Kern - TOPAS 0 Intro

基于Laplacian算子的图像增强
孙增国;韩崇昭
【期刊名称】《计算机应用研究》
【年(卷),期】2007(24)1
【摘要】使用Laplacian算子检测图像的边缘纹理等细节信息,然后以适当比例线性叠加原始图像和细节信息,从而完成图像增强.不同增强方法的比较试验表明,基于Laplacian算子的图像增强方法既能增强图像的高频分量,又能保持图像的低频分量,是图像增强的有效方法.
【总页数】3页(P222-223,240)
【作者】孙增国;韩崇昭
【作者单位】西安交通大学,电子与信息工程学院,陕西,西安,710049;西安交通大学,电子与信息工程学院,陕西,西安,710049
【正文语种】中文
【中图分类】TP391
【相关文献】
1.基于Laplacian算子的扩大比对范围的边缘检测算法 [J], 赵静
2.基于Laplacian算子的小波变换图像融合算法 [J], 谢红;王石川;解武
3.基于Laplacian金字塔和小波变换的医学CT图像增强算法 [J], 吕鲤志;强彦
4.基于Laplacian算子和灰色关联度的图像边缘检测方法 [J], 桂预风;吴建平
5.基于Laplacian算子的同态滤波算法研究 [J], 申晓彦;韩焱
因版权原因，仅展示原文概要，查看原文内容请购买。

Page 1/11Data Sheet 702040Blocks tructureFeature sk S tructured operating and programming layout k S elf-optimi s ation k Ramp functionk Timer functionk Digital input filter withprogrammable filter time con s tant k 1 limit comparator k limit s witchJUMO iTRON 04/08/16/32Compact microproce ss or controller sHou s ing for flu s h-panel mounting to DIN IEC 61554Brief de s criptionThe JUMO iTRON controller s erie s compri s e s univer s al and freely programmable com-pact in s trument s for a variety of control ta sks . It con s i s t s of five model s , with the bezel s ize s 96mm x 96mm, 96mm x 48mm in portrait and land s cape format, 48mm x 48mm and 48mm x 24mm.The controller s feature a clearly readable 7-s egment di s play which, depending on the ver-s ion, i s 10 or 20 mm high, for proce ss value and s etpoint indication or for dialog s . Only three k ey s are needed for configuration. Parameter s etting i s arranged dynamically, and after two operation-free s econd s the value i s accepted automatically. S elf-optimi s ation,which i s provided a s s tandard, e s tabli s he s the optimum controller parameter s by a k ey s tro k e. The ba s ic ver s ion al s o include s a ramp function with adju s table gradient s . A timer function ha s been integrated a s an extra.All controller s can be employed a s s ingle-s etpoint controller s with a limit comparator, or as double-s etpoint controller s . The lineari s ation s of the u s ual tran s ducer s are stored. Pro-tection i s IP66 at the front and IP20 at the bac k . The electrical connection i s by a plug-in connector with s crew terminal s .The input s and output s are s hown in the bloc k s tructure below.JUMO iTRON 08Type 702042JUMO iTRON 04Type 702044JUMO iTRON 08Type 702043JUMO iTRON 32Type 702040JUMO iTRON 16Type 702041Approval s /approval mark s (s ee "Technical data")Data Sheet 702040Page 2/11Technical dataThermocouple inputRe s i s tance thermometer inputStandard s ignal inputMea s urement circuit monitoring 1De s ignation Range 1Mea s urement accuracy Ambienttemperature error Fe-Con L Fe-Con J EN 60584Cu-Con U Cu-Con T EN 60584NiCr-Ni K EN 60584NiCr S i-Ni S i N EN 60584Pt10Rh-Pt S EN 60584Pt13Rh-Pt R EN 60584Pt30Rh-Pt6Rh BEN 60584-200to +900°C -200to +1200°C -200to +600°C -200to +400°C -200to +1372°C -100to +1300°C0— 1768°C 0—1768°C +300— 1820°C≤0.4%≤0.4%≤0.4%≤0.4%≤0.4%≤0.4%≤0.4%≤0.4%≤0.4%100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°CCold junctionPt 100 internal1. The s e range s refer to the ambient temperature of 20°CDe s ignation Connection type RangeMea s urement accuracy Ambienttemperature error Pt 100 EN 607512-/3-wire -200to +850°C ≤0.1%50 ppm/°C Pt 1000 EN 607512-/3-wire -200to +850°C ≤0.1%50 ppm/°C K TY11-62-wire-50to +150°C≤1.0%50 ppm/°CS en s or lead re s i s tance 20Ω max. per lead for 2- and 3-wire circuitMea s urement current 250µALead compen s ationNot required for 3-wire circuit. For 2-wire circuit, lead compen s ation can be implemented in s oftware through proce ss value correction.De s ignation RangeMea s urement accuracy Ambienttemperature error Voltage0—10V , input re s i s tance R E > 100k Ω2—10V , input re s i s tance R E > 100k Ω0—1V , input re s i s tance R E > 10M Ω10,2—1V , input re s i s tance R E > 10M Ω1≤0.1%≤0.1%≤0.1%≤0.1%100 ppm/°C 100 ppm/°C 100 ppm/°C 100 ppm/°C Current4—20mA, voltage drop 3V max.0—20mA, voltage drop 3V max.≤0.1%≤0.1%100 ppm/°C 100 ppm/°C1. for Type 702040/41 with 2 relay output s (option)Tran s ducer Overrange/underrangeProbe /lead s hort-circuit 1Probe/lead breakThermocouple•-•Re s i s tance thermometer •••Voltage 2—10V / 0.2—1V0—10V/ 0—1V •••-•-Current4—20mA 0—20mA•••-•-1. In the event of a fault, the output s move to a defined s tatu s (configurable).= factory s etting •recogni s ed-not recogni s edData Sheet 702040Page 3/11Output sControllerTimerElectrical dataHou s ingA ss ignment Type 702040/41Type 702042/43/44Output 1relayrelay Output 2logic 0/5V or logic input logic 0/5V Output 2 (option)logic 0/12V or logic input logic 0/12V Output 2 (option)relay not po ss ible Output 3not availablerelayTechnical data Relay ratingcontact life n.o. (ma k e) contact 3A at 250VAC re s i s tive load 150 000 operation s at rated loadLogiccurrent limiting load re s i s tance 0/5V 20mAR load 250Ω min .Logiccurrent limiting load re s i s tance 0/12V 20mAR load 600Ω min.= factory s ettingController types ingle-s etpoint controller with limit comparator, double-s etpoint controllerController s tructure s P/PD/PI/PIDA/D converter re s olution better than 15 bitS ampling time210m s ec/250m s ec with activated timer functionAccuracy0.7% ± 10ppm/°CS upply (s witch-mode power s upply)110—240V -15/+10%AC 48—63Hz, or 20—30V AC/DC 48—63Hz, or10—18V DC (Connection to S ELV or PELV)Te s t voltage s (type te s t)to EN 61010, Part 1, March 1994,overvoltage category II, pollution degree 2, for Type 702040/41overvoltage category III, pollution degree 2, for Type 702042/43/44Power con s umption max. 7VA Data bac k upEEPROMElectrical connectionat the rear, via plug-in s crew terminal s ,conductor cro ss -s ection up to 1.5mm 2 (1.0mm 2 for Type 702040/41) or2x 1.5mm 2 (1.0mm 2 for Type 702040/41) with ferrule sElectromagnetic compatibility interference emi ss ion interference immunity EN 61 326Cla ss Bto indu s trial requirement sS afety regulationto EN 61010-1In s tallation height maximum 2000 m above s ea levelCa s e typePla s tic ca s e for panel mounting acc to. IEC 61554 (indoor u s e)Dimen s ion s in mm (for Type)702040702041702042702043702044Bezel s ize48x 2448x 4848x 96(portrait)96x 48(land s cape)96x 96Depth behind panel100100707070Panel cut-out45+0.6x 22.2+0.345+0.6x 45+0.645+0.6x 92+0.892+0.8x 45+0.692+0.8x 92+0.8Ambient/s torage temperature range 0—55°C /-40 to +70°CData Sheet 702040Page 4/11Approval s /approval mark sDi s play and control sSelf-optimi s ation (SO)The s tandard s elf-optimi s ation facility produce s an automatic adju s tment of the controller to the proce ss .S elf-optimi s ation determine s the controller parameter s for PI and PID controller s (proportional band, re s et time, derivative time), a s well a s the cycle time and the filter time con s tant of the digital input filter.Ramp functionClimatic condition s not exceeding 75% rel. humidity, no conden s ationOperating po s ition unre s tricted Protection to EN 60529,IP66 at the front, IP20 at the bac kWeight75g approx.95g approx.145g approx.160g approx.200g approx.Approval mark Te s ting agencyCertificate/certification numberIn s pection ba s i sValid forUL Underwriter Laboratorie s E201387UL 61010-1all device s C S A C S A-Approval232831CAN/C S A-C22.2No. 61010.1-04all device sData Sheet 702040Page 5/11Limit comparatorLimit s witch (extra code)If the limit comparator function i s active, then the s witched s tate will have to be re s et by hand.Precondition: the condition that cau s ed the alarm i s no longer pre s ent (for l k 8: proce ss value < AL). The di s play s how s the alarm s tatu s .The alarm s tatu s will be retained after a power failure.Timer function (extra code)U s ing the timer function, the control action can be influenced by mean s of the adju s table time t i 0. After the timer ha s been s tarted by power ON, by pre ss ing the k ey or via the logic input, the timer s tart value t i 0 i s counted down to 0, either in s tantly or after the proce ss value ha s gone above or below a programmable tolerance limit. When the timer ha s run down, s everal event s are triggered, s uch a s control s witch-off (output 0%) and s etpoint s witching. Furthermore, it i s po ss ible to implement timer s ignalling during or after the timer count, via an output.The timer function can be u s ed in conjunction with the ramp function and s etpoint s witching.Table: Timer function s (u s ing the example of a rever s ed s ingle-s etpoint controller)Data Sheet 702040Page 6/11Tolerance limitThe po s ition of the tolerance limit depend s on the controller type:- S ingle-s etpoint controller (rever s ed, heating): Tolerance limit i s below the s etpoint - S ingle-s etpoint controller (direct, cooling): Tolerance limit i s above the s etpoint - Double-s etpoint controller: Tolerance limit i s below the s etpointIf, during the control proce ss , the proce ss value goe s above/below the tolerance limit, then the timer will be s topped for the duration of the infringement.Di s play and operationThe timer value i s di s played at the operating level and remain s s o permanently (no time-out).Operation i s from the k eypad, when the timer value i s vi s ible in the di s play, or via the logic input. The operating option s compri s e s tart,s top, continue and cancel timer function, and are s hown differently in the di s play.The current timer value and the timer s tart value are acce ss ible and adju s table at any time at a s eparate timer level.Data Sheet 702040Page 7/11Parameter and configurationOperating levelParameter levelConfiguration levelDe s ignation Di s play Factory s ettingValue range S etpointSP /SP1/SP20S PL—S PH Ramp s etpointSPr 0S PL—S PH Timer value/timer s tart valuet i /t i 00 —999.9hDe s ignation Di s play Factory s ettingValue range S etpoint 1SP 10S PL—S PH S etpoint 2SP 20S PL—S PH Limit value for limit comparator AL 0-1999to +9999digit Proportional band 1Pb:100—9999digit Proportional band 2Pb:200—9999digit Derivative time dt 80s ec 0—9999s ec Re s et time rt 350s ec 0—9999s ec Cycle time 1CY 120.0s ec 1.0—999.9s ec Cycle time 2CY 220.0s ec1.0—999.9s ec Contact s pacingdb 00—1000digit Differential (hy s tere s i s ) 1HYS.110—9999digit Differential (hy s tere s i s ) 2HYS.210—9999digit Wor k ing point Y:00%-100to +100%Maximum output Y:1100%0to 100%Minimum output Y:2-100%-100to +100%Filter time con s tant dF 0.6s ec 0.0—100.0s ec Ramp s loperASd—999digitDe s ignation Di s play Factory s etting Value range/s electionTran s ducerC111Pt100Pt100, Pt1000, K TY11-6, T, J, U, L, K , S ,R, B, N, 0 (4)—20mA, 0 (2)—10VDecimal place/unitC112none/°C none, one, two/°C, FController type/output s C113s ee table on next pageLimit comparator function C114no function no function, l k 1—8Ramp functionC115no function no function, °C/min, °C/h Output s ignal on overrange/ underrange C1160% output limit comparator off 0%, 100%, -100% limit comparator on/offLogic inputC117no function k ey / level inhibit,ramp s top, s etpoint s witchingOutput s 1, 2 and 3(only Type 702042/43/44)C118function s a s defined under C113freely configurable(s ee table on next page)Timer functionC120no function s ee de s cription “Timer function”S tart condition for timerC121from k eypad/logic input - power ON - k eypad/logic input- tolerance limitTimer s ignalling C122no function - timer s tart to timer run-down- after run-down for 10s ec - after run-down for 1 min.- after run-down until ac k nowledgementUnit of time (timer)C123mm.ss - mm.ss- hh.mm - hhh.hS tart value of value range SCL0-1999 to +9999 digitData Sheet 702040Page 8/11Controller type/output s (C 113)Expanded configuration option s for the output s on Type 702043/44 (C118)End value of value range SCH 100-1999 to +9999 digit Lower s etpoint limit SPL -200-1999 to +9999 digit Upper s epoint limitSPH 850-1999 to +9999 digit Proce ss value correction OFFS 0-1999 to +9999 digit Differential (hy s tere s i s )HySt 1 0—9999 digitController typeOutput 1Output 2 + 3S ingle s etpoint rever s ed controller limit comparator/timer s ignalling S ingle s etpoint direct controllerlimit comparator/timer s ignallingDouble s etpointcontroller rever s edcontroller direct S ingle s etpoint rever s ed limit comparator/timer s ignalling controller S ingle s etpoint direct limit comparator/timer s ignallingcontrollerDouble s etpoint controller directcontroller rever s ed= factory s ettingOutput 1: Relay (K1)Output 2: Logic (K2)Output 3: Relay 1-s e t p o i n t c o n t r o l l e rFunction s of the output s a s defined under C 113controller output limit comparator timer s ignalling controller output timer s ignalling limit comparator limit comparator controller output timer s ignalling limit comparator timer s ignalling controller output timer s ignalling controller output limit comparator timer s ignalling limit comparator controller output 2-s e t p t .c o n t r o l l e rcontroller output 1controller output 2limit comparator/timer controller output 1limit comparator/timer controller output 2controller output 2controller output 1limit comparator/timer controller output 2limit comparator/timer controller output 1limit comparator/timer controller output 1controller output 2limit comparator/timercontroller output 2controller output 1Data Sheet 702040Page 9/11Dimen s ion sType 702040 / …Type 702043/...Type 702041 / …Type 702044/...Type 702042 / …Typehorizontal vertical 70.2040/418mm min.8mm min.70.2042/43/4410mm min.10mm min.Edge-to-edge mounting(minimum s pacing s of the panel cut-out s)Data Sheet 702040Page 10/11Connection diagram sJUMO iTRON 32, Type 702040, 48mm x 24mm format JUMO iTRON 16, Type 702041, 48mm x 48mm formatJUMO iTRON 08, Type 702042, 48mm x 96mm format (portrait)JUMO iTRON 08, Type 702043. 96mm x 48mm format (land s cape)JUMO iTRON 04, Type 702044, 96mm x 96mm formatStandard ver s ion / Ver s ion with 12V logic outputVer s ion with 2 relay outputs2014-09-01/00357838Data Sheet 702040Page 11/11Order detail sExtra order code s for cu s tomized configuration(2)Ba s ic type exten s ion(3)Input s(1)(2)(3)(4)(5)(6)Type de s ignation 7020../..-...-...-../...,...** Li s t extra code s in s equence, s eparated by comma s(1)Ba s ic type(bezel s ize in mm)40=48x 24, 41 = 48x 48, 42 = 48x 96 (portrait), 43 = 96x 48 (land s cape), 44 = 96x 96(2)Ba s ic typeexten s ion 8899==controller type configurable 1controller type configured to cu s tomer s pecification 2(3)Input s 888999==input s configurable 1input s configured to cu s tomer s pecification 2(4)Output s000=S tandardType 702040/41Type 702042/43/44Output 1relay (n.o. ma k e)relay (n.o. ma k e)Output 2logic 0/5V , optionally configurable a s logic input logic 0/5V Output 3not available relay (n.o. ma k e)Option sType 702040/41Type 702042/43/44113=Output 2(output s 1+3 a s for S tandard)logic 0/12V , optionally configurable a s logic input logic 0/12V 101=Output 2(output 1 a s for S tandard)relay (n.o. ma k e)(logic input i s alway s available)not po ss ible(5)Supply162523===10—18V DC20—30V AC/DC 48—63Hz110—240V AC -15/+10% 48—63Hz (6)Extra code069=UL and C S A approval 210=Timer function220=Timer function + limit s witch 3Delivery package ex-factory for Type 702040/41Type 702042/43/441 mounting frame2 mounting brac k et s1 s eal, 1 Operating In s truction s 70.20401. s ingle-s etpoint with limit comparator, s ee factory s etting s under configuration and parameter level2. s ee extra order code s (below) or factory s etting s under configuration and parameter level3. The linearization s for K TY11-6 and thermocouple B have been deletedController typeOutput 1Output 2 and 310=s ingle s etpoint rever s ed 1controller limit comparator/timer s ignalling 11=s ingle s etpoint direct 2controllerlimit comparator/timer s ignalling 30=double s etpoint controller rever s edcontroller direct 20=s ingle s etpoint rever s ed 1limit comparator/timer s ignalling controller 21=s ingle s etpoint direct 2limit comparator/timer s ignalling controller33=double s etpointcontroller directcontroller rever s ed1. controller output i s active when proce ss value i s below s etpoint, e. g. heating2. controller output i s active when proce ss value i s above s etpoint, e. g. cooling001=Pt1003-wire 040=Fe-Con J 045=Pt13 Rh-Pt R 063=0—10V 003=Pt1002-wire041=Cu-Con U 046=Pt30 Rh-PtRh B 071=2—10V 005=Pt1000 2-wire 042=Fe-Con L 048=NiCr S i-Ni S i N601=K TY11-6 (PTC)006=Pt1000 3-wire 043=NiCr-Ni K 052=0—20mA 039=Cu-Con T044=Pt10Rh-PtS053=4—20mA= factory-s et。

基于单亲遗传模拟退火算法的顶点p-中心问题蒋建林;徐进澎;文杰【期刊名称】《系统工程学报》【年(卷),期】2011(026)003【摘要】针对顶点p-中心问题这一经典的离散选址NP困难问题提出了一种单亲遗传和模拟退火的混合算法.该算法:1)采用单亲遗传算法简化遗传操作过程;2)加入模拟退火策略,增强局部优化能力;3)提出自适应选择法,根据个体的优劣及算法迭代情况来选择个体;4)设计了自适应基因重组操作;5)采取最优保存策略,避免最优解的丢失.数值实验结果表明了该算法对于解决规模较大的顶点p-中心问题的有效性.%Aiming at p - center problem such a NP-hard discrete location problem a new hybrid algorithm with partheno-genetic algorithm ( PGA) and simulated annealing algorithm ( SA) is proposed. First, the hybrid algorithm utilizes PGA to simplify genetic operation process. Second, it adds SA strategy to the PGA to improve the local searching ability. Third, it puts forward an adaptive selection and selects individual according to the value of individual fitness and iteration of algorithm. Fourth, it designs an adaptive gene recombination operation. Fifth, it employs the elitist strategy to avoid the loss of the best solution. Experimental results show that this algorithm is efficient for solving relatively large vertex p-center problem.【总页数】7页(P414-420)【作者】蒋建林;徐进澎;文杰【作者单位】南京航空航天大学理学院,江苏南京210016;南京航空航天大学理学院,江苏南京210016;南京航空航天大学理学院,江苏南京210016【正文语种】中文【中图分类】O221;TP18;TB114.1【相关文献】1.基于P-中心法的农资配送中心选址研究——以辽宁省昌图县为例 [J], 赵小明;王利2.基于改进单亲遗传算法的军事物流中心选址问题研究 [J], 李振东;张启义;陈亮3.基于模拟退火算法的B2C企业配送中心选址问题探讨 [J], 苏兴国;胡玥4.基于模拟退火算法的B2C企业配送中心选址问题探讨 [J], 苏兴国;胡玥5.基于p-中值模型的物流中心选址问题研究 [J], 童旭;梁欢;郑丽娜;王岩焱;罗融宇;;;;;因版权原因，仅展示原文概要，查看原文内容请购买。

topsis熵权方法 r语言
TOPSIS（Technique for Order Preference by Similarity to Ideal Solution）是一种多属性决策分析方法，用于评估候选方案的优劣。

而熵权法是TOPSIS方法中的一种权重确定方法，用于处理权重的不确定性和模糊性。

在R语言中，可以使用相关的包和函数来实现TOPSIS和熵权法。

首先，要使用TOPSIS方法，可以使用R语言中的"TOPSIS"包。

该包提供了一个名为"TOPSIS"的函数，可以通过计算每个候选方案与理想解决方案的接近程度来进行多属性决策分析。

该函数需要输入候选方案的属性数据矩阵、权重向量以及正负理想解决方案的权重向量。

通过调用该函数，可以得到每个候选方案的综合评分，从而进行排序和选择最优方案。

其次，熵权法可以通过R语言中的"entropy"包来实现。

该包提供了一个名为"entropy"的函数，可以使用熵值法来计算每个属性的权重。

该函数需要输入候选方案的属性数据矩阵，通过计算每个属性的熵值和信息增益比来确定权重。

得到属性的权重后，可以将其用于TOPSIS方法中进行多属性决策分析。

总的来说，在R语言中可以通过使用"TOPSIS"包和"entropy"包
来实现TOPSIS方法和熵权法。

通过这些包提供的函数，可以进行多
属性决策分析并得出最优解决方案。

当然，在使用这些方法时，需
要对数据进行预处理和权重的确定，以确保结果的准确性和可靠性。

一种新的静止图象压缩编码算法
黎洪松;全子一
【期刊名称】《电子科学学刊》
【年(卷),期】1995(017)006
【摘要】本文提出了一种新的静止图象压缩编码算法，即ＶＱ＋ＤＰＣＭ＋ＤＣＴ算法，并与ＪＰＥＧ标准的基本系统进行了比较，实验结果表明，新算法的压缩比有较大提高。

【总页数】8页(P561-568)
【作者】黎洪松;全子一
【作者单位】不详;不详
【正文语种】中文
【中图分类】TN919.8
【相关文献】
1.一种基于二维DCT的分形静止彩色图象压缩编码 [J], 朱艳秋;初连禹;陈贺新
2.一种小波变换与矢量量化结合的图象压缩编码算法 [J], 刘宇;郑善贤;江波涛
3.一种自适应图象压缩编码算法 [J], 张基宏;王晖
4.一种新的分形图象压缩编码方法 [J], 尹忠科;杨绍国
5.一种基于二值图象压缩编码前处理的新观点和新方法 [J], 刘子良;王航;王珂;陈贺新
因版权原因，仅展示原文概要，查看原文内容请购买。

基于DSP的数字调制信号神经网络识别算法
胡正新;王自强;聂文华
【期刊名称】《微处理机》
【年(卷),期】2005(26)4
【摘要】DSP处理模块是软件无线电的核心,TMS320C6701芯片以其优越的性能越来越成为其首选.本文研究了在该芯片上实现数字调制信号的人工神经网络识别算法的方法和过程,并通过实验验证了其正确性和合理性.
【总页数】4页(P44-47)
【作者】胡正新;王自强;聂文华
【作者单位】南京大学电子工程系,南京,210093;南京大学电子工程系,南
京,210093;南京大学电子工程系,南京,210093
【正文语种】中文
【中图分类】TP391.4
【相关文献】
1.基于高阶累积量的数字调制信号识别算法 [J], 包锡锐;吴瑛;周欣
2.基于决策树的调制信号自动识别算法及其DSP实现 [J], 孙闽红;赵太飞;郭勇;王茂芝
3.基于近似对数似然比的数字调制信号自动识别算法 [J], 张炜;杨虎;张尔扬
4.基于卷积神经网络的多进制相位调制信号识别算法 [J], 吴佩军; 侯进; 吕志良; 刘雨灵; 徐茂; 张笑语; 陈曾
5.基于卷积神经网络的调制信号识别算法 [J], 邵敏兰;周鸿渐;张浩然
因版权原因，仅展示原文概要，查看原文内容请购买。

Voigt-function model in diffraction line-broadening analysisDavor BalzarMaterials Science and Engineering LaboratoryNational Institute of Standards and TechnologyBoulder, Colorado 80303andPhysics DepartmentUniversity of ColoradoBoulder, CO 80309AbstractDiffraction-line broadening routes are briefly reviewed. Both laboratory and synchrotron x-ray measurements of W and MgO showed that a Voigt function satisfactorily fits the physically broadened line profiles. The consequences of an assumed Voigt-function profile shape for both size-broadened and strain-broadened profiles (“double-Voigt” method) are studied. It is shown that the relationship between parameters obtained by the Warren-Averbach approximation and integral-breadth methods becomes possible. Line-broadening analysis of W and MgO is performed by using the Warren-Averbach and "double-Voigt" approaches and results are compared.To appear in Microstructure Analysis from Diffraction, edited by R. L. Snyder, H. J. Bunge, and J. Fiala, International Union of Crystallography, 1999.List of symbolsa, b, c,m, m N,U, V, W General constantsz General variableA Fourier coefficienta Edge of orthorhombic cell, orthogonal to diffracting planes3D"Apparent" domain size orthogonal to diffracting planesd Interplanar spacinge"Maximum" (upper limit) strainFWHM Full width at half maximum of profilef, F Pure-specimen (physically) broadened profile and its Fourier transformg, G Instrumentally broadened profile and its Fourier transformh, H Observed broadened profile and its Fourier transformhkl Miller indicesI IntensityK Scherrer constant1/2k$/(B$), characteristic integral-breadth ratio of a Voigt functionC GL, L’na, column length (distance between two cells in a real space) orthogonal to 3diffracting planesl Order of reflectionMSS Mean-square strainn Harmonic numberp Column-length distribution functionR Relative errorRMSS Root-mean-square strains2sin2/8 = 1/d, variable in reciprocal spacex Data-sampling variable: either 22 or s$$(22)cos2/8, integral breadth in units of s (Å)-1(Geometrical-aberration profile2<,(L)>Mean-square strain, orthogonal to diffracting planes, averaged over the distance L0“Apparent” strain2Diffraction angle2Bragg angle of K" reflection maximum018X-ray wavelengthT Wavelength-distribution profileS UBSCRIPTSC Denotes Cauchy (Lorentz) functionD Denotes distortion-related parameterf Denotes physically (pure-specimen) broadened profileG Denotes Gauss functiong Denotes instrumentally broadened profileh Denotes observed broadened profilem Denotes a maximum indexS Denotes size-related parameters Denotes surface-weighted parameterV Denotes Voigt functionv Denotes volume-weighted parameter wp Denotes weighted-residual errorO PERATORS’Convolution: g(x)’f(x) = I g(z)f(x-z)d z1 IntroductionPhenomenological line-broadening theory of plastically deformed metals and alloys was developed almost 50 years ago (Warren and Averbach 1950; Warren 1959). It identifies two main types of broadening: the size and strain components. The former depends on the size of coherent domains (or incoherently diffracting domains in a sense that they diffract incoherently to one another), which is not limited to the grains but may include effects of stacking and twin faults and subgrain structures (small-angle boundaries, for instance); and the latter is caused by any lattice imperfection (dislocations and different point defects). The theory is general and was successfully applied to other materials, including oxides and polymers. However, the parameters obtained need a careful assessment of their physical validity and correlation to the particular structural features of the material under study. In different approaches (Krivoglaz and Ryaboshapka 1963; Wilkens 1984; Groma et al. 1988), the effects of simplified dislocation configurations on diffraction-line broadening were modeled. These theories were applied mainly to the plastically deformed copper single crystals and recently to the (’-precipitate-hardened nickel-base superalloy (Kuhn et al. 1991). These microscopic models correctly identify origins of broadening in the terms of physically recognized quantities. As our understanding and ability to model complex systems develops further, it is expected that their application will gradually include other materials where different sources of line broadening usually occur simultaneously. The focus here is to study and compare methods of line-broadening analysis in the frame of phenomenological approaches.The development of this research field began when Scherrer (1918) understood that small crystallites cause broadening of diffraction lines. However, more than quarter of a century elapsed before a more complex and exact theory of line broadening was formulated by Stokes and Wilson (1944). They included the lattice strain as another source of broadening. Shortly thereafter, a new impulse was given to the theory: Stokes (1948) adapted the Fourier-deconvolution method to obtain the purely physical broadened line profiles from the observed pattern. Instead of mere estimates ofeither average size of coherent domains or some measure of strain, through the developments of Bertaut (1949) and Warren and Averbach (1950; 1952), a more detailed analysis of complete line-profile shape became possible. Moreover, Wilson (1962a) introduced the analysis of the variance of profile, and Ergun (1968) the method of successive foldings. All those procedures pushed aside the integral-breadth methods because of the important advantage of model independence. Moreover, with a careful application, it was possible to obtain much more information, such as column-length distribution function, the behavior of strain as a function of the averaging distance in domains, etc. However, they also have very serious drawbacks: in cases of large line overlapping, or weak structural broadening, the Stokes deconvolution method cannot be applied without severe errors. This limits application to a small number of specimens and to cubic crystal systems. Moreover, the mathematical process involved is cumbersome and difficult to apply straightforwardly. This is why, after the development of the Rietveld (1967) refinement and other full-powder-pattern-fitting techniques (Pawley 1981; Toraya 1986), the integral-breadth fitting methods became attractive again. After Langford (1978) introduced a Voigt function in the field of x-ray powder diffraction, it was quickly adopted in the Rietveld analysis (Ahtee et al. 1984), along with its approximations (Young and Wiles 1982). It proved to be satisfactory and flexible enough for most purposes when angle dependence of parameters is modeled properly.Alternatively, although the Stokes method has put severe limitations on the analysis, the Warren-Averbach method of separation of size-strain broadening has remained the least constrained method for analyzing diffraction-line broadening. The parameters obtained through the Warren-Averbach and integral-breadth methods are differently defined, and thus not necessarily comparable. The main intention here is to study these two different courses of line-broadening analysis and show their equivalence under the following circumstances: both size-broadened and strain-broadened line-profiles are modeled with a Voigt function and distance-averaged strain follows the Gauss distribution (Balzar and Ledbetter 1993). The first condition also defines the total physically broadened profileh(x)'g(x)’f(x)%background.g(x)'T(x)’((x).(1)(2)as a Voigt function. Laboratory and synchrotron x-ray diffraction experiments were performed on W and MgO powders to study the feasibility of the simple Voigt-function modeling in line-broadening analysis. The physically broadened line profiles obtained through two different approaches, namely Stokes deconvolution and convoluted-profile fitting, are compared. Furthermore, the Warren-Averbach and integral-breadth methods are applied to the so-obtained respective physically broadened line profiles, and differences are assessed.2 Diffraction-line broadeningBoth instrument and specimen broaden the diffraction lines, and the observed line profile is a convolution (Taupin 1973):Wavelength distribution and geometrical aberrations are usually treated as characteristic of the particular instrument (instrumental profile):To obtain microstructural parameters of the specimen, the physically (specimen) broadened profile f must be extracted from the observed profile h.Origins of specimen broadening are numerous. Generally, any lattice imperfection will cause additional diffraction-line broadening. Therefore, dislocations, vacancies, interstitials, substitutional, and similar defects lead to lattice strain. If a crystal is broken into smaller incoherently diffracting domains by dislocation arrays (small-angle boundaries), stacking faults, twins, large-angle boundaries (grains), or any other extended imperfections, then domain-size broadening occurs.FWHM2(22)'U tan22%V tan2%W.m(22)'a(22)2%b(22)%c.(3)(4)2.1 Instrumental broadeningThe first step before any attempt to analyze diffraction-line broadening is to correct the observed line profiles for instrumental effects. A careful scan of a suitable standard sample, showing minimal physical broadening will define the instrumental contribution to broadening. Thorough recipes for preparing the standard specimen are given elsewhere (Berkum et al. 1995).Previously, to obtain a standard, it was customary to anneal the specimen showing broadened reflections. This was the most desirable approach when the Stokes deconvolution method was applied because the centroids of f and g should be as close as possible. However, very often a material does not give satisfactorily narrow lines. It is becoming more customary to find a suitable certified standard reference material which allows a true comparison of results among different laboratories. Because the lines of standard and studied specimen usually do not coincide, it is required to model the characteristic parameters of the standard's line-profile shapes analytically so that the needed instrumental profile can be synthesized at any angle of interest. Most often, the original Caglioti et al. (1958) relation is used:This function was derived for neutron diffraction. Although not theoretically justified, it was confirmed to satisfactorily model as well the angular variation of the symmetrical part of x-ray diffraction-line-profile width. Contrary to the requirement on the physically broadened line profile, it is most important for the instrumental function to correctly describe the angular variation of parameters, regardless of its theoretical foundation. The asymmetry is most often modeled by the split-Pearson VII function (Hall et al. 1977), where the angular variation of the Pearson-VII “shape”parameter m can be simply defined as (Howard and Snyder 1989):Therefore, for a description of asymmetric instrumental broadening, three parameters from both (3)and (4) for each low-angle and high-angle side of the profiles are needed.2.2 Extraction of physically broadened line profileThe choice of the method to obtain the parameters of pure physically broadened line profiles is of utmost importance for the subsequent line-broadening analysis. Basically, the methods used can be divided in two groups: (i) deconvolution approach where the physically broadened line profile is unfolded from the observed profile using the previously determined instrumental profile; (ii) convolution approach where, contrary to the former, the observed profile is built according to (1) and adjusted to the observed pattern through a least-squares fitting. However, we have a knowledge of h and g, but not of f. Therefore, both the general type and parameters of f are assumed, which introduces a bias in the method. Nevertheless, with the development of Rietveld and similar algorithms, where all the parameters are determined in this way anyhow, this approach can be built into the code in parallel to the structural and other parameters and refined simultaneously, whereas deconvolution procedures are much more complicated to introduce. Another important difference between the two approaches is that deconvolution methods under some circumstances either fail or become unstable and inaccurate. The convolution process is always stable, but, beside the systematic errors introduced by a possible inadequate model of physically broadened line profile, the iterative least-squares minimization procedure can be trapped in a false minimum. Moreover, the smallest reliability index does not necessarily correspond to a physically meaningful solution; for instance, adding more peaks in refinement than actually exist may decrease it. For some illustrations and possible artifacts of profile fitting, see Howard and Preston (1989), and for an example of how a high degree of line overlap may influence size-strain analysis see Balzar (1992).The most used of the first type is the Fourier-transform deconvolution method (Stokes 1948), although there are some novel approaches (Kalceff et al. 1995) using constrained Phillips-Twomey (Twomey 1963) or maximum-entropy deconvolution (Skilling and Bryan 1984) methods. AnotherF(n)'H(n) G(n).f(x)'j n H(n)G(n).(5)(6)(in fact deconvolution) process, although accomplished by simple substraction of the instrumental-profile variance (Wilson 1962a) is not used extensively, and will not be considered here. Likewise, among the methods falling into the second group, the iterative method of successive foldings (Ergun 1968) is used sparingly, and will not be considered. Some new developments can be found in the review by Reynolds (1989).2.2.1 Deconvolution method of StokesFrom (1), it follows that deconvolution can be performed easily in terms of complex Fourier transforms of respective functions:The inverse Fourier transform gives a physically broadened line profile:Hence, the physically broadened profile f is retrieved from the observed profile h without any assumption (bias) on the peak-profile shape. However, (5) may not give a solution if the Fourier coefficients of the f profile do not vanish before those of the g profile (Delhez et al. 1980). Furthermore, if physical broadening is small compared with instrumental broadening, deconvolution becomes too unstable and inaccurate. If the h profile is 20% broader than the g profile, this gives an upper limit of about 1000 Å for the determination of the effective domain size (Schwartz and Cohen 1977). Regardless of the degree of broadening, deconvolution produces unavoidable profile-tail ripples because of truncation effects. To obtain reliable results, errors of incorrect background, sampling, and the standard specimen have to be corrected (Delhez et al. 1986; 1988). The largest conceptual problem, however, is peak overlapping. If the complete peak is not separated, the only possible solution is to try to reconstruct the missing parts. This requires some assumption on theI(x)'I(0)exp&I(x)'I(0)1 $2CB2%x2.$ h C '$g C%$f C$2 h G '$2g G%$2f G.I(x)'I(0)ReerfiB1/2xG%i(7)(8)(9)(10)(11)peak-profile shape, which introduces bias into the method. The application of the strict Stokes method is therefore limited to materials having the highest crystallographic symmetry.2.2.2 Convolution-fitting methodsHere it is required that at least the unknown physically broadened diffraction profile f be approximated with some analytical function. In the past, two commonly used functions were Gaussand Cauchy (Lorentz)From the convolution integral (1), it follows that for Cauchy profilesand for Gauss profilesHowever, the observed x-ray diffraction line profiles cannot be well represented with a simple Cauchy or Gauss function (Klug and Alexander 1974; Young and Wiles 1982). Experience shows that the Voigt function, or its approximations, pseudo-Voigt (Wertheim et al. 1974) and Pearson-VII (Hall et al. 1977) fit very well the observed peak profiles in both x-ray and neutron-diffraction cases. The Voigt function is usually represented following Langford (1978):erfi(z)'exp(&z2)erfc(&i z)$'$G exp(&k2) erfc(k).$2'$C $%$2G.(12)(13)(14)Here, the complex error function is defined asand erfc denotes the complementary error function.Integral breadth of the Voigt function is expressed through its constituent integral breadths (Schoening 1965):Halder and Wagner (1966) showed that the following parabolic expression is a satisfactory approximation:Because convolution of two Voigt functions is also a Voigt function, integral breadths are easily separable conforming to (9) and (10).In case that any of h, g, or f profiles are asymmetric, they cannot be modeled with the discussed functions. Indeed, for Bragg-Brentano geometry, observed diffraction-line profiles are asymmetric toward the low-angle side for small diffraction angles and switch to a slight reverse asymmetry at the highest diffraction angles. Asymmetry is introduced in g by axial beam divergency, specimen transparency and flat surface (Klug and Alexander 1974). However, extrinsic stacking and twin faults may introduce (relatively weak) asymmetry in f also (Warren 1969). Unfortunately, numerical convolutions are usually necessary in these cases, thus consuming calculation time and introducing additional errors. This is why asymmetry is often neglected. However, it may cause large errors in line-broadening analysis. Some examples of numerical convolutions used are the following: Enzo et al. (1988) modeled g with a pseudo-Voigt convoluted with the exponential function and f with a pseudo-Voigt; Howard and Snyder (1989) modeled g with a split-Pearson VII and f with a Cauchy function, whereas Balzar (1992) modeled g also with a split-Pearson VII but f with a Voigt function. There were no attempts yet to assume an asymmetrical f profile.2.2.3 Physically broadened profiles of W and MgOThe Stokes Fourier-deconvolution method followed by the Warren-Averbach analysis of physically broadened line profile is the least biased approach to the analysis of line broadening. The opposite procedure is profile fitting of the convolution of the presumed analytical function, which models the physically broadened line profile and instrumental profile obtained by measuring a suitable standard material and consecutive application of some integral-breadth method. Both approaches have advantages and disadvantages and it is generally assumed that different results are obtained. The critical step is a correction for instrumental broadening. Because of numerous problems associated with the Fourier deconvolution and especially the inevitable ripples at tails of physically broadened profile, there were almost no attempts to fit it with some analytical function. Suortti et al. (1979) fitted a Voigt function to the least-squares deconvoluted profiles of a Ni powder and found good overall agreement. Reynolds (1989) fitted successfully a pseudo-Voigt function to the physically broadened line profiles of chlorite, obtained by the Ergun (1968) method of iterative folding. Although different approaches to deconvolution are possible, we used the Stokes method because of its wide acceptance. Many different integral-breadth routes were used in the literature. We followed our previous studies (Balzar 1992).2.2.3.1 MaterialsTwo "classical" materials for line-broadening analysis, W and MgO, were selected because of their simple cubic structure (the Stokes method is optimally applicable) and different origins of broadening; It is expected that on cold deformation, W shows dominantly strain broadening causedby introduction of numerous dislocations, whereas the thermal decomposition of MgCO gives rise3to size broadening only. W is elastically isotropic and shows well-resolved reflections thus admitting a relatively large number of reflections to be analyzed simultaneously. Likewise, because of negligible strain and low probability of fault formation in MgO, a simultaneous treatment of allaccessible reflections is possible. W powder of nominal size 8-12 µm was ball-milled for differenttimes. After some initial milling time no additional change of peak widths was observed. The fineMgCO powder was decomposed at different temperatures, 350E C and higher. The powder was kept 3at a particular temperature for 3 h and allowed to cool slowly in the furnace to minimize possible lattice defects. As a standard specimen for W, the original untreated W powder was used. It showsminimal line broadening, comparable to the NIST standard reference material LaB. The MgO6standard was prepared by decomposing the MgCO powder at 1300E C for 6 hours. Although it3showed relatively broad reflections (FWHM at 37E 22 was about 0.15E 22 compared to only 0.08E22 for LaB), we used it to get optimal conditions for the Stokes analysis, because the appreciable6difference between standard and specimen peak positions represents a problem, and the needed interpolation of Fourier coefficients of the standard would introduce additional uncertainties. Moreover, we were not interested in the absolute magnitude of results. We chose specimens from each batch (W specimen ball milled for 140 min and MgO specimen decomposed at 550E C) showing approximately 3-4 times broader reflections than the respective standards because the Stokes method has optimum application for this factor in the range 2-6 (Delhez et al. 1980).2.2.3.2 Data analysisWe collected data with both laboratory and synchrotron x-ray sources. Synchrotron-radiationmeasurements were performed on the X3B1 powder-diffraction beamline at the National SynchrotronLight Source (NSLS), Brookhaven National Laboratory (BNL). A Si channel-111-cutmonochromator, flat specimen, Ge 111-cut analyzer crystal, and proportional detector were used.More details on experimental setup and line-profile shape were published elsewhere (Balzar et al.1997a). Laboratory x-ray data were collected using a horizontal goniometer and counted with a solid-state detector. We used a fixed-time counting method with a condition to collect about 10,000 countsat each peak maximum. The raw data were used in the analysis to avoid introducing any bias bysmoothing methods. However, the deconvolution process is very sensitive to noise in the data and itcaused unwanted spurious oscillations in Fourier coefficients, especially for the laboratory x-ray measurements. We measured seven W reflections using Cu K" radiation. The 400 peak was excluded because the high-angle background was not accessible with our goniometer. For the MgO pattern, we analyzed only 220, 400, and 422 reflections because others come in close pairs for the fcc structure. The inevitable line overlapping would require the unsafe estimation of missing tails and introduce unwanted bias in the Fourier deconvolution. Linear background was determined by fitting the first-order polynomial to data before applying either Stokes deconvolution or profile fitting. TheK" elimination prior to the Stokes deconvolution was not performed because it may introduce 2additional errors. In the convolution-fitting approach, the K" component is treated analogously to2K", but with half its intensity and the same profile shape. After correction for the Lorentz-1polarization factor, Stokes deconvolution was performed on a 22 scale to be in accord with the profile fitting. Hence, for both deconvolution and fitting/convolution approaches, there is a systematic error because the subsequent analyses are implemented in Fourier space. Although it is possible to perform both Stokes deconvolution and profile fitting on the s scale, we found the error to be negligible at this or a smaller level of broadening. The origin was taken at the centroid of the standard peak, although there were no substantial shifts of broadened peaks. The convolution-fitting process was performed in the following way: the instrumental profile was obtained by fitting the split-Pearson VII function to the particular line profile of the standard specimen. By convoluting it with the preset physically broadened Voigt function, the "observed" profile is obtained. After adding the previously refined linear background, the parameters of physically broadened Voigt function are adjusted in the least-squares fitting to the observed pattern.2.2.3.3 Fitting of physically broadened line profilesFollowing (6), the physically broadened line profile can be back-synthesized from the Fouriercoefficients obtained by deconvolution. However, because of noise in the raw data, after some harmonic number, Fourier coefficients become unreliable and cause large oscillations in the synthesized profiles (Delhez et al. 1980). Typical plots of synthesized physically broadened line profiles are shown in Figures 1 (110 W at low angle) and 2 (422 MgO at high angle). Conversely, the 110 W physically broadened line profile obtained from the synchrotron data (Figure 3) does not show peak-tail ripples. This is unlikely to be caused by statistical noise only; Figure 4 shows the deconvoluted physically broadened line profile of the weakest 400 MgO line where only 900 counts were collected at the peak maximum. It indicates that generally superior synchrotron resolution and simpler (singlet) wavelength distribution allows for more precise line-broadening studies. To test whether simple analytical functions can successfully approximate a physically broadened profile, least-squares fits of Cauchy, Gauss, and Voigt functions to the profiles in Figures 3 and 4 were performed. It is obvious that the Voigt function shows a superior and overall satisfactory fit. Even more important is how different functions fit profiles. Although the Cauchy function approximates tails quite well, it fails to fit profile shapes close to124128132-390100Synthesized f (2θ) Fitted Voigt functionI n t e n s i t y (a r b i t r a r y u n i t s )2θ (°)-390100Synthesized f (2θ)Fitted Cauchy functionI n t e n s i t y (a r b i t r a r y u n i t s )-390100Synthesized f (2θ) Fitted Gauss functionI n t e n s i t y (a r b i t r a r y u n i t s )4041-380100Synthesized f (2θ) Fitted Voigt functionI n t e n s i t y (a r b i t r a r y u n i t s )2θ (°)-380I n t e n s i t y (a r b i t r a r y -380100Synthesized f (2θ) Fitted Gauss functionI n t e n s i t y (a r b i t r a r y u n i t s )Fig. 2. Cauchy, Gauss, and Voigt-function fits to the Stokes-deconvoluted physically broadened 422 MgO line profile (laboratory x-ray data). Difference patterns plotted around zero intensity on the Fig. 1. Cauchy, Gauss, and Voigt-function fits to the Stokes-deconvoluted physically broadened 110 W line profile (laboratory x-ray data). Difference patterns plotted around zero intensity on the smaller scale.747678-395100Synthesized f (2θ) Fitted Voigt functionI n t e n s i t y (a r b i t r a r y u n i t s )2θ (°)-395100Synthesized f (2θ)Fitted Cauchy functionI n t e n s i t y (a r b i t r a r y u n i t s )-395100Synthesized f (2θ) Fitted Gauss functionI n t e n s i t y (a r b i t r a r y u n i t s )17.617.818.018.2-380100Synthesized f (2θ) Fitted Voigt functionI n t e n s i t y (a r b i t r a r y u n i t s )2θ (°)-380I n t e n s i t y (a r b i t r a r y -380100Synthesized f (2θ) Fitted Gauss functionI n t e n s i t y (a r b i t r a r y u n i t s )Fig. 4. Cauchy, Gauss, and Voigt-function fits to the Stokes-deconvoluted physically broadened 400 MgO line profile (synchrotron x-ray data). Difference patterns plotted around zero intensity on the Fig. 3. Cauchy, Gauss, and Voigt-function fits to the Stokes-deconvoluted physically broadened 110 W line profile (synchrotron x-ray data). Difference patterns plotted around zero intensity on the smaller scale.。

topmpson sampling 算法公式Thompsonsampling算法（也称为贝叶斯采样算法）是一种基于贝叶斯定理的多臂赌博机（multi-armedbandit）问题的解决方案之一。

其基本思想是在每次试验中，在根据当前已经得到的数据更新对每个赌博机的概率分布后，使用每个赌博机的概率分布来生成一个样本，然后选择具有最高样本值的赌博机。

具体来说，设 $X_i$ 表示第 $i$ 个赌博机的奖励值，$N_i$ 表示第 $i$ 个赌博机被选择的次数，$T$ 表示当前试验的总次数，则第 $i$ 个赌博机的奖励期望值 $mu_i$ 可以被估计为：$$mu_i = frac{sum_{t=1}^{T} r_{t} cdotmathbb{1}(I_t=i)}{sum_{t=1}^{T} mathbb{1}(I_t=i)}$$其中 $r_t$ 表示第 $t$ 次试验得到的奖励值，$mathbb{1}(I_t=i)$ 表示第 $t$ 次试验选择的赌博机为第 $i$ 个赌博机。

假设第 $i$ 个赌博机的奖励值服从一个分布 $f_i(cdot)$（比如高斯分布），则可以利用贝叶斯定理来更新第 $i$ 个赌博机的概率分布 $p_i(cdot)$：$$p_i(x) = frac{f_i(x) cdot prod_{t=1}^{T} mathbb{1}(I_t eq i text{ or } r_teq x)}{int f_i(x') cdot prod_{t=1}^{T} mathbb{1}(I_teq i text{ or } r_teq x') dx'}$$其中 $prod_{t=1}^{T} mathbb{1}(I_teq i text{ or } r_teq x)$ 表示第 $i$ 个赌博机未被选中或未得到奖励 $x$ 的概率。

在每次试验中，根据每个赌博机的概率分布 $p_i(cdot)$，可以从每个赌博机的分布中抽取一个样本 $x_i$，然后选择具有最高样本值的赌博机进行下一次试验。

VIAVI//PublicT-BERD/MTS-5800 Portable Network TesterQUICK CARDSONET Bit Error Rate Testing (BERT)This quick card describes how to configure and run a SONET Bit Error Rate Test at the full concatenated line rate. Please note that the T-BERD can also test channelized payloads (DS1, VT1.5, and STS-n). Please refer to the T-BERD 5800 User’s Guide for information.•T-BERD/MTS 5800 equipped with the following:o BERT software release V30.1.0 or greater o C5LSSONSDH test option for OC-3 and OC-12o C525GSONSDH test option for OC-48o C510GSONSDH test option for OC-192•Optical Transceiver supporting the Optical Carrier level to be tested (SFP or SFP+)•LC Attenuators (5dB, 10dB, and/or 15dB)•Cables to match the optical transceiver and the line under test •Fiber optic inspection microscope (P5000i or FiberChek Probe)•Fiber optic cleaning suppliesFigure 1: Equipment Requirements1.Press the Power button to turn on the T-BERD.2.Press the Test icon at the top of the screen to display the Launch Screen .3.Using the Select Test menu, Quick Launch menu, or Job Manager, launch the SONET Bulk BERT test on Port 1 for the desired Optical Carrier level. For Example:SONET ►OC-3►STS-3c Bulk BERT ►P1 Terminate.4.Tap to open the Tools Panel and select .5.Press to continue.Figure 2: Launch ScreenFigure 3: Tools PanelVIAVI//PublicT-BERD/MTS-5800 Portable Network TesterQUICK CARD2•The following Information is needed to configure the test:•Optical wavelength(typically, 1310nm or 1550nm)•Test Pattern(s) (default is 2^23-1 ANSI)•BER Pass/Fail Threshold1.Press the Setup soft key on the top right side of the screen.2.Select the Interface/Connector folder.3.Insert desired SFP into the Port 1 SFP+ slot on the top of the T-BERD.4.Review SFP information in the Connector tab:►Verify that the SFP operates on the required wavelength (1310nm or 1550nm).►Verify that the SFP supports the required optical carrier level (OC-3, OC-12, OC-48, or OC-192).►Note the Min and Max Tx Levels (dBm) and Max Rx Level (dBm) to assess if optical attenuators are required.5.Select the indicated folders and configure your test as follows. Leave all other values at default, unless specified in the work order.6.Press the Results soft key to view the Test Results screen.Figure 4: Work OrderFolder Option Value(s)Interface,SignalClock SourceSelect “Recovered” unless you are testing dark fiber with no SONET equipment PatternPattern Mode ANSI Pattern2^23-1 ANSIFigure 6: Setup, Interface/SignalFigure 7: Setup, PatternFigure 5: Setup, Interface/Connector/SFPVIAVI//PublicT-BERD/MTS-5800 Portable Network TesterQUICK CARD31.Using drop-down menus , select“Payload/BERT ” for the right results display.2.Select the Laser tab in the Actions panel, and press . The button will turn yellow and be relabeled .3.Press the Restart soft key .4.Verify the following:►Level (dBm) is within the Rx Level range of the SFP .►Summary LED is green.►Signal Present LED is green. ►Frame Sync LED is green.►Path Pointer Present LED is green ►Pattern Sync LED is green.►Summary/Status results shows ‘ALL SUMMARY RESULTS OK”5.Allow the test to run for desired duration and verify the following:►Bit/TSE Error Rateresult does notexceedyour required threshold.(0.00E+00 if pass/fail threshold unknown)•Use the VIAVI P5000i or FiberChek Probe microscope to inspect both sides of everyconnection being used (SFP , attenuators, patch cables, bulkheads)►Focus fiber on the screen.►If it appears dirty, clean the fiber end-face and re-inspect.►If it appears clean, run inspection test.►If it fails, clean the fiber and re-run inspection test. Repeat until it passes.•If necessary, insert optical attenuators into theSFP TX and/or RX ports.•Connect the SFP to the port under test using ajumper cable compatible with the line under test.Figure 9: Results, Payload BERTFigure 8: Inspect Before You ConnectStatusTipSignal Present LED not greenCheck your cables.Tx and Rx may be reversed.Path Pointer Present LED not green and AIS-P alarm on There may be no loop or no connectivity to the loop. The wrong payload may be selected (concatenated vs. channelized).RDI-L alarm onThe Tx Level is too high.Add an attenuator between the SFP Tx port and the line under test.Path PointerAdjustments incrementingClock Source is set incorrectly.Change Clock Source to “Recovered .”Figure 10: Troubleshooting TipsVIAVI//PublicT-BERD/MTS-5800 Portable Network Tester QUICK CARD© 2022 VIAVI Solutions, Inc,Product specifications and descriptions in this document are subject to change without notice.Patented as described at /patentsContact UsTo reach the VIAVI office nearest you,visit /contact+1 844 GO VIAVI(+1 844 468-4284)6.In the T-BERD’s Quick Config menu, change “Pattern ” to the next value in the test plan. 7.Press the Restart soft key to reset results.8.Allow test to run for desired duration and verify the following:►Pattern Sync LED is green.►Bit/TSE Error Rate or Round Trip Delay does not exceed your required threshold.►Repeat steps 6 through 8 for all Patterns in the test plan. Patterns may include:▪Delay : MeasuresRound Trip Delay (RTD) instead of Bit Errors.RTD values are shown instead of BERin the “Payload/BERT” results display.Figure 11: Results, Quick Config1.Tap to open the Reports Panel and select . .2.Tap .3.A report will be saved to the T-BERD 5800’s/bert/reportsfolder.Figure 12: Create Report。