Design, Implementation, and Testing of a Hybrid Tool for Network Topology Discovery”, TR-D

合集下载

Meraki MS390 交换机数据手册说明书

MS390 DatasheetOverviewThe Meraki MS390addresses the most demanding enterprise applications by combining the simplicity of the Meraki dashboard with powerful switching hardware. To satisfy high-bandwidth applications and the deployment of high-speed802.11ax/wifi-6access points,the MS390provides multigigabit ports,480G stacking, and modular10/40G uplinks.The MS390delivers resiliency with fast stack convergence and StackPower.The MS390provides Adaptive Policy using an over-the-wire tag which segments traffic into security groups to deliver scalable security.The MS390is integrated under the Meraki dashboard to provide a simply powerful solution to the most demanding wired access applications.Adaptive Policy provides simple&scalable security policies to segment traffic using Security Groups.Security Groups are created in the dashboard using natural language such as“IOT device”&“Guest.”The security policy intent(e.g.,Permit or Deny)is then simply provisioned between Security Groups which results in the segmentation of each group’s traffic.By making security policy management intuitive and scalable relative to legacy IP-address based Access Control Lists, Adaptive Policy empowers operators to confidently secure their network traffic independent of future network changes.By pooling&distributing power across MS390s using a series of StackPower cables,StackPower provides simple and resilient power distribution across the stack.ModelsNumber of Ports Model DescriptionMS390-24-HW24-port GbE switchMS390-24P-HW24-port GbE PoE+ switch24 Ports MS390-24U-HW24-port GbE UPoE switch MS390-24UX-HW24-port mGbE UPoE switch48 Ports MS390-48-HW48-port GbE switchMS390-48P-HW48-port GbE PoE+ switchMS390-48U-HW48-port GbE UPoE switchMS390-48UX-HW36-port 2.5GbE+ 12-port mGbE UPoE switch MS390-48UX2-HW48-port 5GbE UPOE switchFeaturesCategory FeaturesHighlights •Layer-3•40G or 10G modular uplink options on all models •mGig support•Dual Dedicated 120G Hardware Stacking Ports•PoE+ and UPoE Support•StackPower in a ring topology supporting upto 4 switchesManagement •Managed via Cisco Meraki Dashboard•Zero-touch remote provisioning (no staging needed)•Basic configuration capability via local management page•Detailed historical per-port and per-client usage statistics•Operating System, device, and hostname fingerprinting•Automatic firmware upgrades with scheduling control•SNMP and SYSLOG support for integration with other network management solutions*Remote Diagnostics •Email, SMS and Mobile push notification alerts1•Ping, traceroute, cable testing, and link failure detection with alerting •Remote packet capture•Dynamic and interactive network discovery and topology •Combined event and configuration change logs with instant searchStacking •Physically stack up to 8 switches with 480 Gbps of stacking bandwidth on all models•Virtual stacking supports thousands of switch ports in a single logical stack for unified management, monitoring, and configuration•Faster convergence•StackPower in a ring topology supporting upto 4 switchesEthernet Switching Capabilities •802.1p Quality of Service, 8 queues (w/ 6 configurable for DSCP-to-CoS mapping)•802.1Q VLAN and trunking support for up to 4,094 VLANs (1000 active VLANs with STP enabled)•Single Instance of 802.1s Multiple Spanning Tree Protocol (interoperable with RSTP, STP, PVST, RPVST)•STP Enhancements: BPDU guard, Root guard, Loop guard, UDLD•Broadcast storm control•802.1ab Link Layer Discovery Protocol (LLDP) and Cisco Discovery Protocol (CDP)•802.3ad Link aggregation with up to 8 ports per aggregate, Multichassis aggregates supported on stacked switches •Port mirroring•IGMP snooping for multicast filtering•MAC Forwarding Entries: 32KLayer 3•Static routing, OSPFv2•Multicast routing (PIM-ASM)•Warm Spare (VRRP) *•DHCP Server, DHCP RelaySecurity •Integrated multi-factor authentication for Dashboard management•Role-based access control (RBAC) with granular device and configuration control •Corporate wide password policy enforcement•IEEE 802.1X RADIUS and MAB, hybrid authentication and RADIUS server testing •Single-Host/Multi-Domain/Multi-Host/Multi Authentication•Port security: Sticky MAC, MAC whitelisting *•DHCP snooping, detection and blocking, Dynamic ARP Inspection•IPv4 and IPv6 ACLs•Secure Connect *•Adaptive Policy *MS390 LicensingMS390license structure includes two feature tiers:Enterprise and Advanced. The MS390 also introduces a new and simpler license to hardware mapping, specifically 24-port or 48-port licenses. As with all MS, every MS390 license is available in 1, 3, 5, 7, and 10 year terms.MS390 License Structure24-Port Model48-port ModelAdvanced Features LIC-MS390-24A LIC-MS390-48A* Available in a future software releaseEnterprise Features LIC-MS390-24E LIC-MS390-48EThe features available with advanced licensing are:•Adaptive policy *•Greater than 1,000 routes for OSPFContext and Comparisons24 Port Models48 Port ModelsDescriptionMS350-24XMS355-24X2MS390-24UXMS350-48FPMS355-48X2MS390-48UX21GbE RJ4516--4824-mGbE RJ4582424-244810GbE SFP+44Modular 44Modular 40GbE QSFP+n/a 2Modular -2Modular Hardware Stack Port 2x 40G 2x 100G2x 120G2x 40G 2x 100G2x 120GManagement Interface 111111Hot Swap PS Yes, Dual Yes, Dual Yes, Dual Yes, Dual Yes, Dual Yes, Dual Hot Swap Fans Yes, 2x Yes, 3x Yes, 3x Yes, 2x Yes, 3x Yes, 3x Layer 3 Routing Yes Yes Yes YesYes Yes UPoE CapableYes, 740W Yes, 740W Yes, 560W No, 740W(only PoE/PoE+)Yes, 740W Yes, 645W Max Switching Capacity 176 Gbps 640 Gbps 640 Gbps 176 Gbps 688 Gbps 640 Gbp Max Stacking Bandwidth160 Gbps400 Gbps480 Gbps160 Gbps400 Gbps480 Gbps* Available in a future software releaseIn the Co-term licensing model (most existing Organizations), an Organization must either have all MS390 Enterprise or all MS390 Advanced licenses - they cannot be mixed. In the Per-device licensing model, a mix of Enterprise and Advanced can be added to a single Organization, but certain features may require all devices in a Network to have Advanced licenses, e.g. Adaptive Policy.For more information on licensing, refer to Meraki Licensing Models article .Technical Breakdown Interfaces SpecificationsModel InterfacesUplink10/40GbE(SFP+, QSFP+)120G HardwareStack PortDedicatedManagementInterfacePoE/UPoECapabilitiesMS390-24-HW24 x1GbE RJ45Modular21n/a MS390-24P-HW24 x1GbE RJ45Modular21PoE MS390-24U-HW24 x1GbE RJ45Modular21UPoEMS390-24UX-HW 24 x 100M/1G/2.5G/5G/10G RJ45Modular21UPoEMS390-48-HW48 x1GbE RJ45Modular21n/a MS390-48P-HW48 x1GbE RJ45Modular21PoE MS390-48U-HW48 x1GbE RJ45Modular21UPoEMS390-48UX-HW36 x100M/1G/2.5G +12 x100M/1G/2.5G/5G/10GModular21UPoEMS390-48UX2-HW48 x100M/1G/2.5G/5G Modular21UPoE Physical SpecificationsModelDimensions(h x w x d)Weight Mount Type Hot Swap Fans Operating Temperature HumidityAll models are available with modular uplinks that have been listed under the Accessories list. For supported SFP modules please refer the SFP Datasheet.Cabling Best Practices for Multi-Gigabit operations:While Category-5e cables can support multigigabit data rates upto 2.5/5 Gbps, external factors such as noise, alien crosstalk coupled with longer cable/cable bundle lengths can impede reliable link operation. Noise can originate from cable bundling, RFI, cable movement, lightning, power surges and other transient event. It is recommended to use Category-6a cabling for reliable multigigabit operations as it mitigates alien crosstalk by design.W/ Default Power Supply1.73” x 17.5” x17.7”MS390-24-HW16.03 lb (7.27 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90%(4.4 x 44.5 x 44.9cm)1.73” x 17.5” x17.7”16.33 lb (7.4 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90% MS390-24P-HW(4.4 x 44.5 x 44.9cm)1.73” x 17.5” x 19.2”16.63 lb (7.54 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90% MS390-24U-HW(4.4 x 44.5 x 44.8cm)1.73” x 17.5” x 20.2”MS390-24UX-HW18.18 lb (8.25 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90%(4.4 x 44.5 x 51.3cm)1.73” x 17.5” x17.7”16.43 lb (7.45 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90% MS390-48-HW(4.4 x 44.5 x 44.9cm)1.73” x 17.5” x17.7”16.73 lb (7.59 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90% MS390-48P-HW(4.4 x 44.5 x 44.9cm)1.73” x 17.5” x 19.2”MS390-48U-HW17.03 lb (7.72 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90%(4.4 x 44.5 x 48.8cm)1.73” x 17.5” x 22.2”20.50 lb (9.34 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90% MS390-48UX-HW(4.4 x 44.5 x 56.4cm)1.73” x 17.5” x 22.2”20.05 lb (9.09 kg)1U Rack Mount Yes, 3x-5°C to 45°C 5 to 90% MS390-48UX2-HW(4.4 x 44.5 x 56.4cm)PerformanceSwitching Capacity Stacking Bandwidth Forwarding rateModelMS390-24-HW208 Gbps480 Gbps154.76 MppsMS390-24P-HW208 Gbps480 Gbps154.76 Mpps MS390-24U-HW208 Gbps480 Gbps154.76 Mpps MS390-24UX-HW640 Gbps480 Gbps476.19 Mpps MS390-48-HW256 Gbps480 Gbps190.48 Mpps MS390-48P-HW256 Gbps480 Gbps190.48 Mpps MS390-48U-HW256 Gbps480 Gbps190.48 Mpps MS390-48UX-HW580 Gbps480 Gbps431.54 Mpps MS390-48UX2-HW640 Gbps480 Gbps476.19 MppsPower Options and SpecificationsModel Default PowerSupplyHot Swap PowerSupplyAvailable PoE W/Primary PSAvailable PoE W/ SecondaryPS***Power Load(idle/max)MS390-24-HWMA-PWR-350WAC**Yes, Dual n/a n/a79.2 / 99 WMS390-24P-HWMA-PWR-715WAC**Yes, Dual445W720W84.1 / 554.4WMS390-24U-HWMA-PWR-1100WACYes, Dual830W1440W85.4 / 990.3WMS390-24UX-HWMA-PWR-1100WACYes, Dual560W1440W162.7 / 809.9WMS390-48-HWMA-PWR-350WAC**Yes, Dual n/a n/a83.9 / 109.9WMS390-48P-HWMA-PWR-715WAC**Yes, Dual437W1152W92.6 / 555 WMS390-48U-HWMA-PWR-1100WACYes, Dual822W1800W145 / 844.9WMS390-48UX-HWMA-PWR-1100WACYes, Dual490W1590W218.5 / 785.5WMS390-48UX2-HWMA-PWR-1100WACYes, Dual645W1745W157.9 / 843.8W** Upgrade options to715W and 1100W PSU are available.*** The PoE values are provided considering the secondary PS to be the default power supply of the respective model.What's includedModel Package ContentsMS390-24-HW 1 x Power Supply (MA-PWR-350WAC), Rack mount brackets and screw kit,3 x Pre-Installed Fans, Cable guide MS390-24P-HW 1 x Power Supply (MA-PWR-715WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-24U-HW 1 x Power Supply (MA-PWR-1100WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-24UX-HW 1 x Power Supply (MA-PWR-1100WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-48-HW 1 x Power Supply (MA-PWR-350WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-48P-HW 1 x Power Supply (MA-PWR-715WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-48U-HW 1 x Power Supply (MA-PWR-1100WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-48UX-HW 1 x Power Supply (MA-PWR-1100WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guide MS390-48UX2-HW 1 x Power Supply (MA-PWR-1100WAC), Rack mount brackets and screw kit, 3 x Pre-Installed Fans, Cable guideOptional AccessoriesAccessory Description Supported ModelsMA-PWR-350WAC350W AC Power Supply MS390-24-HW, MS390-48-HWMA-PWR-715WAC715W AC Power Supply All ModelsMA-PWR-1100WAC1100W AC Power Supply All ModelsMA-MOD-2X40G 2 x 40G Uplink Module All ModelsMA-MOD-4X10G 4 x 10G Uplink Module All ModelsMA-MOD-8X10G8 x 10G Uplink Module All ModelsMA-CBL-120G-50CM Meraki 120G Stacking Cable, 0.5Meter All ModelsMA-CBL-120G-1M Meraki 120G Stacking Cable, 1Meter All ModelsMA-CBL-120G-3M Meraki 120G Stacking Cable, 3Meter All ModelsMA-CBL-SPWR-30CM Meraki MS390 30CM StackPower Cable All ModelsMA-CBL-SPWR-150CM Meraki MS390 150CM StackPower Cable All Models MA-FAN-16K2System Fan All Models MA-RCKMNT Meraki MS390 Rack Mount Kit All ModelsRegulations and ComplianceElectromagnetic CompatibilityCertifications FCC Part 15 (CFR 47) Class A, ICES-003 Class A, CISPR22 Class A,CNS13438, EN 300 386 V1.6.1,EN 55022 Class A, EN 61000-3-2,EN61000-3-3, KN 32, TCVN 7189 Class A, EN 55032 , CISPR 32 Class A, V-2/2015.04, V-3/2015.04, VCCI-CISPR 32 Class A, CISPR24, EN 300 386 V1.6.1, EN 55024, KN35, TCVN 7317SafetyCAN/CSA-C22.2 No. 60950-1, UL 60950-1, EN 60950-1, IEC 60950-1, AS/NZS 60950.1 Environmental Reduction of Hazardous Substances (RoHS)Warranty Full lifetime hardware warranty with next-day advanced replacement included MTBF RatingModelMTBF at 25°CMS390-24-HW314,790MS390-24P-HW299,000MS390-24U-HW238,410MS390-24UX-HW214,760MS390-48-HW305,870MS390-48P-HW277,770MS390-48U-HW227,410MS390-48UX-HW202,160MS390-48UX2-HW198,647Installation GuideFor instructions on how to install and configure the MS390 series switch please refer the MS390 Series Installation Guide。

Adaptive QoS Routing by Cross-Layer Cooperation in Ad Hoc Networks

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
662
EURASIP Journal on Wireless Communications and Networking and numerical results are discussed in Section 4. Finally, the summary is presented in Section 5. 2. SCHEME OVERVIEW
the physical properties of the transmission channel can lead to problems such as possible failure at high data loads. The network-data-link level approach is more promising because it combines information from both the network and datalink layer [7, 8]. However, due to dynamic changes in network topology and the diﬃculties in predicting link states, indirect or estimation approaches are often used (e.g., using signal strength and link lifetime as routing parameters [9]). The obvious problem with such approaches is that the impacts on QoS performance are hard to quantify, since the cross-layer behaviors of mobile networks are not considered. Therefore, most of the proposed routing schemes for mobile networks are only QoS aware, but do not guarantee QoS. To address this problem, appropriate cross-layer cooperation is required. We propose an adaptive scheme to provide QoS information by factoring the impacts of node mobility and lower-layer link parameters into QoS performance. There are many proposals for QoS routing in the literature [7, 10, 11]. Most approaches tend to focus on only one QoS parameter (e.g., packet loss, end-to-end delay, and bandwidth). For example, while many of the QoS-related schemes are successful in reducing packet loss by adding redundancy in the packet [12, 13, 14], they do this at the expense of end-to-end delay. Because packet loss and end-toend delay are inversely related, it may not be possible to ﬁnd a path that simultaneously satisﬁes the delay, packet loss, and bandwidth constraints. Some proposed QoS routing algorithms [15, 16, 17] do consider multiple metrics, but without considering cross-layer cooperation. Multipath routing is another type of QoS routing that has received much attention, since it can provide load balancing, fault tolerance, and higher aggregate bandwidth [12, 18, 19]. Although this approach decreases packet loss and end-to-end delay, it is only eﬃcient and reliable if a relationship can be found between the number of paths and QoS constraints. In this paper, we propose a cross-layer cooperation mechanism to support adaptive multipath routing with multiple QoS constraints in an ad hoc network. The cross-layer mechanism provides information on link performance for the QoS routing. It treats traﬃc distribution, wireless link characteristics, and node mobility in an integrated fashion. That is, it reﬂects the impacts of lower-layer parameters on QoS performance in higher layers, with emphasis on translating these parameters into QoS parameters for the higher-layer connections. A multiobjective optimization algorithm is used to calculate routing parameters using the cross-layer mechanism. These parameters are adapted to the current network status, determining the number of routing paths and code parity lengths for FEC. In addition, a traﬃc engineering strategy is used to evenly distribute traﬃc over multiple paths. The remainder of the paper is organized as follows. Section 2 provides an overview of our cross-layer routing architecture. Three functions (the routing, the local statistic computation, and the integrated decision-making functions) in the routing mechanism are introduced in Section 3. The layered network models used to support the implementation of these functions are also presented in this part. Simulation

信息与通信工程专业英语

Signal and System
Mathematical Foundations in Information and Communication Engineering
Mathematics plays a fundamental role in information and communication engineering, promoting the theoretical frameworks and analytical tools for system design and analysis
contents
目录
Frontier Technologies in Information and Communication Engineering Practice and Application of Information and Communication Engineering Reading and Writing Skills for Information and Communication Engineering English
01
Fundamentals of Information and Communication Engineering
Summary
Information theory is a branch of engineering that studies the quantification, transmission, and processing of information It provides fundamental principles for the design of communication systems

Z-Active 差分探头系列 P7313、P7380A、P7360A、P7340A 数据手册说明书

Z-Active™Differential Probe FamilyP7313•P7380A •P7360A •P7340A DataSheetFeatures &Bene ﬁtsSignal Fidelity>12.5GHz Bandwidth (P7313,Typical)>8.0GHz Bandwidth (P7380A,Typical)>6.0GHz Bandwidth (P7360A,Typical)>4.0GHz Bandwidth (P7340A,Typical)Extended Linear Dynamic Range 1.25V p-p at 5x Attenuation (P7313)4V p-p at 25x Attenuation (P7313)2V p-p at 5x Attenuation (P7380A,P7360A,P7340A)5V p-p at 25x Attenuation (P7380A,P7360A,P7340A)Low Probe Loading DC Input Resistance 100k ΩDifferential 50k ΩSingle Ended AC LoadingZ min >200Ωout to 10GHz (P7313)Z min >290Ω,4GHz to 8GHz (P7380A,P7360A,P7340A)VersatilityMake Differential or Single-ended (Ground-referenced)Measurements*1Solder-down CapabilityHandheld Probing with Variable Spacing and Compliance Fixtured Probing Interchangeable Tip-Clip™Assemblies Connect to a Variety of Devices Economical TekConnect ®InterfaceApplicationsExamples Include,but are not Limited To:PCI-Express I and II,Serial ATA II,USB 2.0,DDRII,DDRIII,Fireware 1394b,Rambus,XAUI*1For details,please see application note 60W-18344-0,“Making Single-ended Measurements with DifferentialProbes.”1981Data SheetZ-Active™Probing Architecture Leads the Way for High-speed Probing Applications Tektronix has created a revolutionary Z-Active probe architecture that sets the industry benchmark for signalﬁdelity.Tektronix active probe architecture preserves high bandwidth while providing improved connectivity with low loading.The Z-Active architecture is a hybrid approach composed of a distributed attenuator topology feeding an active probe ampliﬁer.The Z-Active probes use a tiny passive probe tip element that is separate from the ampliﬁer,extending the usable reach of the probe.In traditional active probes,adding this much length can introduce signalﬁdelity problems.However this architecture maintains high DC input resistance and presents a higher AC impedance than previous probe architectures.It accomplishes this while providing signiﬁcant length between the probe body and the probe attachment point to the DUT.This architecture provides the best of both worlds:high DC impedance like existing active probes and the stable high-frequency loading of Z0probes.Signal FidelityYou can be conﬁdent in the signalﬁdelity of your measurements because the Z-Active architecture provides:High BandwidthExcellent Step ResponseLow LoadingHigh CMRRExtended Linear Dynamic RangeExtended Linear Dynamic RangeMany of today’s logic signals and serial bus signals require the capabilityto measure up to several volts peak to peak.These voltage levels may easily be viewed with the Z-Active architecture probes(P7380A,P7360A, and P7340A)with the extended linear dynamic range.With a2.0V p-p linear dynamic input range at the5x attenuation setting,you can accurately measure DDR II and III,Firewire1394b,and PCI-Express I and II signals at reduced noise levels.In addition the25x attenuation setting’s linear dynamic input voltage range can be used up to5.0V p-p for accessing even larger signal swings found during transition times.ConnectivityThe Z-Active probe design allows the probe to easily switch between soldered,handheld,orﬁxtured applications.This family of probes uses Tip-Clip™assemblies,an interchangeable probe tip system that enables customers to conﬁgure their probe with the optimal tip for their application.These detachable assemblies make it possible to replace a tip for a fraction of the cost formerly associated with such hardware changes.The several lengths and variable spacing of the assemblies provideﬂexibility for adapting to vias and other test points of differing sizes.With Tektronix Tip-Clip assemblies,Monday’s solder-in probe can become Tuesday’s handheld tool,simply by switching tips. ValueThe combination of the Z-Active architecture and the Tip-Clip assemblies provide superior signalﬁdelity at a cost-effective price.The inexpensive Tip-Clip assemblies enable full-performance solder connections at a very low price per connection.Over the life of a probe this can add up to signiﬁcant savings in the cost of operation.Performance You Can Count OnDepend on Tektronix to provide you with performance you can count on.In addition to industry-leading service and support,this product comes backed by a one-year warranty as standard.Z-Active™Differential Probe Family—P7313•P7380A•P7360A•P7340ACharacteristicsCharacteristic P7340A P7360A P7380A P7313Bandwidth(Typical)>4GHz>6GHz>8GHz>12.5GHzRise Time(10%-90%)(Guaranteed)<100ps<70ps<55ps<40psRise Time(20%-80%)(Typical)<70ps<50ps<35ps<25psAttenuation5x or25x,user selectableDifferential Input Range±1.0V(5x)±2.5V(25x)±0.625V(5x)±2.0V(25x)Linearity Error for Differential Input Dynamic Range(Typical)±0.5%for-0.5V to+0.5V(5x)±1.0%for-0.75V to+0.75V(5x)±2.0%for-1.0V to+1.0V(5x)±0.5%for-1.5V to+1.5V(25x)±1.0%for-2.5V to+2.5V(25x)±2.0%for-3.0V to+3.0V(25x)±0.25%for-0.5V to+0.5V(5x)±0.75%for-0.625V to+0.625V(5x)±0.5%for-1.6V to+1.6V(25x)±1.0%for-2.0V to+2.0V(25x)Operating Voltage Window+5.0V to-3.0V+4.0V to-3.0V Offset Voltage Range+4.0V to-3.0VDC Input Resistance100kΩAC Loading(Differential Z min)>290Ω>200ΩNoise<31nV/√Hz(5x)<75nV/√Hz(25x)CMRR>50dB at1MHz>35dB at1GHz>20dB at4GHz >50dB at1MHz>35dB at1GHz>20dB at6GHz>50dB at1MHz>35dB at1GHz>20dB at8GHz>50dB at1MHz>35dB at1GHz>20dB at6GHz>15dB at12.5GHzNondestructive Input Range±15VInterface TekConnect®Cable Length 1.5m 1.5m 1.2m 1.2m Ordering InformationP7313>12.5GHz Z-Active Differential Probe for TekConnect®Interface. P7380A>8.0GHz Z-Active Differential Probe for TekConnect®Interface.P7360A>6.0GHz Z-Active Differential Probe for TekConnect®Interface.P7340A>4.0GHz Z-Active Differential Probe for TekConnect®Interface.All Include:One-year warranty,plus see Standard Accessories table.3Data SheetStandard AccessoriesDescriptionP7340AP7360AP7380AP7313Reorder Part NumberPouch,Nylon Carrying Case with Inserts1each 1each 1each 1each 016-1952-xx Qty 1Accessory Performance Summary and Reorder Sheet1each 1each 1each 1each 001-1389-xx Qty 1020-2640-xx Qty 1–Opt.L0020-2648-xx Qty 1–Opt.L5User Manual -Printed.Includes Reply Card and CD 1each1each1each1each040-2649-xx Qty 1–Opt.L7BNC (M)-to-Minigrabber Adapter 1each 1each 1each 1each 013-0342-xx Qty 1Anti-static Wrist Strap 1each 1each 1each 1each 006-3415-xx Qty 1Magnifying Glasses 1each 1each 1each 1each 378-0486-xx Qty 1Calibration Data Report 1each 1each 1each 1each Opt.D1Handheld Probe Adapter 1each 1each 1each1each 015-0717-xx 1eachP7313:020-2636-xx 1eachP7380A:020-2557-xx 1eachP7360A:020-2690-xx Accessory Box and Contents1each P7340A:020-2690-xx Attachment Kit1each 1each 1each 1each 016-1953-xx Qty 1Velcro Fastening Strap 10each 10each 10each 10each –Velcro Fastening Dots 10each 10each 10each 10each –Adhesive Tip-Clip Tape*2(Strip of 10)3each 3each 3each 3each –Color Band Kit (2ea.of 5colors)1each 1each 1each 1each 016-1948-xx Qty 1Short Flex,Small Resistor Tip-Clip Assembly 2each 2each 3each 3each 020-2600-xx Qty 10Medium Flex,Small Resistor Tip-Clip Assembly 2each 2each 3each 3each 020-2602-xx Qty 10Long Flex,Small Resistor Tip-Clip Assembly 2each 2each 3each 3each 020-2604-xx Qty 10Variable Spacing Tip-Clip Kit 3each 3each 3each 3each 020-2596-xx (Kit of 3)Square Pin Adapter Tip-Clip 1each 1each 1each 1each 020-2701-xx (Kit of 3)Tip-Clip Ejector*23each 3each 3each 3each –020-2639-xx Qty 10HBW Straight Flex Tip-Clip Assembly –––3each020-2657-xx Qty 5020-2638-xx Qty 10HBW Right-Angle Flex Tip-Clip Assembly –––3each 020-2656-xx Qty 5Wire Replacement Kit–––1each 020-2644-xx Qty 1Short Flex,Large Resistor 1/8W Tip-Clip Assembly––3each –020-2601-xx Qty 10Long Flex,Large Resistor 1/8W Tip-Clip Assembly––3each –020-2605-xx Qty 10Medium Flex,Large Resistor 1/8W Tip-Clip Assembly2each2each3each–020-2603-xx Qty 10*2Tip-Clip Ejectors and Tip-Clip Tape are shipped standard with the 020-xxxx-xx Tip-Clip Assembly Kits.Recommended AccessoriesDescriptionP7360P7380P7313Part NumberProbe Positioner Yes Yes Yes PPM100Probe PositionerYes Yes Yes PPM203B PPM203B,PPM100Adapter Fixture Yes Yes Yes 013-0339-xx P7340A:067-0419-xx P7360A:067-0419-xx P7380A:067-0419-xx Calibration Fixture Yes Yes YesP7313:067-1616-xxDSA8200Series TekConnect ®Probe Interface Yes Yes Yes 80A03Deskew FixtureYes Yes Yes 067-1586-xx Real-time Spectrum Analyzer TekConnect Probe AdapterYes Yes YesRTPA2AZ-Active™Differential Probe Family—P7313•P7380A•P7360A•P7340AService OptionsOpt.CA1–Single Calibration or Functional Veriﬁcation.Opt.C3–Calibration Service3Years.Opt.C5–Calibration Service5Years.Opt.D3–Calibration Data Report3Years(with Opt.C3).Opt.D5–Calibration Data Report5Years(with Opt.C5).Opt.G3–Complete Care3Years(includes loaner,scheduled calibration and more). P7360A,P7380A onlyOpt.G5–Complete Care5Years(includes loaner,scheduled calibration and more). P7360A,P7380A onlyOpt.R3–Repair Service3Years.Opt.R5–Repair Service5Years.Language OptionsOpt.L0–English Manual.Opt.L5–Japanese Manual.Opt.L7–Simpliﬁed Chinese Manual.Additional Service Products Available During Warranty (DW)or Post Warranty(PW)P7313-CA1–Single Calibration or Functional Veriﬁcation.P7313-R1PW–Repair Service Coverage1-year Post Warranty.P7313-R2PW–Repair Service Coverage2-year Post Warranty.P7313-R3DW–Repair Service Coverage3Years(includes product warranty period); 3-year period starts at time of customer instrument purchase.P7313-R5DW–Repair Service Coverage5Years(includes product warranty period); 5-year period starts at time of customer instrument purchase.P7340A-CA1–Single Calibration or Functional Veriﬁcation.P7340A-R1PW–Repair Service Coverage1-year Post Warranty.P7340A-R2PW–Repair Service Coverage2-year Post Warranty.P7340A-R3DW–Repair Service Coverage3Years(includes product warranty period);3-year period starts at time of customer instrument purchase.P7340A-R5DW–Repair Service Coverage5Years(includes product warranty period);5-year period starts at time of customer instrument purchase.P7360A-CA1–Single Calibration or Functional Veriﬁcation.P7360A-R1PW–Repair Service Coverage1-year Post Warranty.P7360A-R2PW–Repair Service Coverage2-year Post Warranty.P7360A-R3DW–Repair Service Coverage3Years(includes product warranty period);3-year period starts at time of customer instrument purchase.P7360A-R5DW–Repair Service Coverage5Years(includes product warranty period);5-year period starts at time of customer instrument purchase.P7380A-CA1–Single Calibration or Functional Veriﬁcation.P7380A-R1PW–Repair Service Coverage1-year Post Warranty.P7380A-R2PW–Repair Service Coverage2-year Post Warranty.P7380A-R3DW–Repair Service Coverage3Years(includes product warranty period);3-year period starts at time of customer instrument purchase.P7380A-R5DW–Repair Service Coverage5Years(includes product warranty period);5-year period starts at time of customer instrumentpurchase.Tektronix is registered to ISO9001and ISO14001by SRI Quality SystemRegistrar.Product(s)complies with IEEE Standard488.1-1987,RS-232-C,and with TektronixStandard Codes and Formats.5Data SheetZ-Active™Differential Probe Family—P7313•P7380A•P7360A•P7340A7Data Sheet Contact Tektronix:ASEAN/Australasia(65)63563900Austria0080022554835*Balkans,Israel,South Africa and other ISE Countries+41526753777Belgium0080022554835*Brazil+55(11)37597627Canada180********Central East Europe and the Baltics+41526753777Central Europe&Greece+41526753777Denmark+4580881401Finland+41526753777France0080022554835*Germany0080022554835*Hong Kong4008205835India0008006501835Italy0080022554835*Japan81(3)67143010Luxembourg+41526753777Mexico,Central/South America&Caribbean52(55)56045090Middle East,Asia,and North Africa+41526753777The Netherlands0080022554835*Norway80016098People’s Republic of China4008205835Poland+41526753777Portugal800812370Republic of Korea00180082552835Russia&CIS+7(495)7484900South Africa+41526753777Spain0080022554835*Sweden0080022554835*Switzerland0080022554835*Taiwan886(2)27229622United Kingdom&Ireland0080022554835*USA180*********European toll-free number.If not accessible,call:+41526753777Updated10February2011For Further Information.Tektronix maintains a comprehensive,constantly expandingcollection of application notes,technical briefs and other resources to help engineers workingon the cutting edge of technology.Please visit Copyright©Tektronix,Inc.All rights reserved.Tektronix products are covered by U.S.and foreign patents,issued and rmation in this publication supersedes that in all previously published material.Speciﬁcation and price change privileges reserved.TEKTRONIX and TEK are registered trademarks ofTektronix,Inc.All other trade names referenced are the service marks,trademarks,or registered trademarksof their respective companies.02Oct201151W-17891-12。

CTCSS Reject High Pass Filters FM 无线通信评估说明书

CTCSS REJECT HIGH PASS FILTERSINFM RADIO COMMUNICATIONSAN EVALUATIONVirgil Leenerts WØINK 8 June 2008The response of the audio voice band high pass filter is evaluated in conjunction with the rejection of the CTCSS tone frequencies. The dual function of this filter in FM radio communications makes for conflicting requirements of passing quality voice audio, while rejecting the CTCSS tone frequencies that are in the low end of the audio band and can be heard if not rejected by the filter. The typical requirement of the filter is that the voice audio from 300 to 3000 Hertz be passed and rejection of the CTCSS band from 67 to 203.5 Hertz.From an ideal world this requirement would not be a problem, but the real world of filters makes it a challenge to design. The real world of filters does not have sharp corners or go from pass to reject in 0 Hertz, thus a lot of issues. So how does the real world work? Filters have a corner frequency and that means that at that frequency, the response is down by –3 dB and is a point on the curve transitioning from flat to the slope toward the reject level of the filter. The slope determines how many Hertz it will take before the filter will attenuate the desired level for the reject frequencies. To illustrate this more, lets say one would desire the response level at 300 Hertz to be the same as at 1000 Hertz, then that means the –3 dB frequency must be lower than 300 Hertz. This can be done but that means that the upper frequency of the CTCSS band must also be lower which for some is not acceptable as the desire is to have a higher CTCSS tone – thus the conflict! ABOUT ELECTRONIC FILTERSIn the electronic world, filters are a big topic so I plan to just point to that which is relevant to this paper. The topic of electronic filters is covered in many textbooks such as “Electronic Filter Design Handbook” by Arthur B. Williams also there are application notes such as Analog Devices AN-649. Filters that have a particular response curve have names like Butterworth, Chebyshev, and Elliptic or Cauer. These response curves are determined by mathematical functions that can be used to predetermine the response curves as needed by the designer. Typically these have been reduced to tables of coefficients, as the calculation is very complex. The implementation of these response curves can be done by many topologies. These topologies can be accomplished with passive components, passive components and amplifiers, switch capacitor devices, and digital methods. The use of op-amps for a filter is generally called an active filter. A particular name for one of the common topologies is “Unity-Gain Sailen-Key Active Filter”. There are a plethora of topologies for the implementation of particular response curves.THE RESPONSE CURVESFigure 1 – Relative Comparative 3 Order Filter Response CurvesAs can be noted in figure 1, the relative response curves show the fundamental differences between the curves. For this application of rejecting a CTCSS tone and at the same time having the lowest flat audio response is the desired goal. For this plot, the frequency has been normalized to be 10 at the –3 dB or corner frequency of the HPF. Examination at this point will easily show that the response of the Chebyshev and the Elliptic filter is flat closer to the corner frequency than the Butterworth. From the rejection of CTCSS tones, the Elliptic filter is easily seen to be better. However the Elliptic filter is not typically used because of the additional complexity of more parts and nonstandard parts. It may be easy to implement an Ellipitic filter with a digital topology, which is not covered by this paper.So what can be done to improve the overall response to meet the requirements of the application? Selection of filter parameters for each type of filter response will change the response curve. The order of the filter will change the steepness or transition band for all the different types. The larger the order number, the response will be steeper. The above nominal response curves are for 3rd order filters. The order refers to the number of poles for the filter; for example, a simple RC network is a 1st order filter. Another factor that will affect steepness is pass-band and or stop-band ripple, this is applicable to the Chebyshev and Elliptic filters. Thus the selection of a filter response along with the appropriate parameters can be made to meet requirements most of the time. A study of filters will show that this can be a complex topic as there are many aspects of filter design such as group delay, impulse and step response that are not covered here and is beyondthe scope of this paper. Another factor in filter design is the input and output impedance requirements for correct filter response. In the RF world, this is often referred to as matching which is important for correct filter response as well as power transfer. In the audio world, matching is for correct filter response and seems to be left out of the filter description in a number of cases.FM RADIO COMMUNICATIONS HIGH PASS FILTERFive high pass filters were evaluated for this paper and the results plotted. The filters measured were – TS-32 & 64, Motorola Micor & MSR2000, and GE. The generator used was an HP3336B Synthesizer and the voltmeter was an HP 34401A. Reference level is 0 dB at 1000 Hz. The specific measurement process for each of the filters is described in the appendices following this paper. The comparison here is limited to the best representative response curve for each filter. In the comparison, the assumed nominal level for acceptable CTCSS tone rejection is –30dB. Also this comparison is not done with listening tests but response level comparison only.Figure 2 – Comparative Response of the Five FiltersThis figure is a little busy but it gives the best picture as to the comparative performance of the filters.Starting with the TS-32, it is a 3rd order Butterworth filter using the “Unity-Gain Sailen-Key Active Filter” topology and shows the smooth transition from pass band to stop band. This filter response shows that it is just barely acceptable and the CTCSS tone will need to be in the low frequency range.The TS-64 is a 3rd order Chebyshev using the “Unity-Gain Sailen-Key Active Filter” topology with a nominal 2db ripple in the pass band. This is a nice improvement over the TS-32 as the nominal pass band response is closer to the corner frequency of the filter, also the stop band for CTCSS moves up and allows for higher tone frequencies to be attenuated.The other filters are a sort of hybrid as they sort of look like an elliptic filter, but as far as I can tell, these are not true elliptic-function filters. The Micor filter consists of inductors and capacitors while the MSR2000 uses an active circuit called a gyrator to simulate an inductor. Other than for the active circuit to simulate an inductor, both are an L-C-L configuration with input and output coupling capacitors. The GE filter combines an active notch filter with a high pass filter. (Note: the schematic and manual calls it a low pass but I feel this is an error, as I do not see the low pass function.)The Micor filter response definitely has a higher frequency of attenuation for the CTCSS tones, but the audio pass band is nominally the same as the TS-64.The MSR2000 and GE have very similar response curves with the GE curve having its pass band flatter closer to the –3dB point. The notch is about 185 Hz and as such has good attenuation up to 200 Hz. But below the notch frequency, the response raises in amplitude and the attenuation in the 100 Hz range is less than the notch frequency attenuation. From a CTCSS stop band perspective, this filter may not be acceptable in the midrange and may account for some folks having CTCSS tone in the 185 Hz range. CONCLUSIONFrom a pass band perspective, the GE CG filter has the lowest response to the corner frequency, but the overall rejection of the CTCSS tone band is somewhat lacking especially in the 100 Hz range. Overall it is the better filter from a response point of view than any of the filters measured, with the MSR2000 a very close second.I designed a 5th Order Chebyshev filter with 1 dB pass band ripple for comparison. Below is the plot of the GE and a Chebyshev filter for a comparison.Figure 3 – Comparison of GE CG and 5th Order Chebyshev FilterThe corner frequency or –3dB point for the 5th Order Chebyshev filter is 275 Hz. From an overall performance point of view, the Chebyshev filter has as good of pass band response as the GE and better attenuation of the CTCSS tone frequency band. Thus it is a good candidate for those wishing good band pass down to at least 300 Hz with good attenuation of CTCSS tones below 200 Hz. CTCSS tones around 160 Hz and below are attenuated more than the GE filter.I would like to thank the amateurs that donated filters to make this response comparison possible.Appendix Index:A – TS-32B – TS-64C – Micor PLD - MSR2000E – GE CGF – 5th Order ChebyshevG – Loudness CurvesThe TS-32 CTCSS Encoder/Decoder by Communication Specialists is no longer available but the schematic is available from their web site. The HPF is of the Sailen-Key Topology and is a typical 3rd Oder Butterworth filter.The –3dB point or corner frequency is at 375 Hertz. Since the filter has high input impedance, varying the impedance of the driving generator in the 0 to 10K ohm range had no effect on the frequency response of the filter.The schematic for the TS-64 CTCSS Encoder/Decoder is available from Communication Specialists web site. The HPF is of the Sailen-Key Topology and is a typical 3rd Order Chebyshev filter with pass band ripple of about 2 dB.The –3db point or corner frequency is 335 Hertz. Note the lower frequency response in the pass band. Since the filter has somewhat lower input impedance, there is a difference in response between a 75 ohm generator and a 5K ohm generator. The difference is not large but is noticeable. Of course, as the generator or source impedance increases, the response will change even more.The response of the Micor PL high pass filter is a not a straight forward measurement because it is inside the deemphasis circuit.ACVoltmeterTLN4294PL Filter0.033 uF0.1 uF0.022 uF0.1 uFInductors6 Henry0.33 uFC2330.033 uF10 k ohms0.22 uFAC Source5In4Out1Filter response measurement circuit that shows connections of the AC Source and Voltmeter along with interface components to filter. The 10k ohm resistor and C233 0.033uF capacitor form the deemphasis circuit. Response measurements were made with the capacitor C233 connected and disconnected with the AC source a flat response generator. Also a response measurement was made from a source simulating adiscriminator output with capacitor C233 connected.The simulator for the discriminator response had its preemphasis corner frequency at 250 Hz. The response curve shows a notch at a nominal frequency of 200 Hz. Ideally, the curve for the flat generator and the discriminator should be the same, but they are not, which shows that the system response curve of the filter may not be that which a system measurement would indicate. The “deemphasis” response shows the typical slope –6 dB per octave with the 10k ohm and 0.033 uF deemphasis circuit. With the deemphasis circuit in place, the discriminator source then shows the response of a transmitted PM signal resulting in a nominal flat audio response. The –3 dB point for the discriminator response is 350 Hz.Appendix D:The Motorola MSR2000 PL filter is similar in design to the Micor filter except for the use of gyrators to simulate the inductors.AC GYRATOR “inductor”U3A&BGYRATOR“inductor”U3C&D Voltmeter0.001 uF MSR2000PL Filter 0.1 uF 0.1 uF0.022 uF C540.015 uF “0.15 uF”15 k ohmsAC SourceFilter response measurement circuit that shows connections of the AC Source and Voltmeter along with interface components to the filter. The schematic for the circuit with the gyrators is in the manual. Measuring the response of the filter with the value of 0.015 uF for C54 as stated in the manual yielded an unreasonable response. I assumed that there may be an error in manual and that the value should be 0.15 uF. If anyone has an actual board, it would be nice to measure the value to verify the correct value for C54. The circuit was duplicated from the schematic in the manual.The frequency response curves for the two values of C54 shows what I feel is an unreasonable response for C54 = 0.015 uF. The response with C54 = 0.15 uF is the one I used in the main part of this paper. The notch frequency is a nominal 180 Hz and the –3 dB corner frequency is 300 Hz. As can be noted, the CTCSS tones around the notch frequency are highly attenuated but around 100 Hz they are attenuated a nominal 28 dB. Even though the notch is effective in allowing for a lower voice band response and higher CTCSS band width, this is accomplished at the expense of less overall attenuation of the CTCSS band.Appendix E:The GE CG filter board # 19C320627G1 is a combo of a notch and high pass filter. In the manual ( LBI30727 ) the unit is described to have a notch and a low pass filter, which Ifeel is in error.Above is the schematic from the manual showing the low pass filter identification. To test this filter, I built a circuit to replicate the driving circuit of this filter from a GE repeater manual.ACJ11 2 3J21 2 3GE Channel Guard Filter19C320627G1+ 10 VDC+ 10 VDC150k ohmAC SourceVoltmeter0.22 uF5.1k ohm+ 2200 uF 25 V2N39041.2k ohm39k ohm1k ohm10 uF +Filter response measurement circuit that shows connections of the AC Source and Voltmeter along with the interface components to the filter.Response curve follows:Overall this is the best curve for low-end pass band response and rejection of CTCSS tones of the five filters measured for this paper. The –3 dB or corner frequency is 253 Hz and at 205.3 Hz, the attenuation is –25 dB. The notch attenuation is –48 dB at 185.5 Hz. The attenuation is 29 dB around 140 Hz where the curve rises below the notch frequency.Appendix F:The 5th Order Chebyshev filter was used as a comparison to the filters measured. A pass band ripple of 1 dB was chosen to make the low end or the pass band response peak as close as possible to the corner frequency. The corner frequency is 275 Hz. The Unity-Gain Sailen-Key topology was used for the filter.AC+-Voltmeter+-AC SourceCapacitors - uF Resistors - k ohms5th Order Chebyshev - 1 dB Pass Band Ripple - HPF14.72263096.49 2.490.010.010.010.020.02As in any higher order filter, the use of precision components is required to obtain the expected response. This filter was constructed with 1% resistors and capacitors. Also due to the high value resistors, the op-amps should be of the high input impedances types such as FET or CMOS input. Op-amp power and biasing are not shown in this schematic.Resistors with 1% tolerance are readily available, but 1% capacitors are not. I did find 2% film capacitors that should work well but they are not surface mount.Some data for the 5th Order Chebyshev filter is 275 Hz for the –3 dB or corner frequency. The attenuation for the stop band is –20 dB @ 220 Hz –30 dB @ 185 Hz.Appendix G:The response measurements of the filters are measured with a voltmeter and as such are voltage response levels and not actual hearing or listening response levels. The folks in the hearing or loudness business of sound have long had response curves that simulate the typical hearing levels of a human ear. These curves have been around since the early 1930’s and have been standardized in recent time by ISO.One of the items to note is that the ear’s sensitivity to lower frequencies starts to decrease in the 500 Hz range depending on the actual loudness. This normal response of the ear has an effect on what folks hear as compared to the electronic measurements. It may be that for good listening the filter should have a more peaking response at the low end to compensate for the normal response of the ear.。

Compressive data gathering for large-scale wireless sensor networks-包括大规模的负载平衡-sjtu-msra

Feng Wu
Microsoft Research Asia No.49 Zhichun Road Beijing, CHINA 10019 Sun
Shanghai Jiao Tong University No.800 Dongchuan Road Shanghai, CHINA 200240
Compressive Data Gathering for Large-Scale Wireless Sensor Networks
Chong Luo
Shanghai Jiao Tong University No.800 Dongchuan Road Shanghai, CHINA 200240 Microsoft Research Asia No.49 Zhichun Road Beijing, CHINA 100190
The need for energy consumption load balancing is also clear because of the required multi-hop data transmission for such large scale sensor networks. Fig. 1 shows such a network where sensors are densely deployed in the region of interest and monitor the environment on a regular basis. A simple but typical example is the highlighted route in Fig. 1. Suppose N sensors, denoted as s1 , s2 , ..., and sN form a multi-hop route to the sink. Let dj denote the readings obtained by node sj . The intuitive way to transmit dj , j = 1, 2, ...N to the sink is through multi-hop relay as depicted in Fig. 2(a). Node s1 transmits its reading d1 to s2 , and s2 transmits both its reading d2 and the relayed reading d1 to s3 . At the end of the route, sN transmits all N readings to the sink. It can be observed that the closer a sensor is to the sink, the more energy is consumed. Clearly, the sensor nodes closer to the data sink will soon run out of energy and lifetime of sensor network will be signiﬁcantly shortened. This paper presents the ﬁrst complete design to apply compressive sampling theory [13][4][7] to sensor data gathering for large-scale wireless sensor networks (WSNs), successfully addressing the two major challenges as outlined above. First, the proposed data gathering is able to achieve substantial sensor data compression without introducing excessive computation and control overheads. With elegant design, the proposed scheme is also able to disperse the communication costs to all sensor nodes along a given sensor data gathering route. This will result in a natural load balancing and extend the lifetime of the sensor network. The basic idea of the proposed compressive data gathering (CDG) is depicted in Fig. 2(b). Instead of receiving individual sensor readings, the sink will be sent a few weighted sums of all the readings, from which to restore the original data. To transmit the ith sum to the sink, s1 multiplies its reading d1 with a random coeﬃcient φi1 and sends the product to s2 . Upon receiving this message, s2 multiplies its reading d2 with a random coeﬃcient φi2 and then sends the sum φi1 d1 + φi2 d2 to s3 . Similarly, each node sj contributes to the relayed message by adding its own product. Finally, the sink receives N j =1 φij dj , a weighted sum of all the readings. This process is repeated using M sets of diﬀerent weights so that the sink will receive M weighted sums. With such design, all nodes transmit M messages and consume same amount of energy. Each node only performs one addition and one multiplication in order to compute one weighted sum. Comparing Fig. 2(a) and Fig. 2(b), careful

基于ANSYS的结构拓扑优化

基于ANSYS的结构拓扑优化林丹益;李芳【摘要】针对拓扑优化技术在现实中的应用问题,将拓扑优化技术应用到自行车车架和多拱拱桥的最优化设计中.开展了各种拓扑优化方法的分析研究,建立了“以单元材料密度为设计变量,以结构的柔顺度最小化为目标函数,体积减少百分比为约束函数”的数学模型；通过采用商用有限元软件ANSYS中的拓扑优化设计模块对自行车车架和多拱拱桥进行了拓扑优化设计,优化结果表明所得拓扑结构清晰,并与实际的自行车车架和多拱拱桥非常相似.研究结果表明,该结构拓扑优化方法正确而有效,具有一定的工程应用前景.%In order to solve the application problems of topological optimization technology in reality, the bicycle frames and multiple arch bridge was investigated.After the analysis of all kinds of methods of topological optimizaiton, the mathematical model that unit material density as design variables, the minimum of structural compliance as the objective function, the volume reduction percentage as the constraint function was established. The topology optimization design module of the commercial finite element software ANSYS was used to the bicycle frame and multiple arch bridge for the topology optimization design.The topological structure is clear and they are very likely to the bicycle frame and multiple arch bridge in reality. The results indicate that the method is correct and effective, it has a certain engineering application prospect.【期刊名称】《机电工程》【年(卷),期】2012(029)008【总页数】5页(P898-901,915)【关键词】拓扑优化;ANSYS;自行车车架;多拱拱桥【作者】林丹益;李芳【作者单位】浙江工业大学机械工程学院,浙江杭州310014;浙江工业大学机械工程学院,浙江杭州310014【正文语种】中文【中图分类】TH112;U4840 引言连续体结构优化按照设计变量的类型和求解问题的难易程度可分为尺寸优化、形状优化和拓扑优化3个层次，分别对应于3个不同的产品设计阶段，即详细设计、基本设计和概念设计3个阶段。

Doherty PA 2004_IMS_Shin

Linear Power Amplifier based on 3-Way Doherty Amplifierwith PredistorterBumjae Shin, Jeonghyeon Cha, Jangheon Kim, Y.Y. Woo, Jaehyok Yi, and Bumman Kim DEPARTMENT OF ELECTRONIC AND ELECTRICAL ENGINEERING AND MICROWAVE APPLICATION RESEARCH CENTER, POHANG UNIVERSITY OF SCIENCE AND TECHNOLOGY, KYOUNGBUK, 790-784, REPUBLIC OF KOREAAbstract This paper presents a 3-way Doherty amplifier with predistorter(PD) for a repeater application. It is implemented using three 60 watts PEP silicon LDMOSFETs and tested using two-tone and one- and two-carrier down-link WCDMA signals. For the two-carrier down-link WCDMA signal, the amplifier provides -49.1 dBc adjacent-channel-leakage-ratio(ACLR) and 10.3 % power-added efficiency(PAE) at an output power 40 dBm which is an improvement of 8.5 dBc in linearity and 2 % in efficiency compard to a similar class-AB amplifier.Ⅰ. IntroductionLinearity is the most important figure of merit for the power amplifiers of CDMA applications, such as IS95, CDMA 2000, WCDMA , and so on. There are many linearity boosting techniques, for example, feedforward, feedback, predistortion technique [1]. Among them, the feedforward technique is still considered to be the most popular and the best performing method. However, it also has many drawbacks, such as complexity, poor efficiency, and large size, which result in cost problems. Recently, digital PD amplifier becomes a very important technique. On the other hand, an analog predistortion technique is a low-cost solution for the moderate performance improvement. It is also a low-power consumption and simple circuit configuration over the feedforward or digital PD [2]. Therefore, for a repeater system which has less stringent linearity requirement and a small power handling than a base station system, the feedforward is not certainly necessary. Microwave Doherty amplifier has been originally proposed to improve the efficiency, but it has been reported that the efficiency and linearity can be improved simultaneously [3]. To enhance the performances, a new load line topology using offset line has been implemented and a linearity enhancement technique has been incorporated by canceling the intermodulations from the carrier and peaking amplifiers [3]. For a microwave N-way Doherty amplifiers with one carrier amplifier and N-1 peaking amplifiers, it has been reported that 2-way Doherty amplifier can deliver a highly enhanced efficiency with some linearity improvement, but 3-way or 4-way Doherty amplifiers improve much more linearity significantly [4]. 3-way Doherty amplifier has a similar efficiency to 3-way class-AB amplifier, but with significantly improved linearity. Moreover, these Doherty amplifiers are very simple and easy to add other linearization techniques, such as predistortion technique.In this paper, we have introduced a microwave 3-way Doherty amplifier with predistorter targeted for a cheap repeater system with 10 watts average power and about -50 dBc ACLR as well as a respectable efficiency. For the amplifier, the dominant harmonic is IMD3 because IMD3 and IMD5 components have been cancelled out simultaneously in the Doherty operation. Therefore, we could adopt a simple 3rd order predistorter. The amplifier has been tested using one- and two-carrier down-link WCDMA signals having 8.6 dB peak-to-average ratio at 0.1 % CCDF and two-tone signals with 5 MHz or 10 MHz spacing. The measured results have been compared with class-AB biased amplifiers as their counterparts.Ⅱ. Design and ImplementationA. 3-way Doherty amplifierThe basic operation principle of the Doherty amplifier has been well described in the literature [1]. The core principle of operation is a load modulation at low power levels by a peaking amplifier. Fig. 1 shows an operational diagram to explain the load modulation mechanism of the 3-way Doherty amplifier.cIpI'Fig. 1. Operational diagram of the 3-way Doherty amplifierIn the figure, I c and I p represent the carrier amplifier and the peaking amplifiers, respectively. From the figure, equation (1) is acquired. In this equation, if the I p becomes zero, theTHIF-52transformed impedance Z c viewed from the current source I c becomes 3R 0 and if the I p becomes two times of I c, Z c becomes R 0.002000,02333,0211/c p c p cc c c p c I I V R c Z I I I I R R R Z Z I I αα⎛⎞′+′⎜⎟==≤≤⎜⎟′′⎝⎠===≤≤+′′+′ (1)Therefore, for the 3-way Doherty amplifier, the load impedance can be modulated from R 0 to 3R 0 according to the value of I p. But, in the actual implementation of the 3-way Doherty amplifier, the load modulation does not occur properly because of the power matching circuits at the amplifiers. Moreover, the impedance viewed from the the common load to the current source I p is not open at a low power level where the peaking amplifier is off. It has been reported that these problems can be solved by the phase offset lines [3].On the other hand, the nonlinear output current of the active devices can be expressed using Taylor series expansion by23123out i i i I gm v gm v gm v =⋅+⋅+⋅+⋅⋅⋅ (2),where v i is an input voltage and gm n ’s are the Nth-order exapansion coefficients of the nonlinear transconductance. The third-order intermodulation distortion(IMD 3) current is mainly generated by the gm 3·v i 3 of (2). The IMD 3 currents generated by the carrier and peaking amplifiers can be cancelled by selecting proper gate biases for the two amplifiers [4]. Fig. 2 presents the large-signal third-order transconductance coefficient (gm 3) curvethrough the gate bias level for general FETs [5].Fig. 2. Large signal gm 3 v.s. gate bias curve of general FETsFor the 3-way Doherty amplifier, which has two peaking amplifiers, the bias of the carrier amplifier is fixed at class-AB or class-A mode and the biases of the peaking amplifiers are adjusted to have perfect IMD 3 cancellation. Generally, the biases of the peaking amplifiers are a deep class-AB mode. For the 3-way Doherty amplifier, the bias point of the peaking amplifier for perfect IMD 3 cancellation is higher than that of 2-way casebecause of two peaking amplifiers used to cancel the IMD 3 of the carrier amplifier. Therefore, the peaking amplifiers of the 3-way Doherty amplifier can be operated more linearly without excessively generating higher order terms and the three amplifiers can be more linear.In this paper, a 2.14 GHz 3-way Doherty amplifier has been implemented using three Motorola’s MRF21060 (60 watts PEP) LDMOSFETs. The inputs have been matched to R 0=50 Ω from their source impedances of Z s =3.547-j3.377 Ω. But their load impedances are matched a little differently from 50 Ω for the optimized performance. The 50 Ω loads are matched to Z L,C =2.841-j1.445 Ω and Z L,P =2.602-j1.930 Ω, the carrier and peaking amplifiers, respectively. Fig. 3 shows a photograph of the implemented 3-way Doherty amplifier.Fig. 3. Photograph of the implemented 3-way Doherty amplifierFrom fig. 3, it is seen that the output matchings of the carrier and peaking amplifiers are somewhat different to improve the linearity maximally. The design process including the offset lines is presented in our previous works [3], [4]. In this experiment, the offset lines of 50 Ω with 0.012λ length are used for both the carrier and peaking amplifiers.B. PredistorterAs shown in the experimental results of the following section, the amplifier performance is slightly off the target, so we have employed a predistorter. Fig. 4 represents the schematic diagram of the implemented predistorter. There are two paths in the predistorter. The upper path represents the fundamental component path and the lower path represents the IM 3 component path. The IM 3 generator consists of 90° hybrid, Schottky diodes, and series RC passive network with a short microstrip delay line. The input signal applied to the IM 3 generator is split into 0° and -90° ports of the 3dB hybrid coupler. The Schottky diodes at 0° port generate IM 3 term and then reflect the IM 3 and fundamental terms into the IN and ISO ports. At the -90° port, the incident signal is also reflected intoIN and ISO ports by the passive network reflector. The reflected fundamental terms of 0° and -90° ports are cancelled out at the ISO port. Accordingly, the IM3 terms can be made on the lower path. The implemented predistoerter generates the IMD3 with a 2dBc/dBm slope according to the output power [2].Delay Line Att.3dB CouplerⅢ. Experimental ResultsThe performance of the 3-way Doherty amplifier with predistorter has been compared with that of a comparable class-AB amplifier using two-tone signals(5 MHz, 10 MHZ spacing) and one-carrier down-link WCDMA signal and two-carrier with 10 MHz spacing. In this experiments, the quiescent drain currents of the carrier and peaking amplifiers are set to 700 mA at V DD=28 V for the class-AB case. For the 3-way Doherty amplifier, the quiescent drain current I D,C of the carrier amplifier is set to 820 mA at V DD=28 V but those (I D,P) of the peaking amplifiers are set to 270 mA at V DD=28 V respectively.Fig. 5 shows the measured ACLRs and PAEs of the class-AB, 3-way Doherty, and 3-way Doherty with predistorter. Fig. 5(a) is for one-carrier WCDMA signal and 5(b) is for two-carrier WCDMA signal with 10 MHz spacing. For the one-carrier WCDMA signal, the ACLRs of the 3-way Doherty and the 3-way Doherty with predistorter are improved by about 10 dB at an output power 40 dBm and the PAEs are improved slightly by about 2 % at the same output power. For the two-carrier WCDMA signal, ACLRs of the 3-way Doherty and the 3-way Doherty with predistorter are improved by 6.8 dB and 8.5 dB at an output power 40 dBm, respectively and the improvement of the PAEs is similar to the one-carrier case. Test results for one- and two-carrier down-link WCDMA signals have been summarized in TableⅠand Ⅱ. As seen by Tables, the ACLR and PAE are more improved by combining the 3-way Doherty amplifier with the predistorter.481216202428323640PAE[%]ACLR[dBc]Pout[dBm](a)481216202428323640PAE[%]ACLR[dBc]Pout[dBm](b)Fig. 5. Measured ACLRs and PAEs of the class AB, the 3-wayDoherty, and the 3-way Doherty with PD. (a) one-carrier down-link WCDMA signal (b) two-carrier down-link WCDMA signalTable Ⅰ. Measured performances of the class AB, the 3-wayDoherty, and the 3-way Doherty with PD at an output power 40dBm. (a) one-carrier down-link WCDMA signal (b) two-carrierdown-link WCDMA signal(a)ACLR[dBc]PAE[%]Class AB -40 8.2Doherty -50.1 10.4Doherty with PD -51 10.4(b)ACLR[dBc]PAE[%] Class AB -40.6 8.2Doherty -47.4 10.3Doherty with PD -49.1 10.3Table Ⅱ. Measured performances with -45 dBc ACLR. (a) one-carrier down-link WCDMA signal (b) two-carrier down-link WCDMA signal(a)(b)I M D [d B c ]Pout[dBm]Pout[dBm]I M D [d B c ](b)Fig. 6. Measured IMD 3s and IMD 5s of the class AB, the 3-way Doherty, and the 3-way Doherty with predistorter. (a) 5 MHz spacing (b) 10 MHz spacingFig. 6 explains the linearity boosting mechanism of the 3-way Doherty amplifier and the 3-way Doherty amplifier with predistorter. For the 3-way Doherty amplifier, IMD 3 and IMD 5 have been improved simultaneously compared to the class-AB case. On the other hand, for the 3-way Doherty amplifier withpredistorter, IMD 3 has been further improved at high powerlevels than the 3-way Doherty amplifier. Fig. 7 shows the power spectra of the class AB and the 3-way Doherty with predistorter at an output power 40 dBm.P out [dBm] PAE[%] Doherty 41.8 13.6Doherty with PD42.6 15.32.112.122.132.142.152.162.17-70-60-50-40-30-20-10Pout=40 dBmPD+DohertyClass ABP S D [d B m , 10d B /d i v .]Freq.[GHz](b) P out [dBm] PAE[%] Doherty 41.2 12.2 Doherty with PD 42.2 14.2Fig. 7. Power spectral densities at an output power 40 dBm for atwo-carrier down-link WCDMA signalⅣ. ConclusionsFor a repeater system application, a 3-way Doherty with predistorter has been proposed. For one- and two-carrier down-link WCDMA signals, the proposed one has ACLR of -51 dBc, -49.1 dBc and PAE of 10.4 %, 10.3 % respectively at output power 40 dBm, which are improvement of 11 dBc, 8.5 dBc in linearity and about 2.2 %, 2.1 % in PAE, respectively, compared to the class-AB case at the same output power. The experimental results show that the 3-way Doherty amplifier with predistorter is a good alternative for a cheap repeater system that requires less stringent linearity requirement and small power handling than a base station system.References[1] S.C. Cripps, RF Power Amplifiers for WirelessCommunications , Artech House Inc., Norwood, MA, 2000. [2] T. Nojima and T.Konno, “Cuber predistortion linearizer forrelay equipment in 800 MHz band land mobile telephone system,” IEEE Trans.Veh. Technol., vol.VT-34, pp. 169-177, Nov. 1985.[3] Y. Yang, J. Yi, Y.Y. Woo, and B. Kim, "Optimum Designfor Linearity and Efficiency of Microwave Doherty Amplifier Using a New Load Matching Technique," Microwave Journal, vol. 44, No. 12, pp. 20-36, Dec. 2001. [4] Youngoo Yang, Jeonghyeon Cha, Bumjae Shin, andBumman Kim, “A Fully Matched N-Way Doherty Amplifier With Optimized Linearity,” IEEE Trans. Microwave Theory Tech., vol. 51, no. 3, pp. 986-992, Mar. 2003.[5] J.C. Pedero and J. Perez, “Accurate simulation of GaAsMESFET’s intermodulation using a new drain-source current model,” IEEE Microwave Teory Tech ., vol. 42, pp.25-33, Jan.1994。

基于电力电子开关的混合式固态断路器的结构及其原理分析

基于电力电子开关的混合式固态断路器的结构及其原理分析摘要：随着电力电子技术的发展，为克服传统断路器操作过程拉弧和产生操作过电压的问题，技术人员正在逐步展开一种融合了传统开关和电力电子开关的优点的低损耗、无弧断路器新技术的研究和应用。

本文简要介绍了目前固态断路器的发展现状，从理论上分析了混合式固态断路器的结构和原理，并对主回路拓扑、放电回路、缓冲吸收回路和测控回路进行分析，实现基于电力电子开关的固态断路器的应用，为该类型断路器研究工作提供理论和设计依据。

关键词：混合式；固态断路器；IGBT；结构原理Abstract：With the development of power electronic technology, in order to overcome the problems of arcing and over-voltage in the operation process of traditional circuit breakers, technicians are gradually carrying out the research and application of a new technology of low loss, arc-free circuit breakers, which integrates the advantages of traditional switches and power electronic switches. This paper briefly introduces the current development of solid-state circuit breakers, theoretically analyzes the structure and principle of hybrid solid-state circuit breakers, analyzes the main circuit topology, discharge circuit, buffer absorption circuit and measurementand control circuit, realizes the application of solid-state circuit breakers based on power electronic switches, and provides theoretical and design basis for the research work of this type of circuit breakers.Key words：Hybrid，Solid state circuit breaker，IGBT，Principle of structure0 引言高压断路器是电力系统中重要一次设备，在系统正常运行时起到接通、断开正常负荷电流的作用，在系统发生故障时，要及时切断几十倍的故障大电流，切除故障区域，保护电力系统稳定安全运行[1-3]。

WebViz A Tool for World-Wide Web Access Log Analysis

WEBVIZ: A TOOL FOR WORLD-WIDE WEB ACCESS LOG ANALYSISJames E. Pitkow & Krishna A. BharatGraphics, Visualization and Usability CenterCollege of ComputingGeorgia Institute of TechnologyAtlanta, GA 30332-0280E-mail {pitkow, kb}@ABSTRACTVarious programs have emerged that provide statistical anal-ysis of World-Wide Web (WWW) access logs. These pro-grams typically detail the number of accesses for a ﬁle, the number of times a site has visited the database, and some programs even provide temporal analysis of requests1. However, these programs are not interactive nor do they provide visualizations of the local database. WebViz was developed with the intention of providing WWW database maintainers and designers with a graphical view of their local database and access patterns. That is, by incorporating the Web-Path paradigm into interactive software, users can see not only the documents (represented visually as nodes) in their database but also the hyperlinks travelled (repre-sented visually as links) by users requesting documents from the database. WebViz further enables uses to selec-tively ﬁlter the access log (i.e. restrict the graphical view by specifying the desired domain names or DSN numbers, directory names, and start and stop times), control bindings to graph attributes (i.e. node size, border width and color as well as link width and color can be bound to frequency and recency information), play back the events in the access log (i.e. re-issue the logged sequence of requests), select a lay-out of nodes and links that best presents the database’s struc-ture, and examine the graph at any instant in time. Clearly, WebViz is a useful WWW database utility given that it can provide the user with graphical information about document accesses and the paths taken by users through the database. Such analyses can facilitate structural and contextual changes resulting in a more efﬁcient use of the document space. This paper details the implementation of WebViz and outlines possible future extensions.KEYWORDSvisualization, HTTP, administration, tools, statistics, access logs1. For the purposes of this paper, the terms accesses and document requests willbe used interchangeably.INTRODUCTIONWorld-Wide Web (WWW) database developers, designers, and maintainers have a potentially formable task in analyz-ing the overall efﬁciency of their database. Following in the footsteps of the all-too-common end-user question: “Where am I?” [Nielson, 1990], comes the database-provider ques-tion: “How are people using our database?” The latter ques-tion requires analyses of the structure of the hyperlinks as well as the content of the documents in the database. The end products of such analyses might include 1) the fre-quency of visits per document, 2) the most recent visit per document, 3) who is visiting which document, 4) the fre-quency of use of each hyperlink and 5) the most recent use of each hyperlink. Granted, this list does not include all potentially useful analyses; rather, it provides a starting point for the development of tools to provide such function-ality. Towards this end, we developed a C++ visualization tool (running on SunOS 4.1.3 and X) called WebViz. The next section describes the underlying concept of WebViz, the Web-Path paradigm.WEB-PATH PARADIGMCollections of hypertext documents can be categorized by the underlying topology of links and nodes [Parunak, 1989]. WWW databases are intrinsically directed cyclic graphs. This can be thought of as a web-like structure. Y et most WWW databases reside on ﬁle systems that are explicitly hierarchical, e.g. UNIX TM, Macintosh, V AX, etc. As a result of this incongruence, problems can arises when one attempts to view such databases. WebViz tackles this prob-lem by displaying the database as a directed graph2, with nodes representing separate documents in the database and links representing the hyperlinks, or paths, between docu-ments. When a user “travels” from a source document to a separate destination document via the hyperlink embedded in the source document, a path is said to have been taken3. This path corresponds to the user clicking on the anchor 2. The screen capture presented does not display arrows at the end of links. Thedata structure WebViz uses, however contains directional information. Arrowsare soon to be implemented.point and retrieving the anchor (See Case One in Figure 1). We refer to this scenario as internal referencing (point of origin coming from within the database).Note that since WWW enables users to enter a database via any document (via a known Uniform Resource Locator, or URL [Berners-Lee 1994]), causality between successive document requests is not always decidable. That is, even though there may exist a path between document A and doc-ument B and the access log records a request for document A followed by a request for document B from the same site, it remains a possibility that 1) the user at the site knew the location of both document A and document B and requested each ﬁle separately (See Case Two in Figure 1), or 2) there were two different users logged onto the same site who hap-pened to request document A and document B individually and in that order (See Case Three in Figure 1). That is the users did not click on the hyperlink in A to get to B. We refer to these scenarios as external referencing (point of ori-gin exists outside the database) and dual referencing (points of origin in same address space). Even though the possibil-ity of other cases exists, WebViz assumes the Case One sce-nario for successive document requests. It is this assumption that underlies the algorithm for determining the paths taken by users in the access log.WebViz uses the Web-Path paradigm to display the relations between the access log and the local database. Speciﬁcally, the program displays the documents of the local database and the connections between the documents as a web-like graph structure. Information is gathered from the access log about the number of times documents have been accessed as well as the recency of these accesses. WebViz further infers 3. This contrasts to hyperlinks which point to different location with in the same document. WebViz does not analyze such information since such events are not captured by Hypertext Transfer Protocol (HTTP) servers.paths travelled by users by assuming that successive accesses by each user were internally referenced. The num-ber of times paths were taken as well as the recency of the traversals are also collected by WebViz for display.To recap, WebViz visualizes the collection of hypertext doc-uments as a directed cyclic graph. The links in this web-like structure are referred to as paths, and represent the hyper-links between documents. Nodes represent separate docu-ments. Documents connected by hyperlinks can be successively accessed either internally or externally. By uti-lizing the Web-Path paradigm, WebViz collects frequency and recency information about documents and paths to drive the visualization. We now move onto an explanation of how WebViz creates the visualization. The following sections are arranged in the order that each stage is invoked during pro-gram execution.INITIALIZATIONWebViz currently parses the National Center for Supercom-puting Application’s (NCSA) Hypertext Transfer Protocol (HTTP) 1.0 server access logs. As demonstrated by other access log analyzers, writing separate parsing routines for other HTTP servers is trivial. The sample access log entries below shows that the time of access, the machine name (either hostname or DSN), and requested ﬁle are logged for each transaction. [Tue Mar 8 10:50:25 1994] GET /gvu/intro_gvu.html HTTP/1.0128.37.132.23 [Tue Mar 8 10:51:31 1994] GET /gvu/agenda.html HTTP/1.0 [Tue Mar 8 10:52:01 1994] GET /gvu/agenda_more.html HTTP/1.0 Initially, lookups tables of hostname to DSN and DSN to hostname mappings are read into two separate hash tables. The intent here is to reduce the time consuming task of look-ing up a machine’s DSN or hostname, since the log can con-HyperLink Known URLs HyperLinktain either type of entry (see above example). Hence, the process of looking up hostnames and DSN numbers, which is network dependent and therefore potentially prohibitively slow, is done precisely once for each machine in the access log. Next, the speciﬁed access log is read into memory into a structure we refer to as the Master Log. With each transac-tion read, the hash tables are ﬁrst consulted to see if the map-ping is know and as a last resort, attempts the look up using the appropriate system calls. Once the entire access log has been processed, the time of the ﬁrst and last entry can be extracted from the Master Log for use in the View Control Window.The View Control Window (see Figure 2) enables the user to determine the content of the visualization. Controls are pro-vided the facilitate the selection of speciﬁc directories,domain names, and start and stop times. The directory selec-tion allows for an arbitrary number of directories to be added to the visualization. As in the above example, lets assume that the user only wants to view the access patterns of the “softviz” and “people” directories, the person would add those directories to the selection list. This permits the user to restrict the contents of the web to only include the ﬁles within the speciﬁed directories, hence avoiding visualizing unnecessary ﬁles and directories. Internally referenced docu-ments are also added to the web, though we plan to make it an option to exclude such connections from the visualization.Thus, even though the user may requests to see only access patterns from a speciﬁc directory, additional ﬁles from other directories may be included into the visualization; however,embedded media (images, sounds, etc.) are not added.Similarly, the domain selector enables the user to restrict the visualization to only machines that have accessed the data-base whose hostname or DSN contain the speciﬁed sub-string. This allows the end user to look at the access patterns from local machines, machines from speciﬁc companies,etc. (In the above example, we have restricted the view to three companies, two speciﬁed by hostname and the other by DSN). Clearly, unless complete or nearly complete DSNs are used, ambiguous results will occur, i.e. numerous machines will match and their all accesses will end up in the visualization. Finally, the user can control the start and stop times used for the visualization. Hence, peak periods can be isolated just as easily as longer periods of time for analysis.To summarize, all the variable attributes recorded in the access log (time, machine making request, and requested ﬁle) are subject to user ﬁltering.Once the user has ﬁnished determining the view, the speciﬁ-cations are used to create a copy of the Master Log. This copy, called the View List, contains only the entries from the Master Log that the user desires to visualize. While this list provides enough information to determine the number of visits to a ﬁle and the times the ﬁle was accessed, it does not provide the when and the number of times the path was trav-elled. This information is gathered by creating an Edge List that contains the source ﬁle, the destination ﬁle, the access times for both ﬁles, and the DSN of the machine traversing the path. A previously stated, an path is considered to have been travelled if there exists a path in the web and the same machine is making the successive requests, disregarding the possibility of external references (Case Two of Figure 1)Figure 2: The View Control WindowFigure 3: The WebViz Control Windowand dual references (Case Three of Figure 1). W e do place a time constraint on the interval between accesses of three days. That is, if the interval is greater than three days, we assume the user requested the document via another hyper-link than the one embedded in the source document. The selection of the three day period was not based on any empirical evidence. Next, the local database is processed. LOCAL DATABASE PROCESSINGThe local database is processed to ascertain the structure of the web. The ﬁles in the database are processed one at a time, with processing proceeding recursively through the ﬁle system hierarchy. For each ﬁle being processed, if a cor-responding node does not already exist in the web, a node is added. Currently, each node contains the ﬁle’s name, size, and last modiﬁcation time, though additional information like owner, number and type of embedded media, etc. could be added to facilitate more sophisticated analyses. Files that do not contain Hypertext Markup Language (HTML) are not processed or added to the web at this stage. This deci-sion reﬂects the implicit assigning of roles in HTML. That is, marked-up ﬁles act as either end documents or as inter-mediary documents with paths to other documents, while non HTML ﬁles can only assume end document roles. For each marked-up ﬁle, the contents are parsed and the URLs that point within the database are extracted, with relatively addressed URLs are simpliﬁed into their full path names. If a node does not already exist for the ﬁle, a node is created and inserted into the web. Regardless of document type, a link is added from the processed document to the anchor, since it can be referenced internally and hence of possible analytical interest. At the end of the local database process-ing stage, the structure of the web has been deﬁned as is ready to be displayed.GRAPH LAYOUTGraph layout is an arduous task in any setting - more so in WebViz since there are multiple, possibly conﬂicting inter-ests:Clarity: The layout must make good use of the available space to present the information in an easy to read fash-ion. Occlusion of nodes by other nodes or edges should be avoided.Natural Structure:Hierarchical graphs present a natu-ral structure for embedding. The hierarchy in the web ought to mirrors the ﬁle system hierarchy of the database as far as possible.Presentation: The graph must look presentable. Center-ing, regular spacing between nodes, staggering of nodes to avoid collinearity contribute to this end. A good pre-sentation will minimize the lengths of edges in theFigure 4: WebViz Screen Dumpgraph. This may be incompatible with a hierarchical embedding since home directories which tend to be high up in the hierarchy have plenty of back edges and are best placed near the center of the graph.A good layout will try and do justice to all these criteria in a judicious fashion. Since there is no clear optimization crite-rion many schemes are possible [Rivlin, 1994; Parunak, 1989]. The one we adopt presently is a randomized scheme with greedy placement of nodes. Besides being computa-tionally cheap and easy to implement, the randomization has an added beneﬁt. If a certain embedding is not found satis-factory the scheme can generate a new graph for the user’s consideration. Speciﬁcally, our algorithm is as follows:1. For each node we compute its “depth” in the UNIX TM ﬁle system hierarchy and use it to sort the nodes. Nodes are embedded in decreasing order of depth. As a result nodes that are high up in the hierarchy and have a lot of references will be placed close to their natural position.2. The available screen space is partitioned into compart-ments of uniform size. The number of compartments is of the same order as the number of nodes in the graph. Each compartment will hold at most one of the nodes to be embedded. Note that the partitioning problem is a more complicated in 2D than it is in 1D.3. Compartments are staggered at regular intervals in the X and Y direction to prevent collinearity.4. For each node, twenty random empty compartments are sampled. The node is embedded in the compartment which minimizes a penalty function. The penalty function weights the following criteria:a) The Euclidean distance from the vertical line that par-titions the screen space into two halves. Note that his cri-terion tries to keep the nodes close to the center of the screen.b) The Euclidean distance from a horizontal line thatrepresents the natural position of nodes for the given depth value. This helps place nodes close to their natural position in the ﬁle system hierarchy.c) The Euclidean distance from all adjacent nodes thathave already been embedded. This minimizes the length of edges, i.e. this function attempts to clusters associated nodes.5. When the graph is drawn, edges (presently straight) are drawn before nodes to prevent occlusion. Occlusion of nodes by other nodes is avoided by the compartment scheme which also ensures moderately good usage of the available space.The embedding produced by this scheme seems balanced and presentable. However, there is always room for other scheme, e.g. a tiered scheme that strictly follows the ﬁle sys-tem hierarchy or a scheme that minimizes edge intersections (see Future Work section below). We have used straight edges rather than curved edges to simplify “picking” and speed-up redraw, since the graph needs to redrawn for ani-mation, as edges change thickness and color with the pas-sage of time.VISUAL MAPPINGEventually in a visualization, data (processed or raw) needs to be mapped to visual (or audio) parameters. In our case the visual parameters are the thickness and color of nodes and links. We render labels in a ﬁxed color to maintain readabil-ity. Thickness has a low resolution (4 levels currently) while color provides a much richer level of detail. The two param-eters in each case are mapped to the either recency or fre-quency of access. Formally:1. The recency of access of a node (link) is the time elapsed since the last access (traversal) of the node (link).2. The frequency of access of a node (link) is the number of accesses (traversals) it has suffered since the beginning of the simulation expressed as a percentage of the maximum number of accesses (traversals) of any node (link) in the graph.Since frequency and recency ranges tend to be large and it is desirable that the sensitivity of the mapping be greater for small values than larger values, we use a quasi-logarithmic function to map from four data ranges called “Quartiles” to the visual parameter range.To simplify computation we use a piecewise-linear curve, consisting of four linear segments. In the case of recency we map the real-time duration since the last access to the visual parameter. In case of frequency it is the number of accesses expressed as percentage of the maximum number of accesses in the log.The four “quartiles” are mapped to colors as shown in Fig-ure 5. This intuitively mimics the non-linear cooling curve of a hot body; from white hot to yellow hot through red hot to blue. The initial variation is rapid and corresponds to the ﬁrst quartile. The variation slows down gradually and never quite reaches the end of the 4th quartile. For the 4th Quartile we need a ﬁnite and reasonable upper bound to get some variation. V alues larger than the upper bound map to the end of the 4th Quartile.In the case of thickness, the thickness values 4, 3, 2 and 1 correspond to the quartiles 1, 2, 3 and 4 respectively. Given the mappings in Table 1, we can understand the rela-tionship between the visual attributes of nodes and links and their access history. A white-hot body is intuitively one that has just been touched (if color is mapped to recency) or very frequently accessed (if color is mapped to frequency). A blue-body was touched a long time ago (if color is mapped to recency) or very infrequently (if color is mapped to fre-quency). In the case of recency there is a real-world corre-spondence. A white-hot body must have been touched within the last minute. A blue body on the other hand is one that has not been touched for at least a day.Thickness and color are typically mapped to separate quan-tities. If they are mapped to the same quantity we get redun-dancy, which can be beneﬁcial too. Redundancy can help reinforce the expressiveness of the graph.Now that the steps leading to and the mechanisms behind the visualization have been explained, we next discuss the actual visualization.THE VISUALIZATIONThe visualization is composed of two separate windows, the WebViz Control Window (see Figure 3) and the actual dis-play window (see Figure 44). The former provides the user with controls to adjust the bindings (the left-most buttons), select a speciﬁc time to view (the Percent Complete slider), control the animation (the play and pause icons), and rear-range the layout (the right-most button). The node and link binding buttons can be bound to either frequency or recency information. In Figure 3, the conﬁguration is such that a node’s width corresponds to how often the document was accessed and the node’s color corresponds to the recency of the node’s last access. Similarly, a link’s width represents recency and the link’s width represents frequency. The user can adjust these bindings as well as change the layout at time. Given an embedding on the screen the user can selecta node or a link to get information about it.4. WebViz displays are color. Unfortunately, the conversion process to incorpo-rate the images into this document degraded the quality of the images, i.e. the screen capture of the display window looks dramatically different in paper thanon screen.Temporal manipulation is achieved by either the slider or by the playback controls. The slider enables the user to select a time to be displayed. the playback controls starting and pausing the re-issuing of the events in the access log. Cur-rently the playback is not synchronized with real-time. The simulation takes discrete steps forward in simulation time, and renders views of the graph in sequence. We are consid-ering making the playback adaptive so that it completes in a user-speciﬁed interval of time.FUTURE WORKWhile WebViz achieves the goal of being a tool for WWW access log visualization, extensions to the interface, layout, and available analyses will make WebViz more robust. With respect to the View Control Window interface, we intend to create a timeline widget with two sliders and a rescale option to enable better time manipulation. Also, the direc-tory and domain selectors could use some reﬁnement. The WebViz Control Window needs controls for varying the speed of the simulation well adjusting the amount of time used to advance each frame. Clearly, since the current layout scheme does not allow for different layouts to be repro-duced, mechanisms that enable the user to add and deleted preferred layouts will be developed. Along these lines, we are currently experimenting with additional layout methods. These include 1) a hierarchical layout that mirrors the orga-nization of the database for a ﬁle system perspective, 2) a derivative of the multi dimensional scaling node placement algorithm [Eick, 1993], 3) a layout that uses a node’s rela-tive out centrality (ROC) as the primary placement determi-nant [Rivlin, 1994], and others. Direct manipulation of a node’s placement will be added for all selected layout algo-rithms.We are also experimenting with more sophisticated analy-ses. For instance, while it is useful to know how many times documents have been accesses and when, presenting the access information based upon “what’s hot and what’s not”, (i.e. collapsing the recency and frequency information for each document and path into a quantitizable number), might also prove useful. This type of information answers the question: “What is the most popular document at a given moment?” To implement this, we are considering adding a connectionist form of short term memory, most likely a gamma based memory [Mozer, 1993]. We are also curious as to the predictive nature of this architecture.Another tools under consideration is based upon initial evi-dence that the most frequently and recently accessed docu-ments in a given time window (i.e. day 0 through day 7) will be accessed on the day immediately following the time win-dow (i.e. day 8) [Recker, 1994; Anderson, 1991]. Hence, once the access pattern has been established for a database, subsequent access could be compared to the expected value and displayed as positive/negative deviations.Finally, given that all salient information is readily avail-able, WebViz can provided tab delineated ﬁle dumps for use by spreadsheets and graphing software. Hence, users can get output for speciﬁc ﬁles, paths, time intervals, etc. in a pointand click manner. While other access log analyzers provide similar functionality as far as document accesses, WebViz provides path analysis in an interactive environment. Thus, users can visualize the data and isolate speciﬁc patterns before deciding upon dumping to ﬁle for close inspection. CONCLUSIONSMotivated by providing useful analyses of WWW access logs, we developed WebViz. Towards this end, we enable database designers and maintainers to visualize the data-base’s document space and re-issue events from the access log. By providing the user with controls to adjust the bind-ings properties of nodes and links, access patterns can be inferred. These accesses patterns can contribute to structural and contextual changes in the database.ACKNOWLEDGEMENTSThe authors extend their appreciation to the Graphics, Visu-alization, and Usability Center and its members for their support and assistance, especially John Stasko. REFERENCESAnderson, John R. & Schooler, Lael J. (1991) Reﬂec-tion of the environment in memory.American Psycho-logical Society, 2, 62. 396-408.Berners-Lee, T. (1994)Uniform Resource Locators. Internet Engineering Task Force Working Draft, 21 March 1994. URL:ftp:///internet-drafts/ draft-ietf-uri-url-03.txtEick, Stephen G. & Willis, Graham J. (1993) Navigat-ing large networks with hierarchies.Proceedings of IEEE Visualization Conference, 1993.204-210. Nielson, Jakob. (1990) The art of navigating through munications of the ACM 33, 3.296-310.Mozer, Michael C. (1993) In A. S. Weigend & N. A. Gershenfeld (Eds.).SFI Studies in the Sciences of Complexity, Proc. Vol XV. Addison-Wesley. Parunak, H. Van Dyke. (1989) Hypermedia topologies and user navigation.Hypertext ‘89 Proceedings.43-50.Recker, Margaret M. & PItkow, James E. (in prepara-tion) Predicting document access in large, multimedia repositories: a www case study.Rivlin, Ehud & Botafogo, Rodrigo & Shneiderman, Ben. (1994). Navigating in hyperspace: designing a structure-based munications of the ACM 37, 2.87-96.。

1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。

1Design,Implementation,andTestingofa

HybridToolforNetworkTopologyDiscovery

AlessioBotta,WalterdeDonato,AntonioPescap´e,andGiorgioVentreUniversityofNapoli“FedericoII”,Italy{a.botta,walter.dedonato,pescape,giorgio}@unina.it

AbstractInthisTechnicalReportwereportouractivitiesontheDesign,Implementation,andTestingofaHybridToolforNetworkTopologyDiscoverywecalledHynetd.

I.INTRODUCTION

Computernetworksarebecomingevermoreubiquitousand,asaconsequence,moreandmorecomplex.Theknowledgeofthetopologyofanetworkallowstoimproveitsmanagementandtoexecutemoreaccuratesimulations.Intheﬁrstcase,substantialadvantagesinfaultmanagement,performanceanalysisandserviceallocationareobtainable.Inthesecondone,thereisabeneﬁtbecauseoftheautomaticgenerationofrealistictopologiesisadifﬁculttask[1].Inaddition,duetodynamicbehaviorandlargesizeofrealnetworktopologies,thediscoveryprocesshastobenecessarilyperformedinanautomaticfashionanditshouldsupplycompleteandcorrectresultsassoonaspossible,generatingasfewtrafﬁcaspossible.Inthepastyears,severaltechniquesandtoolshavebeenproposed.WehaveclassiﬁedtheminActive,Passive,andHybrid.TheﬁrstisbasedontoolssuchasPingandTraceroute,inferringtopologyinformationfromnetworkbehavior,andthesecondoneusesSNMPandtoobtaininformationfromdevices.Usingactivemethodologyintroducestwoadditionalproblems:(i)recognizingtheinterfacesbelongingtothesamenetworkdevice,problemknownasaliasresolution;(ii)reconstructingcorrectsubnetidsandnet-masks.Finally,wereferasHybridatoolthatusesbothmethodologies.Inourpreviouswork[2]weproposedapreliminaryversionofanalgorithmimplementedinHynetd(v.0.1),ahybridtoolthatassumesminimumprerequisitesonthenetwork(i.e.oneormoreaddressrangestobescanned).InthistechnicalreportwepresentthenewHynetdversion(v.0.2)havinganewandamoreefﬁcientarchitectureandcontainingsometechniquestoimprovediscoveryefﬁciencyandperformance.Moreprecisely,weintroducenewapproachesandmodiﬁedtheﬁrstversioninorderto:(i)performthestepsfollowedbythealgorithminaparallelfashion;(ii)reducetheredundancyoftheinformationcollectedand,therefore,(iii)toreducethetrafﬁcoverhead;(iv)improvethealiasresolutionphase;(v)increasetheaccuracyofthelinkreconstructionphase.Todemonstratetheimprovedperformanceofthenovelapproach,weprovideresultsofacarefulcomparisonwithboththeﬁrstversionandwithNetworkView3.5[25],acommercialtopologydiscoverysoftware.Theexperimentalevaluation

March28,2007DRAFT2hasbeencarriedoutintermsofaccuracy,trafﬁcoverhead,anddiscoverytimeovertwonetworkscenarios(smallandlargescaletopologies).Therestofthepaperisorganizedasfollows.Section2presentsareviewofthemostrelevantworksabouttopologydiscovery.InSection3thedesignofHynetd0.2isshown.Section4explainsitsarchitecture,innovations,andimprovements.InSection5wepresentsmallandlargescaleexperimentalanalysis.Finally,Section6endsthepaperwithconclusionsandissuesforresearch.

II.REVIEWINGTHELITERATURE

Since1993manyworksdealingwithTopologyDiscoveryhavebeenpublished.Theydifferintermsofmethod-ologyused,numberofemployedprobes,prerequisitesandexploredprotocollayers.InthissectionwerevisesuchworksaccordingtotheclassiﬁcationofthemethodologiesweproposedinSectionI.First,wereporttheworksusinganactivemethodology.Theauthorsof[3]proposeanalgorithmnamedInedthatusesPingandTraceroutefromasinglesourcepoint.Theyintroduce,fortheﬁrsttime,twotechniquestoresolvethealiasresolutionproblemwhicharebasedonDNSinverselook-upandthesourceaddressofICMPportunreachablepackets(knownasSourceAddresstechnique)respectively.TheSkitterproject[7]facestheproblemofInternettopologydiscoveryattwodifferentlayers,thatareIPandAutonomousSystems(AS).Oneofitscomponents,namelyifﬁnder,performsthealiasresolutionbyusingacombinationoftheSourceAddressandthePingwithRecordRouteoption(PingRRinthefollowing).MoreoverthisistheonlyprojectsupportingtheTracerouteIPoption[8],evenifitisnotwidelysupportedbyoperationaldevices.Mercatorproject[11]triestodiscovertheInternettopologyusingactiveonlymethodologies.Itistheﬁrsttoexploitsourceroutinginordertoincreasethenumberofdiscoveredinterfacesandtorevealtransversallinks.Barfordetal.in[12],byusingtheTraceroutetool,studytheutilityofaddinginformationsourceswhenperformingwide-areameasurementsinthecontextofInternettopologydiscovery.Theyshowthattheutilityofadditionalmeasurementsitesrapidlydeclinesevenaftertheﬁrsttwosites.TheRocketfuelproject[15]isaimedtodiscoverISPtopologiesandpresentsacentralizedarchitecturethatemploysmanyprobes.Itintroducesanewtechniquetoresolvealiases,namedAllyalgorithm,thatexploits‘id’ﬁeldofIPheaderextractedfromICMPportunreachableerrorpackets.In[17]MagoniandHoerdtpresentNec(NetworkCartographer)whoseaimistomaptheheartoftheInternetasfastaspossiblewiththehighestattainableaccuracy.Theirapproachfocusesonroutersandlayer3linksandintroducessomeheuristicstominimizethenumberofIPaddresspairsinvolvedinaliasresolution.In[19]GuneseSarac,afterananalysisofallaliasresolutiontechniquesproposedinliterature,introduceanewmethodthatexploitsthepartialsymmetryofroutesfromasourcetoadestination.ThisapproachdoesnotgenerateadditionaltrafﬁcoverheadbecauseitonlyusesTraceroutediscoveredpaths.Traceroute@home[16]introducestheDoubletreealgorithmwhoseefﬁciencyhasbeenprovedtoreducetrafﬁc