基于28nn Stratix V FPGA的100GbE线路卡设计技术
扫描二维码
随时随地手机看文章
Altera’s 28-nm Stratix V FPGAs include innovations such as an enhanced core architecture, integrated transceivers up to 28 Gbps, and a unique array of integrated hard intellectual property (IP) blocks. With these innovations, Stratix V FPGAs deliver a new class of application-targeted devices optimized for:
■ Bandwidth-centric applications and protocols
■ Data-intensive applications for 40G/100G and beyond
■ High-performance, high-precision digital signal processing (DSP) applications Stratix V FPGAs are available in four variants (GT, GX, GS, and E), each targeted for a different set of applications. For higher volume production, you can prototype with Stratix V FPGAs and use the low-risk, low-cost path to HardCopy® V ASICs.Stratix V FPGAs deliver the industry’s most flexible transceivers with the highest bandwidth from 600 Mbps to 28 Gbps, low Bit Error Ratio (BER), and low power.
Stratix V transceivers have many enhancements to improve flexibility and robustness.
These enhancements include robust analog receive Clock and Data Recovery (CDR), advanced pre-emphasis and equalization for 12.5 Gbps backplanes. In addition, all transceivers are identical with full featured embedded PCS hard IP to simplify the design, lower the power, and save valuable core resources. Stratix V transceivers are designed to be standard compliant for a wide range of protocols and data rates, and are equipped with a variety of signal conditioning features to support backplane, optical module, and chip-to-chip applications
Stratix V FPGA主要特性:
■ Technology
● 28-nm TSMC process technology
● 0.85-V core voltage
■ Low power serial transceivers
● 28 Gbps transceivers on Stratix V GT devices
● 600 Mbps to 12.5 Gbps backplane capability
● Transmit pre-emphasis and de-emphasis
● Dynamic reconfiguration of individual channels
■ General purpose I/Os
● 1.6-Gbps LVDS
● 533-MHz/1066-Mbps external memory interface
● On-chip termination (OCT)
● 1.2-V to 3.3-V interfacing for Stratix V GX/GS/E devices
● 1.2-V to 2.5-V interfacing for Stratix V GT device
■ Embedded HardCopy Block
● PCI Express Gen1/Gen 2 complete protocol stack, ×1/×4/×8 end point and root port
■ Embedded transceiver hard IP
● Interlaken PCS
● Gigabit Ethernet (GbE) and XAUI PCS
● 10G Ethernet PCS
● Serial Rapid I/O (SRIO) PCS
● Common Public Radio Interface (CPRI) PCS
● Gigabit Passive Optical Networking (GPON) PCS
■ Power Management
● Programmable Power Technology
● Quartus® II integrated PowerPlay Power Analysis
■ High performance core fabric
● Enhanced ALM with 4 registers
● Improved routing architecture reduces congestion and improves compile times
■ Embedded memory blocks
● M20K: 20-Kbit with hard ECC
● MLAB: 640-bit
■ Variable precision DSP blocks
● Up to 500 MHz performance
● Natively support signal processing with precision ranging from 9×9 up to 54×54
● New native 27×27 multiply mode
● 64-bit accumulator and cascade for systolic FIRs
● Embedded internal coefficient memory
● Pre-adder/subtractor improves efficiency
● Increased number of outputs allows more independent multipliers
■ Fractional PLLs
● Fractional mode with third-order delta-sigma modulation
● Integer mode
● Precision clock synthesis, clock delay compensation, and zero delay buffering
■ Clock networks
● 717-MHz fabric clocking
● Global, quadrant, and peripheral clock networks
● Unused clock networks can be powered down to reduce dynamic power
■ Configuration
● Serial and parallel flash interface
● Enhanced AES design security features
● Tamper protection
■ High performance packaging
● Multiple device densities with identical package footprints enables seamless migration between different FPGA densities
● FBGA packaging with on-package decoupling capacitors
● Lead and RoHS-compliant lead-free options
■ HardCopy V migration
图1。Stratix V GT/GX/GS FPGA芯片图
表1。Stratix V GT/GX/GS FPGA型号
Stratix V FPGAs address the design challenges for applications in a variety of industries. Expand the sections below for details about specific applications.
Stratix V GT/GX/GS FPGA应用:
1.100-Gb Optical Transport Network (OTN) Multiplexing Transponder
图2.100Gb OTN多路收发器框图
100Gb OTN多路收发器主要特性:
lMulti-standard client interfaces enabled through easy-to-use partial reconfiguration and serial transceivers with continuous data range of 600 Mbps to 12.5 Gbps
lEnhanced clocking flexibility with up to 44 independent transmit clock domains
lIntegrated electronic dispersion compensation (EDC) capability in transceivers to enable direct drive of optical modules (SFP+, SFP, QSFP, CFP)
l28-Gbps transceivers for next-generation optical interfaces
lAdvanced fPLL replacing external voltage-controlled crystal oscillators (VCXOs)
2.100 Gigabit Ethernet (GbE) Line Card
图3.100GbE线路卡框图
100GbE线路卡主要特性:
lHigher system integration through highest density and hard PCS blocks for 40 GbE, 100 GbE, and Interlaken
lHigh-bandwidth data-buffering with up to 1,600-Mbps external memory interfaces
lEfficient implementation of packet processing and traffic management functions
lHigher system performance while staying within your power and cost budget
3.Crossbar and Backplane Switch Fabric
图4.纵横制和背板交换框图
纵横制和背板交换主要特性:
lHighest bandwidth through 66 identical transceivers with continuous data rate from 600 Mbps to 12.5 Gbps
lBuilt-in advanced signal conditioning circuitry for direct drive of 10GBASE-KR backplanes
lFlexible support for various line-card interfaces with partial and dynamic reconfiguration
lOptimized implementation of scheduling functions through high level of integration
4.Military Radar Application
图5.军用雷达应用框图[!--empirenews.page--]
军用雷达主要特性:
lEfficient floating-point multiplication with up to 1,000 GFLOPS
lHigher signal processing bandwidth with up to 1,840 GMACS
lAutomatic single event upset (SEU) detection and correction
lDesign security with enhanced Advanced Encryption Standard (AES) algorithm and 256-bit volatile and non-volatile keys
lProductivity-boosting tools in Quartus® II software, including DSP Builder Advanced Blockset and incremental compilation
5. RF Card and Channel Card
图6.RF卡和通路卡框图
RF卡和通路卡主要特性:
lReduced board space, power, and cost via fewer data channels and higher throughput per channel
lLower system latency and increased system performance and reliability via greater integration
lDesign differentiation using highest DSP- and memory-to-logic ratios
Higher MIMO and bandwidth density compared to competitive offering
6.Studio Video Server
图7.视频服务器框图
视频服务器主要特性:
Best-in-class serial digital interface (SDI) solution
Support for multiple CODECs through user-friendly partial reconfiguration
Optimal memory design with native 10-bit support
Efficient video processing with high multipliers-and memory-to-logic ratios
Complete solution via CODECs and 1080p video framework IP core
采用28nm FPGA的100GbE线路卡设计方案
The components of a 100-GbE line card include:
■ Optical interface—The optical interface unit can consist of multiple SFP+ or XFP modules, or it can be driven by 100G traffic via CFP or QSFP modules.
■ PHY—The PHY unit is the serializer/deserializer (SERDES) component of the line card. The PHY line rate and jitter specifications should be compliant with the optical interface.
■ MAC/PCS—The MAC/PCS unit performs the gearbox, scrambling, and encoding functions based on the protocol. In the case of 40-GbE or 100-GbE implementations, there is a multilane distribution (MLD) function as per the IEEE 802.3ba specifications. In addition, flow control as well as error handling is performed by the MAC.In some cases, the received 10G data from the MAC unit is aggregated before it is passed over to the network processing unit (NPU).
■ NPU—The key function of the NPU is to optimize the performance of packet processing in the evolving functional framework of the line card. Key functions include compression, classification/lookup, modification, and deep-packet inspection. The most common function of the NPU is to interface with a switch fabric device that performs complicated routing of the packets through the network.
■ Traffic manager—The primary function of the traffic manager is to offer a large number of high-speed queues, optimize queue depths, and use sophisticated scheduling mechanisms to meet the QoS requirements of the application. Because NPUs are not designed with QoS in mind, they require excessive processing power and software optimization before they can function as efficiently as a dedicated traffic manager.
图8。100-GbE线路卡元件框图
图9。采用两片Stratix V FPGA的100-GbE线路卡框图
100-GbE线路卡主要特性:
Higher system integration through highest density and hard PCS blocks for 40 GbE, 100 GbE, and Interlaken
High-bandwidth data-buffering with up to 1,600-Mbps external memory interfaces
Efficient implementation of packet processing and traffic management functions
Higher system performance while staying within your power and cost budget