the NAND Flash, when accessed (an AHB access to its space), the software will control the CLE and ALE
signals as well as the device Chip-select through a GPIO dedicated pins. The expansion controller will provide the
flash address/data/command and drive the NAND_WE # \NAND_RE# signals. Each NAND flash access phase will
be considered a full expansion bus access cycle.
Some peripherals may require the support for a single command signal to be used for Read/Write indication, along
with the command-strobing signal. A special configuration bit STRB_MODE is used to select the type of the
provided command signals: separate command signals: 1 for read and 1 for write, or 1 signal to be used for read/
write indications and the other signal to be used as the strobe signal. The active level of these 2 signals is
configurable.
Accesses to the expansion block are decoded internally to assert the appropriate chip-select of the targeted
device. 32-bit AHB input data will be latched in case of write operations. Depending on the peripheral bus size
used, the latched data will be transferred over decoded number of expansion cycles.
The maximum number of expansion cycles to write 32-bit data over an 8-bit bus peripheral will be 4 cycles,
meanwhile the AHB bus is halted until all data bytes are transferred. In case of read operations, the input data from
the peripheral device is latched and formatted over a 32-bit AHB bus according to the peripheral bus size selected.
The read data is duplicated over the 32-bit AHB data bus, that is, if MEM_SIZE is set to 8, the read byte will be
latched over all bytes of AHB data bus.
This block is highly configurable per chip select to accommodate for most common types of memory/flash in the
industry.
Expansion 25-bit generated address to the peripheral = {AHB Address Used Bits, Exp. Cycle Count Bits};
Expansion cycle counter increments on every complete expansion peripheral access cycle.
NOTE: AHB accesses outside the expansion chip-selects allocated memory segments will
cause an AHB Error response. Each chip select address map is configurable. CS0
boundary address map is defined by the CS0 memory base address register and the
CS0 memory segment address register.
NOTE: EXP_CS4 is muxed with the EXP_NAND_WE_N. Chip-select 4 address space will be
used by the NAND chip-select (GPIO pin) when NAND_MODE is enabled (default).
Setting NAND_MODE to 0 and CS4_EN to 1 will disable NAND support, and CS4 pin will
be used to access generic peripheral.
These modes support both synchronous and asynchronous devices. By default, synchronous support is set and
the output expansion clock is provided by setting EXP_CLK_EN global bit to 1 (high). When a asynchronous device
is accessed, EXP_CLK_EN global bit should be set to 0 (low) which will suppress expansion clock output.
Figure 9-2 shows the Normal Mode relative signals and their programmable fields.
Figure 9-2 Normal Mode Timing Diagram (RDY_EN = 0)
CS and the RE or WE command will be asserted, depend on the access type, after a programmed number of
expansion clock cycles from the active address. The expansion clock is a divide down frequency from the AHB
clock, it is generated from dividing down hclk by one of the configured divide integers (3, 4, 5, 6, or 7).
In the read cycle, the data bus is tri-stated and data is expected to be ready sometime after the assertion of the RE
command. The expansion controller will latch the data upon de-assertion of either the RE command or the CS
signal, whichever comes first. The CS is terminated after the expiration of the CS width count or upon detection of
the de-assertion of the RDY signal provided by the device (when enabled by setting EXP_RDY_EN to 1).
During the Write cycle, CS and WE command are asserted some programmable cycles after the valid address.
The external device is expected to latch the data by the de-assertion of the WE command. The data will be held
valid for some programmable cycles after the de-assertion of the CS.
Next Read/Write transaction can only be started after the expiration of the Data-hold time count.
When EXP_RDY_EN bit is set high, detected RDY pulse will be synchronized to AHB clock (HClk) before causing
the de-assertion of the CS and the RE or WE control signals. As was mentioned above, that the read data will be
latched upon de-assertion of the CS or RE signal, whichever comes first. A configuration bit per chip-select is
provided to enable/disable detection of the RDY signal and should be disabled for devices that do not provide the
RDY signals. If case of RDY is enabled and was not detected by the expansion controller before the de-assertion of
the CS, an error response will be indicated on the AHB interface for this transaction.
In Figure 9-3 in regards of the read and write commands, when a read access is being performed, only RE signal
will toggle. Also, when a write command is being performed, only WE command will toggle, not both.
Table 9-7 depicts the list of signals that are valid when the Synchronous mode is set:
Figure 9-4 shows the ALE Mode relative signals and their programmable fields.
Figure 9-4 ALE Mode Timing Diagram
Figure 9-5 shows the Strobe Mode relative signals and their programmable fields.
Figure 9-5 Strobe Mode Timing Diagram
NOTE: In this mode, both EXP_RW# and EXP_STRB# signals shall be switching within read
and write transactions. The read data will be sampled by the expansion block upon de-
assertion of the EXP_CS or EXP_STRB, whichever comes first.
NOTE: For NAND mode, chip select will be used from a GPIO pin. In the Transcede 2xxx
GPIO28 is used to drive nand_cs when in NAND mode.
When NAND mode is disabled, NAND_MODE=0, then exp_nand_we_n can be used as
a generic Flash chip select.
The chip select and latch controls for this mode are provided by a dedicated GPIO pin. NAND_RE/NAND_WE
commands, address and data are initiated by the expansion block as an expansion cycle. Each NAND transaction
should be broken up into several expansion (Normal Mode type) transactions. The software has the ultimate
control over NAND transactions. NAND ECC support is also provided by software only.
Table 9-9 depicts the list of signals that are valid when the NAND mode is set:
Figure 9-6, Figure 9-7, and Figure 9-8 show the timing diagrams respectively for command latch, address latch,
and data latch in the NAND mode.
Figure 9-6 NAND Command Latch Timing Diagram
NAND_CLE (GPIO)
NAND_CS (GPIO)
NAND_WE
NAND_ALE (GPIO)
EXP_Data Command
NAND_CLE (GPIO)
NAND_CS (GPIO)
NAND_WE
NAND_ALE (GPIO)
NAND_CLE (GPIO)
NAND_CS (GPIO)
NAND_ALE (GPIO)
NAND_WE
Each toggle of WE or RE to access the NAND flash will be executed over a full expansion cycle, from the point view
of the expansion bus, it is an expansion cycle in normal mode. The software initiates a transaction in the NAND
allocated address space, which will cause an expansion cycle to be started. During the NAND Address cycle or the
NAND Command cycle, the software will initiate the write transaction to the NAND address space placing the
NAND address or command type value on the data bus. The expansion controller will issue a write access to the
NAND flash, as if accessing a generic device. The software at the meantime will drive the NAND chip-select and
the controls (CLE/ALE) through dedicated GPIO pins.
NOTE: NAND_RE and NAND_WE will not be toggled when there is an access to CS#0-3.
9.6 Clocks
One clock is input to the expansion block running at AHB clock rate (150 MHz). The expansion block has a built in
ExpClk Generator, which is responsible of generating a divided synchronous clock to the external peripherals with
a maximum of 50 MHz. A clock switch selects, glitch-free, between AHB clock and the divider output, and shall
default to AHB Clock upon reset and will switch to the divided AHB clock right after reset ends.
Power down logic shall be default to activate the clock. When reset is asserted, output clock might have glitches
when switching to reference clock.
This clock generator circuitry can be configured to divide the AHB clock with the following divide values: 3-6 and 7.
The internal logic of the expansion bus will be running at AHB clock speed, and the generated synchronous clock
(EXPCLK) will be treated as a synchronizing pulse enable for the programmable timing counters.
Figure 9-9 shows the expansion bus clock generation.
Figure 9-9 Expansion Bus Clock Generation
Divide Ratio
clk
ratio [4:0] switch
9.7 Reset
Asynchronous reset will be used to reset the block’s internal logic. This reset may be a combination of the HW and
SW resets provided by the chip and is synchronized externally to the AHB clock. A self-clear soft reset can be
generated by the block’s internal configuration logic when asserted by software. This self-clear reset will reset the
entire block’s logic synchronously. Configuration registers are not affected by the block’s soft-reset.
9.9 Memories
No memories are instantiated in this block.
All I2C devices must have a unique address to identify it on the bus. Slave devices have a predefined address, but
the lower bits of the address can be assigned to allow for multiples of the same devices on the bus.
The Transcede 2xxx device operates as a master, a slave, or in multi-master mode. It supports all I2C speeds:
standard (100 kbps), fast (400 kbps) and high speed (3.4 Mbps). The Transcede 2xxx I2C interface includes a
programmable clock divider to allow adjustment of data speed over a wide range. However, the Transcede 2xxx
I2C interface does not include an output signal for direct control of a bridge between fast/standard and high-speed
bus segments. Thus, the Transcede 2xxx I2C interface is typically configured to accommodate the slowest device
on the bus.
I2C_SCL
I2C_SDA
I2C_SCL
Tis Tih
I2C_SDA
(input)
Tod Toh
I2C_SDA
(output)
Tsu;sta (Repeated) START condition setup time 160 - ns See note 1 and 2
Thd;sta (Repeated) START condition hold time 160 - ns See note 1 and 2
ss_o ss ss
ss
ss_x
.
.
.
Slave Slave
ss ss
A B
The SPI Controller can connect to any serial slave peripheral device that supports the Motorola Serial Peripheral
Interface (SPI) – a four-wire, full-duplex serial protocol. There are four possible combinations for the serial clock’s
phase and polarity. The clock phase (SCPH) determines whether the serial transfer begins with the falling edge of
the slave select signal or the first edge of the serial clock. The slave select line is held high when the SPI controller
is idle or disabled.
The serial protocols supported by the SPI controller allow for serial slaves to be selected/addressed using either
hardware or software. When implemented in hardware, serial slaves are selected under the control of dedicated
hardware select lines. The number of select lines generated from the serial master is equal to the number of serial
slaves present on the bus. The serial master device asserts the select line of the target serial slave before data
transfer begins. When implemented in software, the input select line of each serial slave can either originate from a
single slave select output pin on the serial master (user must configure the master to have one slave select output)
or be permanently grounded. The main program in the software domain controls the selection of the target slave
device.
The SPI controller does not enforce hardware or software control for serial slave device selection. The user can
configure the SPI controller for either implementation.
SPI_SCLK
Tis Tih
SPI_RXD
Tod Toh
SPI_TXD
In the asynchronous mode, the transmitter and receiver are set to operate at a given baud rate from a table of
standard baud rates. Other than the baud information, no clock is passed between them. The receiver is normally
programmed to automatically determine the baud rate of the transmitter. Absolute baud rate accuracy is not
required because the receiver relocates the center of the start bit of each new word or byte. It is this feature that
enables the asynchronous mode to operate without sharing the actual baud rate clock. The receiver rejects false
bits.
On the Rx direction, there is a read-only register (called RBR) contains the data byte (data character) received on
the serial input port (sin). There are also FIFOs on the receive side and the FIFOs can be enabled or disabled by
software. When FIFOs are disabled, the data in RBR must be read before the next data arrives, otherwise it will be
overwritten, resulting in an overrun error. When FIFOs are enabled, RBR accesses the head of the receive FIFO. If
the receive FIFO is full and RBR is not read before the next data character arrives, then the data already in the
FIFO will be preserved but any incoming data will be lost. An overrun error will also occur.
On the Tx direction, there is a transmit holding register (called THR) that contains the data to be transmitted on the
serial output port (sout). There are also FIFOs on the transmit side and FIFOs can be enabled or disabled by
software. When FIFOs are disabled, a single character can be written to THR when THR is empty. Any additional
writes to the THR before it is empty again causes the previous data in THR to be overwritten. When FIFOs are
enabled, data character can be written to the transmit FIFO when it is not full. Any attempt to write data when FIFO
is full results in the write data being lost.
Thgp
GPIO
(input)
NOTE: The support for 512 timeslots is enabled as a separate mode by software configuration.
When the mode is enabled, only TDM0 interfaces will be active.
NOTE: Slower CLK frequencies can be used for fewer timeslots. For example:
• 32.768 MHz for 512 timeslots per bus.
• 16.384 MHz for 256 timeslots per bus.
• 8.192 MHz for 128 timeslots per bus.
TDM_CK
Tfs Tfh
TDM_FS
(input)
Tdis Tdih
TDM_DR
Tdod Tdoh
TDM_DX
TDM_FS
(output)
Tdoz
– Guard time
– Timeout timers
• Automatic operating voltage class selection
Figure 16-1 USIM Block
In case that no operating class is enabled, the controller performs activation for all three voltage classes (1.8V, 3V
and 5V) in sequence.
When Smart Card Reader performs activation of contacts the lowest enabled voltage class is automatically applied
first. When the ATR sequence is received, the selected voltage class is correct. When the ATR sequence reception
fails, ATRFAIL interrupt is not activated, deactivation is performed and next higher enabled voltage class is applied.
If the ATR sequence reception fails and no other higher class is enabled or the highest Class A was already applied
the ATRFAIL interrupt is activated and the last applied voltage class remains active.
After the automatic voltage class selection is finished the selected class can be read from bits VCC18, VCC33 and
VCC50 in CTRL2 register. If the automatic voltage class selection fails, these bits remain untouched.
There is a delay applied between deactivation of contacts with lower voltage class and activation of contacts with
higher voltage class. This delay should be at least 10 ms. This delay is realized by running of N-bit timer until it is
full. Width of the timer is given by parameter CLASSDELAYWIDTH in SCR parameters file.
Waiting
Activation Cold reset for ATR ATR response
SCVCC
SCVPPEN
Tc
SCRSTN
SCCLK
Td
Reception
ATR response
SCIO mode – Hi-Z
Ta Ta Ta
The first character of the ATR sequence is the TS character which defines the coding convention used by the card.
The TS information is decoded and stored as bits INVLEV and INVORD in the CTRL1 register. Therefore, storing
the TS character in the RX FIFO is not necessary. You can disable it by clearing the TS2FIFO bit in the CTRL1
register. The convention defined in the TS character is used for all transfers after the TS character reception. You
can modify the INVLEV and INVORD values after the ATR sequence reception, when needed.
After the ATR sequence is received, the superior system reads the ATR from the RX FIFO and programs the Smart
Card Reader according to extracted ATR values.
Deactivation
SCVCC
SCVPPEN
SCVPPPP
SCRSTN
SCCLK
SCIO
Ta Ta Ta Ta Ta
SCVCC
Tc
SCRSTN
SCCLK
Td
RPDIF1_MCLK
RPDIF1_RX_FRAME
6 P0_D[0:5]
RF RFIC
Front T2xxx
End AD936x RPDIF1_FBCLK
RPDIF1_TX_FRAME
6 P0_D[6:11]
Figure 17-3 Interfacing with AD936x in Dual Port Full Duplex - CMOS Mode
RPDIF1_MCLK
RPDIF1_RX_FRAME
6 P0_D[0:5]
6 P0_D[6:11]
RF RFIC
Front T2xxx
End AD936x RPDIF1_FBCLK
RPDIF1_TX_FRAME
6 P1_D[0:5]
6 P1_D[6:11]
• RPDIF has two mandatory ports namely TX_FRAME and RX_FRAME which are used to indicate the valid
data in transmission and reception operation respectively.
• RPDIF11_TX_FRAME and RPDIF1_RX_FRAME signals can operate in two modes. Level mode and Pulse
mode.
• RPDIF has two data ports of 12 bit P0 and P1 which enable to operate both in SDR (Single Data Rate) mode
and DDR (Dual Data Rate) mode. Also allows differential signaling in a different mode.
• RPDIF supports two IO types:
– CMOS (single ended signaling)
– LVDS (Differential signaling)
• CMOS IOs support the following modes:
– Single port TDD Mode
– Single port Full Duplex Mode
– Dual port TDD Mode
– Dual bus Full Duplex Mode
• LVDS IOs support FDD and TDD using differential signals..
RPDIF1_MCLK_P
RPDIF1_MCLK_M
RPDIF1_RX_FRAME_P
RPDIF1_RX_FRAME_M
12 RPDIF1_RX_P[5:0], RPDIF1_RX_M[5:0]
RF AD936x T2xxx
Front RPDIF1_FBCLK_P
End RPDIF1_FBCLK_M
RPDIF1_TX_FRAME_P
RPDIF1_TX_FRAME_M
12 RPDIF1_TX_P[5:0], RPDIF1_TX_M[5:0]
RPDIF1_MCLK
RPDIF1_RX
2T2R Timing Shown
_FRAME
P1 D[5:0] Q I Q I Q I Q I Q I Q
RPDIF1_
FBCLK
RPDIF1_TX
2T2R Timing Shown
_FRAME
P0 D[5:0] Q I Q I Q I Q I Q I Q
RPDIF 1 RFIC #1
4
RF-SPI 1 4
MUX
4
SPI 0 SPI0_SS0/1/3 SPI Devices
3
SW Control #2
RPDIF 0
4 RFIC #2
MUX
RF-SPI 0 4
4
SPI 1 SPI1_SS1 SPI Device
tSRD/tSRC
tHRD/tHRC
RPDIF_MCLK
tDDTD/tDDTC
RPDIF_FBCLK
RPDIF_RX_FRAME
RPDIF_DIQ0/1 TI TQ
TI TQ RI RQ RI RQ
[11:0]
RPDIF_TX_FRAME
RPDIF_ENABLE
tDDTXNRX2ENA tDDTXNRX2ENA
RPDIF_TXNRX
tCP tMP
RPDIF_MCLK_P
RPDIF_MCLK_M
RX_FRAME_P
RX_FRAME_M
tSRD/tSRC
RPDIF_RX_P/M IL QL IM IL QL
IM QM QM
[5:0]
tHRD/tHRC
tCP tMP
RPDIF_FBCLK_P
RPDIF_FBCLK_M
tDDTC
RPDIF_TX_FRAME_P
RPDIF_TX_FRAME_M
tDDTD
RPDIF_TX_P/M IL QL IM IL QL
IM QM QM
[5:0]
RPDIF_ENABLE
tDDTXNRX2ENA tDDTXNRX2ENA
RPDIF_TXNRX
Sytem
Timestamp
Counter
Cross Trigger Matrix CTM interface
Timestamp Distribution
CTI
L2
CORTEXA9 INTEGRATION
Cross Trigger Matrix F ETB
CTI
ITM
CORTEX
A9MP
SCU
PTM Replicator
CortexͲA9
ROM
Table
F TPIU
Debug APB
Debug APB
PC based SWD/JTAG
Debug Tool External pin interfaces
JTAG/SWD Trace Port
Emulator Analyzer
Figure 19-2 shows the memory architecture of IRAM. The memory size of IRAM is specified as 256KB.
Figure 19-2 IRAM Architecture
AXI Ͳ
DW_axi_gs –
RD/WR
Read/Write
(64 bit)
word1 Byte7 Byte6 Byte5 Byte4 Byte3 Byte2 Byte1 Byte0 Byte7 Byte6 Byte5 Byte4 Byte3 Byte2 Byte1 Byte0
word0
word3 word2
. 6Kx16 6Kx16 6Kx16 6Kx16 6Kx16 6Kx16 6Kx16 6Kx16 .
. .
pm_mem pm_mem_ pm_mem_ pm_mem_ pm_mem pm_mem pm_mem pm_mem
_r_blk3 r_blk2 r_blk1 r_blk0 _l_blk3 _l_blk2 _l_blk1 _l_blk0
address bit 3
Mux
NOTE: Setting the PCIe or USB interfaces to secure mode can allow external masters to access
internal secure areas. This will most likely break 3GPP TrE compliance.
Specific address ranges that map to DDR, IRAM or the expansion bus may be set to secure mode. Both the IRAM
and expansion bus have corresponding registers to set the low and high address of the secure mode area. Only
one secure window can be set up for each. The DDR controller has four separately configurable ports, each with up
to four secure areas that can be set up.
The Transcede 2xxx also provides a secure timer, which is always in secure mode.
word1 Byte15 Byte14 Byte3 Byte2 Byte1 Byte0 Byte15 Byte14 Byte3 Byte2 Byte1 Byte0 word0
word3 word2
MAP
FP MAP
FP MAP
FP MAP
FP APB
APBreg
reg
0 1 2 3
FIFO
FIFO FIFO
FIFO Master
MasterControl
Control
AXI Bus
• There is only one APB interface module that connects to the two MAPs, and the internal APB bus can be used
for high speed data transfer (Inter MAP transfer) between MAPs.
• MAPDMA can be configured as multi-channel DMA by enabling the FHEAD_EXP module. The combined
group allows independent descriptor chain for each MAP.
• Each MAP has comprehensive debug support. In debug mode, it can be stopped with specified instruction
pointer and step the execution. Also the registers can be read and write in debug mode through APB access.
64 bits 64 bits
IO
Mbank_0 32
2Kx(2x16) 32
Mbank_1
2Kx(2x16)
Mbank_2
2Kx(2x16) Data Path
Mbank_3
2Kx(2x16)
Mbank_4 48
2Kx(2x24) 48
4xMAC
Mbank_5
2Kx(2x24) 16 x 24 - bit
Mbank_6
2Kx(2x24)
Mbank_7
2Kx(2x24)
Program Mem
APB
Control Unit 512x160
Address Unit
8 Interrupt
Bank_sw I/O_Cntrl Event_reg Debug_cnt
8 8 16 24
000-7FF Segment_15
Bank 7
000-7FF Segment_14
000-7FF Segment_13
Bank 6
000-7FF Segment_12
000-7FF Segment_11
Bank 5
000-7FF Segment_10
000-7FF Segment_9
Bank 4
000-7FF Segment_8
000-7FF Segment_7
Bank 3
000-7FF Segment_6
000-7FF Segment_5
Bank 2
000-7FF Segment_4
000-7FF Segment_3
Bank 1
000-7FF Segment_2
000-7FF Segment_1
Bank 0
000-7FF Segment_0
MAP data memory is divided into 16 memory segments and grouped into 8 memory banks, with each two
segments form one memory bank. The word length of segment 0 to 7 is 16 bit and the segment 8 to 15 is 24 bit.
The depth of memory segment is 2K words, and each memory segment has its own address space.
2 KB Write Broadcast
offset1=0xF800h
2 KB Master Control
offset1=0xF000h
~ ... ~
~ ~
offset1=0xE800h
1 KB FHEAD_EXP 0
offset1=0xE400h
1 KB MAPDMA 0
offset1=0xE000h
~ ... ~
~ ~
offset1=0x2000h
2 KB MAP3
offset1=0x1800h
2 KB MAP2
offset1=0x1000h
2 KB MAP1
offset1=0x0800h
2 KB MAP0
offset1=0x0000h
(Base Address)
Every block in the MAP Processor Array is assigned with 2 Kbyte of APB memory space. This includes four MAP
blocks and the Master Control block. The address of APB bus register inside of each block will have:
Block Base Address = Base Address + offset 1
The Base Address of MAP Processor Array APB registers is 0xF3D4_0000.
Besides these blocks of address space assigned to each module, there is a Write Broadcast block at offset1 =
0xF800h that allows a single APB bus write to any address in this block to be translated into simultaneous
broadcast APB write to any specified group of MAP blocks. In order words, a write to Write Broadcast block APB
memory space will assert the "p_sel" of every selected MAP block, and thus enable simultaneous write to more
than one MAP blocks.
NOTE: Data transferred from memory to MAP must be a multiple of 8 bytes. When programming
Buffer descriptors, Bpointer[2:0] and Bcontrol[2:0] must be 0 (8-byte boundary.)
Data transferred from MAP to memory must be a multiple of 8 bytes. When programming
Buffer descriptors, Bpointer[3:0] must be 0 (16-byte boundary) and Bcontrol[2:0] must be
0 (8-byte boundary.) Scatter mode can still be used but buffers need to be at 16-byte
boundary. Historically, Scatter mode was not used.
Figure 21-5 shows the MAP DMA and FFT_MC block diagram.
Figure 21-5 MAP DMA and FFT_MC Block Diagram
5 Sets of Controls
MAP Data In[63:0] MAP Data Out[4:0][63:0]
Valid/Ready Handshake
5:1
64 MAP Master Control APB
Mux
64
Tx FIFO Rx FIFO
Clock Domain Crossing
64x64 512x64
APB
DMA
IRQs
Linked List-Transfer
AXI
Queue
Access Frame Descriptors
BPointer BPointer
BControl BControl
Data Buffers
Table 21-1 Inbound (Tx) Frame and Buffer Descriptors Programming (Continued)
Bit Name Description
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
2 BRESP_N 1: No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 Reserved. Must be 0
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 5 TR_CTRL Tr_Control bits [26:0].
bit [6:5] = load_type,
bit [7] = load_cont,
bit [15:8] = start_bank,
bit [16] = start_seg,
bit [18:17] = bus_size,
bit [19] = data_type,
bit [22:20] = tr_interrupt,
bit [31:23] = start_addr[8:0],
4 WSTAY Stay working at the current MAP (inbound only)
3 XPAN 3G Expander: Two’s complement a data bit to a byte
Note: This bit is for Transcede 2120/2150/2200.
2–0 FPID MAP number to be served. FPID=7 means to select a MAP from the FPenable register (Pool
process.) For Target transfer:
FPID = 0 -> MAP0
FPID = 1 -> MAP1
FPID = 2 -> MAP2
FPID = 3 -> MAP3
Frame Word3 = FControl2. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 29 Not used
28 – 16 TR_CTRL Tr_Control bits [39:27].
bit [17:16] = start_addr[10:9],
bit [28:18] = seg_count,
15 – 8 D2MBdesc Must be zero.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer Words always start at offset x10.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer. Must be a multiple of 8 bytes.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Not used
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes. Must be a multiple of 8 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Table 21-1 Inbound (Tx) Frame and Buffer Descriptors Programming (Continued)
Bit Name Description
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 21-2 Outbound (Rx) Frame and Buffer Descriptors Programming (Continued)
Bit Name Description
30 – 29 Not used
28 – 16 TR_CTRL Tr_control bits [39:27].
bit [17:16] = start_addr[10:9],
bit [28:18] = seg_count.
15 – 8 D2MBdesc Dma2Mem (from I/O to memory) Number of Bpointer/BControl pairs. Buffer Words always
start at offset x10
7– 0 M2DBdesc Must be zero.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer. Must be a multiple of 16 bytes.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Not used
23 – 0 BLEN Number of bytes to be transferred. Zero means none. Maximum 16Meg – 1 bytes. Since the FFT
interface data bus is 64 bits wide, BLEN must be a multiple of 8. If scattering, BLEN must be a
multiple of 16 unless it is for the last buffer.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
NOTE: FEC accelerator for 3G is for Transcede 2120, Transcede 2150, and Transcede 2200.
– 2nd interleaving
– Physical channel mapping
– HS-DSCH bit scrambling
– HS-DSCH HARQ functionality
– HS-DSCH constellation re-arrangement
– HS-SCCH coding, types 1,2,3
– E-AGCH coding
• 3G WCDMA UL
– 2nd de-interleaving
– Physical channel concatenation
– Transport channel de-multiplexing
– Rate de-matching
– Radio frame concatenation
– 1st de-interleaving
– Channel decoding: CC decoder, CTC decoder
– CRC check
– HARQ combining
– Control channel Reed-Muller decoding
NOTE: 3G WCDMA DL and 3G WCDMA UL features are for Transcede 2120, Transcede 2150,
and Transcede 2200.
M1
CRC
checker
HARQ Write
Back Bypass
M5
M4
HARQ BYP CC,CTC LTE
LTE- CTC/CC Tx
sub block CC Tail Uplink
Rx
48bits Soft de- Viterbi Biting / DMA
Uplink M1 M3
Descra Combine nterleaver Decoder Buffer
/ DMA
mbler
LTE Channel
LTE de-
de-
Interleave
rate match HARQ CC,CTC 3G CTC, 3G and 4G 64 bits
64 bits De-mux
Memory sub block Decoder
De-concat de-
3G- CTC/CC
nterleaver
3G 2nd de-
3G de-
interleave
Rate
De-mux
Match Block
De-concat
Decoder
General
CRC
DownLink
Encode
Parameters
Rate Match
127 0 Interleaver
Addr
Scrambler
Control
Control
Size
General
CRC
Frame
Descriptor Decode
Uplink
Data DeRatematch Parameters
Total
Size De-interleave
HARQ
Queue
Access Frame Descriptors
BPointer BPointer
BControl BControl
Data Buffers
NOTE: The memory address of a frame descriptor Word0 must be 16-byte align. The least
significant 4 bits of Frame Head and Frame Next must be 0. Frame and Buffer Words
must be consecutive.
For DMA to start operating, Frame Head must not be 0. For DMA to continue, Frame
Next must not be 0.
When Register FSYN = 1, only 8 Bpointer/BControl pairs are allowed. Both In and
Outbound share the same frame descriptor.
NOTE: In this section, some different naming conventions have been inherited.
• IO = I/O Application such as FEC or FFT
• Inbound = Data are moved from Memory to DMA = Mem2Dma = M2D = Transmit = From Memory to
IO = M2IO
• Outbound = Data are moved from DMA to Memory = Dma2Mem = D2M = Receive = From DMA to
Memory = IO2M
Table 22-4 Mem2Dma (or memory to I/O) Frame and Buffer Descriptor
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 – 3 Reserved
2 BRESP_N 1: No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 FPAUSE Pause after finishing this frame.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Zero.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer Words always start at offset x10.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Table 22-4 Mem2Dma (or memory to I/O) Frame and Buffer Descriptor (Continued)
Bit Name Description
Buffer Word 1 = BControl. Offset: x14
31 – 30 Reserved. Must be 0.
29 – 27 LAST_BYTE_HOLE Beginning invalid bit (or hole) location in the last byte. Bit numbering method: 0 to 7, counting
from left to right. Zero means all bits are valid. If BLEN = 1, Last_Byte_Hole must be greater
than First_Byte_Bit. In this case, Zero means 8 or no holes after the First_Byte_Bit.
26 – 24 FIRST_BYTE_BIT Beginning valid bit location in the first byte. Bit numbering method: 0 to 7, counting from left to
right. If BLEN = 1, Last_Byte_Hole must be greater than First_Byte_Bit.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 22-5 Dma2Mem (or from I/O to memory) Frame and Buffer Descriptor
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 – 3 Reserved
2 BRESP_N No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 FPAUSE Pause after finishing this frame.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Dma2Mem (from I/O to memory) Number of Bpointer/BControl pairs. Buffer Words always
start at offset x10
7– 0 M2DBdesc Zero.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
Table 22-5 Dma2Mem (or from I/O to memory) Frame and Buffer Descriptor (Continued)
Bit Name Description
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 22-6 Common, or 1, Frame and Buffer Descriptor for both Mem2Dma and Dma2Mem (FCOM)
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 – 3 Reserved
2 BRESP_N No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 FPAUSE Pause after finishing this frame.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Dma2Mem Number of Bpointer/BControl pairs. Buffer descriptors start at offset x10 + 8 *
M2DBdesc.
D2MBdesc + M2DBdesc = 8 maximum if Register FSYN = 1.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer descriptors always start at offset x10.
D2MBdesc + M2DBdesc = 8 maximum if Register FSYN = 1.
Mem2Dma Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Mem2Dma Buffer Word 1 = BControl. Offset: x14
31 – 30 Reserved. Must be 0.
29 – 27 LAST_BYTE_HOLE Beginning invalid bit (or hole) location in the last byte. Bit numbering method: 0 to 7, counting
from left to right. Zero means all bits are valid. If BLEN = 1, Last_Byte_Hole must be greater
than First_Byte_Bit. In this case, Zero means 8 or no holes after the First_Byte_Bit.
26 – 24 FIRST_BYTE_BIT Beginning valid bit location in the first byte. Bit numbering method: 0 to 7, counting from left to
right. If BLEN = 1, Last_Byte_Hole must be greater than First_Byte_Bit.
Table 22-6 Common, or 1, Frame and Buffer Descriptor for both Mem2Dma and Dma2Mem (FCOM)
Bit Name Description
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Mem2Dma Buffer Word N = BPointer. Offset: x10 + N*4 (N = even integer)
Mem2Dma Buffer Word N+1 = BControl. Offset: x14 + N*4 (N = even integer)
…..
Dma2Mem Buffer Word 0 = BPointer. Offset: 8 * M2DBdesc + x10
31 – 0 BPOINTER Address pointer of a data buffer.
Dma2Mem Buffer Word 1 = BControl. Offset: 8 * M2DBdesc + x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Dma2Mem Buffer Word M = BPointer. Offset: 8 * M2DBdesc + x10 + M*4 (M = even integer)
Dma2Mem Buffer Word M+1 = BControl. Offset: 8 * M2DBdesc + x14 + M*4 (M = even integer)
Frame Word0 = FNext. Offset: x0
aes_wrapper
Memory
Context
AES
Core
Message
Memory
Inbound
des_wrapper
Packet
Memory
Context
Processor
DES
Core
Message
Memory
Security
Resource
Manager
Slave Interface
Decoder
hmac_adv_wrapper
Decoder
Arbiter
Memory
Context
Memory
Secret
Hashing
Master Interface
Outbound Core
Packet
Processor
Message
Memory
SDMA
AXI
System
Bus
AXI AXI
Master Slave
SPAccͲPDU
Register/ SDMA
Decode
Interface
Sequencer
Cipher Data Hash Data
Buffer Buffer
23.4.1 Overview
The True Random Number Generator (TRNG) generates random data that is statistically equivalent to a uniformly
distributed random data stream. The circuit includes a noise generator which creates non-deterministic random
noise to seed a noise whitener circuit. The output of the noise whitener is the random number.
The TRNG can operate in one of two modes: self-seeding mode (also called random mode) and nonce seeding
mode. In self-seeding mode, the ring oscillators automatically produce a seed that is fed to a whitening circuit. The
initial automatic self-seeding takes on the order of 260,000 clock cycles to complete. After the initial self-seeding
phase, the TRNG will produce pseudo-random numbers at a rate of one 128-bit random number every 512 clock
cycles.
In nonce-seeding mode, a user specified is injected into the noise generation circuit as an alternate way of seeding.
In this case the TRNG is ready to generate random numbers as soon as the nonce write operation is complete.
The host may initiate a nonce TRNG seeding operation at any time after reset. A self-seeding operation may be
initiated whenever another self-seeding operation is not under way.
The supported feature set is listed as below:
• Random Number Generation (TRNG)
– Internal random (re)seed operation
– Host-driven nonce reseed option
– 128-bit random number generation
Ring Ctrl
Ring 5
XOR D Q D Q Accumulator
Ring 4 G G
XOR D Q D Q XOR
Ring 1 G G LFSR
The random noise seed generator is based on a number of ‘relatively prime’ length ring oscillators. Several of these
ring oscillators are mixed together and delivered to accumulators. (The term ‘mixed’ here is used in the modulation
sense of the word. The modulator is this case is a multi-input XOR gate.)
1. Each ring length is based on a different prime number to minimize periodic behaviour.
2. The lengths of rings 0 through 5 are all different and chosen such that the middle frequency is approximately 2
times the system clock frequency. Each ring is chosen to be a different prime number of inverting elements.
The frequencies are chosen to ensure that sufficient edges are seen at the sampling flip flop to ensure the flip
flop enters metastable timing regions a reasonable fraction of the time. The mixing function is a standard logical
exclusive-OR (XOR).
3. Rings 1 through 5 are mixed and delivered to sampling flop #1. The sample flop is designed to be driven into
metastability some fraction of the time and the result is accumulated through two non- overlapping time
windows.
4. Rings 0 through 4 are mixed and delivered to sampling flop #2 This set is designed to both perturb the seed
LFSR while driving the front end LFSR flip flop into metastability some fraction of the time.
Seed bits are created until the whitening element declares it has been completely seeded. However, in order to
maximize entropy, the seed LFSR runs continuously, even after the whitener declares it has been completely
seeded, unless the rings are shut down by an external circuit.
Nonce Reseed
LILI-2 Key Stream Command
Free Random Generator
Noise Generator Seed Cntr
Random Reseed
Command
Bit Cntr Shift Reg
The TRNG can operate in one of 2 modes: random re-seed mode and nonce re-seed mode.
In random re-seed mode, the TRNG uses a set of free running ring oscillators to generate a random seed for a
Pseudo Random Number Generator (PRNG). The random reseed operation takes between 200,000 and
300,000 clock cycles. The PRNG continuously generates 128-bit pseudo random numbers based on the random
seed. The PRNG can be reseed from the ring-based seed generator. After reset the TRNG automatically performs
a random re-seed operation. When a random reseed operation is under way, the host must not initiate a new
random reseed operation.
In nonce reseed mode, the TRNG is “manually” seeded / reseeded via a host defined nonce. After the reseed
operation, the PRNG continuously generates 128-bit pseudo random numbers based on this seed. A new nonce
reseed operation may be initiated at any time. The nonce reseed operation may even be initiated while a random
reseed operation is under way, as this automatically terminates the ongoing random reseed operation.
The TRNG_CTRL register allows the host to select several different modes of operation.
When written with a 1 the RAND_RESEED bit causes the circuit to be (re)seeded from the random seed circuit. A
read of this bit will show the current state of a random reseeding operation.
When written with a 1 the NONCE_RESEED bit causes the circuit to be (re)seeded by the host in parallel from the
TRNG_DATAx registers. A nonce reseed involves the following steps:
1. Write the NONCE_RESEED bit with a 1 to place the TRNG in nonce reseed mode. All other bits should be
zero.
2. Write the TRNG_DATAx registers with the first 128 bits of the 255 bit nonce.
3. Write the TRNG_CTRL register with {NONCE_RESEED, NONCE_RESEED_LD, NONCE_SEED_SELECT}
== 3'b110.
4. Write the TRNG_DATAx registers with the second 127 bits of the 255 bit nonce. The 127 bits are formatted
across the registers as follows: {TRNG_DATA3[30:0], TRNG_DATA2, TRNG_DATA1, TRNG_DATA0}.
5. Write the TRNG_CTRL register with {NONCE_RESEED, NONCE_RESEED_LD, NONCE_SEED_SELECT}
== 3'b111.
6. Clearing the NONCE_RESEED bit completes the nonce reseeding process.
At the completion of either type of (re)seeding operation the TRNG will automatically begin generating the first new
random number in the same manner as if the GEN_NEW_RANDOM bit had been written with a0x1.
The TRNG will assert the TRNG_IRQ_STAT bit and an interrupt (if enabled) upon completion of a random number
generation. The host may now extract the random data from the TRNG_DATAx registers. The TRNG_IRQ_STAT bit
must be acknowledged. The data has no endianess and thus it does not matter whether TRNG_DATA0 is extracted
as the least or most significant word.
Once a random number has been read, writing a 1 to the GEN_NEW_RANDOM bit causes the TRNG to begin
collecting a new random number in the TRNG_DATAx registers. Reading the GEN_NEW_RANDOM bit returns the
current state of the random number generation operation.
clk
I_en
I_addr A B C D E F G H
I_wr
I_data a b c d
I_rd
O_data e f g h
O_ack
NOTE: CRP correlator is for Transcede 2120, Transcede 2150, and Transcede 2200.
All four kinds of modules have the AXI interface for data and control configuration. DMA is used to transfer the data.
The number of copies of each module is highly scalable to the size and capacity of base station.
Figure 24-1 shows top level block diagram, on which, each module has one copy as an example.
Figure 24-1 Top Level Block Diagram of CRP
RPD
AXI
PSC
CORͲC
CORͲT
The Transcede 2120, Transcede 2150, and Transcede 2200 have the following capabilities related to the CRP:
• Number of cell = 1. (It could support multiple cells but with the same amount of total users.)
• Number of Rx antennas = 2
• Number of users supported = 24(each can support one R99 RL and HSPA RL)
• Aggregate L1 T-put for uplink = 11.5 Mbps.
• Number of G-rake users = 8
• Cell size = 500 m
AXI
Controller RPD
The PRACH preamble searcher is consisted of input memory, a controller and two engines for searching even
samples and odd samples. Each engine has its own output memory. The search engines are used for computing
correlations for 32 timing candidates. The even search engine processes 32 timing candidates τ = [0,1,...,31]Tchip
at a time, the odd search engine processes 32 timing candidates τ = ([0,1,...,31] + 0.5) Tchip at a time. The
correlation results are stored into the output buffer for each engine. The controller is responsible to send the data
from input memory and split the even samples to even engine and odd samples to odd engine. After the first run of
32 chip of delay spread is searched, the controller will start the second round of search with timing candidates for
the next 32 chips. This process is repeated until all the timing candidates are completed.
The top level functional diagram of the pilot search correlator (PSC) is shown in Figure 24-3.
Figure 24-3 Top Level Diagram of PSC Unit
Controller
Even Path
samples
AXI Searcher
Input Even
Output
Mem
Mem
Odd Path
samples
Searcher
Odd
The pilot searcher consists of input memory, a controller and two engines for searching even samples and odd
samples.
Command
Command
Correlator queue queue
Engine
Command
queue
Filter Engine
controller
controller
Circular Controller
Buffer
Correlator Output
engine
Correlator memory
Output
I/Q engine memory
Buffer Correlator Output
8 buffer
(16x) engine memory
The COR-C consists of one filter engine and multiple correlator engines. The number of correlator engines is
configurable to match with different requirements. In Transcede 2120, Transcede 2150, and Transcede 2200, the
number of engine is set to one.
The filter engine has an input buffer holding multiple streams data with 32 chips for each stream. It has an
interpolation filter which takes the data from each stream and after interpolation the output data is sent to a buffer.
The data in buffer are broadcasted to correlator engines to despread in parallel. The outputs are sent to the output
memory without combining.
Command
Correlator Engine
queue
Command
queue
Command
Filter Engine Queue
Combiner Engine
controller
controller
Input Controller Chest
Buffer Memory
Correlator
engine
Correlator
I/Q engine
Correlator Combiner Output
Upsample
buffer
8 Engine Core Engine Core Memory
Buffer
The COR-T consists of a filter engine, multiple correlator engines and multiple combiners.
NOTE: The memory address of a frame descriptor Word0 must be 16-byte align. The least
significant 4 bits of Frame Head and Frame Next must be 0. Frame and Buffer Words
must be consecutive.
For DMA to start operating, Frame Head must not be 0. For DMA to continue, Frame
Next must not be 0.
When Register FSYN = 1, only 8 Bpointer/BControl pairs are allowed. Both In and
Outbound share the same frame descriptor.
Queue
Access Frame Descriptors
BPointer BPointer
BControl BControl
Data Buffers
NOTE: In this section, some different naming conventions have been inherited.
• IO = I/O Application such as FEC or FFT
• Inbound = Data are moved from Memory to DMA = Mem2Dma = M2D = Transmit = From Memory to
IO = M2IO
• Outbound = Data are moved from DMA to Memory = Dma2Mem = D2M = Receive = From DMA to
Memory = IO2M
Table 24-2 Mem2Dma (or memory to I/O) Frame and Buffer Descriptor
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Table 24-2 Mem2Dma (or memory to I/O) Frame and Buffer Descriptor (Continued)
Bit Name Description
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 FACK_N 1: No DMAACK to be asserted at the end of this frame.
15 - 12 IRQTID Reserved for TID
11 - 4 IRQTID Task Done ID number when IRQTDONE is active
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
2 BRESP_N 1: No waiting for AXI Bresp. Don’t care if any AXI write is successfully get to the destination.
1 FPAUSE Pause after finishing this frame. Allow the arbiter to give the AXI bus to another link.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Zero.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer Words always start at offset x10.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 24-3 Dma2Mem (or from I/O to memory) Frame and Buffer Descriptor
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 FACK_N 1: No DMAACK to be asserted at the end of this frame.
Table 24-3 Dma2Mem (or from I/O to memory) Frame and Buffer Descriptor (Continued)
Bit Name Description
15 - 12 IRQTID Reserved for TID
11 - 4 IRQTID Task Done ID number when IRQTDONE is active
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
2 BRESP_N No waiting for AXI Bresp. Don’t care if any AXI write is successfully get to the destination.
1 FPAUSE Pause after finishing this frame. Allow the arbiter to give the AXI bus to another link.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Dma2Mem (from I/O to memory) Number of Bpointer/BControl pairs. Buffer Words always
start at offset x10
7– 0 M2DBdesc Zero.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 24-4 Common, or 1, Frame and Buffer Descriptor for both Mem2Dma and Dma2Mem (FCOM)
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 FACK_N 1: No DMAACK to be asserted at the end of this frame.
15 - 12 IRQTID Reserved for TID
11 - 4 IRQTID Task Done ID number when IRQTDONE is active
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
Table 24-4 Common, or 1, Frame and Buffer Descriptor for both Mem2Dma and Dma2Mem (FCOM)
Bit Name Description
2 BRESP_N No waiting for AXI Bresp. Don’t care if any AXI write is successfully get to the destination.
1 FPAUSE Pause after finishing this frame. Allow the arbiter to give the AXI bus to another link.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Dma2Mem Number of Bpointer/BControl pairs. Buffer descriptors start at offset x10 + 8 *
M2DBdesc.
D2MBdesc + M2DBdesc = 8 maximum if Register FSYN = 1.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer descriptors always start at offset x10.
D2MBdesc + M2DBdesc = 8 maximum if Register FSYN = 1.
Mem2Dma Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Mem2Dma Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Mem2Dma Buffer Word N = BPointer. Offset: x10 + N*4 (N = even integer)
Mem2Dma Buffer Word N+1 = BControl. Offset: x14 + N*4 (N = even integer)
…..
Dma2Mem Buffer Word 0 = BPointer. Offset: 8 * M2DBdesc + x10
31 – 0 BPOINTER Address pointer of a data buffer.
Dma2Mem Buffer Word 1 = BControl. Offset: 8 * M2DBdesc + x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Dma2Mem Buffer Word M = BPointer. Offset: 8 * M2DBdesc + x10 + M*4 (M = even integer)
Dma2Mem Buffer Word M+1 = BControl. Offset: 8 * M2DBdesc + x14 + M*4 (M = even integer)
BPDMA
AXI_0
Slave
DMA1LINK BITPC DL3_SCGU
(Full)
MUX
POSTPROC
DMA1READ BITPC_w DL3_SCGU
DL3_SCGU
AXI_1~3
Slaves
DMA1READ BITPC_w DL3_SCGU
(AR, R
Only)
POSTPROC_WRAP
Addr[9:2],
DmaPsel_0~3
Addr Decoder
Addr[12:2]
APB Configuration Registers (2'sCmpl, Bypass, Chn_En, N_Trunc)
BPDMA_4DL_REG
NOTE: The memory address of a frame descriptor Word0 must be 16-byte align. The least
significant 4 bits of Frame Head and Frame Next must be 0. Frame and Buffer Words
must be consecutive.
For DMA to start operating, Frame Head must not be 0. For DMA to continue, Frame
Next must not be 0.
When Register FSYN = 1, only 8 Bpointer/BControl pairs are allowed. Both In and
Outbound share the same frame descriptor.
Queue
Access Frame Descriptors
BPointer BPointer
BControl BControl
Data Buffers
NOTE: In this section, some different naming conventions have been inherited.
• IO = I/O Application such as FEC or FFT
• Inbound = Data are moved from Memory to DMA = Mem2Dma = M2D = Transmit = From Memory to
IO = M2IO
• Outbound = Data are moved from DMA to Memory = Dma2Mem = D2M = Receive = From DMA to
Memory = IO2M
Table 26-1 Mem2Dma (or memory to I/O) Frame and Buffer Descriptor
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 FACK_N 1: No DMAACK to be asserted at the end of this frame.
15 - 12 IRQTID Reserved for TID
11 - 4 IRQTID Task Done ID number when IRQTDONE is active
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
2 BRESP_N 1: No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 FPAUSE Pause after finishing this frame. Allow the arbiter to give the AXI bus to another link.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Zero.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer Words always start at offset x10.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 26-2 Dma2Mem (or from I/O to memory) Frame and Buffer Descriptor
Bit Name Descriptor
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Table 26-2 Dma2Mem (or from I/O to memory) Frame and Buffer Descriptor (Continued)
Bit Name Descriptor
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 FACK_N 1: No DMAACK to be asserted at the end of this frame.
15 - 12 IRQTID Reserved for TID
11 - 4 IRQTID Task Done ID number when IRQTDONE is active
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
2 BRESP_N No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 FPAUSE Pause after finishing this frame. Allow the arbiter to give the AXI bus to another link.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Dma2Mem (from I/O to memory) Number of Bpointer/BControl pairs. Buffer Words always
start at offset x10
7– 0 M2DBdesc Zero.
Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Buffer Word 2 = BPointer. Offset: x18
Buffer Word 3 = BControl. Offset: x1C
Buffer Word 4 = BPointer. Offset: x20
Buffer Word 5 = BControl. Offset: x24
Etc..
Table 26-3 Common, or 1, Frame and Buffer Descriptor for Both Mem2Dma and Dma2Mem (FCOM)
Bit Name Description
Frame Word0 = FNext. Offset: x0
31 – 4 FNext Frame Next Pointer. 0 means this frame is the last frame to be processed.
3–0 Not used
Frame Word1 = FControl0. Offset: x4
31 – 17 Not Used.
16 FACK_N 1: No DMAACK to be asserted at the end of this frame.
Table 26-3 Common, or 1, Frame and Buffer Descriptor for Both Mem2Dma and Dma2Mem (FCOM)
Bit Name Description
15 - 12 IRQTID Reserved for TID
11 - 4 IRQTID Task Done ID number when IRQTDONE is active
3 IRQTDONE IRQ Task Done Enable. Set an interrupt when acknowledging to an external DMAREQ.
2 BRESP_N No waiting for AXI Bresp. Do not care if any AXI write successfully gets to the destination.
1 FPAUSE Pause after finishing this frame. Allow the arbiter to give the AXI bus to another link.
0 IRQEN IRQ enable. Set a level interrupt when DMA completes this frame or if the FDONE bit of this
frame is on.
Frame Word2 = FControl1. Offset: x8
31 – 0 Not used
Frame Word3 = FStatus. Offset: xC
31 FDONE Done a DMA transfer. Set by Hardware when a frame is completed. If this bit is 1 when the
Hardware reads this descriptor and Register DONOSTOP = 0, no transfer happens and the
frame is completed. If DONOSTOP = 1. DMA will happen.
30 – 16 Not used
15 – 8 D2MBdesc Dma2Mem Number of Bpointer/BControl pairs. Buffer descriptors start at offset x10 + 8 *
M2DBdesc.
D2MBdesc + M2DBdesc = 8 maximum if Register FSYN = 1.
7– 0 M2DBdesc Mem2Dma Number of Bpointer/BControl pairs. Buffer descriptors always start at offset x10.
D2MBdesc + M2DBdesc = 8 maximum if Register FSYN = 1.
Mem2Dma Buffer Word 0 = BPointer. Offset: x10
31 – 0 BPOINTER Address pointer of a data buffer.
Mem2Dma Buffer Word 1 = BControl. Offset: x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Mem2Dma Buffer Word N = BPointer. Offset: x10 + N*4 (N = even integer)
Mem2Dma Buffer Word N+1 = BControl. Offset: x14 + N*4 (N = even integer)
…..
Dma2Mem Buffer Word 0 = BPointer. Offset: 8 * M2DBdesc + x10
31 – 0 BPOINTER Address pointer of a data buffer.
Dma2Mem Buffer Word 1 = BControl. Offset: 8 * M2DBdesc + x14
31 – 24 Reserved. Must be 0.
23 – 0 BLEN Number of bytes to be transferred. Maximum: 16 Meg – 1 bytes.
Dma2Mem Buffer Word M = BPointer. Offset: 8 * M2DBdesc + x10 + M*4 (M = even integer)
Dma2Mem Buffer Word M+1 = BControl. Offset: 8 * M2DBdesc + x14 + M*4 (M = even integer)
Offset – 0x40000
Offset – 0x3D800
Offset – 0x0000
Magic number
0x4D535044
Timestamp creation
CRC
Key length
Image Length
NOTE: See Table 2-1, Transcede 2xxx Boot Options, on page 30 for boot options.
The Transcede 2xxx IBR supports SPI devices supporting both polarities for serial clock, as shown in Table 27-4.
The Transcede 2xxx IBR supports two clock freq options, as shown in Table 27-1.
If Debug mode is enabled - Bit0 of eFuse ID Byte-1, then Secure Boot Option will be derived from this bit, as shown
in Table 27-5.
48-bits are required for device unique CID (lot#, wafer# and X-Y coordinates), thus rest of the eFuse ID bits are
divided into Configuration bits, Bonding option bits and Secure bits.
The following diagram illustrates the “eFuse ID” decoding fields.
Figure 27-3 eFuse ID
BIT-0 BSCANMODE
0 Not Blocked
BIT 0 - BSCANMODE 1 Blocked
Global lasttime
BIT-7
program
0 Enabled
1 Disabled
28.1 Interrupts
The Transcede 2xxx provides the following categories of interrupts:
• CevaXC Interrupts
• Cortex-A9 MPCore (Dual ARM) Interrupts
28.2 Timers
hclk
APB
timeout
32bit Register 32bit Up Counter
There are four general-purpose timers in the timer block. Each timer consists of two sections, a register and a
count-up counter. The register is loaded on the rising edge of pstbn when selected. The counters are loaded on the
rising edge of hclk one cycle after the registers are loaded or when the counters reach zero. Otherwise, the count-
up counter is incremented on the rising edge of hclk.
IRQ_MASK
Timeout1
Timer1 Status
Timeout3
Timer3 Status
The IRQA and IRQB mask registers are used to selectively enable the timerout pulses to generate the IRQA and
IRQB interrupts. An interrupt status register in the timer block can be read to check which of the timers had
generated a timeout.
2560 chips
0 1 2 3 ... 14
10 ms
WCDMA LTE
Events can be set to trigger at any Chip and/or Slot within a frame, and/or when a certain number of frames or
super-frames have been reached.
As an example application, timers can be connected to a CPRI controller, which asserts a Frame Synchronization
indication (FSync) at the beginning of each frame received or transmitted. If enabled, the timers will start the
counting process once an FSync is detected, and events are generated at the desired chip, slot, frame, and super
frame locations until the timers are disabled. Event generation can be selected to start after a certain count had
elapsed (or offset) for each of the chip, slot or frame counters. Counter values for chip, slot, frame and super frame
can also be read by software for debug or statistics purposes.
In other application, the Timer can be used to generate a Frame Sync indication to drive Radio controllers. Based
on the controller clock rate, the timer can be programmed to generate a periodic 10ms pulse, which can be slewed
by software as a manual adjustment if needed. The pulse generation can start under software control, or at the
detection of an external sync pulse, like a GPS 1 pps pulse.
28.2.3.2 Limitations
The clocking limitations are listed as below:
• Back-to-back writing to APB registers is not allowed before the previous write data has been transferred to the
counter clock domain (6 clock cycles of APB clock, plus 3 cycles of counter clock). Interrupt control register
does not have this limitation because of its potential higher access rate.
• The FSync Timer block needs to be reseted if its selected clock source is stopped (LOS condition for example).
Switching to another clock will not be possible since a glitch-less mux is used.
• For counter value read operation, it is necessary to wait for the counter select bits to propagate from APB clock
domain to the counter clock domain before the read operation can be initiated. That will be about 3 APB + 3
Counter clocks. Otherwise read value can be of the previously selected counter.
• A delay of 15 counter clocks are needed between the de-assertion of the block reset (configuration bit) and the
next access to the block registers or block operation. This is because the internal reset delay due to the reset
synchronization scheme.
When used as pulse generator, positive phase slewing pulse (pulse delay) is possible only if the Terminal Count is
less than the counter overflow value.
Radio Group
Radio Timers
Esync0 Intrpt Timer0 Clk
Chip Interrupt
FBCLK0 ESync0 Timer1 Clk Chip Error
RPDIF0
Timer Timer1 Sync
Frame Slot Interrupt
MCLK0 Sync Slot Error
Timer2 Clk Timer
No. 0 Frame Interrupt
1pps pulse Timer2 Sync
Frame Error
ESync1
RPDIF1 Timer4 Clk
FBCLK1 Timer
Timer4 Sync
APB Interface
10ms Fsync1
MUX Timer5 Clk
Esync1 Intrpt
Timer5 Sync
Esync2 intrpt
Axi Clk
GND
Chip Errors[5:0]
APB Config & Status
Slot Errors[5:0]
Interface Registers
Frame Errors[5:0]
Super Frame
Errors[5:0]
Super Frame CPU interrupt[3:0]
Frame CPU interrupt[3:0] Super Frame
Interrupts[5:0]
Slot CPU interrupt[3:0] Frame
Chip CPU interrupt[3:0] Interrupts[5:0]
Interrupt
Timer CPU interrupt[3:0] Slot
Masking Interrupts[5:0]
Super Frame interrupt Logic
Chip
Frame interrupt Interrupts[5:0]
Slot interrupt
Chip interrupt Timer
Interrupts[5:0]
Timers interrupt
FSync
Each of the 6 available event timers is assigned to one of the 2 RFs and CPRI interfaces, and to either the RX or Tx
lanes. An external clock source and corresponding frame sync are available as well, which can be used for test
mode or to support an external CPRI interface.
Interrupts from all timers are aggregated to form four interrupts: one dedicated for chip interrupt, one for slot, one
for frame, and one for super-frame, and one for the logic ORing of the three of them. Each of the four interrupt
outputs can be enabled or disabled separately.
Four sets of interrupts are routed to 4 different CPUs. Each of the 4 CPUs can be configured to be interrupted by
one or multiple timers out of the 6 available, forming a 6 to 4 switch matrix, with no restrictions.
After the timers are configured, they should be enabled to start operation. If manual start is not enabled, on the first
FSync detected, the counting process starts, and further FSync are disregarded. Each of the three counters
increments until its TC value is reached, at which point, the counting restarts again.
Two values for Terminal Count can be programmed for each of the Chip, Slot and Frame counter, value A and B
respectively. Initially, value A is used, and when TC is reached and the counter restarts, value B will be used, and
so forth. An indication of which value is currently being used is available for each counter as a status bit.
Figure 28-6 Timer Count Operation
11
0
0
3
0
d
Teminal count
TCFrame = 4095
TCslot = 14
TCchip = 2559
TCsubchip = 7
28.2.3.4 IQ Counter
The IQ counter block monitors the amount of IQ samples being transferred from radio interface (CPRI or JESD) to
internal memory and depending on the preprogrammed count value interrupt a processor.
This block has 7 counters and each counter has associated delay counter before interrupt (int_req_out) is
generated. Figure 28-7 shows the IQ transfer counter top level block diagram.
trans_indicator p3_sel
int_req_out[3:0]
Sync Net p3_penable
int_ack_in[3:0]
p3_pwrite
p3_pwdata[31:0]
IQ GP
Counter p3_paddr[15:0] busmatrix
int_req_out[6:4]
ARM
p3_prdata[31:0]
p3_pready
int_ack_in[6:4]
Sys Config p3_pslverr
28.3 Clock
S1_REFCLKP/M
156. 25 MHz
S0_REFCLKP/M for SGMII/ PCIe Crystal: 12 MHz to UP_XO and
156. 25 MHz for 153. 6 MHz UP_XI PS_XI
PCIe Oscillator: 12/19.2/24/48 MHz
for CPRI 19.2/25/26/40 MHz
to UP_XO only
28.4 Reset
NOTE: All CPUs of the ARM core must have their debug resets inactive during boot up, even if
only CPU0 is being used. All debug reset signals are ANDed internally to form one reset
signal.
ref_dk
rst_event
ref_clock_gated
START_EVENT
RESET_WIDTH EVENT_END
reset_n
Table 29-3 DC Electrical Specification for 3.3V tol and non-tol IO (VDD = 1.65V~3.60V, Vext = 3.0~3.6V, TA
= -40 to 85 oC)
Parameter Condition Min Typ Max Unit
Vtol Tolerant external voltage** VDD Power Off & On 3.6 V
Vih High Level Input Voltage
CMOS Interface 0.7*VDD VDD + 0.3 V
Vil Low Level Input Voltage
CMOS Interface VDD = 2.5V ± 10%, 3.3V ± 10% -0.3 0.7 V
VDD = 1.8V ± 10% -0.3 0.3*VDD
∆V Hysteresis Voltage 0.15 V
lih High Level Input Current
Input Buffer Vin = VDD VDD Power ON -3 3 uA
VDD Power Off & SNS = 0 -5 5
Input Buffer with pull-down Vin = VDD VDD = 3.3V ± 10% 20 45 80 uA
VDD = 2.5V ± 10% 20 40 80
VDD = 1.8V ± 10% 20 40 80
Table 29-3 DC Electrical Specification for 3.3V tol and non-tol IO (VDD = 1.65V~3.60V, Vext = 3.0~3.6V, TA
= -40 to 85 oC) (Continued)
Parameter Condition Min Typ Max Unit
lil Low Level Input Current
Input Buffer Vin = VSS VDD Power ON & Off -3 3 uA
Input Buffer with pull-up Vin = VSS VDD = 3.3V ± 10% -15 -40 -80 uA
VDD = 2.5V ± 10% -15 -40 -80
VDD = 1.8V ± 10% -15 -40 -80
Voh Output High Voltage loh = -1.8mA 0.75*VDD VDD V
Voh Output High Voltage loh = -3.2mA, -7.2mA, -10.8mA 0.8*VDD VDD V
Vol Output Low Voltage IoI = 1.8mA, 3.2mA, 7.2mA, 10.8MA 0 0.2*VDD V
loz Output Hi-Z current -5 5 uA
CIN Input capacitance Any input and Bidirectional buffers 5 pF
Note: **specification is only available on tolerant cells.
Table 29-4 DC Electrical Specification for 5V Tolerant IO (VDD = 3.0V~3.60V, Vext = 4.5~5.5V, TA = -40 to
85 oC)
Parameter Condition Min Typ Max Unit
Vtol Tolerant external voltage** VDD = 3.3V ± 10% 5.5 V
Vih High Level Input Voltage
CMOS Interface VDD = 3.3V ± 10% 0.7*VDD VDD+0.3 V
Vil Low Level Input Voltage
CMOS Interface VDD = 3.3V ± 10% -0.3 0.7 V
lih High Level Input Voltage
Input Buffer Vin = VDD -12 12 uA
Vin = 5V ± 10% -20 20
Input Buffer with pull-down Vin = VDD VDD = 3.3V ± 10% 20 45 80 uA
lil Low Level Input Voltage
Input Buffer Vin = VSS -3 3 uA
Input Buffer with pull-up Vin = VSS VDD = 3.3V ± 10% -15 -40 -80 uA
Voh Output High Voltage loh = -1.8mA 0.75*VDD VDD V
Voh Output High Voltage loh = -3.2mA, -7.2mA, -10.8mA 0.8*VDD VDD V
Vol Output Low Voltage lol = 1.8mA, 3.2mA, 7.2mA, 10.8MA 0 0.2*VDD V
loz Output Hi-Z current -12 12 uA
CIN Input capacitance Any input and Bidirectional buffers 5 pF
Note: **specification is only available on tolerant cells.
Note:
VOD typ value is 350mV when ROUT port of ref_bias block connects to the 4.3K external R. Vod value depends on external R
NOTE: The power ramp should be lower than 0.5 V per microsecond.
To prevent excessive current from power supply during power-up and power-down periods, the power sequences
given in Figure 29-1 should be followed.
Figure 29-1 Power Sequence for Normal IO Mode
Core Power
IO Power
Pre-IO Power
IO Power
Pre-IO Power
Core Power
To prevent shoot-through current from power supply and glitch on the pad during power-up and power-down
periods:
1. Power-up PRE-IO supply (VDDP) first, core supply (VDD), and then power-up IO supply (VDDO).
2. Power-down core supply (VDD) first, IO supply (VDDO), and then power-down PRE-IO supply (VDDP).
Die
Heatspreader
Thermal interface material
Underfill resin
Solder bump
Test board
NOTE: Drawing in the figure is not to scale. This is for illustration only.
To obtain the best package thermal performance, connect the thermal balls to the motherboard ground plane,
using thermal vias shown in Figure 29-3.
Figure 29-3 Connecting Thermal Balls to Motherboard Ground Plane Using Thermal Vias
NOTE: See Transcede 2xxx Thermal Application Considerations (PN: 842xx-HWG-001) for
details.
Figure 29-4 shows test conditions. For more information, see JESD 51-9 standard.
Figure 29-4 Test Conditions
A = 114.5 mm
B = 101.5 mm
Lp (PCB thickness) = 1.6 mm
Lp (component height) = 3.2 mm
Revision Summary
Revision Date Changes
A September, 2013 Changes in 842xx-DSH-001-A
A4 September, 2013 Changes in 842xx-DSH-001-A4
A3 October, 2012 Changes in 8422x-DSH-001-A3
A2 September, 2012 Changes in 8422x-DSH-001-A2
A1 July, 2011 Changes in 8422x-DSH-001-A1
Changes in 842xx-DSH-001-A
• Initial release.
Changes in 842xx-DSH-001-A4
1. Changed the device name from T2200 to T2xxx.
2. Updated Figure 1-1, “Transcede 2xxx Block Diagram,” on page 24 for FEC DMA FIFO.
3. Updated Table 1-2, “Transcede 2xxx Devices,” on page 23. and the whole document for the device part
numbers.
4. Updated I/O in Table 3-2, “DDR Signals,” on page 47,Table 3-5, “Expansion Bus Signals,” on page 50, Table 3-
9, “UART Signals,” on page 54, Table 3-12, “USB Signals,” on page 55 and Table 3-15, “JTAG Signals,” on
page 58.
5. Updated EXP_A_3 in both Table 2-2, “Bootstraps Signals,” on page 31 and Table 3-5, “Expansion Bus Signals,”
on page 50.
6. Updated Table 3-7, “TDM Signals,” on page 53 and Table 3-8, “SPI Signals,” on page 53.
7. Updated Table 3-10, “GPIO Signals,” on page 55 for I/O and PD/PU.
8. Updated Table 3-18, “Test and Misc Signals,” on page 59.
9. Updated I/O values in Table 3-13, “RPDIF 0 (JESD207) Signals,” on page 55 and Table 3-14, “RPDIF 1
(JESD207 LVDS) Signals,” on page 56.
10. Updated the CPRI rate 2 and rate 3 to 1228.8 Mbit/s and 2457.6 Mbit/s in Section 6.1.1, "General Features,"
on page 74.
11. Removed the QoS support described in the last feature of Section 8.1, "Feature List," on page 92.
12. Updated the max clock frequency in Section 11.1, "Feature List—SPI," on page 124.
13. Updated GPIO 10 to GPIO 15 in Table 13-1, “GPIO Muxing,” on page 131.
14. Updated Section 22.4.1, "Feature List," on page 180 for the last feature.
15. Added Section 23.4, "True Random Number Generator (TRNG)," on page 190.
16. Removed Section 28.2.3 Timer Interface and Section 28.2.4 Timing.
17. Added Section 28.2.3, "Radio Frame Synchronization Timers," on page 229.
18. Added Table 28-3, “Reference Clock Definition,” on page 238
19. Updated USB PHY in Table 28-3, “Reference Clock Definition,” on page 238.
20. Updated UP_XI in Figure 28-11, “Clock Muxing Diagram,” on page 239.
21. Updated USB PHY Refclk in Table 28-4, “System Clock Definition,” on page 239
22. Added UP_XO and input range for PS_XI in Table 28-5, “Clock Specifications,” on page 241.
23. Added “UDP checksum offload is only supported by RGMII. SGMII does not support UDP checksum offload”.
24. Added Section 28.2.3.4, "IQ Counter," on page 234.
25. Updated Figure 29-1, “Power Sequence for Normal IO Mode,” on page 249.
26. Updated Table 29-3, “DC Electrical Specification for 3.3V tol and non-tol IO (VDD = 1.65V~3.60V, Vext =
3.0~3.6V, TA = -40 to 85 oC),” on page 245 and Table 29-4, “DC Electrical Specification for 5V Tolerant IO
(VDD = 3.0V~3.60V, Vext = 4.5~5.5V, TA = -40 to 85 oC),” on page 247.
Changes in 8422x-DSH-001-A3
1. Added Transcede 2100 device and changed the title to Transcede 2xxx Data Sheet Preliminary.
2. Updated Table 2-2, “Bootstraps Signals,” on page 31.
3. Updated Section 6.2, "CPRI System Block Diagram," on page 76 for AXI2CPUIF block.
4. Updated Section 8.1, "Feature List," on page 92 for the number of QoS queues.
5. Updated Figure 27-4, “eFuse Layout,” on page 221.
6. Updated Section 28.2.1, "Feature List," on page 227 for the TDM Rx FrameSync timers number.
Changes in 8422x-DSH-001-A2
1. Rearranged the outline of the data sheet contents
2. Added Section 1.3, "Transcede 2xxx Devices," on page 23
3. Updated Table 1-3, “CEVA Group Partition (0xF0),” on page 27 for BP memory size.
4. Updated Section 1.6.1, "CEVA Group Slaves," on page 27, Section 1.6.4, "Radio Group," on page 28,
Section 1.6.5, "System Masters/Slaves Group," on page 28
5. Updated Table 2-1, “Transcede 2xxx Boot Options,” on page 30
6. Updated Table 2-2, “Bootstraps Signals,” on page 31
7. Updated Table 3-4, “RGMII Signals,” on page 50
8. Updated Table 3-5, “Expansion Bus Signals,” on page 50, removing the muxing information for EXP_A_14.
9. Updated Table 3-13, “RPDIF 0 (JESD207) Signals,” on page 55 and Table 3-14, “RPDIF 1 (JESD207 LVDS)
Signals,” on page 56
10. Added Section 3.4, "Unused Interface Termination," on page 61
11. Added Section 5.2, "Supported Standards," on page 73
12. Change CPRI version to 4.1 in Section 5.1, "Feature List," on page 73 and Section 6.2, "CPRI System Block
Diagram," on page 76
13. Updated CPRI feature list and added CPRI DMA in Chapter 6, “Common Public Radio Interface (CPRI),” on
page 74
14. Updated Chapter 9, “Expansion Bus Interface,” on page 104
15. Updated Section 11.2, "Feature List—High Speed SPI," on page 124, Section 11.5, "SPI Timing," on
page 126, Section 11.6, "High Speed SPI DMA," on page 128
16. Added Section 12.2, "Feature List—HS-UART," on page 129
17. Added item 7 in Section 16.3, "Smart Card Session Description," on page 144
18. Removed Section USIM Architecture
19. Updated feature list of Section 17.1, "Features List," on page 149
20. Updated Section 17.3, "RPDIF Data Plane Variations," on page 150
21. Added Section 17.5, "RPDIF Timing Parameters," on page 153
22. Added Section 13.2, "GPIO Muxing," on page 131
23. Updated Chapter 16, “Universal Subscriber Identity Module (USIM) Interface,” on page 143
24. Added Section 19.5, "TrustZone Support," on page 162
25. Added Section 21.6, "MAP Direct Memory Access (DMA) Controller," on page 171
26. Added Section 22.4, "FEC Direct Memory Access (DMA) Controller," on page 180
27. Added Section 24.6, "Memory to Memory DMA Controller for CRP (MDMA-CRP)," on page 200
28. Updated Table 24-1, “MDMA Interface,” on page 201
29. Added DMA descriptions in Chapter 26, “Memory to Memory DMA (MDMA) Controller,” on page 209
30. Updated Chapter 27, “Internal Boot ROM (IBR) Implementing Secure Boot,” on page 216
31. Updated Section 27.4, "Clock and Frequency Options," on page 217
32. Updated Table 27-5, “Extended Boot Option Bits – Secure Boot Opt,” on page 220
33. Updated Section 27.9.1, "eFuse Operations," on page 222
34. Added timers in Chapter 28, “Interrupts, Timers, Clock and Reset,” on page 223
35. Updated Figure 28-11, “Clock Muxing Diagram,” on page 239
36. Updated Table 28-4, “System Clock Definition,” on page 239
37. Added Section 28.3.4, "Clock Specifications," on page 241
38. Updated Section 29.2, "Power Turn-On Sequence and Timing Requirement," on page 248
Changes in 8422x-DSH-001-A1
1. Preliminary release.
General Information:
Telephone: (949) 579-3000
Headquarters - Newport Beach
4000 MacArthur Blvd., East Tower
Newport Beach, CA 92660