Validation Study of LPDDR2 SDRAM Based on TD-LTE Baseband chip

来源 :无线互联科技 | 被引量 : 0次 | 上传用户:wxthaoa
下载到本地 , 更方便阅读
声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架
论文部分内容阅读
  Abstract:In recent years, as China has finished the updating of the fourth generation of network, for guaranteeing the information security of the state, communication chip with complete independent intellectual property right must be possessed to support the advancement of such project. TD-LTE Baseband Chip is a super-large-scale integrated circuit designed basing on SOC, which needs to carry out coding, etc to the transmitted baseband signal, or carry out decoding, etc to the received baseband signal. LPDDR2 SDRAM is used in the chip design process due to its low power dissipation, high capacity and high reliability. As PHY in the controller architecture of LPDDR2 SDRAM adopts hard core design, it cannot be achieved in Virtex-7 2000T prototype verification platform. This design mainly builds on such prototype verification platform to propose the verification scheme of LPDDR2 SDRAM controller in TD_LTE baseband chip, so as to guarantee that prototype verification in FPGA can be carried out by TD_LTE baseband system, and meanwhile high capacity storage space can be provided to the system.
  Key words:LPDDR2 SDRAM;Chip Verification;TD-LTE Baseband Chip;Prototype Verification
  1 Introduction
  With the completion of fourth-generation network update in China, there must be the communication chip with completely independent intellectual property right to promote the project, so that the national information security will be safe. Based on the SOC design, TD-LTE baseband chip performs a coding process on baseband signal to be transmitted or performs a decoding process on baseband signal received. Specially, compile the audio signal to baseband code to be transmitted when sending while interpret the baseband code received into audio signal when receiving. Meanwhile, it also performs a coding process on address information like telephone numbers and websites and texts of SMS and websites. Thus, TD-LTE baseband chip is the essential component in the establishment of fourth-generation network.
  This paper introduces SOC architecture design of TD-LTE baseband chip and validates the whole scheme based on FPGA prototype. Besides, it focuses on LPDDR2 SDRAM authentication scheme based on verification process of FPGA prototype depending on reliability and stability and analyze the implantation and result of the scheme during the process. It turns out that the design could meet the performance index of system.
  2 System Architecture of TD-LTE Baseband Chip   2.1 Overall Architecture of Chip
  Based on the VLSI of SOC design, TD-LTE baseband chip performs a coding process on baseband signal to be transmitted or performs a decoding process on baseband signal received. Mainly includes the following parts:
  The OpenRisc1200 processor: the RISC (Reduced Instruction Set Computer) processor with open source code based on GPL protocol uses the architecture of free open 32/64bit RISC/DSP. The processor completes the transport-level description of RISC/DSP architecture register through description of Verilog HDL language by hardware. With mature technology and good performance, it is generalized that the performance of it is between ARM7 to ARM8 which is applicable to embedded devices. Attaching with the powerful chain-tool such as tool for software development, CPU simulation model, operating system and function library of software application, it also has the perfect standardized protocol interface to speed up the integrated efficiency.
  Direct Memory Access: Realize the communication between hardware devices in various speeds with no relying on large amount pf interrupting loading of CPU. It can copy the data from a section of address space to another section of address space. In the baseband chip, DMA moves mobile external memory data to faster memory region of the chip. In the system, it can speed up mobile efficiency of data.
  Shared memory: Large capacity storage unit can be visited by several processors when there are several processors in the system.
  The system of chip also includes DSP data processing module, radio frequency transmission and receiving module, uplink and downlink processing module, Clock generating control module and universal interface module. Show as follow.
  2.2 Architecture in the system of LPDDR2
  In the chip system, LPDDR SDRAM module connect to the system with 64 bit AXI bus among which the ddr2_ctr_subsystem of LPDDR2 system contains AXI conversion module, AXI protocol to AHB protocol module and AHB module while the cadenc_mc_asic contains control module of LPDDR SDRAM and physical layer switching module. And mobile_ddr2 is the storage particle simulation model of LPDDR2 SDRAM. It can simulate the real running of LPDDR2 SDRAM control system.
  3 Architecture of LPDDR2 and Verified Design
  D-LTE baseband chip system need FPGA prototype verification on the verification platform, but PHY of LPDDR2 SDRAM control system is hard-core which cannot be verified on the FPGA platform and there are available so-dimm slot based on DDR2on the development board. So replace DDR2 control system with LPDDR2 system to provide large capacity storage space for the system.   3.1 Design of DDR2 SDRAM Controller
  There is the need to design DDR2 SDRAM controller with the same data bit, address bit and BANK quantity of LPDDR2 SDRAM to provide storage space of the system to make sure that AXI interface of DDR2 SDRAM controller and AXI interface of LPDDR2 SDRAM controller matches. Structure of control system, see in picture 3.
  The Infrastructure module uses to generate and distribute clock of all the modules in control system. Besides, all the reset signals are controlled by DCM.
  There is the phase locked loop internal. With the reference signal of external input, control the frequency and phase of oscillator signal in internal loop. The PLL could provide accurate comprehensive clock, reduce shake and realize the function of filter. The module also includes Delay calibration module which calibrates the delay of DQS to DQ, guarantee the center position of DQS in DQ valid window to lock and store DQ.
  In Iodelay_ctrl module, the IDELAYCTRL elements of Virtex-7 FPGA have been instantiated. IDELAYCTRL element calibrates the delay in its range continuously to reduce the impact of design leading from deviation of temperature and voltage. There is the need of 200MHz clock input in its regular work.
  clk_ibuf module initializs the input of clock buffer to keep the good characteristics of the clock. memc_ui_top_axi module is the main part of the verification. It contains two parts which are controlle (with AXI4 interface) module and PHY module. Explain them respectively as follows:
  Controlle (with AXI4 interface) module is the main controller of Virtex-7 DDR2 SDRAM controller reference design. It generates all the control signals for DDR2 storage interface and user interface. All the command and control signals are generated based on input burst length and CAS delay.
  PHY module is the physical layer module for the verification. Physicallayer contains IOBs and some elements used for reading and writing the double data rate signal in the memory such as IDDR2 and ODDR2. The module also includes IODELAY element of Virtex-7FPGA. Combine the module with DDR2 IP core control of ES05.
  3.2 Establishment and Simulation of DDR2 SDRAM Simulation Model System
  3.2.1 Establishment of Simulation Model System
  After the completion of above DDR2 controller, there is the need to verify whether DDR2 SDRAM controller can be used accurately with simulation of the model. The testing system mainly contains three parts which are testing excitation, DDR2 SDRAM controller and DDR2 SDRAM simulation model.   3.2.1.1 Testing Excitation
  Testing excitation module mainly includes Traffic Generator module and AXI Wapper module. Traffic Generator module is the system testing program. The module stores information into corresponding FIFO which are user’s data, demand and address information generating from lfsr. When writing operation, the data of FIFOs is read out and sent to physical layer to store in memory. When reading operation, data in memory is read out and wrote in FIFOs. AXI Wapper module is a module for reading and writing transmission conversion of data by AXI4 protocol. With the module, signal, address, controlling information and data does AXI4 protocol transmission.
  3.2.1.2 DDR2 Simulation Model
  The DDR2 SDRAM simulation model is mainly used in the replacement of the real DDR2 SDRAM. During the simulation, the written data can be read out to compare to verify whether the design of DDR2 SDRAM controller is accurate.
  3.2.2 Simulation and Result Analysis
  Use VCS-verilog compiled simulator of SYNOPSYS Company to simulate with the advantage of fast speed of simulation and multiple call methods.
  Start the signal of clock and remain stable of the signal, then drive CKE up. After a moment, issue demand of precharge to clusters and cancel the demand after another moment. Initialize the system, drive up CKE. With the execution of the precharge demand, execute external model register load instruction after 1950ns and configure EMR with initialization of EMR1 and EMR2. Execute the model register load instruction and configure memory parameter such as BL, ODT configuration and CAS delay. Execute PCH instruction and reprecharge all the bank.. Execute the ARF instruction for twice. Execute LMR instruction and set operation parameter. Execute ELMR instruction and set impedance of OCD as windows default by EMR.
  init_calib_complete=1 refers to the completion of DDR2 initialization as picture 5 shows and the normal read and write operations can be done. Picture 6 and picture 7 refers to the read and write operation with 16’h1118 as the initial address and 8 as BL. The reading data and writing data of the same address are similar which means DDR2 SDRAM control system can do regular work.
  4 Integration and Implementation of System
  After simulating with VCS software, the following summarizes the integration of system with synplify_premie software and implementation of system with vivado software to guarantee the regular work on the verification platform.   4.1 Process Description
  4.1.1 Integration
  Integration refers to the simple device description of transformation from higher-level description logic to lower-level description logic. In general, integration is based on RTL code. That is to say, convert the transmission-level model, algorithm, behavior and function description of memory into corresponding netlist file of FPGA/CPLD basic structure which means compiling design input to logical connective including basic logic units of and, or, non, ram and memory.
  4.1.2 Implementation and wiring
  The implementation could only wiring on the integration network table. Configure the integration network table to specific FPGA development platform and select the optimal speed and area. Connect each component correctly and reasonably with connection resources and logic resources of internal chip to implement corresponding logic function.
  4.1.3 Programming and Debugging of Chip
  Generate data stream and download to FPGA verification platform to debug with third-party tools. The embedded online logic analysis is the major debugging aid. It can grasp large number of real-time signals with occupation of little logic resources to determine whether the system works regularly.
  4.2 Integration, Implementation and Platform Debugging
  The integration tool is synplify_premier with the advantage of turning the module that cannot be integrated such as divider module into network table which are available on FPGA platform in ASIC design, and then integrate the network table to provide convenience for integration system. Analysis report on the share of DDR2 SDRAM control system in resources after integration shows as follow.
  Register bits not including I/Os:14578
  Latch bits not including I/Os:14
  RAM/ROM usage summary
  Simple Dual Port Rams (RAM32M):141
  Global Clock Buffers:2 of 32
  Mapping Summary:
  Total LUTs:9197
  Distribution of All Consumed LUTs=SRL+RAM+ LUT+ LUT2+LUT3+LUT4+LUT5+LUT6+LUT6_2-HLUTNM/2
  Distribution of All Consumed Luts 9197=655+564+ 423+1104+1268+1089+1523+4019+21-2938/2
  Number of unique control sets:300
  From the analysis table, we can find that 9197 LUT resources have been used in system circuit while the number of dual-port RAM used is 141. Through the resource quantity of each part, we can conclude the unreasonable use of resource. so that we can optimize them with different emphases to reduce the use of resource and save the logic resources in platform developing.   Because the timing analysis report is too much and too long, here is only small part of it shows as follow. Among which, slack=0.139ns refers to the delay of clock period constraint is 0.139ns more than real delay after the wiring. The system works regularly if the constraint conditions are met.
  When the implementation is done, there is the need of vivado software to translate the integrated network table into the bottom module and hardware primitive of the selected device and reflect the design to the structure of device to wiring. So that, we can implement the design circuit on Virex-7 2000T chip generated from Xilinx. After implementation, the circuit effect and the share in logic resources on the chip show in the following picture.
  After integration and implementation, circuit generates this picture. There are no red connecting lines in the picture which means there are no circuits and devices that have not be wiringed. It refers to that the integration and implementation has done and can enter the process of generating bit file.
  Debug the Verification Platform of DDR2 SDRAM Controller
  The debugging platform, taking Virtex-7 2000T FPGA, XC7V2000TFLG1925 of Xilinx Company as the core, supports design verification of more than 40million logic gate. With So-dimm slot based on DDR2, it can meet the debugging condition. The software integrates with synplify_premier, configures high-speed clock and downloads files with TAI software of S2C company and capture real-time waveform with Debug software in vivado2012.04 to debug.
  Testing structure removes modules that cannot integrate and implement based on simulation model and DDR2 SDRAM controller simulation model to restrict pins. At that time, put pin to high-speed clock logic resource block correspondingly, or it can not do regular work. Integrate and wiring the whole system and then generate the .bit and .ltx file that can be downloaded and debugged with attention that there should be corresponding debugging model added before. Replace DDR2 module that cannot be integrated in the simulation system with DDR2 memory chip of simulation model.
  4.3 simulation and result analysis
  Download with DEBUG of vivado14.4 version software and debug and analyze with FPGA verification platform, and then get the real-time signal graph of AXI interface, shows as follow.
  To verify better, set BL to something that can be generated from FLSR. From the picture, we can find that writing 76 digits and 198 digits when BL is 76 or 198, the digits read out is the same. That means the signal of the verification system is of completeness, the affect that electromagnetic interference have on regular work is acceptable and DDR2 SDRAM controller can do regular work and can be integrated to baseban chip system.
其他文献
摘 要:基于web2.0的虚拟学习社区,完成了具有交互、知识管理、协作共享的虚拟学习社区的建构,从而为广大学生的学习提供了良好的环境,为提高学习效率奠定了坚实的基础。  关键词:web2.0;虚拟学习社区;平台  基于web2.0的虚拟学习社区的应用发展呈燎原之势,它为学习者提供了前所未有的网络体验空间,也为教育和教学提供了更富有创造性的良性环境。  1 基本的设计思想  ⑴共同的规范和准则。社区
期刊
摘 要:文章主要介绍了工业化贮能的三种主要技术,即:电磁贮能、化学贮能和物理贮能。从六个方面比较了工业化储能的技术特征,对典型的储能方式进行了比较。分析了飞轮储能的技术优势,给出了飞轮储能的应用领域。  关键词:工业贮能;飞轮储能;电磁能;电化学能  1 贮能技术分类及特性  ⑴电磁贮能。电磁贮能按照介质不同可以分为两类:超导材料和电容器贮能,即分别应用这两种材质存储磁场能,输送功率的时候不需要任
期刊
摘 要:本文主要以综述的方式简单的介绍了一种用微溶玻璃技术制作新一代压力传感器的原理、结构以及补偿电路,最后总结了微溶压力传感器的特点及优越性,可剔除了用传统工艺方法生产传感器带来的弊端,理论上可以完全替代陶瓷、应变片、溅射薄膜及充油芯体等技术。  关键词:微溶技术;压力传感器;应变片;芯体  Abstract:This article mainly reviews the way of simp
期刊
摘 要:近一段时间以来,一项名为“GSM手机监听”的在互联网上疯狂流传。利用该技术,可以在一定范围内监听GSM基站发出的信号,从而嗅探连接该基站的手机用户发送和接收到的短信甚至是通话信息,具有很强的破坏性。由于网络上大量“技术教程”的出现及相关设备所需成本较低,该技术已被大量滥用,本文介绍了GSM监听的技术细节,通过实验证明了该技术的可行性与破坏力,并根据该技术在网络上的流传情况对其威胁现状做出评
期刊
摘 要:幼儿园的科学活动是幼儿进入小学进行自然科学学习的基础和前奏。在我从事幼儿教育多年的教学实践中,发现当前幼儿的科学活动还存在一些问题,尤其是在准备活动上表现出明显的不足。  关键词:科学活动;物质准备;知识储备;班级管理  1 科学活动尚不完美  纵观目前的幼儿园科学活动,不少存在枯燥与乏味的现象,甚至在科学活动中呈现出小学化的现象。我们目前的幼儿科学活动是不完美的,既没有形成完善的课程体系
期刊
Abstract:Nowadays, the modeling of systems may be quite large, even up to tens of thousands orders. In spite of the increasing computational powers, direct simulation of these large-scale systems may
期刊
摘 要:研究在动态背景下对视频图像序列进行运动目标的检测。主要包括三个步骤,分别是运动估计,运动补偿和目标检测。在运动估计中采用的主要是基于特征点匹配算法。这种算法与传统的块匹配算法最大的好处在于它的数据量少,计算简单迅速而且图像的匹配可靠性更高。最后用计算机视觉库OpenCV进行实现。  关键词:运动目标检测;动态背景;特征点匹配;OpenCV  Moving object detection
期刊
摘 要:通过对基于IPv6的无线自组网的介绍和研究,将传感网和承载网共同组网,可以更好的发挥各自的优势,同时弥补各自的不足。这种网络搭配具有很广泛的应用场景。  关键词:IPv6,自组织无线网络;物联网  The Research on the Application of  the IPv6 based Smart Ad-hoc Network System  Fan Xiaofeng1 Liu
期刊
摘 要:无线定位是车联网中各应用的基础,对实时性和精度要求更高。基于此,本文提出了一个在WAVE系统下基于混合高斯模型的位置指纹定位算法。该算法由离线训练和在线定位两部分组成,采用GMM对离线训练阶段的指纹数据库进行聚类建模处理,并在定位阶段对结果采用多点均值方法,不仅降到了系统定位误差,也减少了定位阶段的运算量,提高了定位的实时性。实验结果表明,该算法有较好的定位精度和实时性,为WAVE环境下的
期刊
摘 要:移动互联网的发展给传统互联网带来了巨大的冲击,导致了互联网商业模式的转变,同时也造成了大量传统互联网用户的流失。本文以典型的社交网络服务人人网为例,分析移动互联网对其所带来的用户流失,移动SNS是否吸引了新的用户资源,从用户的角度探究移动社交网络的发展趋势。  关键词:移动互联网;人人网;用户流失;社交网络  Abstract:The development of Mobile Inter
期刊