技術(shù)培訓(xùn)fdd優(yōu)化原理介紹

1、LTE FDD 優(yōu)化介紹優(yōu)化介紹Dept. / Author / Date內(nèi)內(nèi)容容優(yōu)化總述、流程和工具優(yōu)化總述、流程和工具規(guī)劃參數(shù)的影響和優(yōu)化案例規(guī)劃參數(shù)的影響和優(yōu)化案例RF 測量和評估測量和評估RF 環(huán)境的影響和優(yōu)化環(huán)境的影響和優(yōu)化基于基于KPI指標(biāo)的優(yōu)化指標(biāo)的優(yōu)化峰值速率優(yōu)化峰值速率優(yōu)化Dept. / Author / DateOverviewDept. / Author / Date優(yōu)化是使用一系列的方法，通過具體的方案實施，使網(wǎng)絡(luò)性能達(dá)到一個既定的目標(biāo)。從網(wǎng)絡(luò)建成開始到網(wǎng)絡(luò)的運營階段，優(yōu)化工作都是必不可少的。優(yōu)化項目的確立非常靈活，根據(jù)不同分類和側(cè)重，可以展開各種優(yōu)化：從范圍，重點區(qū)域

2、優(yōu)化，全網(wǎng)優(yōu)化；從時間，集中優(yōu)化，持續(xù)優(yōu)化；從內(nèi)容，專題優(yōu)化，系統(tǒng)優(yōu)化；從方法，路測優(yōu)化，指標(biāo)優(yōu)化；優(yōu)化的確立和開展實施，通常需要結(jié)合多種因素，例如在正式商用前，路測優(yōu)化顯然是一種較適合的方式，商用后，隨著用戶數(shù)的增加，指標(biāo)優(yōu)化逐步成為重要環(huán)節(jié)。LTE建網(wǎng)過程中的絡(luò)優(yōu)化內(nèi)容建網(wǎng)過程中的絡(luò)優(yōu)化內(nèi)容優(yōu)優(yōu)化化流流程程舉舉例例-總總體體項項目目流流程程管管理理圖圖Dept. / Author / Date優(yōu)化目標(biāo)確立方法工具數(shù)據(jù)準(zhǔn)備數(shù)據(jù)問題分析RootCause分析優(yōu)化方案方案實施效果評估舉舉例例：基基于于路路測測/測測試試優(yōu)優(yōu)化化的的流流程程Dept. / Author / DateKPI/基站狀態(tài)

3、檢查參數(shù)一致性檢查路測/測試及數(shù)據(jù)分析問題定位及解決結(jié)果和各階段報告基基于于網(wǎng)網(wǎng)管管指指標(biāo)標(biāo)的的優(yōu)優(yōu)化化(Counter Based)Dept. / Author / Date基于使用場景（優(yōu)化的需求）定制優(yōu)化項目。需要提升KPI指標(biāo)，或某些區(qū)域的特定問題，解決客戶對網(wǎng)絡(luò)質(zhì)量的投訴等，均可以做為優(yōu)化項目的目標(biāo)和啟動點。適于使用網(wǎng)管系統(tǒng)的統(tǒng)計指標(biāo) 。例如，可以通過LTE網(wǎng)絡(luò)的質(zhì)量監(jiān)控，掌握網(wǎng)絡(luò)質(zhì)量的實時變化和趨勢變化，及時增加資源和優(yōu)化調(diào)整，保證用戶的服務(wù)質(zhì)量。Performance supervision of the LTE Networksupervise the overall LTE

4、network performance on cell level and adjacency levelcompare the performance and quality of the LTE network with other technologiescompare the performance and quality of an eNodeB with another eNodeBcheck performance degradations over timemonitor the network usage to enhance the network resources in

5、 timecheck on a high level whether the offered quality is sufficient for the end customersLTE 優(yōu)優(yōu)化化現(xiàn)現(xiàn)場場工工具具測試工具信令工具基站工具內(nèi)內(nèi)容容優(yōu)化總述、流程和工具優(yōu)化總述、流程和工具規(guī)劃參數(shù)的影響和優(yōu)化案例規(guī)劃參數(shù)的影響和優(yōu)化案例RF 測量和評估測量和評估RF 環(huán)境的影響和優(yōu)化環(huán)境的影響和優(yōu)化基于基于KPI指標(biāo)的優(yōu)化指標(biāo)的優(yōu)化峰值速率優(yōu)化峰值速率優(yōu)化Dept. / Author / Date規(guī)規(guī)劃劃參參數(shù)數(shù)的的影影響響和和優(yōu)優(yōu)化化案案例例PCIPRACHTA/TA List鄰區(qū)規(guī)劃優(yōu)化鄰區(qū)規(guī)劃

6、優(yōu)化Dept. / Author / DatePCI的規(guī)劃在LTE的空中接口中，物理小區(qū)標(biāo)識（Physical Cell Identity，PCI）用于小區(qū)識別和同步過程。PCI的規(guī)劃需要特別謹(jǐn)慎，如果相鄰小區(qū)的PCI重復(fù)的話，這兩個小區(qū)間的同步信號和RS干擾將會增加，影響小區(qū)的吞吐量和切換性能。 PCI總數(shù)有504個，分為168組，其中組號對應(yīng)輔同步序列SSS；每組有3個，組內(nèi)本地號對應(yīng)為主同步序列PSS（0，1，2）。PCI = (3 NID1) + NID2PCI規(guī)劃應(yīng)按如下優(yōu)先級進(jìn)行考慮： 避免相同的PCI分配給鄰區(qū)； 避免模3相同的PCI分配給鄰區(qū)，規(guī)避相鄰小區(qū)的PSS序列相同及規(guī)避

7、相鄰小區(qū)RS信號的頻域位置相同 避免模6相同的PCI分配給鄰區(qū)，規(guī)避相鄰小區(qū)RS信號的頻域位置相同； 避免模30相同的PCI分配給鄰區(qū)，規(guī)避相鄰小區(qū)的UL 解調(diào)參考信號序列相同 。通過相應(yīng)規(guī)劃軟件，對上述不同的PCI分配限制設(shè)定權(quán)重，基于Atoll規(guī)劃軟件的覆蓋仿真，可以實現(xiàn)PCI的自動分配。 PCI的規(guī)劃PCI中的主同步（PSS）序列解調(diào)的需要：包含0，1，2三個數(shù)，PCI 分配要避免同一個站內(nèi)PSS相同。模3 不同即可。 PCI序列同樣對應(yīng) RS信號在頻域的位置。 共站及其鄰區(qū)的RS位置重疊時，會相互干擾，影響信道估計過程。上圖給出1、2、4天線時，RS信號分布的位置。所以需要規(guī)劃 PCI

8、控制RS的位置。兩發(fā)射天線時， RS位置有3種，PCI需要滿足模3規(guī)則。一發(fā)射天線時， RS位置有6種，PCI需要滿足模6規(guī)則。RS信號(Reference signals) 用于信道估計，小區(qū)選擇/重選，切換過程。時域位置為0，4和1。 2天線時，每RB保留8個RS。 1天線時，每RB保留4個RS。RS信號分布的位置模3原則和模6原則PCI的規(guī)劃(上行RS信號）共站的不同小區(qū)以及鄰小區(qū)應(yīng)盡量使用正交的上行DM RS信號。正交的上行DM RS 是由ZC 根序列經(jīng)過偏轉(zhuǎn)產(chǎn)生的。它所占的帶寬與UE在上行分配到的資源是一致的(PRB數(shù)*12子載波)。對應(yīng)不同的上行分配RB數(shù)，需要相應(yīng)長度的DM RS

9、信號。也就是需要長度不同的ZC根序列來產(chǎn)生這些DMRS信號（信號序列長度與根序列長度一樣）。由一個ZC根序列經(jīng)過位移產(chǎn)生的序列是正交的。可產(chǎn)生ZC序列的數(shù)量由ZC根序列的長度決定。對于每個可能的RB數(shù)（1-100），系統(tǒng)生成足夠數(shù)量的正交序列（30個或60個），這些序列產(chǎn)生的相互正交的DMRS信號。規(guī)劃的目標(biāo)是本小區(qū)及其鄰區(qū)不使用相同的正交序列。上行解調(diào)RS信號(DM Reference signals) 用于上行信道估計。模30原則PCI的規(guī)劃為了對RS信號序列的管理，系統(tǒng)把可能的RS信號序列分為30組（Group0-Group29），每組包括1-100個RB對應(yīng)的RS 序列。30組表示每種

10、RB分配時，都有30種相互正交的RS序列可用。系統(tǒng)根據(jù)PCI和分組參數(shù)決定每個小區(qū)使用的組號(u)。組號(u) （PCI+”分組參數(shù)”）Mod30 通常各小區(qū)的“分組參數(shù)”設(shè)為一致。所以在鄰區(qū)的范圍內(nèi)只需考慮PCI Mod30不同即可保證UL RS信號的序列不同。上行解調(diào)RS信號序列分組模30原則（續(xù)）Mod30 Planning 影響案例影響案例Remember!: Cyclic shifts of a Fourier transform of an extended ZC sequence are fully orthogonalUL Throughput gets considerabl

11、y affected if UL traffic in neighbour cellFrom 40 Mbps to 22 Mbps in the example Worse ResultBetter ResultUL Throughput still suffers from UL interference in neighbour cell but the effect is lowerPRACH的規(guī)劃優(yōu)化每個小區(qū)Preamble需要64個，及其鄰區(qū)所使用的Preamble 均要求保持正交性，通過ZC序列可以產(chǎn)生這樣的Preamble。所以PRACH規(guī)劃的目標(biāo)：確定各小區(qū)的序列。序列共有83

12、8個（根序列）。使用這些序列，可以產(chǎn)生所需的Preamble。Preamble 由ZC根序列生成。根序列通過不同的位移產(chǎn)生符合要去的序列（構(gòu)成Preamble）。對應(yīng)不同接入距離，每個根序列產(chǎn)生若干序列(Sign. /root seq.)。據(jù)此，該小區(qū)需要一個或幾個根序列。(#root seq. required/cell)PRACH規(guī)劃就是要分配好各小區(qū)使用的根序列。即確定PrachCs（循環(huán)位移配置）和根序列索引。838/64/ PRACH的規(guī)劃優(yōu)化建議宏小區(qū) 覆蓋最大距離為12km，PRACH Preamble序列生成使用的循環(huán)移位配置NCS為11，每個小區(qū)需要8個ZC根序列來生成64個

13、Preamble序列，ZC根序列的復(fù)用度為104個小區(qū)，可以保證100個小區(qū)的范圍內(nèi)的不重復(fù)使用。 PRACH配置索引為3，即PRACH Preamble的格式0，每幀（10ms）1個RACH。 室分小區(qū) 覆蓋最大距離為1.4km，考慮到室分系統(tǒng)中相干帶寬較大，快衰落效應(yīng)強(qiáng)，為保證更好的接入性，需要更多的根序列，故而PRACH Preamble序列生成使用的循環(huán)移位配置NCS可以配置為，例如8，每個小區(qū)需要4個ZC根序列來生成64個Preamble序列，ZC根序列的復(fù)用度為208個小區(qū)。 PRACH配置索引為3，即PRACH Preamble的格式為0，每幀（10ms）1個RACH。 在LTE

14、的早期部署過程中，主要是在主城區(qū)的LTE站點密集部署，對于PRACH的規(guī)劃建議如下： TA的規(guī)劃主要原則1）TA劃分需考慮TA更新的信令負(fù)荷和尋呼區(qū)大小之間的平衡 2） TA劃分需考慮尋呼區(qū)域TA中的尋呼信道不受限 3）每個TA區(qū)建議包含大致包含50100個eNodeB。3GPP TS 36.331 定義一個尋呼消息塊 paging message 可支持最大16條尋呼。 目前 空口的最大配置是每無線幀（10ms）配一個尋呼消息塊（占1個1ms子幀，稱為PO），通常配置為每4個無線幀（4*10ms）配一個尋呼消息塊。在最大配置的情況下，每秒鐘尋呼信道可發(fā)送1600條尋呼。通常配置每秒鐘尋呼信道

15、可發(fā)送400條尋呼?？紤]到尋呼的碰撞和eNB/S1的處理負(fù)荷，建議每秒鐘的最大尋呼發(fā)送量控制在500條。對應(yīng)每小時尋呼量180萬條。4）網(wǎng)絡(luò)建設(shè)初期，TA規(guī)劃可以適當(dāng)大，后期隨著網(wǎng)絡(luò)負(fù)荷的上升，再做優(yōu)化調(diào)整 5）TA不要沿著或者與主干道或者鐵路平行，其邊界也不要穿越密集用戶區(qū) 6）鑒于LTE網(wǎng)絡(luò)在現(xiàn)有2G/3G網(wǎng)絡(luò)基礎(chǔ)上部署，最直接的TA規(guī)劃方案是將TA邊界規(guī)劃成與2G/3G RA或LA邊界重疊。TA應(yīng)盡量規(guī)劃的大一些，從而降低TAU開銷。但如果尋呼負(fù)載過高，應(yīng)縮小TA。在LTE網(wǎng)絡(luò)部署后，應(yīng)通過網(wǎng)管來監(jiān)控尋呼負(fù)載，防止尋呼負(fù)載偏高。TA List規(guī)劃建議1）TA list 是一個到多個TA的

16、組合。2）TA list 在MME中管理。NSN的MME 可支持最大8000個TA list, 每個 TA list可支持最大16個TAI。3）TA list 有助于減小移動性的影響。在一個TA list范圍內(nèi)移動，穿過TA邊界時，不需要做TA更新。4）如果TA list包含多個TA, 每個TA一定要適當(dāng)縮小，因為尋呼的范圍擴(kuò)大為TA list的范圍。TA 大小設(shè)置的原則就需應(yīng)用于TA list的范圍上。目前，在TA 區(qū)域已經(jīng)過優(yōu)化完成劃分后（參考2G,3G 網(wǎng)的移動性數(shù)據(jù)），通常采用TA與TA list一一對應(yīng)。 LTE中TA（Tracking Area）和2G/3G中得RA（Routing

17、 Area）類似。LTE只有PS域（Packet Switch），所以沒有LA（Location Area）的概念。小區(qū)所屬的TA在SIB1（System Information Block 1）中廣播。 MME通過TA知道空閑態(tài)UE的位置信息。鄰區(qū)規(guī)劃和數(shù)量限制鄰區(qū)規(guī)劃和數(shù)量限制1）無線網(wǎng)元主干的參數(shù)結(jié)構(gòu)如下圖。服務(wù)eNB ，服務(wù)小區(qū)，相鄰小區(qū)為級聯(lián)分層展開的方式。 2）服務(wù)eNB與相鄰eNB為一對多的關(guān)系，建立了重要的X2 接口連接。 3）相鄰小區(qū)的設(shè)置與X2接口無關(guān)，最大可添加64個。相鄰小區(qū)參數(shù)可以通過SON自動建立（基于UE測量和非基于測量方式），也可以手工建立。切換依然需要具備相鄰

18、關(guān)系才可以進(jìn)行。4）相鄰小區(qū)的信息不通過系統(tǒng)消息通知UE, UE 根據(jù)收到的信號進(jìn)行測量，上報測量報告。具備SON 功能時，有效的相鄰小區(qū)可以逐步被建立。 5）無X2接口的相鄰小區(qū)可以通過S1 接口切換。LNMMELNBTSLNCELLNRELLNADJ最大64個鄰區(qū)1.321.161.6相鄰eNB,建立X2連接服務(wù)eNB服務(wù)小區(qū)相鄰小區(qū)1.64內(nèi)內(nèi)容容優(yōu)化總述、流程和工具優(yōu)化總述、流程和工具規(guī)劃參數(shù)的影響和優(yōu)化案例規(guī)劃參數(shù)的影響和優(yōu)化案例RF 測量和評估測量和評估RF 環(huán)境的影響和優(yōu)化環(huán)境的影響和優(yōu)化基于基于KPI指標(biāo)的優(yōu)化指標(biāo)的優(yōu)化峰值速率優(yōu)化峰值速率優(yōu)化Dept. / Author /

19、DateRF優(yōu)化時三個主要的測試指示優(yōu)化時三個主要的測試指示Dept. / Author / Date RF環(huán)境是網(wǎng)絡(luò)質(zhì)量的根本，RF優(yōu)化不僅是預(yù)優(yōu)化的重點，也是各個優(yōu)化階段都應(yīng)該重視的。 RF環(huán)境的主要問題，如弱覆蓋，干擾，重疊覆蓋等,需要通過小區(qū)無線信號的相關(guān)指標(biāo)來判斷和定位。 用以衡量評估無線環(huán)境的三個主要測量指標(biāo)SINRRSRQRSRPDetecting interference SINRPresentation / Author / Date Example: SSS-CINR + RS CINR versus top-N RSRP SINR measurements can ind

20、icate interference areas, but it doesnt necessarily see all interference sources:Impacted by network load. Traffic in the neighboring cells will reduce serving cell SINR.Depends on the measurement method (RS or SCH) and toolDepends on PCI planning (RS SINR)Results can be misleading!RS CINRS-SCH CINR

21、Detecting interference RSRQ RSRQ (N x RSRP / RSSI)depends on network load, including own cell load,-3 dB for 1Tx no traffic ,-10.8dB for 1Tx 100% PRB utilization. RSRQ reporting range -3-19.5dBRSRQ depends on serving and neighbour cell load Fluctuates quicklyHence difficult to interpret resultsSimil

22、ar to Ec/N0 in 3GDetecting interference RSRPRSRP is the linear average of received power of RS resource elements. Reporting range -44-140 dBm.RSRP measurement with scanner is the most reliable way to detect areas with possible interference problems and bad dominanceNot impacted by network loadRSRP m

23、easurement appears to be consistent between UEs/scannersThe number of PCIs in e.g. 10 dB power window is a useful indicatorA scanner with good dynamic range and PCI tracking capability neededPresentation / Author / Date Bad dominance!Detecting interference summary Absolute SINR measurement values ca

24、nt be used as a reliable performance indicator.Operators should be educated, not to blindly believe measured SINR values.Relative SINR changes can be used as performance indicator, if the same measurement tool is used all the time. SINR measured from S-SCH and RS behaves differently depending on the

25、 interference situation (intra/inter eNodeB). Detailed SINR measurement methods of the terminals and scanners are not known.RSRQ results depend on serving cell load and be variable in different UEs and other tools. The most robust and reliable measurement quantity seems to be RSRPPresentation / Auth

26、or / Date PUSCH SINR, PUSCH RSSI measurement PUSCH RSSI and PUSCH SINR measurement can be used to detect UL coverage and UL interference problemsInterpretation of counter values depends on UL PC settingsMeasurements are not correlatedRSSISINRUL CL PC upper SINR thrshldUL CL PC lower SINR thrshldUL C

27、L PC upper RSSI thrshldUL CL PC lower RSSI thrshldIdeally all samples are in this boxBTS noise floor for 1PRB bandwidth = -119dBmUL interferencebad UL coverageUL tx pwr too high, generates interferenceUL tx pwr too high, generates interferencePUSCH SINR, PUSCH RSSI measurement Noise rise impacts SIN

28、R versus RSSINot possible measure NR from counter directly.Interference drives counter samples to this regionKPI分分析析覆覆蓋蓋問問題：題：Bad downlink vs good downlinkPresentation / Author / Date CQI = 14Good DL coverageFairly bad DL coverage (or DL interference)need e.g. correlated CQI-RSRP counters to analyze

29、 interference further KPI分分析析覆覆蓋蓋問問題：題： Bad uplink coverage vs good uplink coveragePresentation / Author / Date UE Power Headroom: -1dB = PHR +1dB. UE Power Headroom: -15dB = PHR -13dB. Fairly good UL coverageFairly bad UL coverageOpen-loop UL PC used從從KPI/counters-檢檢查查干干擾擾問問題題的的難難點點 下行干擾定位需要由多個下行干擾

30、定位需要由多個counters 綜合分析，如綜合分析，如CQI-RSRP 的關(guān)聯(lián)分析；無直接的關(guān)聯(lián)分析；無直接counter指示下行干擾。指示下行干擾。無直接無直接counter指示上行噪聲抬升。需要根據(jù)指示上行噪聲抬升。需要根據(jù)BTS LOG數(shù)據(jù)進(jìn)數(shù)據(jù)進(jìn)行分析。行分析。Presentation / Author / Date 內(nèi)內(nèi)容容優(yōu)化總述、流程和工具優(yōu)化總述、流程和工具規(guī)劃參數(shù)的影響和優(yōu)化案例規(guī)劃參數(shù)的影響和優(yōu)化案例RF 測量和評估測量和評估RF 環(huán)境的影響和優(yōu)化環(huán)境的影響和優(yōu)化基于基于KPI指標(biāo)的優(yōu)化指標(biāo)的優(yōu)化峰值速率優(yōu)化峰值速率優(yōu)化Dept. / Author / Date優(yōu)優(yōu)化化原

31、原則則和和優(yōu)優(yōu)化化案案例例 一些因素導(dǎo)致的性能下降無法根除,只能通過無線優(yōu)化和其他方法/技術(shù)減少和控制。這些因素的影響是不能忽視的。 鄰區(qū)間干擾 有負(fù)荷的網(wǎng)絡(luò) 另外一些導(dǎo)致的性能下降的原因，應(yīng)該及時發(fā)現(xiàn)糾正和優(yōu)化 天線安裝位置和方式 天線接駁問題 主控區(qū)控制 過覆蓋 RF優(yōu)化的方法包括站點優(yōu)化，天線物理調(diào)整，鄰區(qū)優(yōu)化（missing neighbour)，參數(shù)優(yōu)化等。Presentation / Author / Date 小小區(qū)區(qū)重重疊疊區(qū)區(qū)干干擾擾的的影影響響( Inter-site,Intra-site) Measuring peak MIMO dual-stream throughpu

32、t in the field can be tricky because of interference An idle cell produces common channel + RS interference to impact peak throughput need to find good interference-free measurement spot.Intra-site cell border, frame-synchronized cellsInter-site cell border, non-frame synchronized cellsImpact on pea

33、k tput from idle neighbour cell interference Measurement example #1, Samsung terminal, 20MHz inter-site and intra-site neighbour are unloaded (no PDSCH traffic)Inter-site interference, adjacent site cell about 5 dB weaker RSRP than serving cellIntra-site interference, adjacent cell about 5 dB weaker

34、 RSRP than serving cellAll neighbour cells attenuated 50dBAll neighbour cells attenuated 50dBIntra-site neighbour frame-synced, no RS interferenceImpact on peak tput from 100% loaded neighbour cell Measurement example #2, Samsung terminal, 20MHz Unloaded and 100% loaded inter-site neighbourUDP downl

35、oad 100MbpsNeighbour site cell attenuated 50dBNeighbour site cell about 6 dB weakerNeighbour site cell about 1 dB weakerneighbouring site cell in idle modeTypical SINR= 15-17 dB at inter-site cell border, unloaded neighbour.負(fù)負(fù)荷荷的的影影響響： idle versus 100% loadExample results (RL10, 10MHz bw, 3HK trial)

36、負(fù)負(fù)荷荷的的影影響響： 0% vs 70% using RL20 LTE819The same dt route driven twice, with the same UE setupLTE819: DL Inter-cell Interference Generation to generate load0% load versus 70% DL loadCompare distribution of throughput and SINR, the same drive test route twice with and without load20MHz OL-MIMO, FTP do

37、wnload, 1UE inside the car, Samsung BT-3710, UE-internal antennasaverage throughput is 58% better without interferenceSelection of drive test route strongly affects result, here only results for one dt route010203040506070809010001throughput MbpsCDFEmpirical CDF 70% OCNG0%

38、 OCNG-505101520253001SINR dBCDFEmpirical CDF 70% OCNG0% OCNGMean = 36MbpsMean = 57Mbps案案例：例：MIMO x-feeder 天天線線接接駁駁問問題題ANTL-1 and ANTL-7 are defined active for sector 1ANTL-3 and ANTL-9 are defined active for sector 2Then the configuration in the upper picture is correctThe

39、 configuration in the lower picture is incorrect and results in sectors overlapping with each other bad throughput due to interferencePresentation / Author / Date 1357911Sector1Sector1MIMO x-feeder, example VDF-IT trial, scanner measurementPresentation / Author / Date Sectors 241 and 242 equally str

40、ong in area where 242 should dominate242241站站點點調(diào)調(diào)整整案案例例 reusing 3G sitesImproving performance by blocking excess cellsOverall SINR is improved due to reduction of inter-cell interferenceLocations with improved SINR are visible on the mapImprovement in throughput is even more significant (see next sl

41、ide)Presentation / Author / Date All cells, before optimizationBlocked cells, after optimizationAve SINR improved from 15.2dB to 17.4dBSINR50Mbps regime FTP throughput suffers greatly from:Packet loss, even very little packet loss!Too small TCP RX and TX windowCheck start from UDP If UDP tput not ok

42、 problem in radio parameters or maybe interference. Use scanner to check that there are no neighbours within 30dB of serving cell. Do not fully trust RS CINR measurement! Serving cell RSRP should be -80dBm for peak tput.Check HSS profile that there is no APN-AMBR or UE-AMBR limitationIf UDP tput ok

43、then its transport or TCP problem, troubleshooting can be narrowed down, problem is either i) packet loss in transport network, or ii) TCP settings of FTP server and/or UE. Change to Linux FTP server and optimize UE TCP settings. Check that you have no packet loss in transport network.Presentation /

44、 Author / Date 檢檢查查 radio interface DL buffer utilization: TTI Trace measurement example- -FTP sender okDRB1 data buffer is full OK! Problem not from FTP sender.檢檢查查 radio interface DL buffer utilization: Qualcomm measurement example- -TCP設(shè)設(shè)置置問問題題或或存存在在丟丟包包UE only scheduled about 600-700 subframes a

45、 second, non-full eNB transmit buffer! Problem with TCP tx timeouts and/or packet loss?FTP server is Windows (forget it) Use Linux as FTP server to no optimization needed to reach single-user 100Mbps FTP tput! Windows-based FTP server does not work off-the-shelf if it works at all. WinXPs own FTP cl

46、ient uses by default too small sending side TCP buffer, need to tunning. Also : 3rd party FTP client may override the Windows default. Hence better to set manually tx buffer size from the client Win7 has rx/tx TCP window autotuning. Manually setting window sizes requires 3rd party FTP client, like n

47、cFTP Linux kernels 2.6.x have TCP window rx/tx autotuning, tuning typically not needed for LinuxPresentation / Author / Date 例例:傳傳輸輸?shù)牡挠坝绊戫?Traffic shaping not activated in transport network Whenever feeding a low bandwidth box with high bandwidth input, check shaping in the transport chain Especiall

48、y if the tx buffer in the low bw box is small packet loss can occur reduces TCP throughput!Presentation / Author / Date A2200FPRODUFPR ODUA2200/A1200315Mbps315Mbpsthin pipe1GbpsData transfer directionActivate shaping here to reduce burstinessBox XBox YBox Zthick pipePacket loss here due to too small tx buffer relative to data burstinessGeneralization:Be alert for packet loss here !Example: A-2200 feeding FlexiPacket Traffic shaping not activated in transport networkPresentation / Author / Date