K4Y5416(/08/04)4UF XDR DRAM 256Mbit XDR DRAM(F-die) 2M x 16(/8/4) bit x 8s Banks Version 1.0 Jan. 2005 Version 1.0 Jan. 2005 Page -1 K4Y5416(/08/04)4UF XDR DRAM Change History Version 0.1( July 2004) - Preliminary - First Copy - Based on the Rambus XDR DRAMTM Datasheet Version 0.81 Version 1.0( Jan. 2005) - Delete "Preliminary" - Based on the Rambus XDR DRAMTM Datasheet Version 0.85 Version 1.0 Jan. 2005 Page 0 K4Y5416(/08/04)4UF XDR DRAM Overview The Rambus XDR DRAM device is a general-purpose high-performance memory device suitable for use in a broad range of applications, including computer memory, graphics, video, and any other application where high bandwidth and low latency are required. The 256Mb XDR DRAM device is a CMOS DRAM organized as 16M words by 16bits. The use of Differential Rambus Signaling Level(DRSL) technology permits 4000/3200/2400 Mb/s transfer rates while using conventional system and board design technologies. XDR DRAM devices are capable of sustained data transfers up to 8000 MB/s. XDR DRAM device architecture allows the highest sustained bandwidth for multiple, interleaved randomly addressed memory transactions. The highly-efficient protocol yields over 95% utilization while allowing fine access granuarity. The device's eight banks support up to four interleaved transactions. Features Highest pin bandwidth available - 4000/3200/2400 Mb/s Octal Data Rate(ODR) Signaling Bi-directional differential RSL(DRSL) - Flexible read/write bandwidth allocation - Minimum pin count Programmable on-chip termination - Adaptive impedance matching - Reduced system cost and routing complexity Highest sustained bandwidth per DRAM device - Up to 8000 MB/s sustained data rate - Eight banks : bank-interleaved transaction at full bandwidth - Dynamic request scheduling - Early-read-after-write support for maximum efficiency - Zero overhead refresh Low Latency - 2.0/2.5/3.33ns request packets - Point-to-point data interconnect for fastest possible flight time - Support for low-latency, fast-cycle cores Low Power - 1.8V VDD - Programmable small-swing I/O signaling(DRSL) - Low power PLL/DLL design - Powerdown self-refresh support - Per pin I/O powerdown for narrow-width operation 0.98us refresh intervals(16K/16ms refresh) Version 1.0 Jan. 2005 Page 1 K4Y5416(/08/04)4UF XDR DRAM Key Timing Parameters/Part Numbers Organization 16Mx16 32Mx8 64Mx4 Bandwidth (1/tBIT)a Latency(tRAC)b Binc,d Part Number 2400 36 A K4Y54164UF-JCA2 3200 35 B K4Y54164UF-JCB3 3200 35 C K4Y54164UF-JCC3 4000 28 C K4Y54164UF-JCC4 2400 36 A K4Y54084UF-JCA2 3200 35 B K4Y54084UF-JCB3 3200 35 C K4Y54084UF-JCC3 4000 28 C K4Y54084UF-JCC4 2400 36 A K4Y54044UF-JCA2 3200 35 B K4Y54044UF-JCB3 3200 35 C K4Y54044UF-JCC3 4000 28 C K4Y54044UF-JCC4 a. Data rate measured in Mbit/s per DQ differential pair. See "Timing Conditions" on page 55 and " Timing Characteristics" on page 58. Note that tBIT=tCYCLE/8 b. Read access time tRAC (= tRCD-R+tCAC) measured in ns. See "Timing Parameters" on page 59. c. Timing parameter bin. See "Timing Parameters" on page 59. This is a measure of the number of interleaved read transactions needed for maximum efficiency (the value Ceiling(tRC-R/tRR-D). For bin A, tRC-R/tRR-D=4, and for bin B, tRC-R/tRR-D=5. d. Bin support is vendor dependent. Version 1.0 Jan. 2005 Page 2 K4Y5416(/08/04)4UF XDR DRAM General Description The timing diagrams in Figure 1 illustrate XDR DRAM device write and read transactions. There are three sets of pins used for normal memory access transactions: CFM/CFMN clock pins, RQ11..0 request pins, and DQ15..0/DQN15..0 data pins. The "N" appended to a signal name denotes the complementary signal of a differential pair. A transaction is a collection of packets needed to complete a memory access. A packet is a set of bit windows on the signals of a bus. There are two buses that carry packets: the RQ bus and DQ bus. Each packet on the RQ bus uses a set of 2 bit-windows on each signal, while the DQ bus uses a set of 16 bit-windows on each signal. In the write transaction shown in Figure 1, a request packet (on the RQ bus) at clock edge T0 contains an activate (ACT) command. This causes row Ra of bank Ba in the memory component to be loaded into the sense amp array for the bank. A second request packet at clock edge T1 contains a write (WR) command. This causes the data packet D(a1) at edge T4 to be written to column Ca1 of the sense amp array for bank Ba. A third request packet at clock edge T3 contains another write (WR) command. This causes the data packet D(a2) at edge T6 to also be written to column Ca2. A final request packet at clock edge T13 contains a precharge (PRE) command. The spacings between the request packets are constrained by the following timing parameters in the diagram: tRCD-W , tCC , and tWRP . In addition, the spacing between the request packets and data packets is constrained by the tCWD parameter. The spacing of the CFM/CFMN clock edges is constrained by tCYCLE. Figure 1 : XDR DRAM Device Write and Read Transactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN WR WR RQ11..0 ACT a0 a1 a2 tRCD-W tCC DQ15..0 tCWD DQN15..0 D(a1) tCYCLE PRE a3 tWRP D(a2) Transaction a: WR a0 = {Ba,Ra} a1 = {Ba,Ca1} a2 = {Ba,Ca2} a3 = {Ba} Write Transaction T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 ACT a0 DQ15..0 DQN15..0 RD a1 tRCD-R RD a2 tCC tRDP tCAC Transaction a: RD tCYCLE PRE a3 a0 = {Ba,Ra} Q(a1) Q(a2) a1 = {Ba,Ca1} a2 = {Ba,Ca2} a3 = {Ba} Read Transaction The read transaction shows a request packet at clock edge T0 containing an ACT command. This causes row Ra of bank Ba of the memory component to load into the sense amp array for the bank. A second request packet at clock edge T5 contains a read (RD) command. This causes the data packet Q(a1) at edge T11 to be read from column Ca1 of the sense amp array for bank Ba. A third request packet at clock edge T7 contains another RD command. This causes the data packet Q(a2) at edge T13 to also be read from column Ca2. A final request packet at clock edge T10 contains a PRE command. The spacings between the request packets are constrained by the following timing parameters in the diagram: tRCD-R , tCC , and tRDP . In addition, the spacing between the request and data packets are constrained by the tCAC parameter. Version 1.0 Jan. 2005 Page 3 K4Y5416(/08/04)4UF XDR DRAM Pinouts and Definitions The following table shows the pin assignment of the center-bonded fanout XDR DRAM Package. The mechanical dimensions of this package are shown on page 71. Note - Pin #1 is at the A1 postion. Table 1: 104ball XDR DRAM Package(Top View) 16 DQ10 DQ0 SDO 15 DQN10 DQN0 RST VDD GND VDD VDD GND DQ1 DQ11 SCK DQN1 DQN11 14 DQ6 DQ12 GND RQ2 RQ5 RQ6 RQ8 CMD DQ13 DQ7 13 DQN6 DQN12 VDD RQ1 VREF RQ7 RQ9 VDD DQN13 DQN7 12 VDD GND GND GND GND VDD 11 GND VTERM VTERM GND GND GND VDD VDD GND GND VDD GND VDD VDD GND GND GND 10 9 8 7 ROW 6 GND GND 5 VDD VDD VTERM 4 DQ14 DQ4 GND RQ3 RSRV CFMN RQ11 RQ4 RSRV CFM RQ10 VDD 3 DQN14 DQN4 RQ0 2 DQ2 DQ8 GND 1 DQN2 DQN8 SDI VDD GND Top View A B C D E F VTERM VDD VDD GND DQ5 DQ15 DQN15 VDD DQN5 VDD DQ9 DQ3 VDD GND VDD DQN9 DQN3 G H J K L COL SAMSUNG SAMSUNG 450 401 040 Top View K4Y54164UF K4Y5017UM - PC K4R JCB3 Chip The pin #1(ROW1, COLA) is located at the A1 position on the top side and the A1 position is marked by the marker " ". Version 1.0 Jan. 2005 Page 4 K4Y5416(/08/04)4UF XDR DRAM Pin Description Table2 summarizes the pin functionality of the XDR DRAM device. The first group of pins provide the necessary supply voltages. These include VDD and GND for the core and interface logic, VREF for receiving input signals, and VTERM for the driving output signals. The next group of pins are used for high bandwidth memory accesses. These include DQ15 ... DQ0 and DQN15 ... DQN0 for carrying read and write data signals, RQ11 ... RQ0 for carrying request signals, and CFM and CFMN for carrying timing information used by the DQ, DQN and RQ signals. The final set of pins comprise the serial interface that is used for control register accesses. These include RST for initializing the state of the device, CMD for carrying command signals, SDI and SDO for carrying register read data, and SCK for carrying the timing information used by the RST, SDI, SDO, and CMD signals. Table 2 : Pin Description Signal VDD I/O Type No. of pins Description - - 22a Supply voltage for the core and interface logic of the device. Ground reference for the core and interface logic of the device. GND - - 24a VREF - - 1 Logic threshold reference voltage for RSL signals. - - 4a Termination voltage for DRSL signals. DQ15..0 I/O DRSLb 16 Positive data signals that carry write or read data to and from the device. DQN15..0 I/O DRSL b 16 Negative data signals that carry write or read data to and from the device. RQ11..0 I RSLb 12 Request signals that carry control and address information to the device. CFM I DIFFCLKb 1 Clock from master -- Positive interface clock used for receiving RSL signals, and receiving and transmitting DRSL signals from the Channel. CFMN I DIFFCLKb 1 Clock from master -- Negative interface clock used for receiving RSL signals, and receiving and transmitting DRSL signals from the Channel. RST I RSLb 1 Reset input -- This pin is used to initialize the device. CMD I RSLb 1 Command input -- This pin carries command, address, and control register write data into the device. SCK I RSLb 1 Serial clock input -- Clock source used for reading from and writing to the control registers. SDI I RSLb 1 Serial data input -- This pin carries control register read data through the device. This pin is also used to initialize the device. SDO O CMOSb 1 Serial data output -- This pin carries control register read data from the device. This pin is also used to initialize the device. RSRV - - 2 Reserved pins -- Follow Rambus XDR system design guidelines for connecting RSRV pins VTERM Total pin count per package 104 a. The exact number of VDD/GND/VTERM pins may vary between XDR DRAM vendors. b. All DQ and CFM signals are high-true; low voltage is logic 0 and high voltage is logic 1. All DQN, CFMN, RQ, RSL, and CMOS signals are low-true; high voltage is logic 0 and low voltage is logic 1. Version 1.0 Jan. 2005 Page 5 K4Y5416(/08/04)4UF XDR DRAM Block Diagram A block diagram of the XDR DRAM device is shown in Figure2. It shows all interface pins and major internal blocks. The CFM and CFMN clock signals are received and used by the clock generation logic to produce three virtual clock signals : 1/tCYCLE, 2/tCYCLE, and 16/tCC. The frequency of these signals are 1x, 2x, and 8x that of the CFM and CFMN signals. These virtual signals show the effective data rate of the logic blocks to which they connect; they are not necessarily present in the actual memory component. The RQ11 ... RQ0 pins receive the request packet. Two 12-bit words are received in one tCYCLE interval. This is indicated by the 2/tCYCLE clocking signal connected to the 1:2 Demux Block that assembles the 24-bit request packet. These 24bits are loaded into a register(clocked by the 1/tCYCLE clocking signal) and decoded by the Decode Block. The VREF pin supplies a reference voltage used by the RQ receivers. Three sets of control signals are produced by the Decode Block. These include the bank(BA) and row(R) addresses for an activate(ACT) command, the bank(BR) and row(REFr) addresses for a refresh activate(REFA) command, the bank(BP) address for a precharge(PRE) command, the bank(BR) adddress for a refresh precharge(REFP) command, and the bank(BC) and column(C and SC) addresses for a read(RD) or write(WR or WRM) command. In addition, a mask(M) is used for a masked write(WRM) command. These commands can all be optionally delayed in increments of tCYCLE under control of delay fields in the request. The control signals of the commands are loaded into registers and presented to the memory core. These registers are clocked at maximum rates determined by core timing parameters, in this case 1/tRR, 1/tPP, and 1/tCC(1/4, 1/4, and 1/2 the frequency of CFM in the 3200 component). These registers may be loaded at any tCYCLE rising edge. Once loaded, they should not be changed until a tRR, tPP, or tCC time later because timing paths of the memory core need time to settle. A bank address is decoded for an ACT command. The indicated row of the selected bank is sensed and placed into the associated sense amp array for the bank. Sensing a row is also referred to as "Opening a page"for the bank. Another bank address is decoded for a PRE command. The indicated bank and associated sense amp array are precharged to a state in which a subsequent ACT command can be applied. Precharging a bank is also called "closing the page" for the bank. After a bank is given an ACT command and before it is given a PRE command, it may receive read(RD) and write(WR) column commands. These commands permit the data in the bank's associated sense amp array to be accessed. For a WR command, the bank address is decoded. The indicated column of the associated sense amp array of the selected bank is written with the data received from the DQ15 ... DQ0 pins. The bank address is decoded for a RD command. The indicated column of the selected bank's associated sense amp array is read. The data is transmitted onto the DQ15 ... DQ0 pins. The DQ15 ... DQ0 pins receive the write data packet(D) for a write transaction. 16 sixteen-bit words are received in one tCC interval. This is indicated by the 16/tCC clocking signal connected to the 1:16 Demux Block that assembles the 16x16-bit write data packet. The write data is then driven to the selected Sense Amp Array Bank. 16 sixteen-bit words are accessed in the selected Sense Amp Array Bank for a read transaction. The DQ15 ... DQ0 pins transmit the read data packet(Q) in one tCC interval. This is indicated by the 16/tCC clocking signal connected to the 16:1 Mux Block. The VTERM pin supplies a termination voltage for the DQ pins. The RST, SCK, and CMD pins connect to the Control Register block. These pins supply the data, address and conrol needed to write the control registers. The read data for these registers is accessed through the SDO/SDI pins. These pins are also used to initialize the device. The control registers are used to transition between power modes, and are also used for calibrating the high speed transmit and receive circuits of the device. The control registers also supply bank(REFB) and row(REFr) address for refresh operations. Version 1.0 Jan. 2005 Page 6 K4Y5416(/08/04)4UF XDR DRAM Figure 2 : 256Mb (8x2Mx16) XDR DRAM Block Diagram 1:2 Demux 12 1/tCYCLE 2/tCYCLE 12 Control Registers 16/ tCC reg 12 RST,SCK,CMD,SDI SDO 4 1 CFM CFMN 1/tCYCLE 12 Power Mode Logic Calibration Logic Refresh Logic Initialization Logic Decode 1/tRR BA,BR,REFB 23 3 1 ... {0..1}*tCYCLE decode ACT delay WIDTH R,REFr 1 ACT ROW 1 PRE 1 Bank 0 PRE 16x16*2 4 ... 1 R/W COL 16x16*26 Sense Amp 0 COL 16x16 8 M Sense Amp Array R/W Sense Amp (23 - 1) 16x16 S[15:0][15:0] 16x16 16x16 WIDTH Byte Mask (WR) Dynamic Width Demux (WR) 16x16 ... SC 1 ... 6 decode 3 C ... BC 23 Bank (2 - 1) 16x16*26 6 1/tCC 3 ... 3 ROW 23 16x16*26*211 ... BP,BR,REFB reg 1/tPP Bank Array Bank 0 ACT 11 ... {0..3}*tCYCLE decode PRE delay 3 reg 11 reg {0..1}*tCYCLE 3 reg RD,WR delay 3 ... 6+4 REFB,REFr ACT logic ... 7 PRE logic Dynamic Width Mux (RD) D[15:0][15:0] ... COL logic Q[15:0][15:0] 16x16 16 ... 2/tCYCLE VREF 1 ... RQ11..0 12 16/tCC 16 1:16 Demux 16:1 Mux 16/tCC 16 16 termination 16 2 VTERM DQ15..0 16 DQN15..0 Version 1.0 Jan. 2005 Page 7 K4Y5416(/08/04)4UF XDR DRAM Request Packets A request packet carries address and control information to the memory device. This section contains tables and diagrams for packet formats, field encodings and packet interactions. Request Packet Formats There are five types of request packets: 1. ROWA -- specifies an ACT command 2. COL -- specifies RD and WR commands 3. COLM -- specifies a WRM command 4. ROWP -- specifies PRE and REF commands 5. COLX -- specifies the remaining commands Table 3 describes fields within different request packet types. Various request packet type formats are illustrated in Figure3. Each packet type consists of 24 bits sampled on the RQ11..0 pins on two successive edges of the CFM/CFMN clock. The request packet formats are distinguished by the OP3..0 field. This field also specifies the operation code of the desired command. In the ROWA packet, a bank address (BA), row address (R), and command delay (DELA) are specified for the activate (ACT) command. In the COL packet, a bank address (BC), column address (C), sub-column address (SC), command delay (DELC), and subopcode (WRX) are specified for the read (RD) and write (WR) commands. In the COLM packet, a bank address (BC), column address (C), sub-column address (SC), and mask field (M) are specified for the masked write (WRM) command. In the ROWP packet, two independent commands may be specified. A bank address (BP) and sub-opcode (POP) are specified for the precharge (PRE) commands. An address field (RA) and sub-opcode (ROP) are specified for the refresh (REF) commands. In the COLX packet, a sub-operation code field (XOP) is specified for the remaining commands. Table 3 : Request Field Description Field Packet Types Description OP3..0 4-bit operation code that specifies packet format. ROWA/ROWP /COL/COLM/COLX (Encoded commands are in Table 4 on page 10.) DELA ROWA Delay the associated row activate command by 0 or 1 tCYCLE . BA2..0 ROWA 3-bit bank address for row activate command. R10..0 ROWA 11-bit row address for row activate command. WRX COL Specifies RD (=0) or WR (=1) command. DELC COL Delay the column read or write command by 0 or 1 tCYCLE . BC2..0 COL/COLM 3-bit bank address for column read or write command. C9..4 COL/COLM 6-bit column address for column read or write command. SC3..0 COL/COLM 4-bit sub-column address for dynamic width (see "Dynamic Width Control" on page 46). M7..0 COLM 8-bit mask for masked-write command WRM. POP2..0 ROWP 3-bit operation code that specifies row precharge command with a delay of 0 to 3 tCYCLE. (Encoded commands are in Table 6 on page 11). BP2..0 ROWP 3-bit bank address for row precharge command. ROP2..0 ROWP 3-bit operation code that specifies refresh commands. (Encoded commands are in Table 5 on page 10). RA7..0 ROWP 8-bit refresh address field (specifies BR bank address, delay value, and REFr load value) XOP3..0 COLX 4-bit extended operation code that specifies calibration and powerdown commands. (Encoded commands are in Table 7 on page 11). Version 1.0 Jan. 2005 Page 8 K4Y5416(/08/04)4UF XDR DRAM Figure 3 : Request Packet Formats T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 tCYCLE ACT a0 RD a1 WRM a2 PRE a3 PDN - DQ15..0 DQN15..0 ROWA Packet COL Packet tCYCLE COLM Packet tCYCLE tCYCLE ROWP Packet tCYCLE COLX Packet tCYCLE CFM CFMN RQ11 OP 3 DEL A OP 3 DEL C OP 3 M 7 OP 3 POP 2 OP 3 rsrv RQ10 OP 2 R 8 OP 2 rsrv M 3 M 6 OP 2 ROP 2 OP 2 rsrv RQ9 R 9 R 7 OP 1 rsrv M 2 M 5 OP 1 ROP 1 OP 1 rsrv RQ8 R 10 R 6 OP 0 rsrv M 1 M 4 OP 0 ROP 0 OP 0 rsrv rsrv R 5 WR X C 7 M 0 C 7 POP 1 RA 7 rsrv rsrv rsrv R 4 C 8 C 6 C 8 C 6 POP 0 RA 6 rsrv rsrv rsrv R 3 C 9 C 5 C 9 C 5 rsrv RA 5 rsrv rsrv rsrv R 2 rsrv C 4 rsrv C 4 rsrv RA 4 rsrv rsrv rsrv R 1 rsrv SC 3 rsrv SC 3 rsrv RA 3 XOP 3 rsrv RQ2 BA 2 R 0 BC 2 SC 2 BC 2 SC 2 BP 2 RA 2 XOP 2 rsrv RQ1 BA 1 rsrv BC 1 SC 1 BC 1 SC 1 BP 1 RA 1 XOP 1 rsrv RQ0 BA 0 rsrv BC 0 SC 0 BC 0 SC 0 BP 0 RA 0 XOP 0 rsrv RQ7 RQ6 RQ5 RQ4 RQ3 Version 1.0 Jan. 2005 Page 9 K4Y5416(/08/04)4UF XDR DRAM Request Field Encoding Operation code fields are encoded within different packet types to specify commands. Table4 through Table7 provides packet type and encoding summaries. Table4 shows the OP field encoding for five packet types. The COLM and ROWA packets each specify a single command : ACT and WRM. The COL, COLX, and ROWP packets each use additional fields to specify multiple commands : WRX, XOP, and POP/ROP, respectively. The COLM packet specifies the masked write command WRM. This is like the WR unmasked write command, except that a mask field M7...0 indicates whether each byte of the write data packet is written or not written. The ROWA packet specifies the row activate command ACT. The COL packet uses the WRX field to specify the column read and column write(unmasked) commands. Table 4 : OP Field Encoding Summary OP [3:0] 0000 0001 0010 Packet Command - No operation. RD Column read (WRX=0). Column C9..4 of sense amp in bank BC2..0 is read to DQ bus after DELC*tCYCLE. WR Column write (WRX=1). Write DQ bus to column C9..4 of sense amp in bank BC2..0 after DELC*tCYCLE. COL COLX Description NOP CALy XOP3..0 specifies a calibrate or powerdown command -- see Table 7 on page 11. PREx POP2..0 specifies a row precharge command -- see Table 6 on page 11. REFy,LRRr ROP2..0 specifies a row refresh command or load REFr register command -- see Table 5 on page 10. 0011 ROWP 01xx ROWA ACT Row activate command. Row R10..0 of bank BA2..0 is placed into the sense amp of the bank after DELA*tCYCLE. 1xxx COLM WRM Column write command (masked) -- mask M7..0 specifies which bytes are written. Encoding of the ROP field in the ROWP packet is shown in Table5. The first encoding specifies a NOPR (no operation) command. The REFP command uses the RA field to select a bank to be precharged. The REFA and REFI commands use the RA field and REFH/M/L registers to select a bank and row to be activated for refresh. The REFI command also increments the REFH/M/L register. The REFP, REFA, and REFI commands may also be delayed by up to 3*tCYCLE using the RA[7:6] field. The LRR0, LRR1, and LRR2 commands load the REFH/M/L registers from the RA[7:0] field. Table 5 : ROP Field Encoding Summary ROP[2:0] Command Description 000 NOPR No operation 001 REFP Refresh precharge command. Bank RA2..0 is precharged. This command is delayed by {0,1,2,3}*tCYCLE (the value is given by the expression (2*RA[7]+RA[6]). 010 REFA Refresh activate command. Row R[10:0] (from REFH/M/L register) of bank RA2..0 is placed into sense amp. This command is delayed by {0,1,2,3}*tCYCLE (the value is given by the expression (2*RA[7]+RA[6]). 011 REFI Refresh activate command. Row R[10:0] (from REFH/M/L register) of bank RA2..0 is placed into sense amp. This command is delayed by {0,1,2,3}*tCYCLE (the value is given by the expression (2*RA[7]+RA[6]). R[10:0] field of REFH/M/L register is incremented after the activate command has completed. 100 LRR0 Load Refresh Low Row register (REFL). RA[7:0] is stored in R[7:0] field. 101 LRR1 Load Refresh Middle Row register (REFM). RA[2:0] is stored in R[10:8] field. 110 LRR2 Load Refresh High Row register -- not used with this device. 111 - Reserved Version 1.0 Jan. 2005 Page 10 K4Y5416(/08/04)4UF XDR DRAM The REFH/M/L registers are also refreshed to as the REFr registers. Note that only the bits that are needed for specifying the refresh row(11 bits in all) are implemented in the REFr registers - the rest are reserved. Note also that the RA2 ... RA0 field that specifies the refresh bank address is also referred to as BR2...0. See "Refresh Transactions" on page37. Table6 shows the POP field encoding in the ROWP packet. The first encoding specifies a NOPP(no operation) command. There are four variations of PRE(precharge) command. Each uses the BP field to specify the bank to be precharged. Each also specifies a different delay of up to 3*tCYCLE using the POP[1:0] field. A precharge command may be specified in addition to a refresh command using the ROP field. Table 6 : POP Field Encoding Summary POP [2:0] Command Description 000 NOPP No operation. 001 - Reserved. 010 - Reserved. 011 - Reserved. 100 PRE0 Row precharge command -- Bank BP2..0 is precharged. This command is delayed by 0*tCYCLE. 101 PRE1 Row precharge command -- Bank BP2..0 is precharged. This command is delayed by 1*tCYCLE. 110 PRE2 Row precharge command -- Bank BP2..0 is precharged. This command is delayed by 2*tCYCLE. 111 PRE3 Row precharge command -- Bank BP2..0 is precharged. This command is delayed by 3*tCYCLE. Table7 shows the XOP field encoding in the COLX packet. This field encodes the remaining commands. The CALC and CALE commands perform calibration operations to ensure signal integrity on the Channel. See "Calibration Transactions" on page 39. The PDN command causes the device to enter a power-down state. See"Power State Management" on page 40. Table 7 : XOP Field Encoding Summary XOP [3:0] Command Command and Description XOP [3:0] Command 0000 - Reserved. 1000 CALC Command and Description Current calibration command. 0001 - Reserved. 1001 CALZ Impedance calibration command. 0010 - Reserved. 1010 CALE End calibration command (CALC). 0011 - Reserved. 1011 - 0100 - Reserved. 1100 PDN Reserved. 0101 - Reserved. 1101 - Reserved. Enter powerdown power state. 0110 - Reserved. 1110 - Reserved. 0111 - Reserved. 1111 - Reserved. Version 1.0 Jan. 2005 Page 11 K4Y5416(/08/04)4UF XDR DRAM Request Field Interactions A summary of request packet interaction is Table8. Each case is limited to request packets with commands that perform memory operations(including refresh commands). This includes all commands in ROWA, ROWP, COL, and COLM packets. The commands in COLX packets are described in later sections. See "Maintenance Operations" on page38. Request packet/command "a" is followed by request/ command "b". The minimum possible spacing between these two packet/command is 0*tCYCLE. However, a larger time interval may be needed because of a resource interaction between the two packet/commands. If the minimum possible sapcing is 0*tCYCLE, then an entry of "No limit" is shown in the table. Note that the spacing values shown in the table are relative to the effective beginning of a packet/command. The use of the delay field with a command will delay the position of the effective packet/command from the position of the actual packet/command. See "Dynamic Request Scheduling" on page18 . Any of the packet/command encoding under one of the four operation types is equivalent in terms of the resource constraints. Therefore. both the horisontal columns(packet "a") and vertical rows(packet "b") of the interaction table are divided into four major groups. The four possible operation types for request packet a and b include : : [A] Active Row ROWA/ACT ROWP/REFA ROWP/REFI : [R] Read Column : [W] Write Column COL/RD COL/WR COLM/WRM : [P] Precharge Row ROWP/PRE ROWP/REFP Table 8 : Packet Interaction Summary Second packet/command to bank Bb First packet/command to bank Ba Activate Row [A] ROWA - ACT Bb ROWP - REFA Bb ROWP - REFI Ba Read Column [R] Write Column [W] Precharge Row [P] COL - RD Bb COL - WR Bb COLM - WRM Bb ROWP - PRE Bb ROWP - REFP Bb Case ARd: No limit Case AWd: No limit Case APd: No limit Case ARs: tRCD-R Case AWs: tRCD-W Case APs: tRAS Case RRd: tCC Read Column [R] Ba,Bb different Case RAd: No limit COL - RD Ba Ba,Bb same Case RAs:b tRDP+tRP Case RRs: tCC Case RWd:a tRW Case RPd: No limit Case RWs: a tRW Case RPs: tRDP Write Column [W] Ba,Bb different Case WAd: No limit Case WRd:c tWR COL - WR Ba Case WAsb:tWRP+tRP Case WRs:c tWR COLM - WRM Ba Ba,Bb same Case WWd: tCC Case WPd: No limit Case WWs: tCC Case WPs: tWRP Precharge Row [P] Ba,Bb different Case PAd: No limit ROWP - PRE Ba Case PAs: tRP ROWP - REFP Ba Ba,Bb same Case PRd: No limit Case PWd: No limit Case PPd: tPP Case PRs:d tRP+tRCD-R Case PWs:d tRP+tRCD-W Case PPs: tRC Figure 5 Figure 6 Activate Row [A] Ba,Bb different Case AAd: tRR ROWA - ACT Ba ROWP - REFA Ba Ba,Bb same Case AAs: tRC ROWP - REFI Ba See Examples: Figure 4 Figure 7 a. tRW is equal to tCC + tRW-BUB,XDRDRAM+ tCAC - tCWD and is defined in Table 18. This also depends upon propagation delay - See "Propagation Delay" on page 27. b. A PRE command is needed between the RD and ACT/REFA commands or the WR/WRM and ACT/REFA commands. c. tRW is defined in Table 18. d. An ACT command is needed between the PRE/REFP and RD commands or the PRE/REFP and WR/WRM commands. Version 1.0 Jan. 2005 Page 12 K4Y5416(/08/04)4UF XDR DRAM The first request is shown along the vertical axis on the left of the table. The second request is shown along the horizontal axis at the top of the table. Each request includes a bank specification "Ba" and "Bb". The first and second banks may be the same, or they may be different. These two subcases for each interaction are shown along the vertical axis on the letf. There are 32 possible interaction cases altogether. The table gives each case a label of the form "xyz", where "x" and "y" are one of the four operation types("A" for Activate, "R" for Read, "W" for Write, or "P" for Precharge) for the first and second request, respectively, and "z" indicates the same bank("s") or different bank("d"). Along the horizontal axis at the bottom of the table are cross references to four figures(Figure4 through Figure7). Each figure illustrates the eight cases in the corresponding vertical column. Thus, Figure4 shows the eight cases when the second request is an activate operation("A"). In the following discussion of the cases, only those in which the interaction interval is greater than tCYCLE will be described. Request Interactions Cases In Figure4. the interaction interval for the AAd case is tRR. This parameter is the row-to-row time and is the minimum interval between activate commands to different banks of a device. The interaction interval for the AAs case is tRC. This is the row cycle time parameter and is the minimum interval between activate commands to same banks of a device. A precharge operation must be inserted between the two activate operations. The interaction interval for the RAs case is tRDP + tRP. A precharge operation must be inserted between the read and activate operation. The minimum interval between a read and a precharge operation to a bank is tRDP. The minimum interval between a precharge and an activate operation to a bank is tRP. The interaction interval for the WAs case is tWDP + tRP. A precharge operation must be inserted between the read and the activate operation. The minimum interval between a write and a precharge operation to a bank is tWDP. The minimum interval between a precharge and an activate operation to a bank is tRP. The interaction interval for the PAs case is tRP. The minimum interval between a precharge and an activate operation to a bank is tRP. In Figure5, the interaction interval for the ARs case is tRCD-R. This is the row-to-column-read time parameter and represents the minimum interval between an activate operation and a read operation to a bank. The interaction interval for the RRd and RRs cases is tCC. This is the column-to-column time parameter and represents the minimum interval between two read operations. The interaction interval for the WRd and WRs cases is tWR. This is the write-to-read time parameter and represents the minimum interval between a write and a read operation to any banks. See "Read/Write Interaction" on page 26. The interaction interval for the PRs case is tRP + tRCD-R. An activate operation must be inserted between the precharge and the read operation. The minimum interval between a precharge and an activate operation to a bank is tRP. The minimum interval between an activate and a read operation to a bank is tRCD-R. In Figure6. the interaction interval for the AWs case is tRCD-W. This is the row-to-column-write timing parameter and represents the minimum interval between an activate operation and a write operation to a bank. The interaction interval for the RWd and RWs cases is tRW. This is the read-to-write time parameter and represents the minimum interval between a read and a write operation to any banks. See "Read/Write Interaction" on page26. The interaction interval for the WWd and WWs cases is tCC. This is the column-to-column time parameter and represents the minimum interval between two write operations. The interaction interval for the PWs case is tRP + tRCD-W. An activate operation must be inserted between the precharge and the write operation. The minimum interval between a precharge and an activate operation to a bank is tRP. The minimum interval between an activate and a write operation to a bank is tRCD-W. In Figure7, the interaction interval for the APs case is tRAS. This parameter is the minimum activate-to-precharge time to a bank. The interaction interval for the RPs and WPs cases are tRDP and tWDP, respectively. These are the read-or write-to-precharge time parameters to a bank. Ths interaction interval for the PPd case is tPP. This parameter is the precharge-to-precharge time and the minimum interval between precharge commands to different banks of a device. The interaction interval for the PPs case is tRC. This is the row cycle time parameter and the minimum interval between precharge commands to same banks of a device. An activate operation must be inserted between the two activate operations. This activate operation must be placed a time tRP after the first, and a time tRAS before the second precharge. Version 1.0 Jan. 2005 Page 13 K4Y5416(/08/04)4UF XDR DRAM Figure 4 : ACT-, RD-, WR-, PRE-to-ACT Packet Interactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 ACT a ACT b tRR tRAS ACT a tRC PRE a tRP ACT b DQ15..0 DQ15..0 DQN15..0 DQN15..0 AAd Case (activate-activate-different bank) a: ROWA Packet with ACT,Ba,Ra Ba =/ Bb b: ROWA Packet with ACT,Bb,Rb T0 T1 T2 T3 T4 T5 T6 T7 T8 AAs Case (activate-activate-same bank) a: ROWA Packet with ACT,Ba,Ra Ba = Bb b: ROWA Packet with ACT,Bb,Rb T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 RD ACT a b DQ15..0 DQ15..0 No DQN15..0 DQN15..0 RD a PRE a tRP tRDP+tRP ACT b limit RAd Case (read-activate-different bank) a: COL Packet with RD,Ba,Ca b: ROWA Packet with ACT,Bb,Rb T0 tRDP T1 T2 T3 T4 T5 RAs Case (read-activate-same bank) a: COL Packet with RD,Ba,Ca b: ROWA Packet with ACT,Bb,Rb Ba =/ Bb T6 T7 T8 T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 DQ15..0 No DQ15..0 DQN15..0 DQN15..0 tWRP tWRP+tRP WR a WR ACT a b tRP ACT b limit WAd Case (write-activate-different bank) a: COL Packet with WR,Ba,Ca Ba =/ Bb b: ROWA Packet with ACT,Bb,Rb T0 PRE a T1 T2 T3 T4 T5 T6 T7 T8 WAs Case (write-activate-same bank) a: COL Packet with WR,Ba,Ca b: ROWA Packet with ACT,Bb,Rb T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 DQ15..0 DQ15..0 DQN15..0 DQN15..0 PRE ACT a b PRE a tRP ACT b No limit PAd Case (precharge-activate-different bank) a: ROWP Packet with PRE,Ba Ba =/ Bb b: ROWA Packet with ACT,Bb,Rb PAs Case (precharge-activate-same bank) a: ROWP Packet with PRE,Ba Ba = Bb b: ROWA Packet with ACT,Bb,Rb Version 1.0 Jan. 2005 Page 14 K4Y5416(/08/04)4UF XDR DRAM Figure 5 : ACT-, RD-, WR-, PRE-to-RD Packet Interactions CFM CFMN CFM T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFMN RQ11..0 RQ11..0 ACT RD a b DQ15..0 DQN15..0 DQ15..0 No limit ACT a RD b tRCD-R DQN15..0 ARd Case (activate-read different bank) a: ROWA Packet with ACT,Ba,Ra Ba =/ Bb b: COL Packet with RD,Bb,Cb T0 T1 T2 T3 T4 T5 T6 T7 T8 ARs Case (activate-read same bank) a: ROWA Packet with ACT,Ba,Ra b: COL Packet with RD,Bb,Cb T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 RD a tCC RD a RD b tCC RD b DQ15..0 DQ15..0 DQN15..0 DQN15..0 RRd Case (read-read different bank) a: COL Packet with RD,Ba,Ca b: COL Packet with RD,Bb,Cb T0 T1 T2 T3 T4 T5 RRs Case (read-read same bank) a: COL Packet with RD,Ba,Ca b: COL Packet with RD,Bb,Cb Ba =/ Bb T6 T7 T8 T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 WR a RD b tWR WR a RD b tWR DQ15..0 DQ15..0 DQN15..0 DQN15..0 WRd Case (write-read different bank) a: COL Packet with WR,Ba,Ca b: COL Packet with RD,Bb,Cb T0 T1 T2 T3 T4 T5 WRs Case (write-read same bank) a: COL Packet with WR,Ba,Ca b: COL Packet with RD,Bb,Cb Ba =/ Bb T6 T7 T8 T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 PRE RD a b DQ15..0 DQ15..0 No DQN15..0 DQN15..0 PRE a tRP tRP+tRCD-R ACT B tRCD-R RD b limit PRd Case (precharge-read different bank) a: ROWP Packet with PRE,Ba Ba =/ Bb b: COL Packet with RD,Bb,Cb PRs Case (precharge-read same bank) a: ROWP Packet with PRE,Ba Ba = Bb b: COL Packet with RD,Bb,Cb Version 1.0 Jan. 2005 Page 15 K4Y5416(/08/04)4UF XDR DRAM Figure 6 : ACT-, RD-, WR-, PRE-to-WR Packet Interactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 ACT WR a b ACT WR a b tRCD-W No limit DQ15..0 DQ15..0 DQN15..0 DQN15..0 AWd Case (activate-write different bank) a: ROWA Packet with ACT,Ba,Ra Ba =/ Bb b: COL Packet with WR,Bb,Cb T0 T1 T2 T3 T4 T5 T6 T7 T8 AWs Case (activate-write same bank) a: ROWA Packet with ACT,Ba,Ra b: COL Packet with WR,Bb,Cb T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 tRW RD a DQ15..D0 DQ15..0 DQN15..0 DQN15..0 WR b Q(a) tCAC RWd Case (read-write-different bank) a: COL Packet with RD,Ba,Ca b: COL Packet with WR,Bb,Cb T0 T1 T2 T3 T4 RD a tCWD T5 D(b) tCC tCYCLE Ba =/ Bb T6 T7 T8 T9 tRW WR b tCAC RWs Case (read-write-same bank) a: COL Packet with RD,Ba,Ca b: COL Packet with WR,Bb,Cb tCWD Q(a) D(b) tCC tCYCLE Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 WR a tCC WR a WR b tCC WR b DQ15..0 DQ15..0 DQN15..0 DQN15..0 WWd Case (write-write different bank) a: COL Packet with WR,Ba,Ca b: COL Packet with WR,Bb,Cb T0 T1 T2 T3 T4 T5 WWs Case (write-write same bank) a: COP Packet with WR,Ba,Ca b: COL Packet with WR,Bb,Cb Ba =/ Bb T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 Ba = Bb tRCD-W PRE WR a b PRE a tRP tRP+tRCD-W ACT WR B b No limit DQ15..0 DQ15..0 DQN15..0 DQN15..0 PWd Case (precharge-write different bank) a: ROWP Packet with PRR,Ba Ba =/ Bb b: COL Packet with WR,Bb,Cb PWs Case (precharge-write same bank) a: ROWP Packet with PRE,Ba Ba = Bb b: COP Packet with WR,Bb,Cb Version 1.0 Jan. 2005 Page 16 K4Y5416(/08/04)4UF XDR DRAM Figure 7 : ACT-, RD-, WR-, PRE-to-PRE Packet Interactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 ACT PRE a b ACT a PRE b tRAS No limit DQ15..0 DQ15..0 DQN15..0 DQN15..0 APd Case (activate-precharge different bank) a: ROWA Packet with ACT,Ba,Ra Ba =/ Bb b: ROWP Packet with PRE,Bb T0 T1 T2 T3 T4 T5 T6 T7 APs Case (activate-precharge same bank) a: ROWA Packet with ACT,Ba,Ra Ba = Bb b: ROWP Packet with PRR,Bb T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 RD PRE a b RD a tRDP PRE b No limit DQ15..0 DQ15..0 DQN15..0 DQN15..0 RPd Case (read-precharge different bank) a: COL Packet with RD,Ba,Ca Ba=/ Bb b: ROWP Packet with PRE,Bb T0 T1 T2 T3 T4 T5 T6 T7 RPs Case (read-precharge same bank) a: COL Packet with RD,Ba,Ca b: ROWP Packet with PRR,Bb T8 T9 Ba = Bb T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 WR PRE a b WR a PRE b tWRP No limit DQ15..0 DQ15..0 DQN15..0 DQN15..0 WPd Case (write-precharge different bank) a: COL Packet with WR,Ba,Ca Ba=/ Bb b: ROWP Packet with PRE,Bb T0 T1 T2 T3 T4 T5 T6 T7 WPs Case (write-precharge same bank) a: COL Packet with WR,Ba,Ca Ba = Bb b: ROWP Packet with PRE,Bb T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 PRE a tPP PRE b PRE a tRP ACT b tRAS tRC PRE b DQ15..0 DQ15..0 DQN15..0 DQN15..0 PPd Case (precharge-precharge different bank) a: ROWP Packet with PRE,Ba Ba # Bb b: ROWP Packet with PRE,Bb PPs Case (precharge-precharge same bank) a: ROWP Packet with PRE,Ba Ba = Bb b: ROWP Packet with PRE,Bb Version 1.0 Jan. 2005 Page 17 K4Y5416(/08/04)4UF XDR DRAM Dynamic Request Scheduling Delay fields are present in the ROWA, COL, and ROWP packet. They permit the associated command to optionally wait for a time of one (or more) tCYCLE before taking effect. This allows a memory controller more scheduling flexibility when issuing request packets. Figure8 illustrates the use of the delay fields. In the first timing diagram, a ROWA packet with an ACT command is present at cycle T0. The DELA field is set to "1". This request packet will be equivalent to a ROWA packet with an ACT command at cycle T1 with the DELA field is set to "0". This equivalence should be used when analyzing request packet interactions. In the second timing diagram, a COL packet with a RD command is present at cycle T0. The DELC field is set to "1". This request packet will be equivalent to a COL packet with an RD command at cycle T1 with the DELC field is set to "0". This equivalence should be used when analyzing request packet interactions. In a similar fashion, a COL packet with a WR command is present at cycle T12. The DELC field is set to"1". This request packet will be equivalent to a COL packet with a WR command at cycle T13 with the DELC field is set to "0". This equivalence should be used when analyzing request packet interactions. In the COL packet with a RD command example, the read data delay, tCAC is measured between the Q read data packet and the virtual COL packet at cycle T1. Likewise, for the example with the COL packet with a WR command, the write data delay, tCWD is measured between the D write data packet and the virtual COL packet at cycle T13. In the third timing diagram, a ROWP packet with a PRE command is present at cycle T0. The DEL field(POP[1:0]) is set to "11". This request packet will be equivalent to a ROWP packet with a PRE command at cycle T1 with the DEL field is set to "10", it will be equivalent to a ROWP packet with a PRE commmand at cycle T2 with the DEL field is set to "01", and it will be equivalent to a ROWP packet with a PRE command at cycle T3 with the DEL field is set to "00". This equivalence should be used when analyzing request packet interactions. In the fourth timing diagram, a ROWP packet with a REFP command is present at cycle T0. The DEL field(RA[7:6] ) is set to "11". This request packet will be equivalent to a ROWP packet with a REFP command at cycle T1 with the DEL field is set to "10", it will be equivalent to a ROWP packet with a REFP command at cycle T2 with the DEL field is set to "01", and it will be equivalent to a ROWP packet with a REFP command at cycle T3 with the DEL field is set to "00". This equivalence should be used when analyzing request packet interactions. The two examples for the REFA and REFI commands are identical to the example just described for the REFP command. The ROWP packet allows two independent operations to be specified. A PRE precharge command uses the POP and BP fields, and the REFP, REFA, or REFI commands use the ROP and RA fields. Both operations have an optional delay field(the POP field for the PRE command and the RA field with the REFP, REFA, or REFI commands). The two delay mechanisms are independent of one another. The POP field does not affect the timing of the REFP, REFA, or REFI commands, and the RA field does not affect the timing of the PRE command. When the interactions of a ROWP packet are analyzed, it must be remembered that there are two independent commands specified, both of which may affect how soon the next request packet can be issued. The constraints from both commands in a ROWP packet must be considered, and the one that requires the longer time interval to the next request packet must be used by the memory controller. Furthermore, the two commands within a ROWP packet may not reference the same bank in the BP and RA fields. Version 1.0 Jan. 2005 Page 18 K4Y5416(/08/04)4UF XDR DRAM Figure 8 : Request Scheduling Examples ACT w/DEL=1 at T0 is equivalent to ACT w/DEL=0 at T1 T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 tCYCLE ACT ACT DEL1 DEL0 DQ15..0 DQN15..0 Note DEL value is specified by DELA field. ROWA/ACT Command WR w/DEL=1 at T12 is equivalent to WR w/DEL=0 at T13 RD w/DEL=1 at T0 is equivalent to RD w/DEL=0 at T1 T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 RD RD DEL1 DEL0 tCYCLE WR WR DEL1 DEL0 DQ15..0 DQN15..0 Q tCAC D tCWD Note DEL value is specified by DELC field. COL/RD and COL/WR Commands PRE w/DEL=3 at T0 is equivalent to PRE w/DEL =2 at T1 or PRE w/DEL=1 at T2 or PRE w/DEL=0 at T3 T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 tCYCLE PRE PRE PRE PRE DEL3 DEL2 DEL1 DEL0 DQ15..0 DQN15..0 Note DEL value is specified by {POP1, POP0} field. ROWP/PRE Command REFP w/DEL=3 at T0 is equivalent to REFP w/DEL=2 at T1 or REFP w/DEL=1 at T2 or REFP w/DEL=0 at T3 T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 REFI w/DEL=3 at T13 is equivalent to REFI w/DEL=2 at T14 or REFI w/DEL=1 at T15 or REFI w/DEL=0 at T16 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 DQ15..0 DQN15..0 REFP REFP REFP REFP DEL3 DEL2 DEL1 DEL0 REFA REFA REFA REFA DEL3 DEL2 DEL1 DEL0 REFI REFI REFI REFI DEL3 DEL2 DEL1 DEL0 tCYCLE REFA w/DEL=3 at T6 is equivalent to REFA w/DEL=2 at T7 or REFA w/DEL=1 at T8 or REFA w/DEL=0 at T9 Note DEL value is specified by {RA7, RA6} field. ROWP/REFP,REFA,REFI Commands Version 1.0 Jan. 2005 Page 19 K4Y5416(/08/04)4UF XDR DRAM Memory Operations Write Transactions Figure9 shows four examples of memory write transactions. A transaction is one or more request packets (and the associated data packets) needed to perform a memory access. The state of the memory core and the address of the memory access determine how many request packets are needed to perform the access. The first timing diagram shows a page-hit write transaction. In this case, the selected bank is already open (a row is already present in the sense amp array for the bank). In addition, the selected row for the memory access matches the address of the row already sensed (a page hit). This comparison must be done in the memory controller. In this example, the access is made to row Ra of bank Ba. In this case, write data may be directly written into the sense amp array for the bank, and row operations(activated or precharge) are not needed. A COL packet with WR command to column Ca1 of bank Ba is presented on edge T0, and a second COL packet with WR command to column Ca1 of bank Ba is presented on edge T2. Two write data packets D(a1) and D(a2) follow these COL packets after the write data delay tCWD. The two COL packets are separated by the column-cycle time tCC. This is also the length of each write data packet. The second timing diagram shows an example of a page-miss write transaction. In this case, the selected bank is already open (a row is already present in the sense amp array for the bank). However, the selected row for the memory access does not match the address of the row already sensed (a page miss). This comparsion must be done in the memory controller. In this example, the access is made to row Ra of bank Ba, and the bank contains a row other than Ra. In this case, write data may not be directly written into the sense amp array for the bank. It is necessary to close the present row (precharge) and access the requested row (activate). A precharge command (PRE to bank Ba) is presented on edge T0. An activate command (ACT to row Ra of bank Ba) is presented on edge T6 a time tRP later. A COL packet with WR command to column Ca1 of bank Ba is presented on edge T7 a time tRCD-W later. A second COL packet with WR command to column Ca2 of bank Ba is presented on edge T9. Two write data packets D(a1) and D(a2) follow these COL packets after the write data delay tCWD. The two COL packets are separated by the column-cycle time tCC. This is also the length of each write data packet. The third timing diagram shows an example of a page-empty wirte transaction. In this case, the selected bank is already closed (no row is present in the sense amp array for the bank). No row comparison is necessary for this case; however, the memory controller must still remember that bank Ba has been left closed. In this example, the access is made to row Ra of bank Ba. In this case, write data may not be directly written into the sense amp array for the bank. It is necessary to access the requested row (activate). An activate command (ACT to row Ra of bank Ba) is presented on edge T0. A COL packet with WR command to column Ca1 of bank Ba is presented on edge T1 a time tRCD-W later. A second COL packet with WR command to column Ca2 of bank Ba is presented on edge T3. Two write data packets D(a1) and D(a2) follow these COL packets after the write data delay tCWD. The two COL packets are separated by the column-cycle time tCC. This is also the length of each write data packet. After the final write command, it may be necessary to close the present row (precharge). A precharge command (PRE to bank Ba) is presented on edge T14 a time tWRP after the last COL packet with a WR command. The decision whether to close the bank or leave it open is made by memory controller and its page policy. The fourth timing diagram shows another example of a page-empty write transaction. This is similar to the previous example except that only a single write command is presented, rather than two write commands. This example shows that even with a minimum length write transaction, tRAS parameter will not be a constraint. The tRAS measures the minimum time between an activate command and a precharge command to a bank. This time interval is also constrained by the sum tRCD-W + tWRP which will be larger for a write transaction. These two constraints (tRAS and tRCD-W + tWRP) will be a function of the memory device's speed bin and the data transfer length (the number of write commands issued between the activate and precharge commands), and the tRAS parameter could become a constraint for future speed bins. In this example, the sum tRCD-W + tWRPs is greater than tRAS by the amount tRAS. Version 1.0 Jan. 2005 Page 20 K4Y5416(/08/04)4UF XDR DRAM Figure 9 : Write Transactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WR a1 tCYCLE WR a2 tCC DQ15..0 DQN15..0 D(a1) tCWD D(a2) Page-hit Write Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 PRE a3 ACT WR a0 a1 tRP DQ15..0 DQN15..0 tRCD-W tCYCLE WR a2 tCC tCWD Transaction a: WR D(a1) a0 = {Ba,Ra} D(a2) a1 = {Ba,Ca1} a2 = {Ba,Ca2} a3 = {Ba} Page-miss Write Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 ACT WR a0 a1 DQ15..0 tRCD-W DQN15..0 WR a2 tCWD tDP tCC D(a1) tCWD tCYCLE PRE a3 tWRP D(a2) Transaction a: WR a0 = {Ba,Ra} a1 = {Ba,Ca1} a2 = {Ba,Ca2} a3 = {Ba} Page-empty Write Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 tRAS ACT WR a0 a1 DQ15..0 tRCD-W DQN15..0 Bb = Ba tRAS tWRP tCWD PRE a3 tRP ACT b0 tCYCLE D(a1) Transaction a: WR Transaction b: WR a0 = {Ba,Ra} b0 = {Bb,Rb} a1 = {Ba,Ca1} b1 = {Bb,Cb1} a2 = {Ba,Ca2} b2 = {Bb,Cb2} a3 = {Ba} b3 = {Bb} Page-empty Write Example - Core Limited Version 1.0 Jan. 2005 Page 21 K4Y5416(/08/04)4UF XDR DRAM Read Transactions Figure10 shows four examples of memory read transactions. A transaction is one or more request packets (and the associated data packets) needed to perform a memory access. The state of the memory core and the address of memory access determine how many request packets are needed to perform the access. The first timing diagram shows a page-hit read transaction. In this case, the selected bank is already open (a row is already present in the sense amp array for the bank). In addition, the selected row for the memory access matches the address of the row already sensed (a page hit). This comparison must be done in the memory controller. In this example, the access is made to row Ra of bank Ba. In this case, read data may be directly read from the sense amp array for the bank and no row operations (actiavate or precharge) are needed. A COL packet with RD command to column Ca1 of bank Ba is presented on edge T0 and a second COL packet with RD command to column Ca2 of bank Ba is presented on edge T2. Two read data packets Q(a1) and Q(a2) follow these COL packets after the read data delay tCAC. The two COL packets are separated by the column-cycle time tCC. This is also the length of each read data packet. The second timing diagram shows an example of a page-miss read transaction. In this case, the selected bank is already open (a row is already present in the sense amp array for the bank). However, the selected row for the memory access does not match the address of the row already sensed(a page miss). This comparison must be done in the memory controller. In this example, the access is made to row Ra of bank Ba, and the bank contains a row other than Ra. In this case, read data may not be directly read from the sense amp array for the bank. It is necessary to close the present row (precharge) and access the requested row (activate). A precharge command (PRE to bank Ba) is presented on edge T0. An activate command (ACT to row Ra of bank Ba) is presented on edge T6 a time tRP later. A COL packet with RD command to column Ca1 of bank Ba is presented on edge T11 a time tRCD-R later. A second COL packet with RD command to column Ca2 of bank Ba is presented on edge T13. Two read data packets Q(a1) and Q(a2) follow these COL packets after the read data delay tCAC. The two COL packets are separated by the column-cycle time tCC. This is also the length of each read data packet. The third timing diagram shows an example of a page-empty write transaction. In this case, the selected bank is already closed (no row is present in the sense amp array for the bank). No row comparison is necessary for this case; however, the memory controller must still remember that bank Ba has been left closed. In this example, the access is made to row Ra of bank Ba. In this case, read data may not be directly read from the sense amp array for the bank. It is necessary to access the requested row (activated). An activate command (ACT to row Ra of bank Ba) is presented on edge T0. A COL packet with RD command to column Ca1 of bank Ba is presented on edge T5 a time tRCD-R later. A second COL packet with RD command to column Ca2 of bank Ba is presented on edge T7. Two read data packets Q(a1) and Q(a2) follow these COL packets after the read data delay tCAC. The two COL packets are separated by the column-cycle time tCC. This is also the length of each read data packet. After the final read command, it may be necessary to close the present row (precharge). A precharge command - PRE to bank Ba - is presented on edge T10 a time tRDP after the last COL packet with a RD command. Whether the bank is closed or left open depends on the memory controller and its page policy. The fourth timing diagram shows another example of a page-empty read transaction. This is similar to the previous example except that it uses one read command instead of two read commands. In this case, the core parameter tRAS may also be a constraint upon when the precharge command may be issued. The tRAS measures the minimum time between an activate command and a precharge command to a bank. This time interval is also constrained by the sum tRCD-R + tRDP and must be set to whichever is larger. These two constraints (tRAS and tRCD-R + tRDP) will be a function of the memory device's speed bin and the data transfer length (the number of read commands issued between the activate and precharge commands). In this example, the tRAS is greater than the sum tRCD-R + tRDP by the amount tRDP. Version 1.0 Jan. 2005 Page 22 K4Y5416(/08/04)4UF XDR DRAM Figure 10 : Read Transactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 RD a1 tCYCLE RD a2 tCC DQ15..0 DQN15..0 Q(a1) tCAC Transaction a: RD Q(a2) a0 = {Ba,Ra} a1 = {Ba,Ca1} a2 = {Ba,Ca2} a3 = {Ba} Page-hit Read Example T0 T1 T2 T3 T4 T5 T6 T8 T7 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 PRE a3 ACT a0 RD a1 tRP DQ15..0 DQN15..0 tRCD-R tCYCLE RD a2 tCC Q(a1) tCAC Transaction a: RD a0 = {Ba,Ra} a1 = {Ba,Ca1} a2 = {Ba,Ca2} Q(a2) a3 = {Ba} Page-miss Read Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 ACT a0 RD a1 tRCD-R DQ15..0 DQN15..0 RD a2 tCC Q(a1) tCAC Transaction a: RD tCYCLE PRE a3 tRDP a0 = {Ba,Ra} Q(a2) a1 = {Ba,Ca1} a2 = {Ba,Ca2} a3 = {Ba} Page-empty Read Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 tRAS ACT a0 RD a1 tRCD-R DQ15..0 DQN15..0 tRDP tRDP Transaction a: RD Transaction b: RD tCYCLE ACT b0 Q(a1) tCAC Bb = Ba tRP PRE a3 a0 = {Ba,Ra} b0 = {Bb,Rb} a1 = {Ba,Ca1} b1 = {Bb,Cb1} a2 = {Ba,Ca2} b2 = {Bb,Cb2} a3 = {Ba} b3 = {Bb} Page-empty Read Example - Core Limited Version 1.0 Jan. 2005 Page 23 K4Y5416(/08/04)4UF XDR DRAM Interleaved Transactions Figure11 shows two examples of interleaved transactions. Interleaved transactions are overlapped with one another; a transaction is started before an earlier one is completed. The timing diagram at the top of the figure shows interleaved write transactions. Each transaction assumes a page-empty access; that is, a bank is in a closed state prior to an access and is precharged after the access. With this assumption, each transaction requires the same number of request packets at the same relative positions. If bank were allowed to be in an open state, then each transaction would require a different number of request packets depending upon whether the transaction was page-empty, page-hit or page-miss. This situation is more complicated for the memory controller and will not be analyzed in this document. In the interleaved page-empty write example, there are four sets of request pins RQ11...0 shown along the left side of the timing diagram. The first three show the timing slots used by each of the three requests packet types (ACT, COL and PRE), and the fourth set (ALL) shows the previous three merged together. This allows the pattern used for allocating request slots for the different packets to be seen more clearly. The slots at {T0, T4, T8, T12 ...} are used for ROWA packets with ACT commands. This spacing is determined by the tRR parameter. There should not be interference between the interleaved transactions due to resource conflicts because each bank address - Ba, Bb, Bc, Bd and Be - is assumed to be different from another. If two of the bank addresses are the same, the later transaction would need to wait until the earlier transaction had completed its precharge operation. Five different banks are needed because the effective tRC (tRC + tRC) is 20*tCYCLE. The slots at {T1, T3, T5, T7, T9, T11, ...} are used for COL packets with WR commands. This frequency of the COL packet spacing is determined by the tCC parameter and by the fact that there are two column accesses per row access. The phasing of the COL packet spacing is determined by the tRCD-W parameter. If the value of tRCD-W required the COL packets to occupy the same request slots as the ROWA packets (this case is not shown), the DELC field in the COL packet could be used to place the COL packet one tCYCLEs earlier. The DQ bus slots at {T7, T9, T11, T13, ...} carry the write data packets {D(a1), D(a2), D(b1), D(b2), ...}. Two write data packets are written to a bank in each transaction. The DQ bus is completely filled with write data; no idle cycles need to be introduced because there are no resource conflicts in this example. The slots at {T14, T18, T22, ...} are used for ROWP packets with PRE commands. This frequency of ROWP packet spacing is determined by the tPP parameter. The phasing of the ROWP packet spacing is determined by the tWRP paramter. If the value of tWRP required the ROWP packets to occupy the same request slots as the ROWA or COL packets already assigned (this case is not shown), the delay field in the ROWP packet could be used to place the ROWP packet one or more tCYCLE earlier. There is an example of an interleaved page-empty read at the bottom of the figure. As before, there are four sets of request pins RQ11...0 shown along the left side of the timing diagram, allowing the pattern used for allocating request slots for the different packets to be seen more clearly. The slots at {T0, T4, T8, T12, ...} are used for ROWA packets with ACT command. This spacing is determined by the tRR parameter. There should not be interference between the interleaved transactions due to resource conflicts because each bank address - Ba, Bb, Bc and Bd - is assumed to be different from another. Four different banks are needed because the effective tRC is 16 * tCYCLE. The slots at {T5, T7, T9, T11, ...} are used for COL packets with RD commands. This frequency of the COL packet spacing is determined by the tCC paramter and by the fact that there are two column accesses per row access. The phasing of the COL packet spacing is determined by the tRCD-R parameter. If the value of tRCD-R required the COL packets to occupy the same request slots as the ROWA packets (this case is not shown), the DELC field in the COL packet could be used to place the packet one tCYCLE earlier. The DQ bus slots at {T11, T13, T15, T17, ...} carry the read data packets {Q(a1), Q(a2), Q(b1), Q(b2), ...}, Two read data packets are read from a bank in each transaction. The DQ bus is completely filled with read data - That is, no idle cycles need to be introduced because there are no resource conflicts in this example. The slots at {T10, T14, T18, T22, ...} are used for ROWP packets with PRE commands. This frequency of the ROWP packet spacing is determined by the tPP parameter. The phasing of the ROWP packet spacing is determined by the tRDP parameter. If the value of tRDP required the ROWP packets to occupy the same request slots as the ROWA or COL packets already assigned (this case is not shown), the delay field in the ROWP packet could be used to place the ROWP packet one or more tCYCLEs earlier. Version 1.0 Jan. 2005 Page 24 K4Y5416(/08/04)4UF XDR DRAM Figure 11 : Interleaved Transactions T0 T1 T2 T3 T4 T5 T6 T7 T8 The effective tRC time is increased by 4 tCYCLE T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 T9 CFM CFMN RQ11..0 (ACT) ACT a0 RQ11..0 (COL) WR a1 DQ15..0 tRCD-W DQN15..0 RQ11..0 (PRE) RQ11..0 (ALL) tRC ACT b0 tRR WR a2 WR b1 ACT c0 WR b2 D(a1) ACT WR a0 a1 D(a2) WR c2 D(b1) WR d1 D(b2) WR d2 D(c1) WR e1 WR ACT WR a2 b0 b1 D(c2) tWRP tWRP D(d2) tRP PRE a3 WR f1 WR e2 D(d1) tCYCLE ACT f0 WR f2 D(e1) D(e1) PRE b3 PRE c3 WR ACT WR PRE WR ACT WR PRE WR ACT WR PRE WR f1 f2 c2 d0 d1 a3 d2 e0 e1 b3 e2 f0 c3 WR ACT WR b2 c0 c1 Transaction a: WR Transaction b: WR Transaction c: WR Transaction d: WR Transaction e: WR Transaction f: WR Bf = Ba WR c1 tRC ACT e0 tCC tCWD Ba,Bb,Bc,Bd,Be are different banks. ACT d0 a0 = {Ba,Ra} b0 = {Bb,Rb} c0 = {Bc,Rc} d0 = {Bd,Rd} e0 = {Be,Re} e0 = {Bf,Rf} a1 = {Ba,Ca1} b1 = {Bb,Cb1} c1 = {Bc,Cc1} d1 = {Bd,Cd1} e1 = {Be,Ce1} f1 = {Bf,Cf1} a2 = {Ba,Ca2} b2 = {Bb,Cb2} c2 = {Bc,Cc2} a2 = {Bd,Cd2} e2 = {Be,Ce2} f2 = {Bf,Cf2} a3 = {Ba} b3 = {Bb} c3 = {Bc} d3 = {Bd} e3 = {Be} f3 = {Bf} Interleaved Page-empty Write Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 (ACT) ACT a0 RQ11..0 (COL) tRR DQ15..0 DQN15..0 tRCD-R ACT a0 Ba,Bb,Bc,Bd are different banks. Be = Ba RD a1 ACT c0 RD a2 ACT d0 RD b1 RD b2 ACT e0 RD c1 RD c2 tCYCLE ACT f0 RD d1 RD d2 RD e1 RD e2 tCAC Q(a1) tCC RQ11..0 (PRE) RQ11..0 (ALL) tRC ACT b0 ACT RD b0 a1 Transaction a: RD Transaction b: RD Transaction c: RD Transaction d: RD Transaction e: RD tRDP Q(a2) tRP PRE a3 PRE b3 Q(b1) Q(b2) PRE c3 Q(c1) Q(c2) PRE d3 RD ACT RD PRE RD ACT RD PRE RD ACT RD PRE RD ACT RD PRE RD a2 c0 b1 a3 b2 d0 c1 b3 c2 e0 d1 c3 d2 f0 e1 d3 e2 a0 = {Ba,Ra} b0 = {Bb,Rb} c0 = {Bc,Rc} d0 = {Bd,Rd} e0 = {Be,Re} a1 = {Ba,Ca1} b1 = {Bb,Cb1} c1 = {Bc,Cc1} d1 = {Bd,Cd1} e1 = {Be,Ce1} a2 = {Ba,Ca2} b2 = {Bb,Cb2} c2 = {Bc,Cc2} a2 = {Bd,Cd2} e2 = {Be,Ce2} a3 = {Ba} b3 = {Bb} c3 = {Bc} d3 = {Bd} e3 = {Be} Interleaved Page-empty Read Example Version 1.0 Jan. 2005 Page 25 K4Y5416(/08/04)4UF XDR DRAM Read/Write Interaction The previous section described overlapped read transactions and overlapped write transactions in isolation. This section will describe the interaction of read and write transactions and the spacing required to avoid channel and core resource conflicts. Figure12 shows a timing diagram (top) for the first case, a write transaction followed by a read transaction. Two COL packets with WR commands are presented on cycles T0 and T2. The write data packets are presented a time tCWD later on cycles T4 and T6. The device requires a time tWR after the second COL packet with a WR command before a COL packet with a RD command may be presented. Two COL packets with RD commands are presented on cycles T11 and T13. The read data packets are returned a time tCAC later on cycles T17 and T19. The time tWR is required for turning around internal bi-directional interconnections (inside the device). This time must be observed regardless of whether the write and read commands are directed to same banks or different banks. A gap tWR-BUB,XDRDRAM will appear on the DQ bus between the end of the D(a2) packet and the beginning of the Q(b1) packet (measured at the appropriate packet reference points). The size of this gap can be evaluated by calculating the difference between cycles T2 and T17 using the two timing paths : tWR-BUB, XDRDRAM = tWR + tCAC - tCWD - tCC In this example, the value of tWR-BUB,XDRDRAM is greater than its minimum value of tWR-BUB,XDRDRAM,MIN. The values of tWR and tCAC are equal to their minimum values. In the second case, the timing diagram displayed at the bottom of Figure12 illustrates a read transaction followed by a write transaction. Two COL packets with RD commands are presented on cycles T0 and T2. The read data packets are returned a time tCAC later on cycles T6 and T8. The device requires a time tRW after the second COL packet with a RD command before a COL packet with a WR command may be presented. Two COL packets with WR commands are presented on cycles T10 and T12. The write data packets are presented a time tCWD later on cycles T13 and T15. The time tRW is required for turning around the external DQ bi-directional interconnections (outside the device). This time must be observed regardless whether the read and write commands are directed to the same banks or different banks. The time tRW depends upon four timing parameters. and may be evaluated by calculating the difference between cycles T2 and T11 using the two timing paths : tRW + tCWD = tCAC + tCC + tRW-BUB, XDRDRAM or tRW = (tCAC - tCWD) + tCC + tRW-BUB, XDRDRAM In this example, the values of tRW, tCAC, tCWD, tCC, and tWR-BUB, XDRDRAM are equal to their minimum values. Figure 12 : Write/Read Interaction T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WR a1 WR a2 tCWD tDR RD b1 tWR DQ15..0 DQN15..0 D(a1) RD b2 tCYCLE tCAC D(a2) tCWD Q(b1) Q(b2) tWR-BUB,XDRDRAM tCC Transaction a: WR Transaction b: RD a1 = {Ba,Ca1} b1 = {Bb,Cb1} a2 = {Ba,Ca2} b2 = {Bb,Cb2} Write/Read Turnaround Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 DQ15..0 DQN15..0 RD a1 RD a2 WR b1 tRW tCAC Q(a1) WR b2 Q(a2) tCC tCYCLE tCWD D(b1) D(b2) tRW-BUB,XDRDRAM Transaction a: WR Transaction b: RD a1 = {Ba,Ca1} b1 = {Bb,Cb1} a2 = {Ba,Ca2} b2 = {Bb,Cb2} Read/Write Turnaround Example Version 1.0 Jan. 2005 Page 26 K4Y5416(/08/04)4UF XDR DRAM Propagation Delay Figure 13 shows two timing diagrams that display the system-level timing relationships between the memory component and the memory controller. The timing diagram at the top of the figure shows the case of a write-read-write command and data at the memory component. In this case, the timing will be identical to what has already been shown in the previous sections; i,e. with all timing measured at the pins of the memory component. This timing diagram was produced by merging portions of the top and bottom timing diagrams in Figure12. The example shown is that of a single COL packet with a write command, followed by a single COL packet with a read command, followed by a second COL packet with a write command. These accesses all assume a page-hit to an open bank. A timing interval tWR is required between the first WR command and the RD command, and a timing interval tRW is required between the RD command and the second WR command. There is a write data delay tCWD between each WR command and the associated write data packet D. There is a read data delay tCAC between the RD command and the associated read data packet Q. In this example, all timing parameters have assumed their minimum values except tWR-BUB, XDRDRAM. The lower timing diagram in the figure shows the case where timing skew is present between the memory controller and the memory component. This skew is the result of the propagation delay of signal wavefronts on the wire carrying the signals. The example in the lower diagram assumes that there is a propagation delay of tPD-RQ along both the RQ wires and the CFM/CFMN clock wires between the memory controller and the memory component (the value of tPD-RQ used here is 1*tCYCLE). Note that in an actual system the tPD-RQ value will be different for each memory component connected to the RQ wires. In addition, it is assumed that there is a propagation delay tPD-D along the DQ/DQN wires between the memory controller and the memory component (the direction in which write data travels, and it is assumed that there is the same propagation delay tPDQ along the DQ/DQN wires between the memory component and the memory controller (the direction in which read data travels). The sum of these two propagation delays is also denoted by the timing parameter tPD,CYC = tPD-D + tPD-Q. As a result of these propagation delays, the position of packets will have timing skews that depend upon whether they are measured at the pins of the memory controller or the pins of memory component. For example, the CFM/CFMN signals at the points of the memory component are tPD-RQ later than at the pins of the memory controller. This is shown by the cycle numbering of the CFM/CFMN signals at the two locations - in this example cycle T1 at the memory controller. All the request packets on the RQ wires will have a tPD-RQ skew at the memory component relative to the memory controller in this example. Because the tPD-D propagation delay of write data matches the tPD-RQ propagation delay of the write command, the controller may issue the write data packet D(a0) relative to the COL packet with the first write command "WR(a0)" with normal write data delay tCWD. If the propagation delays between the memory controller and memory component were different for the RQ and DQ buses (not shown in this example), the write data delay at the memory controller would need to be adjusted. A propagation delay is seen by the read command - that is, the read command will be delayed by a tPD-RQ skew at the memory component relative to the memory controller. The memory componet will return the read data packet Q(b0) relative to this read command with the normal read data delay tCAC (at the pins of the memory componet). The read data packet will be skewed by an additional propagation delay of tPD-Q as it travels from the memory component back to the memory controller. The effective read data delay measured between the read command and the read data at the memory controller will be tCAC + tPD-RQ + tPD-Q. tPD-RQ factor is casued by the propagaion delay of the request packets as they travel from memory controller to memory component. The tPD-Q factor is casued by the propagation delay of the read data packets as they travel from memory componet to memory controller. All timing parameters will be equal to their minimum values except tWR-BUB,XDRDRAM (as in the top diagram), and the timing parameters tRW-BUB,XDRDRAM and tRW. These will be larger than their minimum values by the amount (tPD,CYC - tPD,CYC,MIN), where tPD,CYC = tPD-D + tPD-Q. This may be seen by evaluating the two timing paths between cycle T9 at th controller and cycle T21 at the XDR DRAM: tRW + tPD-RQ + tCWD = tPD-RQ + tCAC + tCC + tRW-BUB,XDRDRAM or tRW = (tCAC - tCWD) + tCC + tRW-BUB,XDRDRAM The following relationship was shown for Figure12. tRW, MIN = (tCAC - tCWD) + tCC + tRW-BUB, XDRDRAM, MIN or (tRW - tRW, MIN) = (tRW-BUB, XDRDRAM - tRW-BUB, XDRDRAM, MIN) In other words, the two timing parameters tRW-BUB,XDRDRAM and tRW will change together. The relationship of this change to the propagation delay tPD,CYC (=tPD-D + tPD-Q) can be derived by looking at the two timing paths from T15 to T21 at the XDR DRAM: tPD-Q + tCC + tRW-BUB,XIO + tPD-D = tCC + tRW-BUB,XDRDRAM or tRW-BUB,XDRDRAM = tRW-BUB,XIO + tPD-D + tPD-Q or tRW-BUB,XDRDRAM = tRW-BUB,XIO + tPD,CYC Version 1.0 Jan. 2005 Page 27 K4Y5416(/08/04)4UF XDR DRAM in a system with minimum propagation delays: tRW-BUB,XDRDRAM,MIN = tRW-BUB,XIO + tPD,CYC,MIN and since tRW-BUB,XIO is equal to tRW-BUB,XIO,MIN in the both cases, the following is true: (tPD,CYC - tPD,CYC,MIN) = (tRW-BUB,XDRDRAM - tRW-BUB,XDRDRAM,MIN) = (tRW - tRW,MIN) In other words, the values of the tRW-BUB,XDRDRAM,MIN and tRW,MIN timing parameters correspond to the value of tPD,CYC,MIN for the system (this is equal to one tCYCLE). As tPD,CYC is increased from this minimum value, tRW-BUB,XDRDRAM and tRW increase from their minimum values by an equivalent amount. Figure 13 : Propagation Delay XDR DRAM T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WR a0 DQ15..0 DQN15..0 RD b0 tWR D(a0) tCWD tCAC tCC tWR-BUB,XDRDRAM T1 T2 T3 T4 T5 T6 T7 tCWD Q(b0) D(c0) tCC tRW-BUB,XDRDRAM a0 = {Ba,Ca0} b0 = {Bb,Cb0} c0 = {Bc,Cc0} Transaction a: WR Transaction b: RD Transaction c: WR Controller T0 tCYCLE WR c0 tRW Write-Read-Write at XDR DRAM (portions of top and bottom timing diagrams of Figure 12 merged) T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WR a0 DQ15..0 DQN15..0 tDRW RD b0 tDWR tCC D(a0) XDR DRAMT -1 T0 T1 T2 T3 T5 T6 T7 T8 T9 tCYCLE tRW-BUB,XIO Q(b0) tPD-Q T4 WR c0 D(c0) T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 CFM CFMN tPD-RQ tCWD tPD-RQ D(a0) Transaction a: WR Transaction b: RD Transaction c: WR tPD-RQ RD b0 a0 = {Ba,Ca0} b0 = {Bb,Cb0} c0 = {Bc,Cc0} Q(b0) tCAC WR c0 tPD-D tCWD tRW-BUB,XDRDRAMD(c0) tCC Write-Read-Write at Controller and XDR DRAM w/ tPD-RQ = tPD-Q = tPD-D = 1*tCYCLE tPD-RQ ... DQ15..0 DQN15..0 tCYCLE tPD-D RQ Controller tPD-D RQ DQ DQ tPD-Q XDR DRAM ... RQ11..0 WR a0 Version 1.0 Jan. 2005 Page 28 K4Y5416(/08/04)4UF XDR DRAM Register Operations Serial Transactions The serial interface consists of five pins. This includes RST, SCK, CMD, SDI and SDO. SDO uses CMOS signaling levels. The other four pins use RSL signaling levels. RST, CMD, SDI and SDO use a timing window which surrounds the falling edge of SCK. The RST pin is used for initialization. Figure14 and Figure15 show examples of a serial write transaction and a serial read transaction. Each transaction starts on cycle S4 and requires 32 SCK edges. The next serial transaction can begin on cycle S36. SCK does not need to be asserted if there is no transaction. Serial Write Transactions The serial device write transaction in Figure14 begins with the Start [3:0] field. This consists of bits "1100" on the CMD pin. This indicates to the XDR DRAM that the remaining 28 bits constitute a serial transaction. The next two bits are the SCMD[1:0] field. This field contains the serial command, the bits 00 in the case of a serial device write transaction. The next eight bits are "00" and the SID[5:0] field. This field contains the serial identification of the device being accessed. The next eight bits are the SADR[7:0] field. This field contains the serial address of the control register being accessed. A single bit "0" follows next. This bit allows one cycle for the access time to the control register. The next eight bits on the CMD pin is the SWD[7:0] field. this is the write data that is placed into the selected control register. A final bit"0" is driven on the CMD pin to finish the serial write transaction. A serial broadcast write is identical except that the contents of the SID[5:0] field in the transaction is ignored and all devices perform the register write. The SDI and SDO pins are not used during either serial write transaction. Serial Read Transactions The serial device read transaction in Figure15 begins with the Start[3:0] field. This consists of bits "1100" on the CMD pin. This indicates that the remaining 28 bits constitute a serial transaction. The next two bits are the SCMD [1:0] field. This field contains the serial command, and the bits "10" in the case of a serial device read transaction. The next eight bits are "00" and the SID [5:0] field. This field contains the serial identification of the device being accessed. The next eight bits are the SADR [7:0] field and contain the serial address of the control register being accessed. A single bit "0" follows next. This bit allows one cycle for the access time to the control register and time to turn on the SDO output driver. The next eight bits on the CMD pin are the sequence "00000000". At the same time, the eight bits on the SDO pin are the SRD [7:0] field. This is the read data that is accessed from the selected control register. Note the output timing convention here: bit SRD [7] is driven from a time tQ,SI,MAX after edge S26 to a time tQ,SI,MIN after edge S27. The bit is sampled in the controller by the edge S27. A final bit "0" is driven on the CMD pin to finish the serial read transaction. A serial forece read is identical except that the contents of the SID [5:0] field in the transaction is ignored and all devices perform the register read. This is used for device testing. Figure16 shows the response of a DRAM to a serial device read transaction when its internal SID [5:0] register field doesn't match the SID [5:0] field of the transaction. Instead of driving read data from an internal register for cycle edges S27 through S34 on the SDO output pin, it passes the input data from the SDI input pin to the SDO output pin during this same period. Table 9: SCMD Field Encoding Summary SCMD[1:0] Command DESCRIPTION 00 SDW Serial device write-one device is written, the one whose SID[5:0] register matches the SID[5:0] field of the transaction. 01 SBW Serial broadcast write - all devices are written, regardless of the contents of the SID [5:0] register and the SID [5:0] transaction field. 10 SDR Serial device read - one device is read, the one whose SID[5:0] register matches the SID[5:0] field of the transaction. 11 SFR Serial forced read - all devices are read, regardless of the contents of the SID[5:0] register and the SID[5:0] transaction field Version 1.0 Jan. 2005 Page 29 K4Y5416(/08/04)4UF XDR DRAM Figure 14 : Serial Write Transaction S0 S2 S4 S6 S8 S10 S12 S14 S16 S18 S20 S22 S24 S26 S28 S30 S32 S34 S36 S38 S40 S42 S44 S46 S48 SCK tCYC,SCK RST Start CMD transaction 2'h0,SID[5:0] SCMD `1' `1' `0' `0' `0' `0'`0' `0' 5 4 3 2 SADR[7:0] 1 0 7 6 5 4 3 2 SWD[7:0] 1 0 `0' 7 6 5 4 3 2 1 0 `0' SDI (input) SDO (output) Figure 15 : Serial Read Transaction -- Selected DRAM S0 S2 S4 S6 S8 S10 S12 S14 S16 S18 S20 S22 S24 S26 S28 S30 S32 S34 S36 S38 S40 S42 S44 S46 S48 SCK tCYC,SCK RST Start CMD transaction 2'h0,SID[5:0] SCMD `1' `1' `0' `0' `1' `0'`0' `0' 5 4 3 2 SADR[7:0] 1 0 7 6 5 4 3 2 8'h00 1 0 `0' `0' `0' `0' `0' `0' '0' `0' `0' `0' SDI (input) SDO (output) SRD[7:0] 7 6 5 4 3 2 1 0 Figure 16 : Serial Read Transaction -- Non-selected DRAM S0 S2 S4 S6 S8 S10 S12 S14 S16 S18 S20 S22 S24 S26 S28 S30 S32 S34 S36 S38 S40 S42 S44 S46 S48 SCK S28 tCYC,SCK RST CMD Start SCMD SDI transaction 2'h0,SID[5:0] `1' `1' `0' `0' `1' `0'`0' `0' 5 4 3 2 SADR[7:0] 1 0 7 6 5 4 3 2 tP,SI 8'h00 1 0 `0' `0' `0' `0' `0' `0' '0' `0' `0' `0' SDI (input) 7 6 SDO (output) 7 6 5 SRD[7:0] 5 4 3 2 1 0 SRD[7:0] 4 3 2 SDO combinational propagation from SDI to SDO 1 0 Version 1.0 Jan. 2005 Page 30 K4Y5416(/08/04)4UF XDR DRAM Register Summary Figure17 through Figure42 show the control register in the memory component. The control registers are responsible for configuring the component's operating mode, for managing power state transactions, for managing refresh, and for managing calibration operations. A control register may contain up to eight bits. Each figure shows defined bits in white and reserved bits in gray. Reserved bits must be written as 0 and must be ignored when read. Write-only fields must be ignored when read . Each figure displays the following register information: 1. Register name 2. Register mnemonic 3. Register address (SADR [7:0] value needed to access it) 4. Read-only, write-only or read-write 5. Initialization state 6. Description of each defined register field Figure17 shows the Serial Identification register. The register contains the SID [5:0] (serial identification field). This field contains the serial identification value for the deice. The value is compared to the SID[5:0] field of a serial transaction to determine if the serial transaction is directed to this device. The serial identification value is set during the initialization sequence. Figure18 shows the Configuration Register. It contains three fields. The first is the WIDTH field. This field allows the number of DQ/DQN pins used for memory read and write accesses to be adjusted. The SLE field enables data to be written into the memory through the serial interface using the WDSL register. Figure19 shows the Power Management Register. It contains two fields. The first is the PX field. When this field is written with a "1", the memory component transactions from powerdown to active state. It is usually unnecessary to write a "0" into this field; this is done automatically by the PDN command in a COLX packet. The PST field indicates the current power state of the memory component. Figure20 shows the Write Data Serial Load Register. It permits data to be written into memory via the Serial Interface. Figure23 shows the Refresh Bank Control Register. It contains two fields: BANK and MBR. The BANK field is read-write and contains the bank address used by self-refresh during the powerdown state. The MBR field controls how many banks are refreshed during each refresh operation. Figure24, Figure25 and Figure26 show different fields of the Refresh Row Register (high, middle and low). This read-write field contains the row address used by self- and auto-refresh. See"Refresh Transactions" on page 37 for more details. Figure28 and Figure29 show the Current Calibration 0 and 1 registers. They contain the CCVALUE0 and CCVALUE1 fields, respectively. These are read-write fields which control the amount of IOL current driven by the DQ and DQN pins during a read transaction. The Current Calibration 0 Register controls the even-numbered DQ and DQN pins, and the Current Calibration 1 controls the odd-numbered DQ and DQN pins. Figure30 and Figure31 show the Impedance Calibration 0 and 1 registers. They contain the ZCVALUE0 and ZCVALUE1 field, respectively. These are read-write fields that control the impedance of the on-chip termination components in the DQ and DQN pins. The Impedance Calibration 0 Register controls the even-numbered DQ and DQN pins, and the Impedance Calibration 1 controls the odd-numbered DQ and DQN pins. Figure 36 through Figure 41 and Figure 43 shows the test registers. This includes the TEST, DLL, PLL0, PLL1, IFT, DA and PARTn registers. These are used during device testing. They are not to be read or written during normal operation. Figure42 shows the DLY register. This is used to set the value of tCAC and tCWD used by the component. See "Timing Parameters" on page59. Figure 17 : Serial Identification (SID) Register 7 6 reserved 5 4 3 2 SID[5:0] 1 0 Serial Identification Register SADR[7:0]: 000000012 Read-only register SID[7:0] resets to 000000002 SID[5:0] - Serial Identification field. This field contains the serial identification value for the device. The value is compared to the SID[5:0] field of a serial transaction to determine if the serial transaction is directed to this device. The serial identification value is set during the initialization sequence. Version 1.0 Jan. 2005 Page 31 K4Y5416(/08/04)4UF XDR DRAM Figure 18 : Configuration (CFG) Register 7 6 rsrv 5 4 3 rsrv SLE rsrv 2 1 0 WIDTH[2:0] Configuration Register SADR[7:0]: 000000102 Read/write register CFG[7:0] resets to 000001002 WIDTH[2:0] - Device interface width field. 0002 - Reserved. 0012 - Reserved 0102 - x4 device width 0112 - x8 device width 1002 - x16 device width 1012, 1102, 1112 - Reserved SLE - Serial Load enable field. 02 - WDSL-path-to-memory disabled 12 - WDSL-path-to-memory enabled Figure 19 : Power Management (PM) Register 7 6 5 4 PST[1:0] 3 2 1 0 PX reserved Power Management Register SADR[7:0]: 000000112 Read/write register PM[7:0] resets to 000000002 PX - Powerdown exit field.(write-one-only, read=zero) 02 - Powerdown entry - do not write zero - use PDN command 12 - Powerdown exit - write one to exit PST[1:0] - Power state field (read-only). 002 - Powerdown (with self-refresh) 012 - Active/active-idle 102 - reserved 112 - reserved Figure 20 : Write Data Serial Load (WDSL) Control Register 7 6 5 4 3 2 1 0 WDSD[7:0] Write Data Serial Load Control Register Read/write register SADR[7:0]: 000001002 WDSL[7:0] resets to 000000002 WDSD[7:0] - Writing to this register places eight bits of data into the serial-to-parallel conversion logic (the "Demux" block of Figure 2). Writing to this register "2x16" times accumulates a full "tCC" worth of write data. A subsequent WR command (with SLE=1 in CFG register in Figure 1) will write this data (rather than DQ data) to the sense amps of a memory bank. The shifting order of the write data is shown in Table 11. Figure 21 : RQ Scan High (RQH) Register 7 6 5 reserved 4 3 2 1 RQH[3:0] 0 RQ Scan High Register SADR[7:0]: 000001102 Read/write register RQH[7:0] resets to 000000002 RQH[3:0] - Latched value of RQ[11:8] in RQ wire test mode. Version 1.0 Jan. 2005 Page 32 K4Y5416(/08/04)4UF XDR DRAM Figure 22 : RQ Scan Low (RQL) Register 7 6 5 4 3 2 1 0 RQL[7:0] RQ Scan Low Register SADR[7:0]: 000001112 Read/write register RQL[7:0] resets to 000000002 RQL[7:0] - Latched value of RQ[7:0] in RQ wire test mode. Figure 23 : Refresh Bank (REFB) Control Register 7 6 5 MBR[1:0] 4 3 2 1 0 BANK[2:0] reserved Read/write register Refresh Bank Control Register REFB[7:0] resets to 000000002 SADR[7:0]: 000010002 BANK[2:0] - Refresh bank field. This field returns the bank address for the next self-refresh operation when in Powerdown power state. MBR[1:0] - Multi-bank and multi-row refresh control field. 002 - Single-bank refresh. 102 - Reserved 012 - Reserved 112 - Reserved Figure 24 : Refresh High (REFH) Row Register 7 6 5 4 3 2 1 0 R[18:16] reserved Refresh High Row Register SADR[7:0]: 000010012 Read/write register REFH[7:0] resets to 000000002 reserved - Refresh row field. This field contains the high-order bits of the row address that will be refreshed during the next refresh interval. This row address will be incremented after a REFI command for autorefresh, or when the BANK[2:0] field for the REFB register equals the maximum bank address for self-refresh. Figure 25 : Refresh Middle (REFM) Row Register 7 6 5 reserved 4 3 2 1 R[10:8] 0 Refresh Middle Row Register Read/write register SADR[7:0]: 000010102 REFM[7:0] resets to 000000002 R[10:8] - Refresh row field. This field contains the middle-order bits of the row address that will be refreshed during the next refresh interval. This row address will be incremented after a REFI command for autorefresh, or when the BANK[2:0] field for the REFB register equals the maximum bank address for self-refresh. Version 1.0 Jan. 2005 Page 33 K4Y5416(/08/04)4UF XDR DRAM Figure 26 : Refresh Low (REFL) Row Register 7 6 5 4 3 2 1 0 R[7:0] Refresh Low Row Register SADR[7:0]: 000010112 Read/write register REFL[7:0] resets to 000000002 R[7:0] - Refresh row field. This field contains the low-order bits of the row address that will be refreshed during the next refresh interval. This row address will be incremented after a REFI command for auto-refresh, or when the BANK[2:0] field for the REFB register equals the maximum bank address for self-refresh. Figure 27 : IO Configuration (IOCFG) Register 7 6 5 4 3 2 1 0 ODF[1:0] reserved IO Configuration Register SADR[7:0]: 000011112 Read/write register IOCFG[7:0] resets to 000000002 ODF[1:0] - Overdrive Function field. 00 - Nominal VOSW,DQ range 01 - reserved 10 - reserved 11 - reserved Figure 28 : Current Calibration 0 (CC0) Register 7 6 5 4 3 2 1 0 CCVALUE0[5:0] reserved Current Calibration 0 Register SADR[7:0]: 000100002 Read/write register CC0[7:0] resets to vvvvvvvv2 (vendor-dependent reset value) CCVALUE0[5:0] - Current calibration value field. This field controls the amount of current drive for the even-numbered DQ and DQN pins. Figure 29 : Current Calibration 1 (CC1) Register 7 6 5 4 3 2 1 0 CCVALUE1[5:0] reserved Current Calibration 1 Register SADR[7:0]: 000100012 Read/write register CC1[7:0] resets to vvvvvvvv2 (vendor-dependent reset value) CCVALUE1[5:0] - Current calibration value field. This field controls the amount of current drive for the odd-numbered DQ and DQN pins. Figure 30 : Impedance Calibration 0 (ZC0) Register 7 6 5 4 3 2 1 0 ZCVALUE0[5:0] reserved reserved Read/write register Impedance Calibration 0 Register ZC0[7:0] resets to 000000002 SADR[7:0]: 000100102 reserved Figure 31 : Impedance Calibration 1 (ZC1) Register 7 6 reserved 5 4 3 2 ZCVALUE1[5:0] reserved 1 0 Impedance Calibration 1 Register Read/write register SADR[7:0]: 000100112 ZC1[7:0] resets to 000000002 reserved Version 1.0 Jan. 2005 Page 34 K4Y5416(/08/04)4UF XDR DRAM Figure 32 : Current Fuse Setting 0 (FZC0) Register 7 6 5 4 3 2 1 0 FZCVALUE0[5:0] reserved reserved Current Fuse Setting Register SADR[7:0]: 000101002 Read-only register FZC0[7:0] resets to vvvvvvvv (vendor-dependent reset value) reserved Figure 33 : Current Fuse Setting 1 (FZC1) Register 7 6 5 4 3 2 1 0 FZCVALUE1[5:0] reserved reserved Current Fuse Setting Register SADR[7:0]: 000101012 Read-only register FZC1[7:0] resets to vvvvvvvv (vendor-dependent reset value) reserved Figure 34 : Read Only Memory 0 (ROM0) Register 7 6 5 4 3 VENDOR[3:0] reserved 2 1 0 MASK[3:0] Read Only Memory 0 Register Read-only register SADR[7:0]: 000101102 ROM0[7:0] resets to vvvvmmmm MASK[3:0] - Version number of mask (00012 is first version). VENDOR[3:0] - Vendor number for component: 0000 - reserved 0100-1111-reserved 0001 - Toshiba 0010 - Elpida 0011 - SEC Figure 35 : Read Only Memory 1 (ROM1) Register 7 6 5 BB[1:0] 4 3 2 RB[2:0] 1 0 CB[2:0] Read Only Memory 1 Register SADR[7:0]: 000101112 Read-only register ROM0[7:0] resets to bbrrrccc CB[2:0] - Column address bits: #bits = 6 +CB[2:0] RB[2:0] - Row address bits: #bits = 10 +RB[2:0] BB[2:0] - Bank address bits: #bits = 2 +BB[2:0] These three fields indicate how many column, row, and bank address bits are present. An offset of {6,10,2} is added to the field value to give the number of address bits. Figure 36 : TEST Register 7 6 5 4 WTL WTE 3 2 1 0 reserved TEST Register SADR[7:0]: 000110002 Read/write register TEST[7:0] resets to 000000002 WTE - Wire Test Enable WTL - Wire Test Latch Figure 37 : DLL Register 7 6 5 4 3 reserved 2 1 0 DLL Register SADR[7:0]: 000110012 Read/write register DLL[7:0] resets to 000000002 TBD Version 1.0 Jan. 2005 Page 35 K4Y5416(/08/04)4UF XDR DRAM Figure 38 : PLL0 Register 7 6 5 4 3 2 1 0 reserved PLL0 Register SADR[7:0]: 000110102 Read/write register PLL0[7:0] resets to 000000002 TBD Figure 39 : PLL1 Register 7 6 5 4 3 2 1 0 reserved PLL1 Register SADR[7:0]: 000110112 Read/write register PLL1[7:0] resets to 000000002 TBD Figure 40 : IFT Register 7 6 5 4 3 2 1 0 reserved IFT Register SADR[7:0]: 000111002 Read/write register IFT[7:0] resets to 000000002 TBD Figure 41 : DA Register 7 6 5 4 3 2 1 0 reserved DA Register SADR[7:0]: 000111012 Read/write register DA[7:0] resets to 000000002 TBD Figure 42 : Delay (DLY) Control Register 7 6 5 4 3 2 CWD[3:0] 1 0 CAC[3:0] DLY Register SADR[7:0]: 000111112 Read/write register DLY[7:0] resets to 001101102 CAC[3:0] - Programmed value of tCAC timing parameter: 01102 - tCAC = 6*tCYCLE 10002 - tCAC = 8*tCYCLE 01112 - tCAC = 7*tCYCLE others - Reserved. CWD[3:0] - Programmed value of tCWD timing parameter: 00112 - tCWD = 3*tCYCLE 01002 - tCWD = 4*tCYCLE others - Reserved. Figure 43 : Partner-Definable (PART0-PARTF) Registers 6 5 4 3 2 1 0 reserved 7 6 5 4 3 2 1 PART0 Register SADR[7:0]: 100000002 Read/write register PART0[7:0] resets to 000000002 PART1 Register SADR[7:0]: 100000012 Read/write register PART1[7:0] resets to 000000002 0 7 6 5 4 3 reserved 2 1 *** *** reserved 0 PARTF Register SADR[7:0]: 100011112 *** 7 Read/write register PARTF[7:0] resets to 000000002 Note - The partner-definable registers should not be written or read; doing so will produce undefined results. Version 1.0 Jan. 2005 Page 36 K4Y5416(/08/04)4UF XDR DRAM Maintenance Operations Refresh Transactions Figure44 contains two timing diagrams showing examples of refresh transactions. The top timing diagram shows a single refresh operation. Bank Ba is assumed to be closed (in a precharged state) when a REFA command is received in a ROWP packet on clock edge T0. The REFA command causes the row addressed by the REFr register (REFH/REFM/REFL) to be opened (sensed) and placed in the sense amp array for the bank. Note that the REFA and REFI commands are similar to the ACT command functionally; both specify a bank address and delay value, and both cause the selected bank to open (to become sensed.). The difference is that the ACT command is accompanied by a row address in the ROWA packet, while the REFA and REFI commands use a row address in the REFr register (REFH/REFM/REFL). After a time tRAS, a ROWP packet with REFP command to bank Ba is presented. This causes the bank to be closed (precharged), leaving the bank in the same state as when the refresh transaction began. Note that the REFP command is equivalent to the PRE command functionally; both specify a bank address and delay value, and both cause the selected bank to close (to become precharged). After a time tRP, another ROWP packet with REFA command to bank Bb is presented (banks Ba and Bb are the same in this example). This starts a second refresh cycle. Each refresh transaction requires a total time tRC = tRAS + tRP, but refresh transactions to different banks may be interleaved like normal read and write transactions. Each row of each bank must be refreshed once in every tREF interval. This is shown with the fourth ROWP packet with a REFA command in the top timing diagram. Interleaved Refresh Transaction The lower timing diagram in Figure44 represents one way a memory controller might handle refresh maintenance in a real system. A series of eight ROWP packets with REFA commands (except for the last which is a REFI command) are presented starting at edge T0. The packets are spaced with intervals of tRR. Each REFA or REFI command is addressed to a different bank (Ba through Bh) but uses the same row address from the REFr (REFH/REFM/REFL) register. The eighth REFI command uses this address and then increments it so the next set of eight REFA/REFI commands will refresh the next set of rows in each bank. A series of eight ROWP packets with REFP commands are presented effectively at edge T10 (a time tRAS after the first ROWP packet with a REFA command). The packets are spaced with intervals of tPP. Like the REFA/REFI commands, each REFP command is addressed to a different bank (Ba through Bh). This burst of eight refresh transactions fully utilizes the memory component. However, other read and write transactions may be interleaved with the refresh transactions before and after the burst to prevent any loss of bus efficiency. In other words, a ROWA packet with ACT command for a read or write could have been presented at edge T4 (a time tRR before the first refresh transaction starts at edge T0). Also, a ROWA packet with ACT command for a read or write could have been presented at edge T36 (a time tRR after the last refresh transaction starts at edge T32). In both cases, the other request packets for the interleaved read or write accesses (the precharge commands and the read or write commands) could be slotted in among the request packets for the refresh transaction. Version 1.0 Jan. 2005 Page 37 K4Y5416(/08/04)4UF XDR DRAM Figure 44 : Refresh Transactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN tRAS REFA RQ11..0 tRP REFP a0 a1 tRC DQ15..0 DQN15..0 REFA REFA b0 c0 tCYCLE tREF Bb = Ba Bc/Rc = Ba/Ra T0 T1 Transaction a: REF Transaction b: REF Transaction c: REF T2 T3 T4 T5 a0 = {Ba,REFR} a0 = {Ba,REFR} c0 = {Bc,REFR} T6 T7 T8 T9 a1 = {Ba} b1 = {Bb} c1 = {Bc} Refresh Transaction T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 (ACT) REFA tRR a0 REFA REFA REFA REFA REFA b0 c0 d0 e0 f0 RQ11..0 (PRE) RQ11..0 (ALL) REFP REFP REFP REFP a1 b1 c1 d1 REFA REFA REFA REFP REFA REFP REFA REFP REFA REFP a0 b0 c0 a1 d0 b1 e0 c1 f0 d1 DQ15..0 DQN15..0 T24 T25 T26 T27 T28 T29 T30 T31 T32 T33 T34 T35 T36 T37 T38 T39 T40 T41 T42 T43 T44 T45 T46 T47 CFM CFMN This REFI increments REFR RQ11..0 (ACT) REFA REFI REFA g0 h0 i0 RQ11..0 (PRE) RQ11..0 (ALL) tCYCLE REFP REFP REFP REFP e1 f1 g1 h1 REFA REFP REFA REFP REFA REFP REFP g0 e1 h0 f1 i0 g1 h1 DQ15..0 DQN15..0 Ba,Bb,Bc,Bd, Be,Bf,Bg and Bh are different banks. Bi = Ba Transaction a: REF Transaction b: REF Transaction c: REF Transaction d: REF Transaction e: REF Transaction f: REF Transaction g: REF Transaction h: REF Transaction i: REF a0 = {Ba,REFR} b0 = {Bb,REFR} c0 = {Bc,REFR} d0 = {Bd,REFR} e0 = {Be,REFR} f0 = {Bc,REFR} g0 = {Bd,REFR} h0 = {Be,REFR} i0 = {Ba,REFR+1} a1 = {Ba} b1 = {Bb} c1 = {Bc} d1 = {Bd} e1 = {Be} f1 = {Bf} g1 = {Bg} h1 = {Bh} i1 = {Bi} Interleaved Refresh Example Version 1.0 Jan. 2005 Page 38 K4Y5416(/08/04)4UF XDR DRAM Calibration Transactions Figure45 shows the calibration transaction diagrams for the XDR DRAM device. There is one calibration operation supported: calibration of the output current level IOL, each DQi and DQNi pin. The output current calibration sequence is shown in the upper diagram. It begins when a period of tCMD-CALC is observed after the last RQ packet (with command "CMD a" in this example). No request packets should be issued in this period. A COLX packet with a "CALC b" command is then issued to start the current calibration sequence. A period of tCALCE is observed after this packet. No request packets should be issued during this period. A COLX packet with a "CALE c" command is then issued to end the current calibration sequence. A period of tCALE-CMD is observed after this packet. No request packets should be issued during this period. The first request packet may then be issued (with command "CMD d" in this example). A second current calibration sequence must be started within an interval of tCALC. In this example. the next COLX packet with a "CALC e" command starts a subsequent sequence. The dynamic termination calibration sequence is shown in the lower diagram. Note that this memory component does not use this sequence; termination calibraion is performed during the manufacturing process. However, the termination sequence shown will be issued by the controller for those memory component which do use a periodic calibration mechanism. It begins when a period of tCMD-CALZC is observed after the packet edge T0(with command CMDa in this example). No request packets should be issued in this period. A COLX packet with a CALZ command is then issued at edge T3 to start the current calibration sequence. A second period of tCALZE is ovserved after this packet. No request packets should be issued during this period. A COLX packet with a CALE command is then issued at dege T6 to end the current calibration sequence. A third period of tCALEis observed after this pakets. No request packets should be issued during this period. The first request pakcet may be issued at edge T12(with command CMDd in this example). CMD A second current calibration sequence must be started within an interval of tCALZ. In this example, the next COLX pakcet with a CALZ command occurs at edge T20. Note that the labels for the CFM clock edges(of the form Ti) are not to scale, and are used to identify events in the diagrams. Figure 45 : Calibration Transactions T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 CMD CALC a tCALCE, tCALE-CMD, CALE c b tCMD-CALC DQ15..0 DQN15..0 CMD CALC d e tCALC Packet a: Any CMD Packet b: CALC Packet c: CALE Packet d: Any CMD Packet e: CALC T0 T1 T2 tCYCLE T3 T4 T5 Current Calibration Transaction T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFM CFMN CFMN RQ11..0 RQ11..0 DQ15..0 DQN15..0 CMD CALZ a b tCALZE, CALE tCALE-CMD, c tCMD-CALZ CMD CALZ d e tCYCLE tCALZ Packet a: Any CMD Packet b: CALZ Packet c: CALE Packet d: Any CMD Packet e: CALZ Termination Calibration Transaction Version 1.0 Jan. 2005 Page 39 K4Y5416(/08/04)4UF XDR DRAM Power State Management Figure46 shows power state transition diagrams for the XDR DRAM device. There are two power states in the XDR DRAM: Powerdown and Active. Powerdown state is to be used in applications in which it is necessary to shut down the CFM/CFMN clock signals. In this state, the contents of the storage cells of the XDR DRAM will be retained by an internal state machine which performs periodic refresh operations using the REFB and REFr control registers. The upper diagram shows the sequence needed for Powerdown entry. Prior to starting the sequence, all banks of XDR DRAM must be precharged so they are left in a closed state. Also, all 23 banks must be refreshed using the current value of the REFr registers, and the REFr registers must not be incremented with the REFI command at the end of this special set of refresh transactions. This ensures that no matter what value has been left in the REFB register, no row of any bank will be skipped when automatic refresh is first started in Powerdown. There may be some banks at the current row value in the REFr registers that are refreshed twice during the Powerdown entry process. After the last request packet (with the command CMDa in the upper diagram of the figure), an interval of tCMD-PDN is observed. No request packets should be issued during this period. A COLX packet with the PDN command is issued after this interval, causing the XDR DRAM to enter Powerdown state after an interval of tPDN-ENTRY has elapsed (this is the parameter that should be used for calculating the power dissipation of the XDR DRAM). The CFM/CFMN clock signals may be removed a time tPDN-CFM after the COLX packet with the PDN command. Also, the termination voltage supply may be removed (set to the ground reference) from the Vterm pins a time tPDN-CFM after the COLX pakcet with the PDN command. The voltage on the DQ/DQN pins will follow the voltae on the Vterm pins during Powerdonwn entry. When the XDR DRAM is in Powerdown, an internal frequency source and state machine will automatically generate internal refresh transactions. It will cycle through all 23 state combinations of the REFB register. When the largest value is reached and the REFB value wraps around, the REFr register is incremented to the next value. The REFB and REFr values select which bank and which row are refreshed during the next automatic refresh transaction. The lower diagram shows the sequence needed for Powerdown exit. The sequence is started with a serial broadcast write (SBW command) transaction using the serial bus of the XDR DRAM. This transaction writes the value "00000001" to the Power Management (PM) register (SADR = "00000011") of all XDR DRAMs connected to the serial bus. This sets the PX bit of the PM register, causing the XDR DRAMs to return to Active power state. The CFM/CFMN clock signals must be stable a time tCFM-PDN before the end of the SBW transaction. Also, the termination voltage supply must be restored to its normal operating point (VTERM,DRSL) on the Vterm pins a time tCFM-PDN before the end of the SWB transaction. The voltage on the DQ/DQN pins will follow the voltage on the Vterm pins during Powerdown exit. The XDR DRAM will enter Active state after an interval of tPDN-EXIT has elapsed from the end of the SBW transaction (this is the parameter that should be used for calculating the power dissipation of the XDR DRAM). The first request packet may be issued after an interval of tPDN-CMD has elapsed from the end of the SBW transaction, and must contain a "REFA" command in a ROWP packet. In this example, this packet is denoted with the command "REFA 1". No other request packets should be issued during this tPDN-CMD interval. All "n" banks (in the example, n=23) must be refreshed using the current value of the REFr registers. The "nth" refresh transaction will use a "REFI" command to inrement the REFr register (instead of a "REFR" command). This ensures that no matter what value has been left in the REFB register, no row of any bank will be skipped when normal refresh is restarted in Active state. There may be some banks at the current row value in the REFr registers that are refreshed twice during the Powerdown exit process. Note that during the Powerdown state an internal time source keeps the device refreshed. However, during the tPDN-CMD interval, no internal refresh operations are performed. As a result, an additional burst of refresh transactions must be issued after the burst of "n" transactions described above. This second burst consists of "m" refresh transactions: m = ceiling[23*211*tPDN-CMD/tREF] Where "211" is the number of rows per bank, and "23" is the number of banks. Every "nth" refresh transaction (where n=23) will use a "REFI" command (to increment the REFr register) instead of a "REFA" command. Version 1.0 Jan. 2005 Page 40 K4Y5416(/08/04)4UF XDR DRAM Figure 46 : Power State Management CFM CFMN No signal CMD RQ11..0 tCYCLE tPDN-CFM PDN ba a Powerdown State... DQ15..0 DQN15..0 tPDN-ENTRY tCMD-PDN Transaction a: Last precharge command Transaction b: PDN S0 S2 S4 S6 S8 Powerdown Entry S10 S12 S14 S16 S18 S20 S22 S24 S26 S28 S30 S32 S34 SCK RST CMD tCYC,SCK Power-up transaction Start 2'h0,SID[5:0] SCMD `1' `1' `0' `0' `0' `0' `0' `0' 5 4 3 2 SADR[7:0] 1 0 7 6 5 4 3 2 SWD[7:0] 1 0 `0' 7 6 5 4 3 2 1 0 `0' SDI (input) SDO (output) CFM CFMN No signal tCFM-PDN RQ11..0 tCYCLE tPDN-EXIT ....Powerdown State DQ15..0 DQN15..0 tPDN-CMD CFM CFMN RQ11..0 DQ15..0 DQN15..0 REFA REFA 1 2 REFI REFP n n-2 REFP REFP n-1 n tCYCLE tPDN-CMD Transaction 1: REFA Transaction 2: REFA The final REFA/REFI command increments the REFr register Transaction n-1: REFA Transaction n: REFI Powerdown Exit Version 1.0 Jan. 2005 Page 41 K4Y5416(/08/04)4UF XDR DRAM Initialization Figure47 shows the topology of the serial interface signals of a XDR DRAM system. The three signals RST, CMD, and SCK are transmitted by the controller and are received by each XDR DRAM device along the bus. The signals are terminated to the VTERM supply through termination components at the end farthest from the controller. The SDI input of the XDR DRAM device furthest from the controller is also terminated to VTERM. The SDO output of each XDR DRAM device is transmitted to the SDI input of the next XDR DRAM device (in the direction of the controller). This SDO/SDI daisy chain topology continues to the controller, where it ends at the SRD input of the controller. All the serial interface signals are low-true. All the signals use RSL signaling circuits, except for the SDO output which uses CMOS signaling circuits. Figure 47 : Serial Interface Systems Topology RST CMD SCK RST CMD SCK SRD SDO Controller SDI VTERM RST CMD SCK ... SDO XDR DRAM SDI RST CMD SCK ... SDO XDR DRAM SDI XDR DRAM Figure48 shows the initialization timing of the serial interface for the XDR DRAM [k] device in the system shown above. Prior to initialization, the RST is held at zero. The CMD input is not used here, and should also be held at zero. Note that the inputs are all sampled by the negative edge of the SCK clock input. The SDI input for the XDR DRAM[0] device is zero, and is unknown for the remaining devices. On negative SCK edge S8 the RST input is sampled one. It is sampled one on the next four edges, and is sampled zero on edge S12 a time tRST-10 after it was first sampled one. The state of the control registers in the XDR DRAM device are set to their reset values after the first edge (S8) in which RST is sampled one. Figure 48 : Initialization Timing for XDR DRAM [k] Device S0 S2 S4 S6 S8 S10 S12 S14 S16 S18 S20 S22 S24 S26 S28 S30 S32 S34 S36 S38 S40 S42 S44 S46 S48 SCK tCYC,SCK tRST-10 RST `0' `0' `0' `0' `1' `1' `1' `1' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' tRST-SDI,00 = k * tCYC,SCK CMD SDI (input) `x' `x' `x' `x' `x' `1' `1' `1' `1' `1' `1' `1' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' SDO (output) `x' `x' `x' `x' `x' `1' `1' `1' `1' `1' `1' `1' `1' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' `0' tRST-SDO,11 tSDI-SDO,00 The SDI inputs will be sampled one within a time tRST-SDO,11 after RST is first sampled one in all the XDR DRAMs except for XDR DRAM [0]. XDR DRAM [0]'s SDI input will always be sampled zero. XDR DRAM [k] will see its RST input sampled zero at S12, and will then see its SDI input sampled zero at S16 (after SDI had previously been sampled one). This interval (measured in tCYC,SCK units) will be equal to the index [k] of the XDR DRAM device along the serial interface bus. In this example, k is equal to 4. This is because each XDR DRAM device will drive its SDO output zero around the SCK edge a time tSDI-SDO,00 after its SDI input is sample zero. In other words, the XDR DRAM [0] device will see RST and SDI both sampled zero on the same edge S12 (tRST-SDI,00 will be 0 *tCYC,SCK units), and will drive its SDO to zero around the subsequent edge (S13). Version 1.0 Jan. 2005 Page 42 K4Y5416(/08/04)4UF XDR DRAM The XDR DRAM [1] device will see SDI sampled zero on edge S13 (tRST-SDI,00 will be 1*tCYC,SCK units), and will drive its SDO to zero around the subsequent edge (S14). The XDR DRAM [2] device will see SDI sampled zero on edge S14 (tRST-SDI,00 will be 2* tCYC,SCK units), and will drive its SDO to zero around the subsequent edge (S15). This continues until the last XDR DRAM device drives the SRD input of the controller. Each XDR DRAM device contains a state machine which measures the interval tRST-SDI,00 between the edges in which RST and SDI are both sampled zero, and uses this value to set the SID [5:0] field of the SID (Serial Identification) register. This value allows directed read and write transactions to be made to the individual XDR DRAM devices. Table 10 summarizes the range of the timing parameters used for initialization by the serial interface bus. Table 10 : Initialization Timing Parameters Symbol Parameter Min Max Unit Figure (s) tRST,10 Number of cycles between RST being sampled one and RST being sampled zero 2 - tCYC,SCK - tRST-SDO,11 Number of cycles between RST being sampled one and SDO being driven to one 1 1 tCYC,SCK - tRST,SDI,00 Number of cycles between RST being sampled zero (after being sampled one for tRST,10,MIN or more cycles) and SDI being sampled zero. This will be equal to the index [k] of the XDR DRAM device along the serial interface bus 0 63 tCYC,SCK - tSDI-SDO,00 Number of cycles between SDI being sampled one (after RST has been sampled one for tRST,10,MIN or more cycles and is then sampled zero) and SDO being driven to zero 1 1 tCYC,SCK - Asynchronous reset interval. 20 - tCYC,SCK - tRST-SCK XDR DRAM Initialization Overview [1] Apply voltage to VDD, VTERM, and VREF pins. VTERM and VREF voltages must be less or equal to VDD voltage at all times. Wait a time interval tCOREINIT. Power-on reset circuit in XDR DRAM places XDR DRAM into low-power state. [2] Assert RST, SCK, SDI and CMD to logical zero, Then: - Pulse SCK to logical one, then to logical zero four times. - Assert RST to logical one. Reset circuit places XDR DRAM into low-power state(identical to power-on reset) - Perform remaining initialization sequence in Figure 48. [3] XDR DRAM has valid Serial ID and all registers have default values that are defined in Figure17 through Figure42. [4] Perform broadcast or directed register writes to adjust registers which need a value different from their default value. [5] Perform Powerdown Exit sequence shown in Figure46. This includes the activity from SCK cycle S0 through the final REFP command. [6] Perform termination/current calibration. The CALZ /CALE sequence shown in Figure 45 is issued 128 times. After this, each sequence is issued once every tCALZ or tCALC interval. [7] Condition the XDR DRAM banks by performing a REFA/REFI activate and REFP precharge operation to each bank eight times. This can be interleaved to save time. The row address for the activate operation will step through eight successive values of the REFr registers. The sequence between cycles T0 and T32 in the Interleaved Refresh Example in Figure 44 could be performed eight times to satisfy this conditioning requirement. XDR DRAM Pattern Load with WDSL Register The XDR memory system requires a method of deterministically loading pattern data to XDR DRAMs before beginning Receive Timing Calibration (RX TCAL). The method employed by the XDR DRAMs to achieve this is called Write Data Serial Load (WDSL). A WDSL packet sends one-byte of serial data which is serially shifted into a holding register within the XDR DRAM. Initialization software sends a sequence of WDSL packets, each of which shifts the new byte in and advances the shifter by 8 positions. In this way, XDR DRAMs of varying widths can be loaded with a single command type. Each sequence of WDSL packets will load one full column of data to the internal holding register of the target XDR DRAM. Depending upon the ratio of native device width to programmed width, there may be more than one sub-column per column. After loading a full column, a series of WR commands will be issued to sequentially transfer each sub-column of the column to the XDR DRAM core(s), based upon the SC [3:0] bits. Version 1.0 Jan. 2005 Page 43 K4Y5416(/08/04)4UF XDR DRAM Table 11 : WDSL-to-Core Mapping (First Generation x16/x8/x4 XDR DRAM, BL = 16) WDSL Core Word Load Order DQ Pins Used x16 x8 x4 Core Word x4 x8 x16 WD[n][15:0] DQ0 DQ0 DQ0 WD[0][15:0] DQ1 DQ1 DQ1 DQ2 DQ2 DQ2 DQ3 DQ3 DQ3 SC[3:2] SC[3:2] SC[3:2] SC[3:2] SC[3:2] SC[3:2] SC[3:2] =xx = 0x = 1x = 00 = 01 = 10 = 11 LOGICAL VIEW OF XDR DRAM Word Written (1 = Written, 0 = Not Written) WDSL Word 8 1 1 0 1 0 0 0 WD[1][15:0] WDSL Word 7 1 1 0 1 0 0 0 WD[2][15:0] WDSL Word 12 1 1 0 1 0 0 0 WD[3][15:0] WDSL Word 3 1 1 0 1 0 0 0 DQ0 DQ4 DQ4 WD[4][15:0] WDSL Word 10 1 1 0 0 1 0 0 DQ1 DQ5 DQ5 WD[5][15:0] WDSL Word 5 1 1 0 0 1 0 0 DQ2 DQ6 DQ6 WD[6][15:0] WDSL Word 14 1 1 0 0 1 0 0 DQ3 DQ7 DQ7 WD[7][15:0] WDSL Word 1 1 1 0 0 1 0 0 DQ0 DQ0 DQ8 WD[8][15:0] WDSL Word 9 1 0 1 0 0 1 0 DQ1 DQ1 DQ9 WD[9][15:0] WDSL Word 6 1 0 1 0 0 1 0 DQ2 DQ2 DQ10 WD[10][15:0] WDSL Word 13 1 0 1 0 0 1 0 DQ3 DQ3 DQ11 WD[11][15:0] WDSL Word 2 1 0 1 0 0 1 0 DQ0 DQ4 DQ12 WD[12][15:0] WDSL Word 11 1 0 1 0 0 0 1 DQ1 DQ5 DQ13 WD[13][15:0] WDSL Word 4 1 0 1 0 0 0 1 DQ2 DQ6 DQ14 WD[14][15:0] WDSL Word 15 1 0 1 0 0 0 1 DQ3 DQ7 DQ15 WD[15][15:0] WDSL Word 0 1 0 1 0 0 0 1 PHYSICAL VIEW OF XDR DRAM DQ6 DQ2 DQ2 DQ4 DQ0 DQ0 DQ1 DQ1 DQ5 DQ3 DQ3 DQ7 Word Written (1 = Written, 0 = Not Written) DQ14 WD[14][15:0] WDSL Word 15 1 0 1 0 0 0 1 DQ6 WDSL Word 14 1 1 0 0 1 0 0 DQ10 WD[10][15:0] WDSL Word 13 1 0 1 0 0 1 0 DQ2 WD[6][15:0] WDSL Word 12 1 1 0 1 0 0 0 DQ12 WD[12][15:0] WDSL Word 11 1 0 1 0 0 0 1 WD[2][15:0] DQ4 WD[4][15:0] WDSL Word 10 1 1 0 0 1 0 0 DQ8 WD[8][15:0] WDSL Word 9 1 0 1 0 0 1 0 DQ0 WD[0][15:0] WDSL Word 8 1 1 0 1 0 0 0 DQ1 WD[1][15:0] WDSL Word 7 1 1 0 1 0 0 0 DQ9 WD[9][15:0] WDSL Word 6 1 0 1 0 0 1 0 DQ5 WD[5][15:0] WDSL Word 5 1 1 0 0 1 0 0 DQ13 WD[13][15:0] WDSL Word 4 1 0 1 0 0 0 1 DQ3 WDSL Word 3 1 1 0 1 0 0 0 0 WD[3][15:0] DQ11 WD[11][15:0] WDSL Word 2 1 0 1 0 0 1 DQ7 WDSL Word 1 1 1 0 0 1 0 0 WDSL Word 0 1 0 1 0 0 0 1 WD[7][15:0] DQ15 WD[15][15:0] Version 1.0 Jan. 2005 Page 44 K4Y5416(/08/04)4UF XDR DRAM Table 12 : Core Data Word-to-WDSL Formata DQ Serialization Order CFM/PCLK Cycle Cycle 0 Cycle 1 Symbol (Bit) Time t0 t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 t11 Bit Transmitted on DQ pins t12 t13 t14 t15 D0 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12 D13 D14 D15 WDSL Byte/Bit Transfer Order Core Word Core Word WD[n][15:0] WDSL Byte Order SWD Field of Serial Packet WDSL Byte 0 7 Bit Transmitted on CMD D15 pin 6 5 4 D11 D7 D3 3 WDSL Byte 1 2 D14 D10 1 0 7 6 5 4 3 2 1 0 D6 D2 D13 D9 D5 D1 D12 D8 D4 D0 a. Applies for first generation x16/x8/x4 XDR DRAM with BL=16 Version 1.0 Jan. 2005 Page 45 K4Y5416(/08/04)4UF XDR DRAM Special Feature Description Dynamic Width Control This XDR DRAM device includes a feature called dynamic width control. This permits the device to be configured so that read and write data can be accessed through differing widths of DQ pins. Figure 49 shows a diagram of the logic in the path of the read data (Q) and write data (D) that accomplishes this. The read path is on the right of the figure. There are 16 sets of S signals (the internal data bus connecting to the sense amps of the memory core), with 16 signals in each set. When the XDR DRAM device is configured for maximum width operation (using the WIDTH [2:0] field in the CFG register), each set of 16 S signals goes to one of the 16 DQ pins (via the Q[15:0][15:0] read bus) and are driven out in the 16 time slots for a read data packet. When the XDR DRAM device is configured for a width that is less than the maximum, some of the DQ pins are used and the rest are not used. The SC [3:0] field of the COL request packets which S[15:0] [15:0] signals are passed to the Q[15:0] [15:0] read bus and driven as read data. Figure 50 shows the mapping from the S bus to the Q bus as a function of the WIDTH [2:0] register field and the SC[3:0] field of the COL request packet. There is a separate table for each valid value of WIDTH [2:0]. In each table, there is an entry in the left column for each valid value of SC[3:0]. This field should be treated as an extension of the C[9:4] column address field. The right hand column shows which sets of S[15:0] [15:0] signals are mapped to the Q read data bus for a particular value of SC[3:0]. For example, assume that the WIDTH [2:0] value is "010", indicating a device width of x4. Looking at the appropriate table in Figure50, it may be seen that in the SC [3:0] field, the SC [1:0] sub-column address bits are not used. The remaining SC [3:0] address bit(s) selects one of the 64-bit blocks of S bus signals, causing them to be driven onto the Q [3:0] [15:0] read data bus, which in turn is driven to the DQ3...0/DQN3...0 data pins. The Q[15:4] [15:0] signals and DQ15...4/DQN15....4 data pins are not used for a device width of x4. The write path is shown on the left side of Figure 49. As shown, there are 16 sets of S signals(the internal data bus connecting to the sense amps of the memory core), with 16 signals in each set. When the XDR DRAM device is configured for maximum width operation (using the WIDTH [2:0] field in the CFG register), each set of 16 S signals is driven from one fo the 16 DQ pins (via the D[15:0] [15:0] write bus) from each of the 16 time slots for a write data packet. Figure 50 also shows the mapping from the D bus to the S bus as a function of the WIDTH[2:0] register field and the SC[3:0] field of the COL request packet. There is a separate table for each valid value of WIDTH[2:0]. In each table, there is an entry in the left column for each valid value of SC[3:0]. This field should be treated as an extension of the C[9:4] column address field. The right hand column shows which set of S[15:0] [15:0] signals are mapped from the D read data bus for a particular value of SC[3:0]. For example, assume that the WIDTH[2:0] value is "001", indicating a device width of x2. Looking at the appropriate table in Figure50, it may be seen that in the SC[3:0] field, the SC[0] sub-column address bit is not used. The remaining SC [3:0] address bit(s) selects one of the 32-bit blocks of S bus signals, causing them to be driven from the D [1:0] [15:0] write data bus, which in turn is driven from the DQ1... 0/DQN1... 0 data pins. The D[15:2] [15:2] signals and DQ15... 2/DQN15...2 data pins are not used for a device width of x2. Figure 49 : Multiplexes for Dynamic Width Control S[15:0][15:0] 16x16 16x16 8 M[7:0] 4+3 WIDTH[2:0] SC[3:0] Byte Mask (WR) 16x16 D1[15:0][15:0] Dynamic Width Demux (WR) 16x16 4+3 Dynamic Width Mux (RD) WIDTH[2:0] SC[3:0] 16x16 D[15:0][15:0] Q[15:0][15:0] Version 1.0 Jan. 2005 Page 46 K4Y5416(/08/04)4UF XDR DRAM The block diagram in Figure 49 indicates that the Dynamic Width logic is positioned after the serial-to-parallel conversion (demux block) in the data receiver block and before the parallel-to-serial conversion (mux block) in the data transmitter block (see also the block diagram in Figure2). The block diagram is shown in this manner so the functionality of the logic can be made as clear as possible. Some implementations may place this logic in the data receiver and transmitter blocks, performing the mapping in Figure 50 on the serial data rather than the parallel data. However, this design choice will not affect the functionality of the Dynamic Width logic; it is strictly an implementation decision. Figure 50 : D-toS and S-to-D Mapping for Dynamic Width Control WIDTH[2:0]=000 (x1 device width) WIDTH[2:0]=001 (x2 device width) WIDTH[2:0]=011 (x8 device width) 0000 S[0][15:0] 000x S[1:0][15:0] 0xxx S[7:0][15:0] 0001 S[1][15:0] 001x S[3:2][15:0] 1xxx S[15:8][15:0] 0010 S[2][15:0] 010x S[5:4][15:0] SC[3:0] D[7:0][15:0] Q[7:0][15:0] 0011 S[3][15:0] 011x S[7:6][15:0] 0100 S[4][15:0] 100x S[9:8][15:0] 0101 S[5][15:0] 101x S[11:10][15:0] 0110 S[6][15:0] 110x S[13:12][15:0] 0111 S[7][15:0] 111x S[15:14][15:0] 1000 S[8][15:0] SC[3:0] D[1:0][15:0] Q[1:0][15:0] 1001 S[9][15:0] 1010 S[10][15:0] 1011 S[11][15:0] 1100 S[12][15:0] 00xx S[3:0][15:0] 1101 S[13][15:0] 01xx S[7:4][15:0] 1110 S[14][15:0] 10xx S[11:8][15:0] 1111 S[15][15:0] 11xx S[15:12][15:0] SC[3:0] D[0][15:0] Q[0][15:0] SC[3:0] D[3:0][15:0] Q[3:0][15:0] WIDTH[2:0]=010 (x4 device width) WIDTH[2:0]=100 (x16 device width) xxxx S[15:0][15:0] SC[3:0] D[15:0][15:0] Q[15:0][15:0] Version 1.0 Jan. 2005 Page 47 K4Y5416(/08/04)4UF XDR DRAM Write Masking Figure 51 shows the logic used by the XDR DRAM device when a write-masked command (WRM) is specified in a COLM packet. This masking logic permits individual byte of a write data packet to be written or not written according to the value of an eight bit write mask M [7:0]. In Figure 51, there are 16 sets of 16 bit signals forming the D1[15:0] [15:0] input bus for the Byte Mask block. These are treated as 2 x 16 8-bit bytes: D1 [15] [15.8] D1 [15] [7:0] ... D1 [1] [15:8] D1 [1] [7:0] D1 [0] [15:8] D1 [0] [7:0] The eight bits of each byte is compared to the value in the byte mask field (M[7:0]). If they are not equal (NE), then the corresponding write enable signal (WE) is asserted and the byte is written into the sense amplifier. If they are equal, then corresponding write enable signal (WE) is deasserted and the byte is not written into the sense amplifier. In the example of Figure 51, a WRM command performs a masked write of a 64 byte data packet to all the memory devices connected to the RQ bus (and receiving the command). It is the job of the memory controller to search the 64 bytes to find an eight bit data value that is not used and place it into the M [7:0] field. This will always be possible because there are 256 possible 8-bit values and there are only 64 possible values used in the bytes in the data packet. Figure 51 : Byte Mask Logic S[15][15:8] S[15][7:0] 8 WE-MSB [15] 1 NE 8 Compare 8 8 8 8 Compare 8 8 8 D1[15][15:8] S[0][7:0] 8 WE-MSB [0] 1 NE 8 WE-LSB [0] 1 NE 8 Compare 8 8 8 Compare 8 8 D1[15][7:0] D1[15][15:8] M[7:0] S[0][15:8] WE-LSB [15] 1 NE D1[0][15:8] 8 8 D1[15][7:0] D1[0][7:0] 8 D1[0][15:8] D1[0][7:0] S[15:0][15:0] 16x16 16x16 8 M[7:0] 4+3 WIDTH[2:0] SC[3:0] Byte Mask (WR) 16x16 D1[15:0][15:0] Dynamic Width Demux (WR) 16x16 4+3 Dynamic Width Mux (RD) WIDTH[2:0] SC[3:0] 16x16 D[15:0][15:0] Q[15:0][15:0] Version 1.0 Jan. 2005 Page 48 K4Y5416(/08/04)4UF XDR DRAM Note that other systems might use a data transfer size that is different than the 64 bytes per tCC interval per RQ bus that is used in the example in Figure 51. Figure 52 shows the timing of two successive WRM commands in COLM packets. The timing is identical to that of two successive WR commands in COL packets. The one difference is that the COLM packet includes a M[7:0] field that indicates the reserved bit pattern (for the eight bits of each byte) that indicates that the byte is not to be written. This requires that the alignment of bytes within the data packet be defined, and also that the bit numbering within each byte be defined (note that this was not necessary for the unmasked WR command). In the figure, bytes are contained within a single DQ/DQN pin pair - this is necessary so the dynamic width feature can be supported. Thus, each pin pair carries two bytes of each data packet. Byte[0] is transferred earlier than byte[1], and bit[0] of each byte (corresponding to M [0]) is transferred first, followed by the remaining bits in succession). Figure 52 : Write-Masked (WRM) Transaction Example T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WRM a1 DQ15..0 DQN15..0 WRM a2 tCYCLE RD a1 tCC tCWD D(a1) D(a2) tCAC Q(a1) Bit- and Byte-numbering convention for write and read data packets. Byte [16+0] Byte [0] DQ0 DQN0 [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] Byte [1] [1] [2] [3] [4] [5] [6] [7] [8] [9] Byte [15] DQ15 DQN15 [10] [11] [12] [13] [14] [15] ... [0] Byte [16+1] ... DQ1 DQN1 [10] [11] [12] [13] [14] [15] [0] [1] [2] [3] [4] Byte [16+15] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] Version 1.0 Jan. 2005 Page 49 K4Y5416(/08/04)4UF XDR DRAM Multiple Bank sets and the ERAW Feature Figure 55 shows a block diagram of a XDR DRAM in which the banks are divided into two sets (called the even bank set and the odd bank set) according to the least-significant bit of the bank address field. This XDR DRAM supports a feature called "Early Read After Write" (hereafter called "ERAW") The logic that accepts commands on the RQ11...0 signals is capable of operating these two bank sets independently. In addition, each bank set connects to its own internal "S" data bus (called S0 and S1). The receive interface is able to drive write data onto either of these internal data buses, and the transmit interface is able to sample read data from either of these internal data buses. These capabilities will permit the delay between a write column operation and a read column operation to be reduced, thereby improving performance. Figure 53 shows the timing previously presented in Figure12, but with the activity on the internal S data bus included. The writeto-read parameter tWR ensures that there is adequate turnaround time on the S bus between D (a2) and Q (c1). When ERAW is supported with odd and even bank sets, the tWR,MIN parameter must be obeyed when the write and read column operations are to the same bank set, but a second parameter tWR-D permits earlier column operations to the opposite bank set. Figure54 shows how this is possible because there are two internal data buses S0 and S1. In this example, the four columns read operations are made to the same bank Bb, but they could use different banks as long as they all belonged to the bank set that was different form the bank set containing Ba (for the column write operations). Figure 53 : Write/Read Interaction - No ERAW Feature T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WR a1 WR a2 RD c1 tWR DQ15..0 DQN15..0 D(a1) D(a2) tCWD RD c2 tCYCLE tCAC Q(c1) tWR-BUB,XDRDRAM Q(c2) tCC turnaround tCC S[15:0] [15:0] D(a1) D(a2) Q(c1) a1 = {Ba,Ca1} c1 = {Bc,Cc1} Transaction a: WR Transaction c: RD Q(c2) a2 = {Ba,Ca2} c2 = {Bc,Cc2} Figure 54 : Write/Read Interaction - ERAW Feature T0 T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 WR a1 RD b1 WR a2 tWR-D DQ15..0 DQN15..0 D(a1) tCWD tCAC RD b2 RD b3 D(a2) Q(b2) D(a1) Q(b1) Transaction a: WR Transaction b: RD Transaction c: RD Q(b3) Q(b4) Q(c1) tCC turnaround tCC tWR-BUB,XDRDRAM S1[15:0] [15:0] tCYCLE RD c1 Q(b1) S0[15:0] [15:0] Bank Restrictions Bb is in different bank set than Ba Bc is in same bank set as Ba RD b4 D(a2) Q(b2) a1 = {Ba,Ca1} b1 = {Bb,Cb1} c1 = {Bc,Cc1} Q(c1) Q(b3) Q(b4) a2 = {Ba,Ca2} b2 = {Bb,Cb2} b3 = {Bb,Cb3} b4 = {Bb,Cb4} Version 1.0 Jan. 2005 Page 50 K4Y5416(/08/04)4UF XDR DRAM Figure 55 : XDR DRAM Block Diagram with Bank Sets RQ11..0 12 1:2 Demux Reg COL decode 6 3 ... 1 1 ... COL logic 16x16 COL ... WR even 16x16*26 COL 16x16 ... WR odd Sense Amp Array Sense Amp 0 3 Sense Amp (2 -2) ... ... 16x16*26 R/W 6 6 ... ... COL 3 Bank (2 -2) R/W 1 6 6 Bank 0 PRE ... 16x16 S0[15:0][15:0] RD odd RD even 16x16 16x16 Byte Mask (WR) Dynamic Width Demux (WR) 16x16 Dynamic Width Mux (RD) Q[15:0][15:0] D[15:0][15:0] 16x16 16 16 1:16 Demux 16:1 Mux 16/tCC 16 ... ...... 16x16 S1[15:0][15:0] ... R/W ROW PRE 1 ... COL 1 ROW 1 1 R/W ACT 16x16*26 ... Sense Amp(23-1) PRE logic 16x16*26 Sense Amp Array Sense Amp 1 ... ... 1 PRE 16x16*26 ... 11 11 1 ACT 1 11 ROW 11 PRE Bank(23-1) ACT logic ... 1 ACT ROW Even Bank Array Bank 0 16x16*26*211 1 ... ACT 16x16*26 3 ... 16x16*26*211 Bank 1 ACT decode 3 11 ... Odd Bank Array Bank 0 PRE decode 16/tCC 16 16 DQ15..0 16 DQN15..0 Version 1.0 Jan. 2005 Page 51 K4Y5416(/08/04)4UF XDR DRAM Simultaneous Activation When the XDR DRAM supports multiple bank sets as in Figure 55, another feature may be supported, in addition to ERAW. This feature is simultaneous activation, and the timing of several cases is shown in Figure 56. The tRR parameter specifies the minimum spacing between packets with activation commands in XDR DRAMs with a single bank set, or between packets to the same bank set in a XDR DRAM with multiple bank sets. The tRR-D parameter specifies the minimum spacing between packets with activation commands to different bank sets in a XDR DRAM with multiple bank sets. In Figure 56, Case 4 shows an example when both tRR and tRR-D must be at least 4*tCYCLE. In such a case, activation commands to different bank sets satisfy the same constraint as activation commands to the same bank set. In Figure 56, Case 2 shows an example when tRR must be at least 4*tCYCLE and tRR-D must be at least 2*tCYCLE. In such a case, an activation command to one bank set may be inserted between two activation commands to a different bank set. In Figure 56, Case 1 shows an example when tRR must be at least 4*tCYCLE and tRR-D must be at least 1*tCYCLE. As in the previous case, an activation command to one bank set may be inserted between two activation commands to a different bank set. In this case, the middle activation command will not be symmetrically placed relative to the two outer activation commands. In Figure 56, Case 0 shows an example when tRR must be at least 4*tCYCLE and tRR-D must be at least 0*tCYCLE. This means that two activation commands may be issued on the same CFM clock edge. This is only possible by using the delay mechanism in one of the two commands. See "Dynamic Request Scheduling" on page 18. In the example shown, the packet with the REFA command is received one cycle before the command with the ACT command, and the REFA command includes a one cycle delay. Both activation commands will be issued internally to different bank sets on the same CFM clock edge. Figure 56 : Simultaneous Activation -- tRR-D Cases Case 4: tRR-D = 4*tCYCLE REFA & ACT have same tRR T0 T1 T2 T3 T4 T5 T6 T7 T8 Case 2: tRR-D = 2*tCYCLE REFA fits between two ACT T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 ACT REFA ACT ACT tRR-D tRR-D REFA tRR-D DQ15..0 DQN15..0 tRR T1 T2 T3 T4 T5 tCYCLE note - REFA is directed to bank set different from two ACT Case 0: tRR-D = 0*tCYCLE REFA simultaneous with ACT (REFA uses delay=1*tCYCLE) Case 1: tRR-D = 1*tCYCLE REFA fits between two ACT T0 ACT T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 DQ15..0 DQN15..0 ACT REFA REFA ACT ACT ACT tCYCLE tRR-D tRR-D tRR note - REFA is directed to bank set different from two ACT tRR note - REFA is directed to bank set different from ACT at T12 Version 1.0 Jan. 2005 Page 52 K4Y5416(/08/04)4UF XDR DRAM Simultaneous Precharge When the XDR DRAM supports multiple bank sets as in Figure55, another feature may be supported, in addition to ERAW. This feature is simultaneous precharge, and the timing of several cases is shown in Figure57. The tPP parameter specifies the minimum spacing between packets with precharge commands in XDR DRAMs with a single bank set, or between packets to the same bank set in a XDR DRAM with multiple bank sets. The tPP-D parameter specifies the minimum spacing between packets with precharge commands to different bank sets in a XDR DRAM with multiple bank sets. In Figure57, Case4 shows an example when both tPP and tPP-D must be at least 4*tCYCLE. In such a case, precharge commands to different bank sets satisfy the same constraint as precharge commands to the same bank set. In Figure57, Case2 shows an example when tPP must be at least 4*tCYCLE and tPP-D must be at least 2*tCYCLE. In such a case, a precharge command to one bank set may be inserted between two precharge commands to a different bank set. In Figure57, Case1 shows an example when tPP must be at least 4*tCYCLE and tPP-D must be at least 1*tCYCLE. As in the previous case, a precharge command to one bank set may be inserted between two precharge commands to a different bank set. In this case, the middle precharge command will not be symmetrically placed relative to the two outer precharge commands. In Figure57, Case0 shows an example when tPP must be at least 4*tCYCLE and tPP-D must be at least 0*tCYCLE. This means that two precharge commands may be issued on the same CFM clock edge. This is only possible by using the delay mechanism in one of the two commnads. See "Dynamic Request Scheduling" on page 18. It is also possibly by taking advantage of the fact that two independent precharge commands may be encoded within a single ROWP packet. In the example shown, the ROWP packet contains both a REFP command and a PRE command. Both precharge commands will be issued internally to different bank sets on the same CFM clock edge. Figure 57 : Simultaneous Precharge -- tPP-D Cases Case 4: tPP-D = 4*tCYCLE REFP & PRE have same tRR T0 T1 T2 T3 T4 T5 T6 T7 T8 Case 2: tPP-D = 2*tCYCLE REFP fits between two PRE T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 PRE REFP PRE PRE tPP-D tPP-D REFP PRE tPP-D DQ15..0 DQN15..0 note - REFP is directed to bank set different from two PRE tPP Case 0: tPP-D = 0*tCYCLE REFP simultaneous with PRE Case 1: tPP-D = 1*tCYCLE REFP fits between two PRE T0 T1 T2 T3 T4 T5 tCYCLE T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 CFM CFMN RQ11..0 DQ15..0 DQN15..0 PRE REFP PRE PRE PRE tCYCLE REFP tPP-D tPP-D tPP note - REFP is directed to bank set different from two PRE tPP note - REFP is directed to bank set different from PRE at T12 Version 1.0 Jan. 2005 Page 53 K4Y5416(/08/04)4UF XDR DRAM Operating Conditions Electrical Conditions Table13 summarizes all electrical conditions (temperature and voltage conditions) that may be applied to the memory component. The first section of parameters is concerned with absolute voltage, storage and operating temperatures, and the power supply, reference, and termination voltage. The second section of parameters determines the input voltage levels for the RSL RQ signals. The high and low voltages must satisfy a symmetry parameter with respect to the VREF, RSL. The third section of parameters determines the input voltage levels for the RSL SI(serial interface) signals. The high and low voltages must satisfy a symmetry parameter with respect to the VREF, RSL. The fourth section of parameters determines the input voltage levels for the CFM clock signals. The high and low voltages are specified by a common-mode value and a swing value. The fifth section of parameters determines the input voltage levels for the write data signals on the DRSL DQ pins. The high and low voltages are specified by a common-mode value and a swing value. Table 13 : Electrical Conditions Minimum Maximum Unit VIN,ABS Symbol Voltage applied to any pin (except VDD) with respect to GND Parameter - 0.3 1.5 V VDD,ABS Voltage on VDD with respect to GND - 0.5 2.3 V TSTORE Storage temperature - 50 100 C TJ Junction temperature under bias during normal operation VDD Supply voltage applied to VDD pins during normal operation C 100 1.8 - 0.09 1.8 + 0.09 V VTERM,RSL - 0.450 - 0.025 VTERM,RSL - 0.450 + 0.025 V 1.2 - 0.06 1.2 + 0.06 V VREF,RSL RSL - Reference voltage applied to VREF pina VTERM,DRSL DRSL - Termination voltage applied to VTERM pins VIL,RQ RSL RQ inputs -low voltage VREF,RSL - 0.45 VREF,RSL - 0.15 V VIH,RQb RSL RQ inputs -high voltage VREF,RSL + 0.15 VREF,RSL + 0.45 V RA,RQ RSL RQ inputs - data asymmetry: RA,RQ = (VIH,RQ-VREF,RSL)/(VREF,RSL-VIL,RQ) TBD TBD V VIL,SI RSL Serial Interface inputs -low voltage VREF,RSL - 0.45 VREF,RSL - 0.20 V RSL Serial Interface inputs -high voltage VREF,RSL + 0.20 VREF,RSL + 0.45 V TBD TBD V VIH,SI b RA,SI VICM,CFM VISW,CFM VICM,DQ VISW,DQ RSL Serial Interface inputs - data asymmetry: RA,SI = (VIH,RQ-VREF,RSL)/(VREF,RSL-VIL,RQ) CFM/CFMN input - common mode: VICM,CTM = (VIH,CFMb+VIL,CTM)/2 CFM/CFMN input - high-low swing: VISW,CFM = (VIH,CTMb VIL,CTM) DRSL DQ inputs - common mode: VICM,DQ = (VIH,DQb+VIL,DQ)/2 DRSL DQ inputs - high-low swing: VISW,DQ = (VIH,DQb - VIL,DQ) VTERM,DRSL-0.150 VTERM,DRSL-0.075 0.15 0.30 VTERM,DRSL-0.150 VTERM,DRSL-0.025 0.05 0.30 V V V V a. VTERM,RSL is typically 1.2V0.06V. It connects to the RSL termination components, not to this DRAM component. b. VIH is typically equal to VTERM,RSL or VTERM,DRSL (whichever is appropriate) under DC conditions in a system. Version 1.0 Jan. 2005 Page 54 K4Y5416(/08/04)4UF XDR DRAM Timing Conditions Table14 summarizes all timing conditions that may be applied to the memory component. The first section of parameters is concerned with parameters for the clock signals. The second section of parameters is concerned with parameters for the request signals. The third section of parameters is concerned with parameters is concerned with parameters for the write data signals. The fourth section of parameters is concerned with parameters for the serial interface signals. The fifth section is concerned with all other parameters, including those for refresh, calibration, power state transitions, and initialization. Table 14 : Timing Conditions Symbol Parameter and Other Conditions Minimum Maximum Units Figure(s) tCYCLE or tCYC,CTM CFM RSL clock - cycle time -4000 -3200 -2400 2.00 2.50 3.33 3.83 3.83 3.83 ns ns ns Figure 58 tR,CFM, tF,CFM CFM/CFMN input - rise and fall time - use minimum for test. 0.08 0.20 tCYCLE Figure 58 tH,CFM, tL,CFM CFM/CFMN input - high and low times 40% 60% tCYCLE Figure 58 tR,RQ, tF,RQ RSL RQ input - rise/fall times (20% - 80%) - use minimum for test. 0.08 0.26 tCYCLE Figure 59 tS,RQ, tH,RQ RSL RQ input to sample points (set/hold) @ 2.50 ns > tCYCLE 2.00 ns @ 3.33 ns > tCYCLE 2.50 ns @ 3.83 ns tCYCLE 3.33 ns tbd 0.200 tbd ns ns ns Figure 59 tIR,DQ, tIF,DQ DRSL DQ input - rise/fall times (20% - 80%) - use minimum for test. 0.020 0.074 tCYCLE Figure 60 tS,DQ, tH,DQ DRSL DQ input to sample points (set/hold) @ 2.50 ns > tCYCLE 2.00 ns @ 3.33 ns > tCYCLE 2.50 ns @ 3.83 ns tCYCLE 3.33 ns tbd 0.065 tbd - ns ns ns Figure 60 tDOFF,DQ DRSL DQ input delay offset (fixed) to sample points -0.08 +0.08 tCYCLE Figure 60 tCYC,SCK Serial Interface SCK input - cycle time 20 - ns Figure 62 tR,SCK, tF,SCK Serial Interface SCK input - rise and fall times - 5.0 ns Figure 62 tH,SCK, tL,SCK Serial Interface SCK input - high and low times 40% 60% tCYC,SCK Figure 62 tIR,SI, tIF,SI Serial Interface CMD,RST,SDI input - rise and fall times - 5.0 ns Figure 62 tS,SI,tH,SI Serial Interface CMD,SDI input to SCK clock edge - set/hold time 5 - ns Figure 62 tDLY,SI-RQ Delay from last SCK clock edge for register write to first CFM edge with RQ packet containing a command which uses the value in the register. Also, delay from first CFM edge with RQ packet containing a command which modifies register value to the first SCK clock edge for register read to this register. 10 - tCYC,SCK - tREF Refresh interval. Every row of every bank must be accessed at least once in this interval with a ROW-ACT, ROWP-REF or ROWP-REFI command. - 16 ms Figure 44 tREFA-REFA,AVG Average refresh command interval. ROWP-REFA or ROWP-REFI commands must be issued at this average rate. This depends upon tREF and the number of banks and the number of rows: tREFI = tREF/(NB*NR) = tREF/(23*211). ns - NREFA,BURST Refresh burst limit. The number of ROWP-REFA or ROWP-REFI commands which can be issued consecutively at the minimum command spacing. - 128 commands - tBURST-REFA Refresh burst interval. The interval between a burst of NREFA,BURST,MAX ROWP-REFA or ROWP-REFI commands and the next ROWP-REFA or ROWP-REFI command. - TBD tCYCLE - tCOREINIT Interval needed for core initialialization after power is applied. - 1.5 ms - tCALC Current calibration interval - 100 ms Figure 45 Delay between packet with any command and CALC/CALZ packet tCMD-CALC, tCMD-CALZ w/ PRE or REFP command w/ any other command 4 16 - tCYCLE Figure 45 tCALCE, tCALZE 12 - tCYCLE Figure 45 Figure 45 Delay between CALC/CALZ packet and CALE packet - tREFA-REFA,AVG = 488 tCALE-CMD Delay between CALE packet and packet with any command 24 - tCYCLE tCMD-PDN Last command before PDN entry 16 - tCYCLE Figure 46 tPDN-CFM RSL CFM/CFMN and VTERM stable after PDN entry 16 - tCYCLE Figure 46 tCFM-PDN RSL CFM/CFMN and VTERM stable before PDN exit 16 - tCYCLE Figure 46 tPDN-CMD First command after PDN exit (includes lock time for CFM/CFMN) 4096 - tCYCLE Figure 46 Version 1.0 Jan. 2005 Page 55 K4Y5416(/08/04)4UF XDR DRAM Operating Characteristics Electrical Characteristics Table15 summarizes all electrical parameters (temperature, current and voltage) that characterize this memory component. The only exception is the supply current values(IDD) under different operating conditions covered in the Supply Current Profile section. The first section of parameters is concerned with the thermal characteristics of the memory component. Ther second section of parameters is concerned with the current needed by the RQ pins and VREF pin. The third section of parameters is concerned with the current needed by the DQ pins and voltage levels produced by the DQ pins when driving read data. This section is also concerned with the current needed by the VTERM pin, and with the resistance levels produced for the internal termination components that attach to the DQ pins. The fourth section of parameters determines the output voltage levels and the current needed for the serial interface signals. Table 15 : Electrical Characteristics Symbol Parameter Minimum Maximum Units - TBD C/Watt RSL RQ or Serial Interface input current @ ( VIN= VIH,RQ,MAX) -10 10 uA IREF,RSL VREF,RSL current @ VREF,RSL,MAX flowing into VREF pin -10 10 uA VOSW,DQ DRSL DQ outputs - high-low swing: VOSW,DQ=(VIH,DQVIL,DQN) or (VIH,DQN-VIL,DQ) (See Figure 27) ODF = 00 ODF=01 ODF=10 ODF=11 0.225 TBD TBD TBD 0.350 TBD TBD TBD V RTERM,DQ DRSL DQ outputs - termination resistance 40.0 60.0 VOL,SI RSL serial interface SDO output - low voltage 0.0 0.25 V VOH,SI RSL serial interface SDO output - high voltage VTERM,RSL - 0.25 VTERM,RSL V JC Junction-to-case thermal resistance II,RSL Version 1.0 Jan. 2005 Page 56 K4Y5416(/08/04)4UF XDR DRAM Supply Current Profile In this section, Table16 summarizes the supply current (IDD) that characterizes this memory component. This parameter is shown under different operating conditions. Table 16 : Supply Current Profile Symbol Power State and Steady State Transaction Rates Maximum @tCYCLE= 2.00 ns Maximum @tCYCLE= 2.50 ns Maximum @tCYCLE= 3.33 ns Units IDD,PDN Device in PDN, self-refresh enabled. a TBD 9500 TBD uA IDD,STBY Device in STBY. This is for a device in STBY with no packets on the Channela TBD 340 TBD mA IDD,REF Device in STBY and refreshing rows at the tREF,MAX period.a TBD - TBD mA TBD 1120 TBD mA TBD 1200 TBD mA IDD,WR IDD,RD ACT command every tRR, PRE command every tPP, WR command every tCC.a ACT command every tRR, PRE command every tPP, RD command every tCCa,b ITERM,DRSL,NONE No RD nor WR commands issued. c TBD 35 TBD mA ITERM,DRSL,RD RD command every tCC.c TBD 100 TBD mA ITERM,DRSL,WR WR command every tCC.c TBD 200 TBD mA a. IDD current @ VDD,MAX flowing into VDD pins b. This does not include the IOL,DQ sink current. The device dissipates IOL,DQ*VTERM,DQ in each DQ/DQN pair when driving data. c. ITERM,DRSL current @ VTERM,DQ,MAX flowing into VTERM pins Version 1.0 Jan. 2005 Page 57 K4Y5416(/08/04)4UF XDR DRAM Timing Characteristics Table 17 summarizes all timing parameters that characterize this memory component. The only exceptions are the core timing parameters that are speed-bin dependent. Refer to the Timing Parameters section for more information. The first section of parameters pertains to the timing of the DQ pins when driving read data. The second section of parameters is concerned with the timing for the serial interface signals when driving register read data. The third section of parameters is concerned with the time intervals needed by the interface to transition between power states. Table 17 : Timing Characteristics Symbol Parameter and Other Conditions tQ,DQ DRSL DQ output delay (variation across 16 Q bits on each DQ pin) from drive points - output delay @ 2.50 ns > tCYCLE 2.00 ns @ 3.33 ns > tCYCLE 2.50 ns @ 3.83 ns tCYCLE 3.33 ns tQOFF,DQ DRSL DQ output delay offset (a fixed value for all 16 Q bits on each DQ pin) from drive points - output delay tOR,DQ, tOF,DQ DRSL DQ output - rise and fall times (20%-80%). Minimum Maximum Units Figure(s) tbd -0.065 tbd tbd +0.065 tbd ns ns ns Figure 61 0.00 +0.20 tCYCLE Figure 61 Figure 61 0.02 0.04 tCYCLE tQ,SI Serial SCK-to-SDO output delay @ CLOAD,MAX = 15 pF 2 15 ns Figure 63 tP,SI Serial SDI-to-SDO propagation delay @ CLOAD,MAX = 15 pF - 15 ns Figure 63 tOR,SI, tOF,SI Serial SDO output rise/fall (20%-80%) @ CLOAD,MAX = 15 pF - 10 ns Figure 63 tPDN-ENTRY Time for power state to change after PDN entry - 16 tCYCLE Figure 46 - tCYCLE Figure 46 tPDN-EXIT Time for power state to change after PDN exit 0 Version 1.0 Jan. 2005 Page 58 K4Y5416(/08/04)4UF XDR DRAM Timing Parameters Table18 summarizes the timing parameters that characterize the core logic of this memory component.. These timing parameters will vary as a function of the component's speed bin. The four sections deal with the timing intervals between packets with, respectively, row-row commands, row-column commands, column-column commands, and column-row commands. Table 18 : Timing Parameters Symbol Parameter and Other Conditions Min (A) Min (B) Min (C) Units Figure(s) tRC Row-cycle time: interval between successive ROWA-ACT or ROWP-REFA or ROWP-REFI activate commands to the same bank. tRC tRC-R, 2tCC = tRCD-R + tCC+ tRDP + tRPa tRC-W, 2tCC, noERAW = tRCD-W + tCC+ tWRP +tRPa tRC-W, 2tCC, ERAW = tRCD-W + tCC+ tWRP + tRPa 16 16 19 23 20 20 24 28 24 24 24 28 tCYCLE Figure 4 Figure 7 tRAS Row-asserted time: interval between a ROWA-ACT or ROWP-REFA or ROWP-REFI activate command and a ROWP-PRE or ROWP-REFP precharge command to the same bank. Note that tRAS,MAX is 64 us for all timing bins. 10 13 17 tCYCLE Figure 4 Figure 7 tRP Row-precharge time: interval between a ROWP-PRE or ROWP-REFP precharge command and a ROWA-ACT or ROWP-REFA or ROWP-REFI activate command to the same bank. 6 7 7 tCYCLE Figure 4 Figure 7 tPP Precharge-to-precharge time: interval between successive ROWP-PRE or ROWP-REFP precharge commands to different banks. tPP tPP-Db 4 1 4 1 4 1 tCYCLE Figure 4 Figure 7 tRR Row-to-row time: interval between ROWA-ACT or tRR ROWP- REFA or ROWP-REFI activate commands tRR-Dc to different banks. 4 4 4 4 4 4 tCYCLE Figure 4 Figure 7 tRCD-R Row-to-column-read delay: interval between a ROWA-ACT activate command and a COL-RD read command to the same bank. 5 7 7 tCYCLE Figure 4 Figure 7 tRCD-W Row-to-column-write delay: interval between a ROWA-ACT activate command and a COL-WR or COL-WRM write command to the same bank. 1 3 3 tCYCLE Figure 4 Figure 7 tCAC Column access delay: interval from COL-RD read command to Q read data 6 7 7 tCYCLE Figure 10 tCWD Column write delay: interval from a COL-WR or COLM-WRM write command to D write data. 3 3 3 tCYCLE Figure 9 tCC Column-to-column time: interval between successive COL-RD commands, or between successive COL-WR or COLM-WRM commands. 2 2 2 tCYCLE Figure 4 Figure 7 tRW-BUB, Read-to-write bubble time: interval between the end of a Q read data packet and the start of D write data packet (the end of a data packet is the time interval tCC after its start). 3 3 3 tCYCLE Figure 13 Write-to-read bubble time: interval between the end of a D writed data and the start of Q read data packet (the end of a data packet is the time interval tCC after its start). 3 3 3 tCYCLE Figure 13 tRW Read-to-write time: interval between a COL-RD read command and a COL-WR or COLM-WRM write command.d 8 9 9 tCYCLE Figure 12 tWR Write-to-read time: interval between a COL-WR or tWR COLM-WRM write command and a COL-RD read tWR-De command. 9 2 10 2 10 2 tCYCLE Figure 12 tRDP Read-to-precharge time: interval between a COL-RD read command and a ROWP-PRE precharge command to the same bank. 3 4 4 tCYCLE Figure 4 Figure 7 tWRP Write-to-precharge time: interval between a COL-WR or COLM-WRM write command and a ROWP-PRE precharge command to the same bank. 10 12 12 tCYCLE Figure 4 Figure 7 tDR Write data-to-read time: interval between the start of D write data and a COL-RD read command to the same bank. 6 7 7 tCYCLE Figure 12 tDP Write data-to-precharge time: interval between D write data and ROWP-PRE precharge command to the same bank. 7 9 9 tCYCLE Figure 9 XDRDRAM tWR-BUB, XDRDRAM tLRRn-LRRn Interval between ROWP-LRRn command and a subsequent ROWP-LRRn command. f 16 20 24 tCYCLE Table5 tREFx-LRRn Interval between ROWP-REFx command and a subsequent ROWP-LRRn command. 16 20 24 tCYCLE Table5 tLRRn-REFx Interval between ROWP-LRRn command and a subsequent ROWP-REFx command. 16 20 24 tCYCLE Table5 a. The tRC,MIN parameter is applicable to all transaction types (read, write, refresh, etc.). Read and write transactions may have an additional limitation, depending upon how many column accesses (each requiring tCC) are performed in each row access (tRC). The table lists the special cases (tRC-R, 2tCC, tRCW, 2tCC, noERAW, tRC-W, 2tCC, ERAW) in which two column accesses are performed in each row access. Note that tRC-W, 2tCC, ERAW uses a relaxed value of tRCD-W that is equal to tRCD-R,MIN. All other parameters are minimum. b. tPP-D is the tPP parameter for precharges to different bank sets. See "Simultaneous Precharge" on page 53. c. tRR-D is the tRR parameter for activates to different bank sets. See "Simultaneous Activation" on page 52. d. See "Propagation Delay" on page 27. e. tWR-D is the tWR parameter for write-read accesses to different bank sets. See "Multiple Bank Sets and the ERAW Feature" on page 50. Also, note that the value of tWR-D may not take on the values {3,5,7} within the range{tWR-D,MIN, ... tWR,MIN-1}. tDWR-D may assume any value tWR,MIN. f. ROWP-LRRn includes the commands {ROWP-LRR0,ROWP-LRR1,LOWP-LRR2}, ROWP-REFx includes the commands {ROWP-REFA,ROWP-REFI, LOWP-REFP}, Version 1.0 Jan. 2005 Page 59 K4Y5416(/08/04)4UF XDR DRAM Receive/Transmit Timing Clocking Figure58 shows a timing diagram for the CFM/CFMN clock pins of the memory component. This diagram represents a magnified view of these pins. This diagram shows only one clock cycle. CFM and CFMN are differential signals: one signal is the complement of the other. They are also high-true signals - a low voltage represents a logical zero and a high voltage represents a logical one. There are two crossing points in each clock cycle. The primary crossing point includes the high-voltage-to-low-voltage transition of CFM (indicated with the arrowhead in the diagram). The secondary crossing point includes the low-voltage-to-high-voltage transition of CFM. All timing events on the RSL signals are referenced to the first set of edges. Timing events are measured to and from the crossing point of the CFM and CFMN signals. In the timing diagram, this is how the clock-cycle time (tCYCLE or tCYC, CFM), clock-low time (tL, CFM) and clock-high time (tH, CFM) are measured. Because timing intervals are measured in this fashion, it is necessary to constrain the slew rate of the signals. The rise (tR, CFM) and fall time (tF, CFM) of the signals are measured from the 20% and 80% points of the full-swing levels. 20% = VIL, CFM + 0.2*(VIH, CFM - VIL, CFM) 80% = VIL, CFM + 0.8*(VIH, CFM - VIL, CFM) Figure 58 : Clocking Waveforms tCYCLE or tCYC,CFM tL,CFM logic 1 tH,CFM CFM VIH,CFM 80% CFMN 20% VIL,CFM logic 0 tR,CFM tF,CFM Version 1.0 Jan. 2005 Page 60 K4Y5416(/08/04)4UF XDR DRAM RSL RQ Receive Timing Figure59 shows a timing diagram for the RQ11...0 request pins of the memory component. This diagram represents a magnified view of the pins and only a few clock cycle (CFM and CFMN are the clock signals). Timing events are measured to and from the primary CFM/CFMN crossing point in which CFM makes its high-voltage-to-low-voltage transition. The RQ11...0 signals are lowtrue: a high voltage represents a logical zero and a low voltage represents a logical one. Timing events on the RQ11... 0 pins are measured to and from the point that the signal reaches the level of the reference voltage VREF, RSL. Because timing intervals are measured in this fashion, it is necessary to constrain the slew rate of the signals. The rise (tR, RQ) and fall time (tF, RQ) of the signals are measured from the 20% and 80% points of the full-swing levels. 20% = VIL, RQ + 0.2*(VIH, RQ - VIL, RQ) 80% = VIL, RQ + 0.8*(VIH, RQ - VIL, RQ) There are two data receiving windows defined for each RQ11...0 signal. The first of these (labeled "0") and a set time, tS,RQ, and a hold time, tH,RQ, measured around the primary CFM/CFMN crossing point. The second (labeled "1") has a set time (tS, RQ) and a hold time (tH, RQ) measured around a point 0.5*tCYCLE after the primary CFM/CFMN crossing point. Figure 59 : RSL RQ Receive Waveform tCYCLE CFM CFMN [1/2]*tCYCLE tH,RQ tS,RQ RQ0 tS,RQ 0 logic 0 VIH,RQ 80% VREF,RSL 1 20% VIL,RQ logic1 tF,RQ ... tR,RQ tH,RQ [1/2]*tCYCLE tH,RQ tS,RQ RQ11 0 tR,RQ tS,RQ tH,RQ 1 tF,RQ logic 0 VIH,RQ 80% VREF,RSL 20% VIL,RQ logic 1 Version 1.0 Jan. 2005 Page 61 K4Y5416(/08/04)4UF XDR DRAM DRSL DQ Receive Timing Figure60 shows a timing diagram for receiving write data on the DQ/DQN data pins of the memory component. This diagram represents a magnified view of the pins and only a few clock cycles are shown (CFM and CFMN are the clock signals). Timing events are measured to and from the primary CFM/CFMN crossing point in which CFM makes its high-voltage-to-low -voltage transition. The DQ15...0/DQN15...0 signals are high-true: a low voltage represents a logical zero and a high voltage represents a logical one. They are also differential - timing events on the DQ15...0/DQN15... 0 pins are measured to and from the point that each differential pair crosses. Because timing intervals are measured in this fashion, it is necessary to constrain the slew rate of the signals. The rise time (tIR, and fall time (tIF, DQ) of the signals are measured from the 20% and 80% points of the full-swing levels. DQ) 20% = VIL, DQ + 0.2*(VIH, DQ - VIL, DQ) 80% = VIL, DQ + 0.8*(VIH, DQ - VIL, DQ) There are 16 data receiving windows defined for each DQ15...0/DQN15... 0 pin pair. The receiving windows for a particular DQi/DQNi pin pair is referenced to an offset parameter tDOFF,DQi (the index "i" may take on the values {0, 1, .. , 15} and refers to each of the DQ15... 0/DQN15... 0 pin pairs). The tDOFF,DQi parameter determines the time between the primary CFM/CFMN crossing point and the offset point for the DQi/DQNi pin pair. The 16 receiving windows are placed at times tDOFF,DQi + (j/8)*tCYCLE (the index "j" may take on the values {0, 1, .. , 15} and refers to each of the receiving windows for the DQi/DQNi pin pair). The offset values tDOFF,DQi for each of the 16 DQi/DQNi pin pairs can be different. However, each is constrained to lie inside the range {tDOFF,MIN, tDOFF,MAX}. Furthermore, each offset value tDOFF,DQi is static and will not change during system operation. Its value can be determined at initialization. The 16 receiving windows (j = 0 ... 15) for the first pair DQ0/DQN0 are labeled "0" through "15". Each window has a set time (tS, and a hold time (tH, DQ) measured around a point tDOFF,DQ0 + (j/8) *tCYCLE after the primary CFM/CFMN crossing point. DQ) The 16 receiving windows (j = 0 ... 15) for the each of the other pairs DQi/DQNi are also labeled "0" through "15". Each window has a set time (tS, DQ) and a hold time (tH, DQ) measured around a point tDOFF,DQi + (j/8)*tCYCLE after the primary CFM/CFMN crossing point. Version 1.0 Jan. 2005 Page 62 K4Y5416(/08/04)4UF XDR DRAM Figure 60 : DRSL DQ Receive Waveforms tCYCLE CFM ... CFMN i = {0,1,2,3,4,5,...15} tDOFF,MAX j = {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15} tDOFF,MIN [(j)/8]*tCYCLE tDOFF,DQ0 tS,DQ logic 1 VIH,DQ 80% tH,DQ DQ0 0 1 2 3 4 5 ... 6 ... j 14 15 20% VIL,DQ logic 0 DQN0 tIF,DQ ... tIR,DQ [(j)/8]*tCYCLE tDOFF,DQi logic 1 VIH,DQ 80% tH,DQ tS,DQ DQi 0 1 2 3 4 5 ... 6 ... j 14 15 20% VIL,DQ logic 0 DQNi ... tIR,DQ tIF,DQ [(j)/8]*tCYCLE tDOFF,DQ15 logic 1 VIH,DQ 80% tH,DQ tS,DQ DQ15 0 1 2 3 4 5 6 DQN15 tIR,DQ tIF,DQ ... j ... 14 15 20% VIL,DQ logic 0 Version 1.0 Jan. 2005 Page 63 K4Y5416(/08/04)4UF XDR DRAM DRSL DQ Transmit Timing Figure61 shows a timing diagram for transmitting read data on the DQ15...0/DQN15...0 data pins of the memory component. This diagram represents a magnified view of these pins and only a few clock cycles are shown (CFM and CFMN are the clock signals). Timing events are measured to and from the primary CFM/CFMN crossing point in which CFM makes its high-voltageto-low-voltage transition. The DQ15...0/DQN15...0 signals are high-true: a low voltage represents a logical zero and a high voltage represents a logical one. They are also differential - timing events on the DQ15...0/DQN15...0 pins are measured to and from the point that each differential pair crosses. Because timing intervals are measured in this fashion, it is necessary to constrain the slew rate of the signals. The rise (tOR, DQ) and fall time (tOF, DQ) of the signals are measured from the 20% and 80% points of the full-swing levels. 20% = VOL, DQ + 0.2*(VOH, DQ - VOL, DQ) 80% = VOL, DQ + 0.8*(VOH, DQ - VOL, DQ) There are 16 data transmit windows defined for each DQ15...0/DQN15...0 pin pair. The transmitting windows for a particular DQi/DQNi pin pair is referenced to an offset parameter tQOFF,DQi (the index "i" may take on the values {0, 1, .., 15} and refers to each of the DQ15... 0/DQN15...0 pin pairs). The tQOFF,DQi + tQ,DQ,MAX expression determines the time between the primary CFM/CFMN crossing point and the offset point for the DQi/DQNi pin pair. The offset values tQOFF,DQi for each of the 16 DQi/DQNi pin pairs can be different. However, each is constrained to lie inside the range {tQOFF,MIN, tQOFF,MAX}. Furthermore, each offset value tQOFF,DQi is static; its value will not change during system operation. Its value can be determined at initialization time. The 16 transmit windwos (j = 0 ... 15} for the first pair DQ0/DQN0 are labeled "0" through "15". Each window begins at the time (tQOFF,DQ0 + tQ,DQ,MAX +((j+0.5)/8)*tCYCLE) and ends at the time (tQOFF,DQ0 + tQ,DQ,MIN +((j+1.5)/8)*tCYCLE) measured after the primary CFM/CFMN crossing point. Note that when no read data is to be transmitted on the DQ/DQN pins(and no other component is transmitting on the external DQ/DQN wires), then the voltage level on the DQ/DQN pins will follow the voltage reference value VTERM,DRSL on the VTERM pin. The logical value of each DQ/DQN pin pair in this no-drive state will be "1/1"; when read data is driven, each DQ/DQN pin pair will have either the logical value of "1/0" or "0/1". Version 1.0 Jan. 2005 Page 64 K4Y5416(/08/04)4UF XDR DRAM Figure 61 : RSL DQ Transmit Waveforms tCYCLE CFM ... CFMN i = {0,1,2,3,4,5,...15} tQOFF,MAX j = {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15} tQOFF,MIN [(j+0.5)/8]*tCYCLE [(j-0.5)/8]*tCYCLE DQ0 tQOFF,DQ0 0 1 2 3 4 6 7 ... 8 ... j 14 15 20% VOL,DQ logic "0" DQN0 tOR,DQ tOF,DQ ... [(j+0.5)/8]*tCYCLE [(j-0.5)/8]*tCYCLE DQi 1 2 3 4 5 6 ... 7 ... j 14 15 20% VOL,DQ logic "0" DQni tOR,DQ logic "1" VOH,DQ 80% tQ,DQ,MIN tQ,DQ,MAX tQOFF,DQi 0 tOF,DQ ... [(j+0.5)/8]*tCYCLE [(j-0.5)/8]*tCYCLE DQ7 logic "1" VOH,DQ 80% tQ,DQ,MIN tQ,DQ,MAX tQOFF,DQ7 0 1 2 3 4 5 6 DQN7 tOR,DQ tOF,DQ 7 ... logic "1" VOH,DQ 80% tQ,DQ,MIN tQ,DQ,MAX j ... 14 15 20% VOL,DQ logic "0" Version 1.0 Jan. 2005 Page 65 K4Y5416(/08/04)4UF XDR DRAM Serial Interface Receive Timing Figure62 shows a timing diagram for the serial interface pins of the memory component. This diagram represents a magnified view of the pins only a few clock cycles. The serial interface pins carry low-true signals: a high voltage represents a logical zero and a low voltage represents a logical one. Timing events are measured to and from the VREF,RSL level. Because timing intervals are measured in this fashion, it is necessary to constrain the slew rate of the signals. The rise time (tR,SCK and tRI,SI) and fall time (tF,SCK and tIF,SI) of the signals are measured from the 20% and 80% points of the full-swing levels. 20% = VIL,SI + 0.2 *(VIH,SI - VIL,SI) 50% = VIL,SI + 0.5 *(VIH,SI - VIL,SI) 80% = VIL,SI + 0.8 *(VIH,SI - VIL,SI) There is one receiving window defined for each serial interface signal (RST, CMD and SDI pins). This window has a set time (tS, RQ) and a hold time (tH, RQ) measured around the falling edge of the SCK clock signal. Figure 62 : Serial Interface Receive Waveforms tCYC,SCK logic 0 VIH,SI tH,SCK tL,SCK 80% SCK VREF,RSL 20% VIL,SI tF,SCK logic 1 tR,SCK tS,SI tH,SI logic 0 VIH,SI 80% RST CMD SDI VREF,RSL 20% VIL,SI tIR,SI tIF,SI logic 1 Version 1.0 Jan. 2005 Page 66 K4Y5416(/08/04)4UF XDR DRAM Serial Interface Transmit Timing Figure63 shows a timing diagram for the serial interface pins of the memory component. This diagram represents a magnified view of the pins and only a few clock cycles are shown. The serial interface pins carry low-true signals: a high voltage represents a logical zero and a low voltage represents a logical one. Timing events are measured to and from the VREF,RSL level. Because timing intervals are measured in this fashion, it is necessary to constrain the slew rate of the signals. The rise time (tOR,SI) and fall time (tOF,SI) of the signals are measured from the 20% and 80% points of the full-swing levels. 20% = VOL,SI + 0.2*(VOH,SI - VOL,SI) 50% = VOL,SI + 0.5*(VOH,SI - VOL,SI) 80% = VOL,SI + 0.8*(VOH,SI - VOL,SI) There is one transmit window defined for the serial interface data signal (SDO pins). This window has a maximum delay time (tQ, from the falling edge of the SCK clock signal and a minimum delay time (tQ,SI,MIN) from the next falling edge of the SCK clock signal. SI,MAX) When the memory component is not selected during a serial device read transaction, it will simply pass the information on the SDI input to the SDO output. This combinational propagation delay parameter is tP,SI. The tCYC,SCK will need to be increased during a serial read transaction (relative to the tCYC,SCK value for a serial write transaction) because of the accumulated propagation delay through all of the XDR DRAM devices on the serial interface. During Initialization, when the serial identification is determined, the SDI-to-SDO path is registered, so the tCYC,SCK value can be set to the same value as for serial write transactions. See "Initialization" on page42. Figure 63 : Serial Interface Transmit Waveforms tCYC,SCK logic 0 tH,SCK tL,SCK VIH,SI 80% SCK VREF,RSL tF,SCK 20% VIL,SI logic 1 tR,SCK tQ,SI,MAX tQ,SI,MIN logic 0 VOH,SI 80% tP,SI VREF,RSL SDO 20% VOL,SI tOR,SI tOF,SI Combinational propagation from SDI to SDO when the device is not selected during a serial device read transaction. logic 1 logic 0 VIH,SI 80% SDI VREF,RSL 20% VIL,SI logic 1 Version 1.0 Jan. 2005 Page 67 K4Y5416(/08/04)4UF XDR DRAM Package Description Package Parasitic Summary Table19 summarizes inductance, capacitance, and resistance values associated with each pin group for the memory component. Most of the parameters have maximum values only, however some have both maximum and minimum values. The first group of parameters are for the CFM/CFMN clock pair pins. They include inductance, capacitance, and resistance values. The second group of parameters are for the RQ request pins. They include inductance, mutual inductance, capacitance, and resistance values. There are also limits on the spread in inductance and capacitance values allowed in any one memory component. The third group of parameters are specific to the DQ data pins and include inductance, mutual inductance, capacitance, and resistance values. There are limits on the spread in inductance and capacitance values allowed in any one memory component. The fourth group of parameters are for the serial interface pins. They include inductance and capacitance values. Version 1.0 Jan. 2005 Page 68 K4Y5416(/08/04)4UF XDR DRAM Table 19 : Package RSL Parasitic Summary Symbol Parameter and Other Conditions Minimum Maximum Units LVTERM VTERM pin - effective input inductance per four bits - 2.2 nH LI ,CFM CFM/CFMN pins - effective input capacianceb -2400 -3200 -4000 - tbd 5.0 tbd nH CI ,CFM CFM/CFMN pins - effective input capacianceb -2400 -3200 -4000 1.8 - tbd 2.4 tbd pF RI ,CFM CFM/CFMN pins - effective input resistance 4 15 LI ,RQ RSL RQ pins - effective input inductanceb -2400 -3200 -4000 - tbd 5.0 tbd nH CI ,RQ RSL RQ pins - effective input capacitanceb -2400 -3200 -4000 tbd 1.8 tbd tbd 2.4 tbd pF RI ,RQ RSL RQ pins - effective input resistance 4 15 L12,RQ Mutual inductance between adjacent RSL RQ signals - 0.6 nH LI,RQ Difference in LI,RQ between any RSL RQ pins of a single device - 1.8 nH CI,RQ Difference in CI between CFM/CFMN average and RSL RQ pins of single device -0.06 +0.06 pF ZPKG,DQ -2400 DRSL DQ pins - package differential impednce note - package trace length should be less than 10mm long. -3200 -4000 tbd 70 tbd tbd 130 tbd CI ,DQ DRSL DQ pins - effective input capacitancea -2400 -3200 -4000 - tbd 2.0 tbd pF CI,DQ -2400 Difference in CI between DQi and DQNi of each DRSL paira -3200 -4000 - tbd 0.06 tbd pF RI ,DQ DRSL DQ pins - effective input resistance 4 25 LI ,SI Serial Interface effective input inductance - 8.0 nH CI ,SI Serial Interface effective input capacitance RST, SCK, CMD SDI,SDO 1.7 - 3.0 7.0 pF pF a. This is the effective die input capacitance, and does not include package capacitance. b. CFM/RQ/SI should include package capacitance/Impedance, only DQ deos not include pacage capacitance. This value is a combination of the device I/O circuitry and package capacitance&inductance Version 1.0 Jan. 2005 Page 69 K4Y5416(/08/04)4UF XDR DRAM Figure 64 : Equivalent Circuits for Package Parasitic Pad RQ Pin CI,RQ LI,RQ L12,RQ RQ Pin L12,RQ RQ Pin RI,RQ GND Pin Pad Pad CI,DQ CI,DQ RI,DQ RI,DQ ZPKG,DQ/2 DQ Pin ZPKG,DQ/2 DQN Pin GND Pin Pad Pad CI,CFM CI,CFM RI,CFM RI,CFM ZPKG,CFM/2 CFM Pin ZPKG,CFM/2 CFMN Pin GND Pin Pad LI,SI SCK,CMD,RST Pin SDI,SDO Pin CI,SI GND Pin Version 1.0 Jan. 2005 Page 70 K4Y5416(/08/04)4UF XDR DRAM Package Mechanical Drawing Figure65 illustrates the XDR DRAM device package and Table20 summarizes the mechanical parameters for that package. Table 20 : XDR DRAM Package Mechanical Parameters Symbol Parameter Min Max Unit Min Max e1 Ball pitch (x-axis) 1.27 1.27 mm Symbol Parameter E Package total thickness 0.93 1.13 Unit mm e2 Ball pitch (y-axis) 0.80 0.80 mm E1 Ball height 0.30 0.40 mm A Package body length 13.9 14.1 mm d Ball diameter 0.40 0.50 mm D Package body width 14.4 14.6 mm Figure 65 : XDR DRAM Package Mechanical Drawing (Bottom View) E1 1 E1 2 3 4 1 5 2 6 3 7 4 8 5 9 6 10 7 11 12 13 14 15 16 A16 P L K J H G Top F E D C Top B A d N M L e1 K d J H D G e1 F E D C Bottom D B A Bottom e2 E E e2 A A Bottom Bottom Top Version 1.0 Jan. 2005 Page 71 K4Y5416(/08/04)4UF XDR DRAM Table of Contents Overview Features Key Timing Parameters/Part Numbers General Description Pinouts and Definitions Pin Description Block Diagram Request Packets Request Packet Formats Request Field Encoding Request Field Interaction Request Interaction Cases Dynamic Request Scheduling Memory Operations Write Transactions Read Transactions Interleaved Transactions Read/Write Interaction Propagation Delay Register Operations Serial Transactions Serial Write Transaction Serial Read Transaction Register Summary Maintenance Operations Refresh Transactions Interleaved Refresh Transactions Calibration Transactions Power State Management Initialization XDR DRAM Initialization Overview XDR DRAM Pattern Load with WDSL Reg Special Feature Description Dynamic Width Control Write Masking Multiple Bank Sets and the ERAW Feature Simultaneous Activation Simultaneous Precharge Operating Conditions Electrical Conditions Timing Conditions Operating Characteristics Electrical Characteristics Supply Current Profile Timing Characteristics Timing Parameters Receive/Transmit Timing Clocking RSL RQ Receive Timing DRSL DQ Receive Timing DRSL DQ Transmit Timing Serial Interface Receive Timing Serial Interface Transmit Timing Package Description Package Parasitic Summary Package Mechanical Drawing 1 1 2 3 4 5 6 8 8 10 12 13 18 20 20 22 24 26 27 29 29 29 29 31 37 37 37 39 40 42 43 43 46 46 48 50 52 53 54 54 55 56 56 57 58 59 60 60 61 62 64 66 67 68 68 71 Version 1.0 Jan. 2005 Page 72 K4Y5416(/08/04)4UF XDR DRAM Copyright (c) Jan. 2005, Samsung Electronics. All rights reserved. Rambus and Rambus logo are trademarks or registered trademarks of Rambus Inc. XDR is trademark of Rambus Inc. This document contains advanced information that is subject to change by Samsung Electronics without notice Document Version 0.1ver. Samsung Electronics Co. Ltd. San #16 Banwol-ri, Taean-Eup Hwasung-City, Gyeonggi-Do, KOREA Telephone: 82-31-208-6369 Fax: 82-31-208-6799 http://www.intl.samsungsemi.com Version 1.0 Jan. 2005 Page 73