MEMORY系列之“DDR with ECC”

ECC是“Error Checking and Correcting”的简写,中文名称是“错误检查和纠正“。ECC是一种能够实现“错误检查和纠正“的技术,ECC内存就是应用了这种技术的内存,一般多应用在服务器及工作站上,这将使整个电脑系统在工作时更趋于安全稳定。

带ECC功能的DDR连接如下图所示:

 

不带ECC的32bit数据带宽连接方式如下:

不带ECC的64bit数据带宽连接方式如下:

以下是带ECC的连接方式,主要分为以下几种:

1、Using x16 Component

 

 

When using a x16 component for ECC, the lower byte lane (DQ[7:0]) must beused for the ECC bits. DDR4 memory is capable of per-DRAM addressability(PDA), and this function is used during the DDR4 device initialization sequencefor VREFDQ calibration. PDA is enabled by DQ0, so the lower byte lane must beused and connected to the controller.

The unused upper byte lane should be terminated asfollows:

  • DQ[15:8] can be left floating (ODT will terminate to VDDQ)

  • UDM_n/UDBI_n should be terminated to VDDQ (allows DM or DBI to beenabled)

  • UDQS_t should be terminated to VDDQ

  • UDQS_c should be terminated to VSSQ

2、Using x8 Component

 

 

Using an additional x8 component for ECC results inslightly lower power and will use slightly less board space than a x16component. In addition, there is no unused byte lane to terminate. But, matching the address configuration ofthe x16 data components begins with a x8 ECC component of the same density asthe x16, with half of the bank groups disabled.

Half of the bank groups are disabled by tying BG1to VSSQ. This enables command/address parity. Because the unused BG1 input onthe x16 devices is treated as a zero, the parity generated by both the x16 andthe x8 devices will be the same.

 

3、Using DDR4 x16 DDP

 

The DDR4 standard includes a special configurationfor a x16 dual die package (DDP). This device contains two x8 die connected asa single rank. Although more expensive than two discrete x8 DDR4 packages, theDDP occupies considerably less board area. This makes it an excellent optionwhere higher density is needed, but board space is constrained. Because the DDPuses x8 die, BG1 is required on the x16 DDP package. (A second ZQ is alsorequired on the board; for suggestions on designing a board that accepts bothstandard x16 DDR4 and DDP x16 DDR4 devices, see TN-40-40: DDR4 Point-to-PointDesign Guide.)

When adding ECC to a data bus using x16 DDPdevices, the most economic choice is to use an additional discrete x8 componentfor the ECC device. No special connections are needed, as all devices use BG1,and all DQ signals (byte lanes) are used.

 

As the tables show, using x16 data components witha x8 component for ECC results in the lowest power and the least board area.Using a x16 component for ECC simplifies the BOM (and thus the procurementprocess) at the cost of small increases in power and board area, plus a slightadded complexity in terminating an unused byte lane.

The x16 DDP provides the same density and power asusing all x8 components, but at the same component area as the standard x16with the x8 ECC component. Thus, you get twice the density in the same boardarea.

  • 4
    点赞
  • 28
    收藏
    觉得还不错? 一键收藏
  • 5
    评论
• 重点内容 – 64KB 专用 RAM – 275-MHz, 500-MHz, 600-MHz, or 720-MHz – 仿真/调试 ARM® Cortex™-A8 32-位RISC 微控制器• JTAG • NEON™ SIMD 协处理器• 嵌入式跟踪模块 • 具有单错检测(奇偶校验)的32KB/32KB • 嵌入式跟踪缓冲器 L1 指令/数据高速缓存– 中断控制器(高达128 个中断请求) • 具有错误纠正码(ECC) 的256KB L2 高速缓• 片载存储器(共享L3 RAM) 存– 64 KB 通用片载存储器控制器(OCMC) RAM – 支持移动双倍速率同步动态随机存储器– 所有主机均可访问 (mDDR)(低功耗DDR (LPDDR))/DDR2/DDR3 – 支持快速唤醒保持 – 支持通用存储器(NAND,NOR,SRAM,等)支• 外部存储器接口(EMIF) 持高达16 位ECC – mDDR/DDR2/DDR3 控制器: – SGX530 3D 图形引擎 • mDDR: 200-MHz 时钟频率(400-MHz 数据速– LCD 控制器 率) – 可编程实时单元和工业用通信子系统(PRU- • DDR2: 266-MHz 时钟(532-MHz 数据速ICSS) 率) – 实时时钟(RTC) • DDR3: 303-MHz 时钟(606-MHz 数据速 – 最多2 个具有集成物理层的USB 2.0 高速OTG 率) 端口• 16-位数据总线 – 支持最多2 个端口的10/100/1000 以太网交换机• 1GB 全部可寻址空间 – 串口包括: • 支持1 x 16 或者2 x 8 存储器器件配置 • 2 个控制器局域网端口(CAN) • 支持快速唤醒保持 • 6 个UART,2 个McASPI,2 个McSPI,和3 个 I2C 端口 – 通用存储器控制器 (GPMC) • 具有多达7 芯片(NAND,NOR,复– 12 位逐次逼近寄存器(SAR) ADC 用NOR,SRAM等) 选择的灵活8/16-位异步 – 3 个32 位增强型捕捉模块(eCAP) 存储器接口 – 3 个增强型高分辨率PWM模块(eHRPWM) • 使用BCH 编码以支持4-位,8-位,或者16- – 加密硬件加速器(AES,SHA,PKA,RNG) 位ECC • 使用海明码(Hamming)以支持1-位ECC • MPU 子系统– 错位定位器模块(ELM) – 275-MHz,500-MHz,600-MHz,或者720-MHz • 与GPMC 联合使用,使用BCH 算法以定位 ARM® Cortex™-A8 32-位RISC 微处理器由校验多项式生成的数据错误的地址 – NEON™ SIMD 协处理器• 基于BCH 算法,每512 字节块错误定位支持 – 具有单错检测(奇偶校验)的32KB L1 指令高4-位,8-位,和16-位 速缓存• 可编程实时单元和工业用通信子系统(PRU-ICSS) – 具有单错检测(奇偶校验)的32KB 数据高速缓– 支持的协议包括 存EtherCAT®,PROFIBUS,PROFINET,Ether – 含纠错码(ECC)的256KB L2 高速缓存Net/IP™,和其它更多协议 – 176KB 片载启动ROM – PRU-ICSS 内的外设 1 Please be aware that an important notice concerning availabi

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 5
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值