External Memory Interface Handbook Volume 3: Reference Material: For UniPHY-based Device Families

Download PDF
ID 683841
Date 3/06/2023
Public
Document Table of Contents
Document Table of Contents x
1. Functional Description—UniPHY 2. Functional Description— Intel® MAX® 10 EMIF IP 3. Functional Description—Hard Memory Interface 4. Functional Description—HPS Memory Controller 5. Functional Description—HPC II Controller 6. Functional Description—QDR II Controller 7. Functional Description—RLDRAM II Controller 8. Functional Description—RLDRAM 3 PHY-Only IP 9. Functional Description—Example Designs 10. Introduction to UniPHY IP 11. Latency for UniPHY IP 12. Timing Diagrams for UniPHY IP 13. External Memory Interface Debug Toolkit 14. Upgrading to UniPHY-based Controllers from ALTMEMPHY-based Controllers
1. Functional Description—UniPHY x
1.1. I/O Pads 1.2. Reset and Clock Generation 1.3. Dedicated Clock Networks 1.4. Address and Command Datapath 1.5. Write Datapath 1.6. Read Datapath 1.7. Sequencer 1.8. Shadow Registers 1.9. UniPHY Interfaces 1.10. UniPHY Signals 1.11. PHY-to-Controller Interfaces 1.12. Using a Custom Controller 1.13. AFI 3.0 Specification 1.14. Register Maps 1.15. Ping Pong PHY 1.16. Efficiency Monitor and Protocol Checker 1.17. UniPHY Calibration Stages 1.18. Document Revision History
1.5. Write Datapath x
1.5.1. Leveling Circuitry
1.7. Sequencer x
1.7.1. Nios® II-Based Sequencer 1.7.2. RTL-based Sequencer
1.7.1. Nios® II-Based Sequencer x
1.7.1.1. Nios® II-based Sequencer Function 1.7.1.2. Nios® II-based Sequencer Architecture 1.7.1.3. Nios® II-based Sequencer SCC Manager 1.7.1.4. Nios® II-based Sequencer RW Manager 1.7.1.5. Nios® II-based Sequencer PHY Manager 1.7.1.6. Nios® II-based Sequencer Data Manager 1.7.1.7. Nios® II-based Sequencer Tracking Manager 1.7.1.8. Nios® II-based Sequencer Processor 1.7.1.9. Nios® II-based Sequencer Calibration and Diagnostics
1.8. Shadow Registers x
1.8.1. Shadow Registers Operation
1.9. UniPHY Interfaces x
1.9.1. The DLL and PLL Sharing Interface 1.9.2. The OCT Sharing Interface
1.9.1. The DLL and PLL Sharing Interface x
1.9.1.1. Sharing PLLs or DLLs 1.9.1.2. About PLL Simulation
1.9.2. The OCT Sharing Interface x
1.9.2.1. Modifying the Pin Assignment Script for QDR II and RLDRAM II
1.13. AFI 3.0 Specification x
1.13.1. Bus Width and AFI Ratio 1.13.2. AFI Parameters 1.13.3. AFI Signals
1.13.3. AFI Signals x
1.13.3.1. AFI Clock and Reset Signals 1.13.3.2. AFI Address and Command Signals 1.13.3.3. AFI Write Data Signals 1.13.3.4. AFI Read Data Signals 1.13.3.5. AFI Calibration Status Signals 1.13.3.6. AFI Tracking Management Signals
1.14. Register Maps x
1.14.1. UniPHY Register Map 1.14.2. Controller Register Map
1.15. Ping Pong PHY x
1.15.1. Ping Pong PHY Feature Description 1.15.2. Ping Pong PHY Architecture 1.15.3. Ping Pong PHY Operation
1.15.2. Ping Pong PHY Architecture x
1.15.2.1. Ping Pong Gasket 1.15.2.2. Ping Pong PHY Calibration
1.16. Efficiency Monitor and Protocol Checker x
1.16.1. Efficiency Monitor 1.16.2. Protocol Checker 1.16.3. Read Latency Counter 1.16.4. Using the Efficiency Monitor and Protocol Checker 1.16.5. Avalon® CSR Slave and JTAG Memory Map
1.17. UniPHY Calibration Stages x
1.17.1. Calibration Overview 1.17.2. Calibration Stages 1.17.3. Memory Initialization 1.17.4. Stage 1: Read Calibration Part One—DQS Enable Calibration and DQ/DQS Centering 1.17.5. Stage 2: Write Calibration Part One 1.17.6. Stage 3: Write Calibration Part Two—DQ/DQS Centering 1.17.7. Stage 4: Read Calibration Part Two—Read Latency Minimization 1.17.8. Calibration Signals 1.17.9. Calibration Time
1.17.4. Stage 1: Read Calibration Part One—DQS Enable Calibration and DQ/DQS Centering x
1.17.4.1. Guaranteed Write 1.17.4.2. DQS Enable Calibration 1.17.4.3. Centering DQ/DQS
2. Functional Description— Intel® MAX® 10 EMIF IP x
2.1. Intel® MAX® 10 EMIF Overview 2.2. External Memory Protocol Support 2.3. Intel® MAX® 10 Memory Controller 2.4. Intel® MAX® 10 Low Power Feature 2.5. Intel® MAX® 10 Memory PHY 2.6. Calibration 2.7. Sequencer Debug Information 2.8. Register Maps 2.9. Document Revision History
2.5. Intel® MAX® 10 Memory PHY x
2.5.1. Supported Topologies 2.5.2. Read Datapath 2.5.3. Write Datapath 2.5.4. Address and Command Datapath 2.5.5. Sequencer
2.6. Calibration x
2.6.1. Read Calibration 2.6.2. Write Calibration
3. Functional Description—Hard Memory Interface x
3.1. Multi-Port Front End (MPFE) 3.2. Multi-port Scheduling 3.3. MPFE Signal Descriptions 3.4. Hard Memory Controller 3.5. Hard PHY 3.6. Document Revision History
3.2. Multi-port Scheduling x
3.2.1. Port Scheduling 3.2.2. DRAM Burst Scheduling 3.2.3. DRAM Power Saving Modes
3.4. Hard Memory Controller x
3.4.1. Clocking 3.4.2. Reset 3.4.3. DRAM Interface 3.4.4. ECC 3.4.5. Bonding of Memory Controllers Data Return Bonding FIFO Ready Bonding Latency Impact Bonding Controller Usage Bonding Configurations and Parameter Requirements
3.5. Hard PHY x
3.5.1. Interconnections 3.5.2. Clock Domains 3.5.3. Hard Sequencer 3.5.4. MPFE Setup Guidelines 3.5.5. Soft Memory Interface to Hard Memory Interface Migration Guidelines 3.5.6. Bonding Interface Guidelines
4. Functional Description—HPS Memory Controller x
4.1. Features of the SDRAM Controller Subsystem 4.2. SDRAM Controller Subsystem Block Diagram 4.3. SDRAM Controller Memory Options 4.4. SDRAM Controller Subsystem Interfaces 4.5. Memory Controller Architecture 4.6. Functional Description of the SDRAM Controller Subsystem 4.7. SDRAM Power Management 4.8. DDR PHY 4.9. Clocks 4.10. Resets 4.11. Port Mappings 4.12. Initialization 4.13. SDRAM Controller Subsystem Programming Model 4.14. Debugging HPS SDRAM in the Preloader 4.15. SDRAM Controller Address Map and Register Definitions 4.16. Document Revision History
4.4. SDRAM Controller Subsystem Interfaces x
4.4.1. MPU Subsystem Interface 4.4.2. L3 Interconnect Interface 4.4.3. CSR Interface 4.4.4. FPGA-to-HPS SDRAM Interface
4.5. Memory Controller Architecture x
4.5.1. Multi-Port Front End 4.5.2. Single-Port Controller
4.5.2. Single-Port Controller x
4.5.2.1. Command Generator 4.5.2.2. Timer Bank Pool 4.5.2.3. Arbiter 4.5.2.4. Rank Timer 4.5.2.5. Write Data Buffer 4.5.2.6. ECC Block 4.5.2.7. AFI Interface 4.5.2.8. CSR Interface
4.6. Functional Description of the SDRAM Controller Subsystem x
4.6.1. MPFE Operation Ordering 4.6.2. MPFE Multi-Port Arbitration 4.6.3. MPFE SDRAM Burst Scheduling 4.6.4. Single-Port Controller Operation
4.6.4. Single-Port Controller Operation x
4.6.4.1. Command and Data Reordering 4.6.4.2. Bank Policy 4.6.4.3. Write Combining 4.6.4.4. Burst Length Support 4.6.4.5. ECC 4.6.4.6. Interleaving Options 4.6.4.7. AXI-Exclusive Support 4.6.4.8. Memory Protection 4.6.4.9. Example of Configuration for TrustZone
4.6.4.5. ECC x
4.6.4.5.1. Byte Writes 4.6.4.5.2. ECC Write Backs 4.6.4.5.3. User Notification of ECC Errors
4.10. Resets x
4.10.1. Taking the SDRAM Controller Subsystem Out of Reset
4.12. Initialization x
4.12.1. FPGA-to-SDRAM Protocol Details
4.12.1. FPGA-to-SDRAM Protocol Details x
4.12.1.1. Avalon-MM Bidirectional Port 4.12.1.2. Avalon-MM Write-Only Port 4.12.1.3. Avalon-MM Read Port 4.12.1.4. AXI Port
4.13. SDRAM Controller Subsystem Programming Model x
4.13.1. HPS Memory Interface Architecture 4.13.2. HPS Memory Interface Configuration 4.13.3. HPS Memory Interface Simulation 4.13.4. Generating a Preloader Image for HPS with EMIF
4.13.4. Generating a Preloader Image for HPS with EMIF x
4.13.4.1. Creating a Project in Platform Designer (Standard) 4.13.4.2. Creating a Top-Level File and Adding Constraints
4.14. Debugging HPS SDRAM in the Preloader x
4.14.1. Enabling UART or Semihosting Printout 4.14.2. Enabling Simple Memory Test 4.14.3. Enabling the Debug Report 4.14.4. Writing a Predefined Data Pattern to SDRAM in the Preloader
4.14.3. Enabling the Debug Report x
4.14.3.1. Analysis of Debug Report
5. Functional Description—HPC II Controller x
5.1. HPC II Memory Interface Architecture 5.2. HPC II Memory Controller Architecture 5.3. HPC II Controller Features 5.4. External Interfaces 5.5. Top-Level Signals Description 5.6. Sequence of Operations 5.7. Document Revision History
5.2. HPC II Memory Controller Architecture x
5.2.1. Backpressure Support 5.2.2. Command Generator 5.2.3. Timing Bank Pool 5.2.4. Arbiter 5.2.5. Rank Timer 5.2.6. Read Data Buffer and Write Data Buffer 5.2.7. ECC Block 5.2.8. AFI and CSR Interfaces
5.3. HPC II Controller Features x
5.3.1. Data Reordering 5.3.2. Pre-emptive Bank Management 5.3.3. Quasi-1T and Quasi-2T 5.3.4. User Autoprecharge Commands 5.3.5. Address and Command Decoding Logic 5.3.6. Low-Power Logic 5.3.7. ODT Generation Logic 5.3.8. Burst Merging 5.3.9. ECC
5.3.9. ECC x
5.3.9.1. Partial Writes 5.3.9.2. Partial Bursts
5.4. External Interfaces x
5.4.1. Clock and Reset Interface 5.4.2. Avalon® -ST Data Slave Interface 5.4.3. AXI Data Slave Interface 5.4.4. Controller-PHY Interface 5.4.5. Memory Side-Band Signals 5.4.6. Controller External Interfaces
5.4.3. AXI Data Slave Interface x
5.4.3.1. Enabling the AXI Interface 5.4.3.2. AXI Interface Parameters 5.4.3.3. AXI Interface Ports
5.5. Top-Level Signals Description x
5.5.1. Clock and Reset Signals 5.5.2. Local Interface Signals 5.5.3. Controller Interface Signals 5.5.4. CSR Interface Signals 5.5.5. Soft Controller Register Map 5.5.6. Hard Controller Register Map
6. Functional Description—QDR II Controller x
6.1. Block Description 6.2. Avalon® -MM and Memory Data Width 6.3. Signal Descriptions 6.4. Document Revision History
6.1. Block Description x
6.1.1. Avalon® -MM Slave Read and Write Interfaces 6.1.2. Command Issuing FSM 6.1.3. AFI
7. Functional Description—RLDRAM II Controller x
7.1. Block Description 7.2. User-Controlled Features 7.3. Avalon® -MM and Memory Data Width 7.4. Signal Descriptions 7.5. Document Revision History
7.1. Block Description x
7.1.1. Avalon® -MM Slave Interface 7.1.2. Write Data FIFO Buffer 7.1.3. Command Issuing FSM 7.1.4. Refresh Timer 7.1.5. Timer Module 7.1.6. AFI
7.2. User-Controlled Features x
7.2.1. Error Detection Parity 7.2.2. User-Controlled Refresh
8. Functional Description—RLDRAM 3 PHY-Only IP x
8.1. Block Description 8.2. Features 8.3. RLDRAM 3 AFI Protocol 8.4. Document Revision History
9. Functional Description—Example Designs x
9.1. UniPHY-Based Example Designs 9.2. Document Revision History
9.1. UniPHY-Based Example Designs x
9.1.1. Synthesis Example Design 9.1.2. Simulation Example Design 9.1.3. Traffic Generator and BIST Engine 9.1.4. Creating and Connecting the UniPHY Memory Interface and the Traffic Generator in Platform Designer
9.1.3. Traffic Generator and BIST Engine x
9.1.3.1. Read and Write Generation 9.1.3.2. Address and Burst Length Generation 9.1.3.3. Traffic Generator Signals 9.1.3.4. Traffic Generator Add-Ons 9.1.3.5. Traffic Generator Timeout Counter
9.1.4. Creating and Connecting the UniPHY Memory Interface and the Traffic Generator in Platform Designer x
9.1.4.1. Notes on Configuring UniPHY IP in Platform Designer
10. Introduction to UniPHY IP x
10.1. Release Information 10.2. Device Support Levels 10.3. Device Family and Protocol Support 10.4. UniPHY-Based External Memory Interface Features 10.5. System Requirements 10.6. Intel® FPGA IP Core Verification 10.7. Resource Utilization 10.8. Document Revision History
10.7. Resource Utilization x
10.7.1. DDR2, DDR3, and LPDDR2 Resource Utilization in Arria V Devices 10.7.2. DDR2 and DDR3 Resource Utilization in Arria II GZ Devices 10.7.3. DDR2 and DDR3 Resource Utilization in Stratix III Devices 10.7.4. DDR2 and DDR3 Resource Utilization in Stratix IV Devices 10.7.5. DDR2 and DDR3 Resource Utilization in Arria V GZ and Stratix V Devices 10.7.6. QDR II and QDR II+ Resource Utilization in Arria V Devices 10.7.7. QDR II and QDR II+ Resource Utilization in Arria II GX Devices 10.7.8. QDR II and QDR II+ Resource Utilization in Arria II GZ, Arria V GZ, Stratix III, Stratix IV, and Stratix V Devices 10.7.9. RLDRAM II Resource Utilization in Arria® V Devices 10.7.10. RLDRAM II Resource Utilization in Arria®  II GZ, Arria® V GZ, Stratix®  III, Stratix®  IV, and Stratix® V Devices
11. Latency for UniPHY IP x
11.1. DDR2 SDRAM LATENCY 11.2. DDR3 SDRAM LATENCY 11.3. LPDDR2 SDRAM LATENCY 11.4. QDR II and QDR II+ SRAM Latency 11.5. RLDRAM II Latency 11.6. RLDRAM 3 Latency 11.7. Variable Controller Latency 11.8. Document Revision History
12. Timing Diagrams for UniPHY IP x
12.1. DDR2 Timing Diagrams 12.2. DDR3 Timing Diagrams 12.3. QDR II and QDR II+ Timing Diagrams 12.4. RLDRAM II Timing Diagrams 12.5. LPDDR2 Timing Diagrams 12.6. RLDRAM 3 Timing Diagrams 12.7. Document Revision History
13. External Memory Interface Debug Toolkit x
13.1. User Interface 13.2. Setup and Use 13.3. Operational Considerations 13.4. Troubleshooting 13.5. Debug Report for Arria V and Cyclone V SoC Devices 13.6. On-Chip Debug Port for UniPHY-based EMIF IP 13.7. Example Tcl Script for Running the Legacy EMIF Debug Toolkit 13.8. Document Revision History
13.1. User Interface x
13.1.1. Communication 13.1.2. Calibration and Report Generation
13.2. Setup and Use x
13.2.1. General Workflow 13.2.2. Linking the Project to a Device 13.2.3. Establishing Communication to Connections 13.2.4. Selecting an Active Interface 13.2.5. Reports
13.5. Debug Report for Arria V and Cyclone V SoC Devices x
13.5.1. Enabling the Debug Report for Arria V and Cyclone V SoC Devices 13.5.2. Determining the Failing Calibration Stage for a Cyclone V or Arria V HPS SDRAM Controller
13.6. On-Chip Debug Port for UniPHY-based EMIF IP x
13.6.1. Access Protocol 13.6.2. Command Codes Reference 13.6.3. Header Files 13.6.4. Generating IP With the Debug Port 13.6.5. Example C Code for Accessing Debug Data
14. Upgrading to UniPHY-based Controllers from ALTMEMPHY-based Controllers x
14.1. Generating Equivalent Design 14.2. Replacing the ALTMEMPHY Datapath with UniPHY Datapath 14.3. Resolving Port Name Differences 14.4. Creating OCT Signals 14.5. Running Pin Assignments Script 14.6. Removing Obsolete Files 14.7. Simulating your Design 14.8. Document Revision History
1. Functional Description—UniPHY
2. Functional Description— Intel® MAX® 10 EMIF IP
3. Functional Description—Hard Memory Interface
4. Functional Description—HPS Memory Controller
5. Functional Description—HPC II Controller
6. Functional Description—QDR II Controller
7. Functional Description—RLDRAM II Controller
8. Functional Description—RLDRAM 3 PHY-Only IP
9. Functional Description—Example Designs
10. Introduction to UniPHY IP
11. Latency for UniPHY IP
12. Timing Diagrams for UniPHY IP
13. External Memory Interface Debug Toolkit
14. Upgrading to UniPHY-based Controllers from ALTMEMPHY-based Controllers

3.4.5. Bonding of Memory Controllers

Bonding is a feature that allows data to be split between two memory controllers, providing the ability to service bandwidth streams similar to a single 64-bit controller. Bonding works by dividing data buses in proportion to the memory widths, and always sending a transaction to both controllers. When signals are returned, bonding ensures that both sets of signals are returned identically.

Bonding can be applied to asymetric controllers, and allows controllers to have different memory clocks. Bonding does not attempt to synchronize the controllers. Bonding supports only one port. The Avalon® port width can be varied from 64-bit to 256-bit; 32-bit port width is not supported.

The following signals require bonding circuitry:

Read data return—This bonding allows read data from the two controllers to return with effectively one ready signal to the bus master that initiated the bus transaction.

Write ready—For Avalon® -MM, this is effectively bonding on the waitrequest signal.

Write acknowledge—Synchronization on returning the write completed signal.

For each of the above implementations, data is returned in order, hence the circuitry must match up for each valid cycle.

Bonded FIFO buffers must have identical FIFO numbers; that is, read FIFO 1 on controller 1 must be paired with Read FIFO 1 on controller 2.

Data Return Bonding

Long loop times can lead to communications problems when using bonded controllers. The following effects are possible when using bonded controllers:

  • If one memory controller completes its transaction and receives new data before the other controller, then the second controller can send data as soon as it arrives, and before the first controller acknowledges that the second controller has data.
  • If the first controller has a single word in its FIFO buffer and the second controller receives single-word transactions, the second controller must determine whether the second word is a valid signal or not.

To accommodate the above effects, the hard controller maintains two counters for each bonded pair of FIFO buffers and implements logic that monitors those counters to ensure that the bonded controllers receive the same data on the same cycle, and that they send the data out on the same cycle.

FIFO Ready

FIFO ready bonding is used for write command and write data buses. The implementation is similar to the data return bonding.

Bonding Latency Impact

Bonding has no latency impact on ports that are not bonded.

Bonding Controller Usage

Arria V and Cyclone V devices employ three shared bonding controllers to manage the read data return bonding, write acknowledge bonding, and command/write data ready bonding.

The three bonding controllers require three pairs of bonding I/Os, each based on a six port count; this means that a bonded hard memory controller requires 21 input signals and 21 output signals for its connection to the fabric, and another 21 input signals and 21 output signals to the paired hard memory controller.

Note: The hard processor system (HPS) hard memory controller cannot be bonded with another hard memory controller on the FPGA portion of the device.

Bonding Configurations and Parameter Requirements

Intel has verified hard memory controller bonding between two interfaces with the following configuration:

  • Same clock source
  • Same memory clock frequency
  • Same memory parameters and timings (except interface width)
  • Same controller settings.
  • Same port width in MPFE settings

Bonding supports only one port. The Avalon® port width can be varied from 64-bits to 256-bits; a 32-bit port width is not supported.

Level Two Title