ETSI TR V7.0.0 ( ) Technical Report

Transcription

1 TR V7.0.0 ( ) Techncal Report Unversal Moble Telecommuncatons System (UMTS); Feasblty study on nterference cancellaton for UTRA FDD User Equpment (UE) (3GPP TR verson Release 7)

2 1 TR V7.0.0 ( ) Reference DTR/TSGR v700 Keywords UMTS 650 Route des Lucoles F Sopha Antpols Cedex - FRACE Tel.: Fax: Sret AF 742 C Assocaton à but non lucratf enregstrée à la Sous-Préfecture de Grasse (06) 7803/88 Important notce Indvdual copes of the present document can be downloaded from: The present document may be made avalable n more than one electronc verson or n prnt. In any case of exstng or perceved dfference n contents between such versons, the reference verson s the Portable Document Format (PDF). In case of dspute, the reference shall be the prntng on prnters of the PDF verson kept on a specfc network drve wthn Secretarat. Users of the present document should be aware that the document may be subject to revson or change of status. Informaton on the current status of ths and other documents s avalable at If you fnd errors n the present document, please send your comment to one of the followng servces: Copyrght otfcaton o part may be reproduced except as authorzed by wrtten permsson. The copyrght and the foregong restrcton extend to reproducton n all meda. European Telecommuncatons Standards Insttute All rghts reserved. DECT TM, PLUGTESTS TM, UMTS TM, TIPHO TM, the TIPHO logo and the logo are Trade Marks of regstered for the beneft of ts Members. 3GPP TM s a Trade Mark of regstered for the beneft of ts Members and of the 3GPP Organzatonal Partners. LTE s a Trade Mark of currently beng regstered for the beneft of ts Members and of the 3GPP Organzatonal Partners. GSM and the GSM logo are Trade Marks regstered and owned by the GSM Assocaton.

3 2 TR V7.0.0 ( ) Intellectual Property Rghts IPRs essental or potentally essental to the present document may have been declared to. The nformaton pertanng to these essental IPRs, f any, s publcly avalable for members and non-members, and can be found n SR : "Intellectual Property Rghts (IPRs); Essental, or potentally Essental, IPRs notfed to n respect of standards", whch s avalable from the Secretarat. Latest updates are avalable on the Web server ( Pursuant to the IPR Polcy, no nvestgaton, ncludng IPR searches, has been carred out by. o guarantee can be gven as to the exstence of other IPRs not referenced n SR (or the updates on the Web server) whch are, or may be, or may become, essental to the present document. Foreword Ths Techncal Report (TR) has been produced by 3rd Generaton Partnershp Project (3GPP). The present document may refer to techncal specfcatons or reports usng ther 3GPP denttes, UMTS denttes or GSM denttes. These should be nterpreted as beng references to the correspondng delverables. The cross reference between GSM, UMTS, 3GPP and denttes can be found under

4 3 TR V7.0.0 ( ) Contents Intellectual Property Rghts...2 Foreword...2 Foreword...5 Introducton Scope References Abbrevatons Recever methods Two-branch nterference mtgaton One-branch nterference mtgaton etwork scenaros Interference modellng General Statstcal measures Interference profle based on medan values Interference profles based on weghted average throughput gan General db geometry db geometry Interference profles based on feld data Summary Transmtted code/power characterstcs General Transmtted code and power characterstc n case of HSDPA Common channels for servng and nterferng cells Servng cell Transmtted code and power characterstcs for HSDPA+R"99 scenaro Transmtted code and power characterstcs for HSDPA-only scenaro Interferng cells Transmtted code and power characterstcs for HSDPA+R"99 scenaro Transmtted code and power characterstcs for HSDPA-only scenaro Model for the power control sequence generaton Lnk performance characterzaton General Overvew Smulaton results Types 2 and 2 - medan DIP values Types 3 and 3 - medan DIP values Weghted DIPS: geometres -3 & 0 db Revsed DIP: geometry -3 db Power control Feld based DIP Types 2 / 2 recevers: weghted & revsed DIPS Appendx System performance characterzaton General Frst system-level study (Ercsson) Smulaton setup Smulaton results Second system-level study (oka)...61

5 4 TR V7.0.0 ( ) Smulaton setup for second study Smulaton results for second study Conclusons Recever mplementaton ssues Conclusons...67 Annex A: Change hstory...69 Hstory...70

6 5 TR V7.0.0 ( ) Foreword Ths Techncal Report has been produced by the 3 rd Generaton Partnershp Project (3GPP). The contents of the present document are subject to contnung work wthn the TSG and may change followng formal TSG approval. Should the TSG modfy the contents of the present document, t wll be re-released by the TSG wth an dentfyng change of release date and an ncrease n verson number as follows: Verson x.y.z where: x the frst dgt: 1 presented to TSG for nformaton; 2 presented to TSG for approval; 3 or greater ndcates TSG approved document under change control. y the second dgt s ncremented for all changes of substance,.e. techncal enhancements, correctons, updates, etc. z the thrd dgt s ncremented when edtoral only changes have been ncorporated n the document. Introducton A study tem for further mproved mnmum performance requrements for UMTS/HSDPA UE (FDD) was approved at the 3GPP RA #30 meetng [1]. Ths techncal report summarzes the work that RA4 has accomplshed n ths study tem to assess the feasblty of both one-branch and two-branch nterference cancellaton/mtgaton UE recevers. These recevers attempt to cancel the nterference that arses from users operatng outsde the servng cell. Ths type of nterference s also referred to as 'other-cell' nterference. In past lnk level evaluatons, ths type of nterference has been modelled as AWG, and as such can not be cancelled. The study tem has developed models for ths nterference n terms of the number of nterferng ode Bs to consder, and ther powers relatve to the total other cell nterference power, the latter ratos referred to as Domnant Interferer Proporton (DIP) ratos. DIP ratos have been defned based on three crtera; medan values of the correspondng cumulatve densty functons, weghted average throughput gan, and feld data. In addton, two network scenaros are defned, one based solely on HSDPA traffc (HSDPA-only), and the other based on a mxture of HSDPA and Rel. 99 voce traffc (HSDPA+R99). Interference aware recevers, referred to as type 2 and type 3, were defned as extensons of the exstng type 2 and type 3 recevers, respectvely. The basc recever structure s that of an LMMSE sub-chp level equalzer whch takes nto account not only the channel response matrx of the servng cell, but also the channel response matrces of the most sgnfcant nterferng cells. HSDPA throughput estmates are developed usng lnk level smulatons, whch nclude the other-cell nterference model plus OCS models for the servng and nterferng cells based on the two network scenaros consdered. In addton, system level performance s assessed to determne the gans that nterference cancellaton/mtgaton recever mght provde n throughput and coverage. Complexty ssues assocated wth mplementng these types of recevers are also dscussed. The content of each specfc clause of the report s brefly descrbed as follows. Clause 1 of ths document defnes the scope and objectves of ths feasblty study. Clause 4 descrbes the recever methods that can be appled to one-branch and two-branch Interference Cancellaton (IC) recevers. The reference recevers for the type 2 and type 3 are defned, both of whch are based on LMMSE sub-chp level equalzers wth nterference-aware capabltes. Clause 5 descrbes the two network scenaros that were defned and used to generate the nterference statstcs, whch were then used to develop the nterference models descrbed n clause 6. Clause 6 defnes the nterference models/profles that were developed n order to assess the lnk level performance of IC recevers. The DIP rato s defned as a key statstcal measure, whch forms the bass of the three types of nterference profles consdered.

7 6 TR V7.0.0 ( ) Clause 7 defnes the code and power characterstcs of the sgnals transmtted by the servng and nterferng cells for the two network scenaros defned n clause 5. These latter defntons essentally defne the sgnal characterstcs of the desred user, the common channels and the OCS for both servng and nterferng cells. Clause 8 summarzes the lnk level smulaton results based on the assumptons developed n clauses 6 and 7, whle clause 9 summarzes the system level performance characterzaton. Clause 10 dscusses the possble recever mplementaton losses for a two-branch, sub-chp based LMMSE equalzer wth nterference aware capabltes. Fnally, clause 11 provdes the relevant conclusons that can be taken from ths study.

8 7 TR V7.0.0 ( ) 1 Scope The objectve of ths study s to evaluate the feasblty and potental performance mprovements of nterference cancellaton/mtgaton technques for UTRA FDD UE recevers, based on realstc network scenaros. Scope of the work ncludes: - Determne realstc network scenaros. - Determne sutable nterference models for 'other cell' nterference. - Evaluate the feasblty of two-branch nterference cancellaton recevers through lnk and system level analyss and smulatons. - Evaluate feasblty of one-branch nterference cancellaton recevers through lnk and system level analyss and smulatons. 2 References The followng documents contan provsons whch, through reference n ths text, consttute provsons of the present document. References are ether specfc (dentfed by date of publcaton, edton number, verson number, etc.) or non-specfc. For a specfc reference, subsequent revsons do not apply. For a non-specfc reference, the latest verson apples. In the case of a reference to a 3GPP document (ncludng a GSM document), a non-specfc reference mplctly refers to the latest verson of that document n the same Release as the present document. [1] RP , "ew Study Item Proposal: Further Improved Performance Requrements for UMTS/HSDPA UE", Cngular Wreless, RA #30. [2] R , "Reference structure for nterference mtgaton smulatons wth HSDPA and recever dversty", oka, RA4 #39. [3] R , "Mnutes of Ad Hoc on Further Improved Performance Requrements for UMTS/HSDPA UE (FDD)", oka, RA4 #38. [4] R , "Analyss for smulaton scenaro defnton to nterference mtgaton studes", oka, RA4#38. [5] R , "etwork Scenaros and Assocated Interference Profles for Evaluaton of Generalzed Interference Cancellaton (IC) Recevers", Cngular, RA4 #38. [6] TR v4.0.0, "Physcal layer aspects of UTRA Hgh Speed Downlnk Packet Access (Release 4)". [7] TR V6.0.0 (2004-3), "Feasblty Study for Enhanced Uplnk for UTRA FDD (Release 6)". [8] R , "Throughput smulaton results for Type 3 and Type 3 recevers wth shadow fadng and realstc DIP values for Ior/Ioc=0 db", InterDgtal, RA4 #40. [9] R , "Some observatons on DIP values as a functon of network geometres", TensorComm, RA4 #40. [10] R , "Analyss smulaton results for scenaro defnton to nterference mtgaton studes", oka, RA4 #39. [11] R , "HSDPA etwork Scenaro and Assocated Interference Profle for Evaluaton of Generalzed Interference Cancellaton (IC) Recevers", Cngular/AT&T, RA4 #39.

9 8 TR V7.0.0 ( ) [12] R , "Observatons on Other-Cell Interference Modellng", Motorola, RA4 #39. [13] R , "Interferer Statstcs for the UE IC Study Item," Qualcomm, RA4 #39. [14] TR , "Feasblty Study on Sngle Antenna Interference Cancellaton (SAIC) for GSM etworks (Release 6)". [15] R , "Mnutes of Wednesday Evenng Ad Hoc on Interference Mtgaton", Cngular/AT&T, RA4 #39. [16] 3GPP RA4 reflector e-mal, "Correcton n DIP and AWG/Ioc values shown n R ", May 31, [17] R , "Mnutes of Interference Cancellaton Ad Hoc", Cngular/AT&T, RA4 #40. [18] R , "Throughput smulaton results for type 3 and type 3 recevers based on an alternatve method for determnng DIP values", AT&T Labs, Inc. & Cngular Wreless, RA4 #41. [19] R , Throughput smulaton results for type 3 and type 3 recevers for Ior/Ioc = -3 db based on an alternatve method for determnng DIP values, AT&T Labs, Inc. & Cngular Wreless, RA4 #42. [20] R , "HSDPA Smulaton Results for Type 3 wth Weghted DIP Values", Motorola, RA4 #41. [21] R , Mnutes of nterm conference call on nterference cancellaton study tem, AT&T Labs, Inc. & Cngular Wreless, RA4 #42. [22] R , "Interference data collecton on a lve UMTS network", Orange, RA4 #41. [23] R , "Interference Statstcs Based on Feld Measurements from an Operatonal UMTS/HSDPA Market", Cngular, AT&T and TensorComm, RA4 #41. [24] R , "Scenaro defnton for nterference cancellaton evaluaton based on measurements taken n 3 UK's operatonal network", 3, RA4 #41. [25] R , "Further Thoughts on Scenaro Defnton for Studyng Lnk Performance of Generalzed IC Recevers", Ercsson, RA4 #39. [26] R , "Code Structure of Servng and Interferng Base Statons", Motorola, RA4 #39. [27] R , "Detals of the Code Structure and Power Allocaton for the HSDPA UE IC Case", Qualcomm, RA4 #39. [28] R , "Modellng of transmsson for nterference mtgaton studes", oka, RA4 #39. [29] R , "Modellng of the code structure n servng and nterferng base staton for HSDPA", oka, RA4, 39. [30] R , "Dscusson of realstc scenaros for Interference Cancellaton", Ercsson, RA4 #40. [31] R , "Modellng of power control behavour for OCS", oka, RA4 #42. [32] R , Smulaton Results for Type 2 and 2 Recevers for HSDPA+R99 Scenaro, Motorola, RA4 #40. [33] R , Smulaton Results for Type 3 and 3 Recevers for HSDPA+R99 Scenaro, Motorola, RA4 #40. [34] R , Smulaton results for Interference Cancellaton (IC0 study tem, Ercsson, RA4 #40. [35] R , HSDPA type 3 recever smulaton results, Fujtsu, RA4 #40. [36] R , HSDPA type 2 recever smulaton results, Fujtsu, RA4 #40. [37] R , Smulaton results for agreed medan DIP values, InterDgtal Communcatons Corporaton, RA4 #40.

10 9 TR V7.0.0 ( ) [38] R , Smulaton Results for HSDPA Type 2 and Type 3 Recevers, Intel Corp., RA4 #40. [39] R , Smulaton Results for Type 2 and 2 Recevers for HSDPA Scenaro, Motorola, RA4 #40. [40] R , Smulaton Results for Type 3 and 3 Recevers for HSDPA Scenaro, Motorola, RA4 #40. [41] R , Intal smulaton results for HSDPA+R99 scenaro, oka, RA4 #40. [42] R , Intal Smulaton Results for Type 3 Recever n HSDPA+R99 scenaro, Panasonc, RA4 #40. [43] R , Mnutes of Interference Cancellaton Ad Hoc, Cngular Wreless / AT&T, RA4 #40. [44] R , Smulaton results for HSDPA-Only scenaro usng Type 3 and Type 3 recevers, Ercsson, RA4 #41. [45] R , Ideal smulaton results for HSDPA+R99 scenaro, oka, RA4 #41. [46] R , Ideal smulaton results for HSDPA-only scenaro, oka, RA4 #41. [47] R , Smulaton Results for Type 3 Recever n HSDPA scenaro, Panasonc, RA4 #41. [48] R , Smulaton Results for Type 3 and 3 Recevers for the HSDPA scenaro, Ior/Ioc=- 3and 0 db, InterDgtal, RA4 #41. [49] R , Smulaton results for Type 3 and Type 3 Recevers for HSDPA + R99 Scenaro, Agere Systems, RA4 #41. [50] R , Smulaton results for Type 3 and Type 3 Recevers for HSDPA Scenaro, Agere Systems, RA4 #41. [51] R , HSDPA type3 recever smulaton results, Fujtsu, RA4 #41. [52] R , Smulaton Results for HSDPA Type 3 Recevers, Intel Corp., RA4 #41. [53] R , HSDPA Smulaton Results for Type 3 wth Weghted DIP Values, Motorola, RA4 #41. [54] R , Addtonal Lnk Level Smulaton Results for Type 3 and Type 3 Recevers, AT&T/Cngular, RA4 #42. [55] R , Evaluaton of Power Control Algorthms, AT&T/Cngular, RA4 #42. [56] R , HSDPA Smulaton Results for Type 3 / 3 Recevers, Motorola, RA4 #42. [57] R , Smulaton results for HSDP-only scenaro usng Type 3 and Type 3 recevers, LG Electroncs, RA4 #42. [58] R , Smulaton results for HSDPA+R'99 scenaro usng Type 3 and Type 3 recevers, LG Electroncs, RA4 #42. [59] R , Smulaton results for Type 3 recever for Ior/Ioc = -3dB, Fujtsu, RA4 #42. [60] R , Smulaton Results for Type 3 and 3 Recevers for the HSDPA scenaro, Ior/Ioc=-3, InterDgtal, RA4 #42. [61] R , Smulaton Results for HSDPA Type 3 Recever at Ior/Ioc=-3dB, Marvell, RA4 #42. [62] R , Smulaton results for HSDPA-only scenaro, oka, RA4 #42. [63] R , Smulaton results for R'99 scenaro, oka, RA4 #42. [64] R , HSDPA Smulaton Results for Type 2 /2 Recevers, Tensorcomm, RA4 #42.

11 10 TR V7.0.0 ( ) [65] R , HSDPA Smulaton Results for Type 3 / 3 Recevers wth Power Control, Motorola, RA4 #42. [66] R , "Text proposal for nterference cancellaton SI TR, Secton 9", Ercsson, oka, RA4 #42. [67] TS V7.1.0 ( ), "User Equpment (UE) rado transmsson and recepton (Release 7)". [68] TR V4.0.0 ( ), "Physcal layer aspects of UTRA Hgh Speed Downlnk Packet Access (Release 4)". [69] TR V3.2.0 ( ), "UMTS; Selecton procedures for choce of rado transmsson technologes of the UMTS (UMTS verson 3.2.0)". [70] Globetrotter GT MAX 7.2 Ready Data Card from Opton. [71] R , "Smulaton Assumptons for Rx Dversty + LMMSE Equalzer Enhanced HSDPA Recever (Type 3), Qualcomm, RA4 #36. 3 Abbrevatons For the purposes of the present document, the followng abbrevatons apply: DIP IC LMMSE UE UTRA Domnant Interferer Proporton Interference Cancellaton Lnear Mnmum Mean Squared Error User Equpment UMTS Terrestral Rado Access 4 Recever methods In ths clause we gve the system equatons for the LMMSE chp-level equalzer wth and wthout receve dversty for evaluatng the benefts for nterference mtgaton [2]. In the assumptons used n earler work for enhanced performance requrements Type 2 and Type 3 the nterference structure was assumed to be whte and the varance to be deally known. In the structure presented n followng clauses the nterference structure s now assumed to be colored and the covarance matrx s structured based on deal knowledge of the channel matrces of the nterferng base statons. Ths enables the evaluaton of benefts of nterference mtgaton n the equalzer structure whle the approach to derve (estmate) the nterference covarance matrx does not need to be defned. 4.1 Two-branch nterference mtgaton The receved sgnal s assumed to be expressed as a sum of "own" sgnal, nterferng sgnals and the whte nose: receved sgnal BS j r { = M d + M d j + n { vector j = 1 whte own sgnal , (1) coloured nose j where M j = {0,..., } represents the channel matrx correspondng to BS j, contanng the contrbuton from, BS j ( H ) ( H H = Δ j H 2) j 1 both receve antenna branches. The M where H equals channel-matrx for the -th recever antenna. As a general concept, the equalzer conssts of two FIR flters w 1 and w 2 of length F s : nose T [ w ( 0) w ( F 2) w ( F 1) ], = 1, 2 w = L s s (2)

12 TR V7.0.0 ( ) 11 3GPP TR verson Release 7 where the s s the number of samples per chp and F s the length of the equalzer n unts of chps. The sampled receved vectors at two antennas are denoted by [ ], 1,2, ) 1) (( ) ( 1) 1) (( ) ( = = F D m r m r D m r m T s s s L L r (3) where D s a delay parameter ( ' 0 L F D + ). The equalzaton operaton amounts to obtanng the fltered sgnal ) ( ) ( ) ( m m m y T T r w r w + =. (4) The receved sgnal r (m) can be expressed as ) ( ) ( ) ( ) ( m m m m j j j T T BS n d H d H r + + = =, (5) where T L L L L s s s s s s s s s s s s s s s s s s s = ) ' ( ) ' ( ) ' ( ) ' ( h h h h H L O M O L L (6) s the (F+L') x F s channel-matrx for the -th antenna wth ( ) ( ) ( ) + + = + s s s s s s s s L L h h h L h h h L h h h s ' ) ( (0) 2 1) ' ( 2) (2 2) ( 1 1) ' ( 1) (2 1) ( 1) ' ( L M O M M L L h (7) where s L L = ' s the delay spread normalzed by the chp nterval. Moreover, [ ] T j j j j L F D m d m d D m d m ) 1 ' ( ) ( ) ( ) ( = L L d (8) s the m-th subsequence of the transmtted chp-rate sequence, and n (m) 1 s the correspondng nose vector. Under the assumptons that the nose s colored and the total transmt power from own cell s 1, the LMMSE equalzer taps can be calculated as follows { D own matrx H H rr D H D H matrx sg receved rr H δ δ δ = = ) ( cov ) ( ) ( ) ( ) ( M H H C H H C w w (9) where the notaton Xδ D means the D-th column of the matrx X. The rr C s based on the known propagaton channels only,.e. } {0,..., BS j j = M s based on the deal channel coeffcents. Thus the rr C matrx s constructed as: + + = = BS j j n H j j H rr I P ) ( ) ( σ M M M M C, (10)

13 12 TR V7.0.0 ( ) where the P j, j = {1,..., BS} represent the transmsson power of the BS j takng nto account that the transmsson power of the own base staton s normalzed to unty 1. σ 2 n s the varance of the nose vector n (m), whch s assumed to be same for =1,2. The above equaton also assumes that the noses at dfferent antennas are ndependent. 4.2 One-branch nterference mtgaton In prevous clause 4.1, the system equatons were presented assumng two recever branches. These equatons can be modfed for sngle branch recever. The sgnal model n equaton (1) can be used by assumng that the channel matrx equals receve antenna equalzer. j j H M = Δ ( H1 ) for sngle Usng the same assumptons, the LMMSE equalzer taps can be calculated as follows w 1 0 H 1 = C { rr ( H1 ) δ D receved sg covmatrx (11) where the notaton Xδ D means the D-th column of the matrx X. The C rr s based on the known propagaton channels j only,.e. M, j = {0,... BS} s based on the deal channel coeffcents. Thus, the C rr matrx s constructed as: BS 0 0H = M M + PjM j = 1 j H 2 ( M ) + σ n I j C rr, (12) where the j, j {1,..., BS} power of the own base staton s normalzed to unty. σ 2 n s the varance of the nose vector n (m). P = represent the transmsson power of the BS j takng nto account that the transmsson 5 etwork scenaros To estmate the lnk gan that UE Interference Cancellaton (IC) recevers mght provde for UMTS/HSDPA downlnks t s necessary to frst defne the network scenaros under whch the recevers must operate. A network scenaro for downlnk performance evaluaton s typcally defned n terms of ode B transmt characterstcs, UE receve characterstcs, traffc mx, nter-ste dstance, path loss model, etc. Once the network scenaro(s) s defned one can then determne the assocated nterference profle/model that wll be used n the actual lnk level characterzaton. Ths clause descrbes the network scenaros agreed to n ths study, whle the followng clause defnes the nterference models that were developed based on system level smulatons of these network scenaros. Two network scenaros have been defned n ths feasblty study as shown n Table 5.1, wth one scenaro focusng on HSDPA-only traffc, and the second scenaro focusng on HSDPA + Release 99 voce traffc. Table 5.1: etwork Scenaros Traffc etwork Scenaro 1 etwork Scenaro 2 HSDPA-only HSDPA + Release 99 voce The man system level assumptons are dentcal for each scenaro, and are summarzed n Table 5.2. Ths amounts to defnng two network scenaros whch are dentcal except for the traffc assumed. The system parameters and ther assocated values provded n Table 5.2 were ntally defned n [3], whch summarzed the results of an ad-hoc meetng 1 = BS j = 1 I oc Iˆ 2 orj + σ n

14 13 TR V7.0.0 ( ) held durng TSG RA WG4 #38. These assumptons were based on the mergng of nformaton provded n [4] and [5]. The vast majorty of these assumptons are based on pror work wthn 3GPP RA WG4 ncludng [6] and [7]. In some of these latter studes a second nter-ste dstance of 2800 m was also consdered n addton to the 1000 m specfed n Table 5.2, but snce we are prmarly nterested n nterference-lmted envronments the group felt that the 1000 m condton alone was suffcent. For HSDPA traffc the full-buffer traffc assumpton was made to ensure that all cells were fully loaded. Also, snce the purpose of these system level smulatons was to generate statstcs to accurately characterze the nterference n the system, a round-robn packet scheduler was recommended for the system smulatons to ensure that all UEs had an equal chance of beng scheduled. Ths type of scheduler ensured that when the system smulator was executed over many teratons, that nterference statstcs were collected unformly over the entre smulated area. Choosng a scheduler such as "Max C/I" would skew the generated statstcs because a Max C/I scheduler tends to schedule UEs that are closer n to the cell ste (due to better C/I at closer-n locatons). System level smulatons were then conducted based on the above assumptons for the purposes of collectng nterference statstcs. Statc system level smulators were deemed suffcent for ths exercse, and are preferred over dynamc smulators snce they are typcally easer to develop and requre less computaton tme. For every "teraton" n the statc smulator UEs are randomly dstrbuted across the smulated area and the relevant statstcs collected. From these collected statstcs certan key measures are developed, whch provde some nsght nto how well an nterference cancellaton recever mght work. These key measures and the resultng nterference modellng requred for lnk level performance characterzaton are dscussed n the next clause. Table 5.2: System level assumptons for network scenaros Parameter Assumpton as n [5] Cellular layout Hexagonal grd, 19 stes wth 3 sectors Ste to ste dstance 1000 m Propagaton Model L= Log 10(R km) Std. of slow fadng 8 db Correlaton between sectors 1.0 Correlaton between stes 0.5 Carrer frequency 2000MHz MCL 70 db BS antenna gan 14dB 2 BS antenna pattern θ A( θ ) = Am θ BS total TX power UE antenna gan UE nose fgure mn 12, where θ 3dB θ s defned as the angle between the drecton of nterest and the boresght of the antenna, θ 3dB s the 3dB beamwdth n degrees, and A m s the maxmum attenuaton. Front-to-back rato, A m, s set to 20dB. θ used s 70 degrees. 3dB 20W 0dB 9dB 6 Interference modellng 6.1 General In ths clause we defne the nterference models/profles that were developed n order to assess the lnk level performance of Interference Cancellaton (IC) recevers. Clause 6.2 defnes a number of statstcal measures that were defned durng the study, and whch provde useful nsght nto understandng the complex nterference envronment. One of these measures, referred to as the Domnant Interferer Proporton (DIP) rato, was agreed to n [3] as a key parameter for defnng the nterference profles. System level smulatons were conducted to generate results for the statstcal measures defned n clause 6.2. Based on these smulaton results nterference profles were developed, whch were used n the lnk level performance characterzaton descrbed n clause 8. For the HSDPA-only network scenaro, the workng group defned the followng types of nterference profles:

15 14 TR V7.0.0 ( ) ) Interference profle based on medan values ) Interference profles based on weghted average throughput gan ) Interference profles based on feld data Intally, the group defned an nterference profle based on medan DIP values. However, after the ntal lnk level characterzaton, there were some n the group that thought ths profle was too pessmstc. Ths led the group to explore other methods that mght be more representatve of the gans that an IC recever would actually provde. Subsequently, profles condtoned on geometry were defned based on the "weghted average throughput gan" method as descrbed n [8]. There were some that even thought that ths latter method was too pessmstc when compared to feld data [9], but the majorty of the group felt that t was a good compromse between the profle based on medan values and one based on feld data. Clauses 6.3, 6.4, and 6.5 present the nterference characterzaton results leadng to the development of the above three types of nterference profles respectvely. For the HSDPA + R99 network scenaro, the group decded to use the same nterference profles as the HSDPA-only network scenaro to assess lnk level performance of IC recevers. Fnally, clause 6.6 presents a summary of all the nterference profles developed for ths study tem. 6.2 Statstcal measures etwork nterference statstcs are computed usng the followng defned measures. Geometry G s defned as G = Iˆ I or1 oc = BS Iˆ Iˆ or1 orj j=2 +, where Î orj s the average receved power from the j-th strongest base staton (Î or1 mples servng cell), s the thermal nose power over the receved bandwdth, and BS s the total number of base statons consdered ncludng the servng cell. The Domnant Interferer Proporton (DIP) defnes the rato of the power of a gven nterferng base staton over the total other cell nterference power. It was defned n [3], and can be wrtten as, DIP Iˆ or ( + 1) =, I oc where I oc = BS Iˆ orj +. j= 2 ote that power from the servng cell, Î or1, s never ncluded n any DIP calculaton. Results for the Domnant Interferer Rato (DIR) were also presented n the workng group meetngs to characterze the nterference envronment. However, n [1] t was agreed that DIP ratos would be used to defne the nterference profles and to serve as the nterface between system level smulaton results and lnk level performance characterzaton. Hence, results for DIR statstcs are not ncluded n ths feasblty study report. The reader s referred to references [10] and [11] for DIR defntons and results. 6.2 Interference profle based on medan values Ths clause presents nterference characterzaton results leadng to the development of the nterference profle based on medan DIP values. Ths clause frst presents geometry statstcs obtaned from system level smulatons. Ths s followed by results from contrbutons, whch attempt to determne the number of nterferng base statons whch should be consdered for proper lnk level characterzaton. These latter results ndcated that fve nterferng cells should be modelled n the nterference profle. DIP rato statstcs are presented after that showng uncondtonal DIP CDFs and

16 15 TR V7.0.0 ( ) condtonal medan DIP values, the latter condtoned on varous geometry values. Ths led to the group selectng an nterference profle defned by a sngle set of medan DIP values for all geometres. Fgures 6.1 to 6.4 show the cumulatve dstrbuton functons (CDFs) of geometry (Î or1 /I oc ) generated by varous companes for the HSDPA-only network scenaro. The maxmum value of geometry s lmted to 17 db due to the 20 db front-to-back rato of the antenna specfed n clause 5. These fgures show good agreement between results. For example the medan value s about -2.5 db for all of the curves. Ths close agreement verfes to some extent, proper operaton of each company"s statc system level smulator. Fgure 6.1: Geometry CDF [10] Fgure 6.2: Geometry CDF [11]

17 16 TR V7.0.0 ( ) Fgure 6.3: Geometry CDF [12] CDF Geometry Geometry db Fgure 6.4: Geometry CDF [13] In order to decde the approprate number of nterferers to model for lnk level characterzaton, t was agreed [3] to ntally evaluate nterference statstcs for the eght strongest nterferng cells. There s a trade-off - a larger number of modeled nterferers n the profle makes lnk level characterzaton smulatons and eventual testng more complex, but t also makes the nterference model more accurate. After revewng results for measured statstcs, the group decded [14] that an approprate trade-off between complexty and accuracy can be acheved by defnng the nterference profle wth fve strongest nterferng base statons plus a fltered AWG component to model the resdual nterference. Fgures 6.5 to 6.8 present results generated by varous companes to show the contrbuton of the eght strongest nterferng cells to the total nterference n the system. Here, the term total nterference refers to I oc as defned n Clause 6.2. It can be observed that the fve strongest nterferers contrbute a large majorty of the total nterference.

18 17 TR V7.0.0 ( ) Fgure 6.5: 8 Strongest Interferers [9] Medan Percentage Contrbuton of 8 Strongest Interferers to Total Interference 45% 40% 38.7% 35% % of Total Interference 30% 25% 20% 15% 10% 5% 18.7% 10.0% 6.2% 4.2% 3.0% 2.2% 1.7% 0% Interferer Fgure 6.6: 8 Strongest Interferers [10]

19 18 TR V7.0.0 ( ) Fgure 6.7: 8 Strongest Interferers [11] Ratos Medan Interferer Ratos at Varous Geometres Geometry = -3 db Geometry = 0 db Geometry = 5 db Geometry = 10 db Ior1/Ic Ior2/Ic Ior3/Ic Ior4/Ic Ior5/Ic Ior6/Ic Ior7/Ic Ior8/Ic Fgure 6.8: 8 Strongest Interferers [12] The group evaluated uncondtonal DIP values for the eght strongest nterferng cells, as well as condtonal DIP values condtoned on -3 db, 0 db, 5 db, and 10 db values of geometry. Fgures 6.9 to 6.12 show CDFs of uncondtonal DIP for the eght strongest nterferers. Fgures 6.13 to 6.15 show medan values of condtonal DIP for dfferent values of geometry. Based on these DIP results at that tme the group decded that snce there was not a large varablty n DIP values for dfferent geometres, the group could smplfy the number of smulaton scenaros by defnng an nterference profle wth a sngle set of medan DIP values for all geometres.

20 19 TR V7.0.0 ( ) Fgure 6.9: Uncondtonal DIP CDFs [10] Fgure 6.10: Uncondtonal DIP CDFs [11]

21 20 TR V7.0.0 ( ) Fgure 6.11: Uncondtonal DIP CDFs [12] CDF Interferer Profles Ior1/Ic Ior2/Ic Ior3/Ic Ior4/Ic Ior5/Ic Ior6/Ic Ior7/Ic Ior8/Ic db Fgure 6.12: Uncondtonal DIP CDFs [13]

22 21 TR V7.0.0 ( ) Medan DIP (db) db 0 db 5 db 10 db All Geometry (Ior/Ioc) DIP1 DIP2 DIP3 DIP4 DIP5 Fgure 6.13: Condtonal Medan DIPs [10] Fgure 6.14: Condtonal Medan DIPs [12]

23 22 TR V7.0.0 ( ) Medan DIP Condtoned on Geometry Medan DIP (db) db 0 db 5 db 10 db All Geometry (Ior/Ioc) DIP1 DIP2 DIP3 DIP4 DIP5 DIP6 DIP7 DIP8 Fgure 6.15: Condtonal Medan DIPs [11] Thus, an nterference profle was defned on the bass of averagng uncondtonal medan DIP values submtted by four companes as shown n Table 6.1. It was agreed [15] that the nterference profle would consst of the averaged set of fve medan DIP values and one resdual nterferer to model the remanng nterference. It was also agreed that the resdual nterferer would be modeled as fltered AWG. Based on the DIP values shown n Table 6.1, the rato AWG/I oc should be set to -5.8 db, whch s equvalent to about 26% of the total other cell nterference power. The AWG source should be fltered usng the pulse shapng flter defned n TS to nsure correct spectral propertes. These medan DIP values plus the resdual AWG were to be used wth each of the geometres consdered n the ntal lnk level characterzaton. The geometry values used n that ntal characterzaton were -3 db, 0 db, 5 db and 10 db, see clause 8. Table 6.1: Interference Profle Based on Averaged Set of Uncondtonal Medan DIP Values [14] Cngular Qualcomm Motorola oka Average DIP DIP DIP DIP DIP AWG/I oc % 21% 28% 35% 26% 6.3 Interference profles based on weghted average throughput gan General Upon revewng the ntal lnk level performance results for the nterference profle based on medan DIP values, some companes expressed concern that these values were too conservatve, and led to under-estmaton of the benefts of IC recevers. Ths led to the development of an alternatve method for calculatng DIP values based on what s called the "weghted average throughput gan" as descrbed n [8]. Ths method develops multple sets of DIP ratos, the resultng throughputs of whch are averaged to fnd an average throughput gan. The set of DIP ratos closest to ths average s then selected as the nterference profle. Two profles were ultmately defned, one for 0 db geometry, and the other for

24 23 TR V7.0.0 ( ) -3 db geometry. The remander of ths clause descrbes the methodology used to defne these two nterference profles along wth ther assocated values. ote snce the ntal lnk level gans were neglgble for the hgher geometres (5 and 10 db), the group agreed to focus on performance at the lower geometres, whch s ntutvely where an IC recever s gong to provde beneft db geometry The 0 db geometry profle based on weghted average throughput gan was defned based on a methodology presented n [8] and explaned further as follows. In the statc system smulator, UEs were randomly placed throughout the smulated cells of nterest. All of the randomly placed UEs wth a geometry of near I or1 /I oc = 0dB (±0.2 db) were chosen and ther DIP values were saved. Ths process was repeated for multple realzatons, untl a sgnfcant number of samples were obtaned. Then, the saved DIP values were sorted by DIP 1 and then bnned n 5-percentle bands. One random sample was drawn from each 5-percentle band to obtan a total of 20 representatve DIP rato sets. Table 6.2 shows the 20 representatve DIP rato sets that were used to defne the nterference profle for the 0 db case. Table 6.2: DIP ratos for I or1 /I oc = 0dB [8] # I or1/i oc DIP 1 DIP 2 DIP 3 DIP 4 DIP 5 I oc Lnk level smulatons were conducted for each of the above 20 representatve sets of DIP ratos to obtan lnk level throughputs for each set. The average throughput gan over all 20 sets was then calculated. The DIP rato set whose ndvdual throughput gan was closest to the average throughput gan was then chosen as the DIP rato set for the nterference profle. For the data n Table 6.2, the DIP rato set correspondng to row #14 was found to be the one wth throughput gan closest to the average. The correspondng DIP values for ths row are repeated n Table 6.3. These values were used n a second round of lnk level characterzaton for the 0 db geometry case as descrbed n clause 8. Table 6.3 Interference Profle Based on Weghted Average Throughput for 0 db Geometry [8] DIP 1 [db] DIP 2 [db] DIP 3 [db] DIP 4 [db] DIP 5 [db] db geometry In the methodology used to defne the nterference profle for the 0 db geometry case n clause a random sample was drawn from each of the 20 5-percentle bns to obtan the 20 sets of representatve DIP ratos. It was ponted out n [18] that due to ths random draw, the nterference profle defned n clause was not repeatable by other companes. If repeatablty s desred, an alternatve method of obtanng 20 representatve DIP rato sets based on bn-

25 24 TR V7.0.0 ( ) averagng was proposed n [18]. Accordng to ths alternatve method the DIP values calculated for each 5 percentle nterval are based on the average of all of the values that fall wthn that bn. For example, for the I or1 /I oc = -3 db case, all UEs whose DIP 1 value s equal to -3 db (±0.2 db) are sorted accordng to DIP 1 and sampled at 5 percentle ntervals to yeld 20 groups. The 20 representatve DIP values are the average of the DIP values observed by all UEs that fall wthn each of these 20 groups. The 20 sets of DIP values, calculated usng the bn-averagng method for the -3 db geometry case are shown n Table 6.4. Table 6.4 DIP values for Ior/Ioc = -3 db, sorted on 5 th percentle ncrements. [19] Bn # Ior/Ioc DIP 1 DIP 2 DIP 3 DIP 4 DIP It was shown n [18], that both the random draw and bn-averagng methods produce the same throughput gans for the 0 db geometry case and thus, ether method was deemed acceptable for ths case. However, ths was not found to be the case for the -3 db geometry condton, where n [19] there was shown to be a sgnfcant dfference between the throughput gans for the two methods. A sgnfcant dfference was also observed n [20] where throughput gans were compared usng just the random draw method. Based on all of ths, the group decded [21] to adopt the alternatve method based on bn-averagng for the -3 db geometry case, but to leave the nterference profle defned for the 0 db case n clause unchanged snce there was no sgnfcant dfference n lnk level throughput results for ths latter case. Applyng the "weghted average throughput gan" method to the data of Table 6.4 results n the selecton of row #10 as the nterference profle for the -3 db geometry case. For clarty, the selected DIP values of row #10 are repeated n Table 6.5 below. ote n [19] the DIP rato set s actually selected based on the weghted average throughput as opposed to the weghted average throughput gan, but the two methods were found to be nearly equvalent for the data analyzed, and n fact the former method gave a more consstent answer. Table 6.5: Interference Profle based on Weghted Average Throughput for I or1 /I oc = -3 db [19] DIP 1 [db] DIP 2 [db] DIP 3 [db] DIP 4 [db] DIP 5 [db] Interference profles based on feld data The nterference profles defned n clauses 6.3 and 6.4 are all based on the use of statc system level smulators. These smulators are based on a homogeneous layout of hexagonal cells wth unformly dstrbuted users. Thus, they fal to capture a number of real-world effects ncludng non-homogenety of cells, buldngs/terran, and non-unform dstrbuton of users, just to name a few. Even wth these shortcomngs, system level smulatons are stll extremely

26 25 TR V7.0.0 ( ) valuable snce the results developed are typcally repeatable and one can precsely control the envronment. However, t s also very mportant to consder actual feld data when attemptng to determne the feasblty of an advanced UE recever that s attemptng to cancel nterference from other cells as s beng consdered n ths study tem. To ths end, several contrbutons were submtted durng ths effort, whch descrbe a number of feld measurements [9] [22] [23] [24]. These measurements provde addtonal nsght nto how well an IC recever mght actually perform n a real network. In addton, nterference profles condtoned on geometry were defned based on one of the sets of feld data. Lnk level characterzaton usng ths latter set s descrbed n clause 8. The followng brefly descrbes some of the man observatons that can be drawn from the feld data plus the specfcs of the feld-based nterference profles. In [22] nterference data collected n a lve UMTS network n Pars s descrbed. The major observatons from these measurements are as follows: - For mobles at the cell edge, there are n general no more than 3 nterferng cells seen by the UE. In about 65% of the tme, there are only 2 nterferers detected. - DIP 1 values are farly hgh (when compared to other DIP values) and not too spread out, rangng between 0 and -4 db, at geometry Ior/Ioc = 0. - DIP 2 and DIP 3 values are more spread out. - There are not enough 4 th nterferers detected to nclude n a meanngful statstcal analyss. - A 3-D representaton of data confrmed the spread of DIP 2 n partcular over a large range of values: from -2dB to about -13 db. In [23] feld measurements are provded for an operatonal UMTS/HSDPA network n parts of greater and downtown Chcago. The results from these measurements ndcate that for the 0 db geometry case that most of the nterferng energy when measured at the medan ponts of the DIP CDF curves s contaned n the frst two nterferng cells (78%). For -3 db geometry, most of the nterferng energy s n the frst three nterferng cells (concdentally 78%). Feld measurements of DIP values recorded n central London are presented n [24]. DIP measurements were taken condtoned on the followng values of geometry: -3, 0, 5, and 10 db. Based on these measurements, DIP profles were defned as shown n Table 6.6. These values are based on takng the medan value of each respectve DIP value for each of the geometres consdered. Even though ths approach (takng the medan value) s thought to be conservatve, the values n Table 6.6 are stll more optmstc from an IC performance perspectve than the other profles prevously defned, see clause 6.6. If one were to apply the "weghted average throughput gan" method to ths feld data, the results would be even more optmstc. Lnk level results based on the DIP ratos correspondng to -3 and 0 db geometres are provded n clause 8. One of the conclusons that can be drawn from all of the feld measurements s that most of the nterference s contaned n the frst two nterferng cells wth some energy n the thrd dependng upon geometry. Very lttle energy was detected n the fourth and beyond, and thus, the use of fve nterferng cells n the smulaton-based profles may be bt of an over kll. The second concluson s that feld-based profles are more optmstc than the smulaton-based profles and thus, performance n the real world (at least n those locatons where the feld data was collected) should be better than that predcted by the smulatons. Table 6.6: Interference Profles Based on Feld Data [24] Î or1/i oc DIP -3dB 0dB 5dB 10dB DIP DIP DIP DIP DIP Summary In summary, Table 6.7 shows the nterference profles that have been defned as part of ths feasblty study to assess lnk level performance of IC recevers. The top entry reflects the medan DIP values, whch are to be used for all geometres consdered. The next entry defnes the two DIP profles that were defned based on the weghted average throughput gan method for the 0 db and -3 db geometres, respectvely. The last entry shows the DIP entres based on feld data where we have lmted the geometres to 0 and -3 db once snce there s where gan for IC recevers s

27 26 TR V7.0.0 ( ) expected. It s nterestng to note when comparng these profles that the medan profle s actually qute close to both of the profles condtoned on -3 db geometry, and how really close the latter two are to each other. Ths suggests that the medan profle probably should have only been used for the -3 db geometry condton, and that t s mportant to condton the DIP ratos on geometry to obtan meanngful results. Lnk level performance results based on these nterference profles are presented n clause 8. Table 6.7: Summary of Defned Interference Profles Profle DIP1 DIP2 DIP3 DIP4 DIP5 Based on medan values Based on weghted average throughput gan 0 db geometry db geometry Based on feld data 0 db geometry db geometry Transmtted code/power characterstcs 7.0 General Ths clause descrbes the modellng methods and assumed characterstcs of desred and nterferng sgnals for ths study. 7.1 Transmtted code and power characterstc n case of HSDPA In the followng clauses the modellng of code and power characterstcs for servng and nterferng cells s presented. The text s based on [29] amng to merge the proposals presented n documents [25][26][27][28] accountng also the dscussons held durng RA4#39. Also addtonal changes proposed n [30] and [31] were accounted as agreed n nterm teleconference held between RA4 meetngs #41 and #42 [21]. For modellng the transmtted code and power doman characterstcs n case of HSDPA, two dfference scenaros are determned; the "HSDPA-only" and "HSDPA+R"99". The scenaros descrbed n ths clause are separated by the used HS-PDSCH power allocaton and modellng of the assocated dedcated channels. For "HSDPA-only", 66% HS- PDSCH power allocaton s assumed wth assocated channels modeled as F-DPCH, and for "HSDPA+R"99" 50% and 25% allocatons are assumed together wth dedcated channels assumed as DPCH Common channels for servng and nterferng cells The common downlnk channels and correspondng powers used n RA4 HSDPA demodulaton requrements wth sngle transmt antenna are lsted n the Table C.8 of TS [2]. Smlar defntons exst also for open and closed loop transmt dversty requrements n Tables C.9 and C.10 n [2]. Table 7.1 below summarzes the common downlnk physcal channels for sngle transmt antenna case. As these fgures can be consdered to be qute representatve, t s seen that these could be used also for the evaluaton, for both, servng cell and nterferng cells, n case of sngle transmt antenna.

28 27 TR V7.0.0 ( ) Table 7.1: Downlnk Physcal Channels transmtted durng a connecton for HSDPA Physcal Channel Power rato OTE P-CPICH P-CPICH_Ec/Ior = -10 db Use of P-CPICH or S-CPICH as phase reference s specfed for each requrement and s also set by hgher layer sgnallng. P-CCPCH P-CCPCH_Ec/Ior = -12 db When BCH performance s tested the P- CCPCH_Ec/Ior s test dependent SCH SCH_Ec/Ior = -12 db Ths power shall be dvded equally between Prmary and Secondary Synchronous channels PICH PICH_Ec/Ior = -15 db Servng cell In ths clause the defnton of transmtted code and power characterstcs are gven for the servng cell Transmtted code and power characterstcs for HSDPA+R"99 scenaro The assumed downlnk physcal channel code allocatons for HSDPA+R"99 scenaro s gven n Table 7.2. Table 7.3 summarzes the power allocatons of dfferent channels for the servng cell n "HSDPA+R"99" scenaro for 50 % or 25 % HS-PDSCH power allocaton. Ten HS-PDSCH codes have been reserved for user of nterest n Table 7.2. Dependng on the used fxed reference channel defnton, H-SET3 or H-SET6, part of these may be left unused. In total 46 SF=128 codes have been reserved for other users channels (OCS). For HSDPA+R"99 scenaro the (assocated) dedcated channels of other users are modeled as DPCH. The amount of users present s dependent on the power remanng avalable after HSDPA allocaton. For HSDPA power allocaton of 50%, 18 users were ftted to cell. Correspondngly wth HSDPA allocaton of 25%, 34 users were ftted to the cell. The defnton of the other user orthogonal channels and channel powers are gven n Table 7.4 and Table 7.5. Power control behavor of the other users s ntroduced as descrbed n Clause Table 7.2: Downlnk physcal channel code allocaton for HSDPA+R"99 Channelzaton Code at SF=128 ote 0, 1 P-CPICH, P-CCPCH and PICH on SF= SF=128 codes free for OCS HS-PDSCH codes at SF= SF=128 codes free for OCS Table 7.3: Summary of modellng approach for the servng cell n HSDPA+R"99 scenaros Servng cell Common channels (-7.1dB) As gven n Table 7.1 HS-PDSCH transport format H-SET3 or H-SET6 HS-PDSCH power allocaton [E c/i or] 0.5 (-3dB) 0.25 (-6dB) Other users channels (-5.16dB) Set as gven n Table (-2.58dB) Set as gven n Table 7.5 OTE: The values gven n decbel are only for nformaton.

29 28 TR V7.0.0 ( ) Table 7.4: Defnton of 18 other users orthogonal channels on downlnk scenaro wth 50% HS- PDSCH power allocaton Channelzaton Ec/Ior Channelzaton Ec/Ior Channelzaton Ec/Ior Code Cch,SF,k Code Cch,SF,k Code Cch,SF,k Cch,128, Cch,128, Cch,64, Cch,128, Cch,128, Cch,128, Cch,128, Cch,128, Cch,128, Cch,128, Cch,64, Cch,128, Cch,128, Cch,128, Cch,128, Cch,128, Cch,128, Cch,128, Table 7.5: Defnton of 34 other users orthogonal channels on downlnk scenaro wth 25% HS- PDSCH power allocaton Channelzaton Ec/Ior Channelzaton Ec/Ior Channelzaton Ec/Ior Code Cch,SF,k Code Cch,SF,k Code Cch,SF,k C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,64, C ch,64, C ch,64, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, C ch,128, Transmtted code and power characterstcs for HSDPA-only scenaro The assumed downlnk physcal channel code allocatons for the HSDPA-only scenaro s gven n Table 7.6. In the Table 7.7 power allocatons for the servng cell s presented for HSDPA-only scenaro. For HSDPA-only scenaro, as n Table 7.6, 14 codes are made avalable for the HS-DSCH as all the assocated dedcated channels use F-DPCH. Dependng on the used fxed reference channel defnton, H-SET3 or H-SET6, part of these may be left unused. In order to permt comparable smulatons to be performed usng exstng FRC defntons, an addtonal code multplexed user s ntroduced to the servng cell of HSDPA-only scenaro. As H-SET6 requres a maxmum of 10 codes and H-SET3 requres 5 codes, addtonal code multplexed user s defned n case of the servng cell. Ths addtonal code multplexed user utlzes the rest of the avalable codes assumed to be avalable for HSDPA. The code channels ntended for the addtonal code multplexed user shall have equal power and common modulaton. The power per code for H-SET6 shall be ether 0.04 (-14 db) when HS-DSCH Ec/Ior allocated for the DUT s 50%, or (-9.9 db) when HS-DSCH Ec/Ior allocated for the DUT s 25%. The respectve per code power allocatons for H-SET3 are (-17.5 db) and (-13.4 db), see Tables 7.8 and 7.9. Used common modulaton (QPSK or 16QAM) should be randomly selected wth equal probablty. The defnton of other users orthogonal channels s gven n Table The channelzaton code ndces, C ch,256,x and C ch,256,y gven at same row are consdered as par. At any gven symbol nstant, only symbol from ether code channel s transmtted wth the Ec/Ior gven n the last column of the same row. The other code channel s DTX"ed. The code channel transmtted s selected randomly wth even probablty. Ths s done to account the structure of the F-DPCH. Table 7.6: Downlnk physcal channel code allocaton for HSDPA-only scenaro Channelzaton ote Code at SF=128 0 P-CPICH, P-CCPCH and PICH on SF= SF=128 codes free for OCS HS-PDSCH codes at SF= SF=128 codes free for OCS

30 29 TR V7.0.0 ( ) Table 7.7: Summary of the modellng approach for the servng cell n HSDPA-only scenaros Servng cell Common channels (-7.1dB) As gven n Table 7.1 HS-PDSCH transport format H-SET3 or H-SET6 for user of nterest. Addtonal other HSDPA users code allocaton s based on Table 7.8 for H- SET3 or Table 7.9 for H-SET6. Total HS-PDSCH power allocaton [E c/i or] 0.66 (-1.8dB) HS-PDSCH power allocaton for DUT (of [0.5, 0.25] the total) [E c/i or] ([-3dB, -6dB]) Other users dedcated channels 0.14 (-8.54dB) Set accordng to Table OTE: The values gven n decbel are only for nformaton. Table 7.8: Defnton of addtonal code multplexed users orthogonal channel for HSDPA-only scenaro (H-SET3) Channelzaton Ec/Ior [db] Channelzaton Ec/Ior [db] Channelzaton Ec/Ior [db] Code Cch,SF,k Code Cch,SF,k Code Cch,SF,k Cch,16,6 ote1 Cch,16,7 ote1 Cch,16,8 ote1 Cch,16,9 ote1 Cch,16,10 ote1 Cch,16,11 ote1 Cch,16,12 ote1 Cch,16,13 ote1 Cch,16,14 ote1 OTE 1: Used common modulaton should be randomly selected for codes wth equal probablty. The code channels shall have equal power, ether (-17.5 db) when HS-DSCH E c/i or allocated for the DUT s -3dB or (-13.4 db) when HS-DSCH E c/i or allocated for the DUT s -6dB. Table 7.9: Defnton of addtonal code multplexed users orthogonal channel for HSDPA-only scenaro (H-SET6) Channelzaton Ec/Ior [db] Channelzaton Ec/Ior [db] Channelzaton Ec/Ior [db] Code Cch,SF,k Code Cch,SF,k Code Cch,SF,k Cch,16,11 ote2 Cch,16,12 ote2 Cch,16,13 ote2 Cch,16,14 ote2 OTE 2: Used common modulaton should be randomly selected for codes wth equal probablty. The code channels shall have equal power, ether 0.04 (-14 db) when HS-DSCH E c/i or allocated for the DUT s -3dB or (-9.9 db) when HS-DSCH E c/i or allocated for the DUT s -6dB. Table 7.10: Defnton of other users orthogonal channels on downlnk for HSDPA-only scenaro Channelzaton Channelzaton Ec/Ior Code Cch,SF,x Code Cch,SF,y Cch,256,4 Cch,256, Cch,256,5 Cch,256, Cch,256,249 Cch,256, Cch,245,8 Cch,256, Cch,256,9 Cch,256, Cch,256,11 Cch,256, Cch,256,240 Cch,256, Cch,256,242 Cch,256, Interferng cells In ths clause the defnton of transmtted code and power characterstcs are gven for the nterferng cells Transmtted code and power characterstcs for HSDPA+R"99 scenaro For the nterferng cells n HSDPA+R"99 scenaro, same downlnk physcal channel code allocatons are assumed as gven n Table 7.2. The modellng s summarzed n Table 7.11 for HSDPA+R"99 scenaro for nterferng cells.

31 30 TR V7.0.0 ( ) Table 7.11: Summary of modellng approach for the nterferng cells n HSDPA+R"99 scenaros wth power allocaton of 50% and 25% Interferng cell(s) Common channels (-7.1dB) As gven n Table 7.1 HS-PDSCH transport Selected randomly from Table format Independent for each nterferer. HS-PDSCH power 0.5 allocaton [E c/i or] (-3dB) Other users channels (-5.16dB) Set accordng to Table 7.4. OTE: The values gven n decbel are only for nformaton. The HS-PDSCH transmsson for nterferng cells s modelled to have randomly varyng modulaton and number of codes to model the actual dynamc system behavour to some extent. The predefned modulaton and code allocatons are gven n Table The transmsson from each nterferng cell s randomly and ndependently selected every HSDPA sub-tti among the four optons gven n the table. Table 7.12: Predefned nterferer transmsson for HSDPA+R"99 scenaro # Used modulaton and number of HS-PDSCH codes 1 QPSK wth 5 codes 2 16QAM wth 5 codes 3 QPSK wth 10 codes 4 16QAM, wth 10 codes The modellng of the other users dedcated channels s done n same way as n the case of the servng cell. The defnton of the other users" orthogonal channels and channel powers are gven n Table 7.4. As fxed HSDPA power allocaton (50%) s assumed for nterferng cells, only one defnton set s enough Transmtted code and power characterstcs for HSDPA-only scenaro Same downlnk physcal channel code allocatons as for servng cell, gven Table 7.6, are used for the nterferng cells n HSDPA-only scenaro. The modellng of the transmsson of the nterferng cells for HSDPA-only scenaro s summarzed n Table Table 7.13: Summary of modellng approach for the nterferng cells n HSDPA-only scenaros Interferng cell(s) Common channels (-7.1dB) As gven n Table 7.1 HS-PDSCH transport Selected randomly from Table format Independent for each nterferer. Total HS-PDSCH power 0.66 allocaton [E c/i or] (-1.8dB) Other users channels 0.14 (-8.54 db) Set accordng to the Table 7.10 ote: The values gven n decbel are only for nformaton. Smlarly as n case of HSDPA+R"99 scenaro the HS-PDSCH transmsson s modeled as havng varyng modulaton and allocaton of codes. Ths s done by selectng a code and modulaton format from a group of predefned sets, as gven n Table Three dfferent optons are determned, wth one opton ncludng code multplexng. In case of the code multplexng for the nterferng HS-DSCH transmsson (e.g. opton #3) the power s dvded equally between the two assumed users havng dfferent modulaton.

32 31 TR V7.0.0 ( ) Table 7.14: Predefned nterferer transmsson for HSDPA-only scenaro # Used modulaton and number of HS-PDSCH codes 1 QPSK wth 14 codes 2 16QAM, wth 14 codes 3 QPSK wth 7 codes and 16QAM wth 7 codes Model for the power control sequence generaton In ths clause the modellng of power control behavor for the other users channels s gven. In the assumed approach the power of each user, at slot n, equals P n db. The power s vared randomly, ether by ncreasng or decreasng t by 1dB steps n each slot,.e. n n th user at tme nstant n can be determned as Pn 1 n P P 1 + Δ, where Δ { 1, + 1}. The probablty of Δ havng a value of +1 for the = 0.5 Prn ( = + 1) = 0.5 ( Pn 1 P0 ) L where, s the transmtted power at tme nstant n-1 and ntal transmt power values Δ, (1) P 0 s the ntal value of the th users power n db. The P 0 for dfferent users are gven n Tables 7.4 and 7.5 n absolute values. L s a scalng factor whch can be used to determne the range to whch the varaton of power s confned. The value of L s set to10, leadng to varance of 5 db. In order to mnmze the mpact to test equpment due to power control modellng the total OCS power s ntended to be normalzed. Thus the power control sequence for each user s generated as descrbed above. The actual appled transmt power Pˆ n (n lnear doman) for the th user would stll be normalzed by the total sum power of dfferent users, defned as ˆ. (2) Pln, n Pn = Pln, 0 P ln, n where Pln, n s the transmtted power at tme nstant n and Pln,0 s the ntal value of the th user"s power n lnear doman. The normalzaton has mplcatons to the power varaton of each user. 8 Lnk performance characterzaton 8.0 General The purpose of ths clause s to summarze the lnk-level smulaton results that were provded by the companes that supported ths study tem. Emphass s on results that were based on agreed set of smulaton parameters. However, a complete lst of results s gven n the appendx of ths clause. 8.1 Overvew Ths overvew s meant to be a bref synopss of the framework n whch lnk-level smulatons were performed. More detaled nformaton regardng such topcs as code modellng, system scenaros, and nterference modellng are avalable elsewhere n ths report. As descrbed n clause 6 system smulatons were performed by partcpatng companes n order to reach an agreement on lnk-level smulaton parameters. The man focus of ths work was determnng the number of nterferng ode-bs and assocated set of DIP values and geometres to be used. Based on these smulaton studes t was decded to set the number of nterferng ode-bs to fve. Also, a sngle set of medan DIP values would be used for selected geometres. See Table 8.1 for these and other referenced DIP values. In the

33 32 TR V7.0.0 ( ) course of ths study t was proposed that medan DIP values mght gve a pessmstc assessment of the potental gan that a LMMSE recever wth nterference cancellaton capablty mght provde. It was decded that the DIP values should be modfed to take nto account the potental gans at low geometres. Consequently a 2nd set of DIP values were accepted for lnk-level smulaton purposes. In ths case the DIP values are a functon of the operatng geometry; 0 or -3 db. Afterwards, there was dscusson regardng the method of generaton of these DIP values. Subsequently t was decded to accept a modfed DIP for geometry -3 db. Contrbutons that nvestgated the DIP ssue based on feld measurements were also presented; smulatons usng these derved values were presented. Table 8.1: DIP values (db) used durng progresson of SI Evoluton DIP1 DIP2 DIP3 DIP4 DIP5 Status Medan Obsolete Weghted G= Actve Weghted G= Obsolete Revsed Weghted G= Actve Feld Measurement G= Informaton Feld Measurement G= Informaton 8.2 Smulaton results The followng clauses show the average relatve gans as demonstrated by a type 2 over a type 2 recever or a type 3 over a type 3 recever. (As noted n clause 8.2.1, t was eventually decded that ths SI would concentrate on the type 3 / 3 confguraton. However, the type 2 / 2 results are gven for completeness; smlarly lke smulaton results for the obsolete DIP sets.) For a gven smulaton condton, average relatve gan s defned as the rato of the FRC throughput of the type 2 or 3 recever to the throughput of the type 2 or 3 recever, respectvely. Ths "normalzed" numercal result allows for a readly dscernable evaluaton of the effcacy of the type 2 / 3 recever. However, ths sngle quantty may not tell the complete story. If for example, the relatve gan for a gven channel condton s shown to be 7.0 for a 16QAM confguraton, whle the relatve gan s 1.3 for QPSK under the same condton, the utlty of the 16QAM value s of lttle practcal mportance f the actual throughput values are 7 Kb/s and 1 Kb/s for the type and non type recevers. Whereas, f for the same channel condton the QPSK throughput values are 195 Kb/s and 150 Kb/s for the dfferent recevers, these values mght be consdered substantal and useful gans. Therefore, the correspondng average FRC throughput values are shown - expressed as rato of ntegers - for all relatve gans Types 2 and 2 - medan DIP values Tables 8.2 and 8.3 show the relatve average gans and throughputs of the type 2 / 2 recevers for the varous sgnal levels and channel condtons as ndcated for the HSet-6 HSDPA+R99 QPSK and 16QAM scenaros, respectvely. These sets of statstcs are based on the contrbutons of sx companes; see Appendx, Tables 8.A1-8.A2 3 Table 8.2: HSet-6 QPSK gans and throughputs for types 2 / 2 recevers for HSDPA+R99 wth medan DIP values QPSK HSet6 PB3 VA30 Sgnal Level G (db) G(dB) G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / / / / / It should be noted that the relatve gans and rato of throughput values may not be equal. Ths s due to the fact that average of a set of ratos, s not equal to the average of the numerators dvded by the average of the denomnators. 3 Due to the nature of ths secton, where a consderable number of contrbutons are condensed nto a sngle statstc, drect reference s not made to the contrbuton(s). Rather, the raw data of the contrbutons are gven n a set of tables n the appendx; the tables gve the RA4 TD numbers for the ndvdual values, ths TD may be cross-referenced n the Reference secton for the complete ttle and date.

34 33 TR V7.0.0 ( ) Table 8.3: HSet-6 16 QAM gans and throughputs for types 2 / 2 recevers for HSDPA+R99 wth medan DIP values 16QAM HSet6 PB3 VA30 Sgnal Level G (db) G(dB) G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / / / / / 2260 Tables 8.4 and 8.5 gve smulaton results smlar to above but for HSet-3. In ths case the statstcs are based on two contrbutng companes; see Appendx, Table 8.A4 and 8.A5. Table 8.4: HSet-3 QPSK gans and throughputs for types 2 / 2 recever for HSDPA+R99 wth medan DIP values QPSK HSet3 PB3 VA30 Sgnal Level G (db) G(dB) G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / / / / / 1546 Table 8.5: HSet-3 16QAM Gans and throughputs for types 2 / 2 recever for HSDPA+R99 wth medan DIP values 16QAM HSet3 PB3 VA30 Sgnal Level G (db) G(dB) G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / / / / / 1682 From the above tables t can be observed that for geometres of 5 or 10 db the gans are generally small, wth all gans less than 1.10 except for one case where the gan s The average of the gans for geometres 5 and 10 db s Ths average s based on all modulaton types, sgnal levels, channels, and scenaros. The average relatve gan for geometry 0 db s 1.10 when the average s computed only over the better performng modulaton scheme for a gven channel and sgnal level; ths s always QPSK n these cases Types 3 and 3 - medan DIP values Tables 8.6 and 8.7 gve the relatve gans and absolute ratos for recever types 3 and 3 for HSet-6 and the ndcated modulaton wth the use of the medan DIP values. These statstcs are based on appendx Tables 8.A10 and 8.A11. Table 8.6: HSet-6 QPSK gans and throughputs for types 3 / 3 recevers for HSDPA+R99 wth medan DIP values QPSK HSet6 PB3 VA30 Sgnal Level G (db) G(dB) G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / / / / /3209

35 34 TR V7.0.0 ( ) Table 8.7: HSet-6 16QAM gans and throughputs for types 3 / 3 recevers for HSDPA+R99 wth medan DIP values 16QAM HSet6 PB3 VA30 Sgnal Level G (db) G(dB) G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / / / / / 3607 From the tables for types 3 and 3 recevers smulated under the medan DIP set, the average gan computed over geometres 5 and 10 db s Based on these results, ncludng the type 2 / 2 recevers, t was decded to lmt the study tem to lower geometres, specfcally -3 and 0 db. The average relatve gan for geometry 0 db s 1.13 when the average s computed only over the better performng modulaton scheme for a gven channel and sgnal level; lke the types 2 / 2 case, ths s for QPSK modulaton. It was decded that only the type 3 / 3 case would be further studed. Ths decson was not based solely on the gan of 1.13 of the type 3 / 3 compared to the 1.10 gan of the type 2 / 2, as the performance gans are somewhat comparable. However, when the absolute throughput values at geometry 0 db are compared for both recever groups, the larger throughputs of the 3 recever nfluenced the decson to lmt the study tem to the 3 / 3 case Weghted DIPS: geometres -3 & 0 db As ndcated n the Overvew of ths clause the study group decded to use a set of weghted DIP values rather than medan DIP values n order to more realstcally determne the potental gan of the type 3 recever under low geometry condtons. Tables 8.8 and 8.9 show the results when these weghted DIPs were used for smulaton. In ths case these tables represent the combned results of the HSDPA+R99 and HSDPA only scenaros. Specfcally, there were 6 ndependent contrbutons for the HSDPA+R99 scenaro, and 2 to 4 ndependent contrbutons, dependent upon the channel condton and geometry, for the HSDPA only scenaro. In order to maxmze the number of samples for averagng the results of both scenaros were combned. However, f a company contrbuted to HSDPA+R99 and HSDPA scenaro, then just the HSDPA only scenaro was used n the averagng. Because of the very close values for HSDPA+R99 and HSDPA only scenaros as observed when a gven company submtted for both sets - ths and pror smulatons - t s beleved that ths s vald statstc. See appendx Tables 8.A16 through 8.A21 Tables 8.8 and 8.9 show the substantal relatve gan that a type 3 recever may realze over a type 3 at the low geometres of 0 and -3 db. Also, t should be noted that these gans are at very useful throughput values. Agan QPSK shows the better performance, though at geometry 0 db and Ec/Ior = -3 db, the dfference between QPSK and 16QAM s farly close. Table 8.8: HSet-6 QPSK gans and throughputs for types 3 / 3 recevers wth weghted DIP values QPSK HSet6 PB3 VA30 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / 1498 Table 8.9: HSet-6 16QAM gans and throughputs for types 3 / 3 recevers wth weghted DIP values 16QAM HSet6 PB3 VA30 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / 1225

36 35 TR V7.0.0 ( ) Revsed DIP: geometry -3 db As dscussed n Clause t was decded to use a revsed weghted DIP value for geometry -3 db. Tables 8.10 and 8.11 show the average throughput and gan values based on the contrbutons that ncorporated ths modfed DIP set. Agan, these statstcs are based on a combned set of HSDPA and HSDPA+R99 results. These smulatons also nclude the modfed set of OCS codes that were concurrently agreed to when the revsed DIP for geometry -3 db was establshed. The tables show that for geometry -3 db there s a small reducton n the average relatve gans due to the revsed DIP; compare to Tables 8.8 and 8.9.Whereas for a geometry of 0 db there s a lesser degradaton n the gans, especally for the QPSK values whch tend to represent the more useful modulaton scheme for these sgnal levels and geometres. (Only at geometry 0 db and Ec/Ior = - 3 db does 16QAM approach the QPSK average throughput values.) Because the DIP value was not changed for 0 db t s to be expected that there should be lttle change n the pror results as shown n Tables 8.8 and 8.9 compared to those of 8.10 and 8.11 at 0 db. Probably most of the change s due to the sample sze for geometry 0 db. In ths case there were only 3 and 2 sets of ndependent smulaton results for PB3 and VA30, respectvely. Whereas for geometry - 3 db there are 7 and 6 sets for PB3 and VA30. See appendx Tables 8.A22-8.A31 for the complete set of smulaton results, whch nclude smlar contrbutons for H-Set 3. Table 8.10: HSet-6 QPSK gans and throughputs for types 3 / 3 recevers wth revsed DIP values QPSK HSet6 PB3 VA30 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / 1514 Table 8.11: HSet-6 16QAM gans and throughputs for types 3 / 3 recevers wth revsed DIP values Power control 16QAM HSet6 PB3 VA30 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / 1254 As explaned n Clause 7 t was decded to nvestgate two methods to model the effects of power control. These two methods are referred to "normalzed" and "un-normalzed" power control. Based on the contrbutons by partcpatng companes the smulaton results showed that power control had relatvely lttle effect on the average throughput and relatve gans; see appendx Table 8.A32-8.A39, For example, based on appendx Table 8.A32, for H-Set 6 QPSK HSDPA+R99 scenaro wth geometry -3 db and propagaton channel PB3 the average un-normalzed throughput relatve to no power control was 0.93, and for normalzed t was For the same condtons for geometry 0 db the relatve un-normalzed throughput s 0.97, whle the normalzed s ote: ths s based on only two ndependent sets of smulaton results, but there s a strong smlar consstency throughout the submtted data 4. Consequently the relatve throughput gans due to nterference cancellaton are only neglgbly modfed by power control. Ths s due to the structure of the reference recever, the LMMSE. Ths statement of course may not hold n general when other types of recevers may be subjected to power controlled sgnals. Fnally, t s reasonable to conjecture that the un-normalzed power control nduces nstances of larger nterference compared to the normalzed power control. Therefore, there s somewhat greater degradaton due to un-normalzed power control Feld based DIP As outlned n clause 6.4 there were contrbutons addressng the modellng of DIP values based on feld measurements. Subsequently, smulaton results utlzng these feld measurements were contrbuted. Table 8.1 gves the DIP sets for geometres 0 and -3 db based on certan feld measurements. Tables 8.12 and 8.13 gve the smulated average throughput and relatve gans for the stated condtons usng these DIP values. The relatve gan results for the QPSK H- 4 The reason for the small sample sze of averaged data s due to the lack of multple data sets that share the same set of condtons.

37 36 TR V7.0.0 ( ) Set 6 n Table 8.12 are wthn 15% or less of the smlar values for the weghted revsed DIP values shown n Table 8:10. The same comment s true regardng the correspondng 16QAM values shown n Tables 8.13 and The results n ths clause are based on appendx Tables 8.A40 and 8.A41. Table 8.12: QPSK gans and throughputs for types 3 / 3 recevers wth feld based DIP values for propagaton channel PB3 and HSDPA only scenaro QPSK H-Set 6 H-Set 3 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / 1217 Table 8.13: 16QAM gans and throughputs for types 3 / 3 recevers wth feld based DIP values for propagaton channel PB3 and HSDPA only scenaro 16QAM H-Set 6 H-Set 3 Sgnal Level G (db) G (db) Ec/Ior (db) A 939 / / / Types 2 / 2 recevers: weghted & revsed DIPS As ths SI progressed there was a group decson to lmt ths SI to types 3 and 3 recevers [R ]. However, there was one set of smulaton results that reported throughput values and relatve gans for types 2 / 2 recevers usng the revsed DIP for geometry -3 db, and the weghted DIP for geometry 0 db. Tables 8.14 and 8.15 show the results for H-Set 6 under the stated condtons. A comparson of these tables wth Tables 8.10 and 8.11 show smlar relatve gans for condtons that lead to non-neglgble throughput values; that s prmarly correspondng to QPSK and Ec/Ior = - 3dB condtons. See appendx Tables 8.A42-8.A45 for the complete set of data. Table 8.14: HSet-6 QPSK gans and throughputs for types 2 / 2 recevers wth revsed DIP values QPSK HSet6 PB3 VA30 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) / / / / / / / / 757 Table 8.15: HSet-6 16QAM gans and throughputs for types 2 / 2 recevers wth revsed DIP values 16QAM HSet6 PB3 VA30 Sgnal Level G (db) G (db) G (db) G (db) Ec/Ior (db) A 0 / / 13 A 0 / 0 A 1 / / / 223 A 3 / / Appendx The tables contaned n ths appendx represent the totalty of the smulatons results contrbuted by the partcpatng partes. It also shows the values that were used for computng the statstcs n clauses 8.2. The followng tables are contaned heren: Table 8.A1 Table 8.A2 Type 2 / 2, HSDPA + R99, PB3, H-Set6, Medan DIP Type 2 / 2, HSDPA + R99, VA30, H-Set 6, Medan DIP

38 37 TR V7.0.0 ( ) Table 8.A3 Table 8.A4 Table 8.A5 Table 8.A6 Table 8.A7 Table 8.A8 Table 8.A9 Table 8.A10 Table 8.A11 Table 8.A12 Table 8.A13 Table 8.A14 Table 8.A15 Table 8.A16 Table 8.A17 Table 8.A18 Table 8.A19 Table 8.A20 Table 8.A21 Table 8.A22 Table 8.A23 Table 8.A24 Table 8.A25 Table 8.A26 Table 8.A27 Table 8.A28 Table 8.A29 Table 8.A30 Table 8.A31 Table 8.A32 Table 8.A33 Table 8.A34 Table 8.A35 Table 8.A36 Table 8.A37 Table 8.A38 Table 8.A39 Table 8.A40 Table 8.A41 Table 8.A42 Table 8.A43 Table 8.A44 Table 8.A45 Type 2 / 2, HSDPA + R99, PA3, H-Set 6, Medan DIP Type 2 / 2, HSDPA + R99, PB3, H-Set 3, Medan DIP Type 2 / 2, HSDPA + R99, VA30, H-Set 3, Medan DIP Type 2 / 2, HSDPA + R99, PA3, H-Set 3, Medan DIP Type 3 / 3, HSDPA + R99, PB3, H-Set 3, Medan DIP Type 3 / 3, HSDPA + R99, VA30, H-Set 3, Medan DIP Type 3 / 3, HSDPA + R99, PA3, H-Set 3, Medan DIP Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Medan DIP Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Medan DIP Type 3 / 3, HSDPA + R99, PA3, H-Set 6, Medan DIP Type 3 / 3, HSDPA only, PB3, H-Set 6, Medan DIP Type 3 / 3, HSDPA only, VA30, H-Set 6, Medan DIP Type 3 / 3, HSDPA only, PB3, H-Set 3, Medan DIP Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Weghted DIPS Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Weghted DIPS Type 3 / 3, HSDPA, PB3, H-Set 6, Weghted DIPS Type 3 / 3, HSDPA, VA30, H-Set 6, Weghted DIPS Type 3 / 3, Combned HSDPA & HSDPA+R99 PB3, H-Set 6, Weghted DIPS Type 3 / 3, Combned HSDPA & HSDPA+R99, VA30, H-Set 6, Weghted DIPS Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA + R99, PB3, H-Set 3, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA + R99, VA30, H-Set 3, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA only, PB3, H-Set 6, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA only, VA30, H-Set 6, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA only, PB3, H-Set 3, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA only, VA30, H-Set 3, Revsed G= -3 DIP & Codes Type 3 / 3, Combned HSDPA & HSDPA + R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes Type 3 / 3, Combned HSDPA & HSDPA + R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA + R99, PB3, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA + R99, VA30, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA only, PB3, H-Set 6, Revsed G= -3 DIP & Codes wth Unnormalzed Power Control Type 3 / 3, HSDPA only, VA30, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA only, PB3, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA only, VA30, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Type 3 / 3, HSDPA only, PB3, H-Set 6, Feld Derved DIP Values Type 3 / 3, HSDPA only, PB3, H-Set 3, Feld Derved DIP Values Type 2 / 2, HSDPA+R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes Type 2 / 2, HSDPA+R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes Type 2 / 2, HSDPA+R99, PB3, H-Set 3, Revsed G= -3 DIP & Codes Type 2 / 2, HSDPA+R99, VA30, H-Set 3, Revsed G= -3 DIP & Codes

39 38 TR V7.0.0 ( ) Table 8.A1: Type 2 / 2, HSDPA + R99, PB3, H-Set6, Medan DIP Rx Type Gan Gan Gan Gan Reference Motorola A 1.17 R Average A 1.17 Ercsson A 1.16 R Fujtsu R Intel R Motorola R oka R Tensorcomm R Average gan G= 5 db Ercsson R Fujtsu R Intel R Motorola R oka R Tensorcomm R Average gan G = 10 db Ercsson R Fujtsu R Intel R Motorola R oka R Average gan

40 39 TR V7.0.0 ( ) Table 8.A2: Type 2 / 2, HSDPA + R99, VA30, H-Set 6, Medan DIP Rx Type Gan Gan Gan Gan Reference Motorola A 1.33 R Average gan A 1.33 Ercsson R Fujtsu A 0.74 R Intel R Motorola R oka A 1.86 R Tensorcomm A 1.41 R Average gan G= 5 db Ercsson R Fujtsu R Intel R Motorola R oka R Tensorcomm R Average gan G = 10 db Ercsson R Fujtsu R Intel R Motorola R oka R Average gan Table 8.A3: Type 2 / 2, HSDPA + R99, PA3, H-Set 6, Medan DIP Rx Type Gan Gan Gan Gan Reference Fujtsu R G= 5 db Fujtsu R G = 10 db Fujtsu R

41 40 TR V7.0.0 ( ) Table 8.A4: Type 2 / 2, HSDPA + R99, PB3, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference Ercsson R Fujtsu R Average gan G= 5 db Ercsson R Fujtsu R Average gan G = 10 db Ercsson R Fujtsu R Average gan Table 8.A5: Type 2 / 2, HSDPA + R99, VA30, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference Ercsson R Fujtsu R Average gan G= 5 db Ercsson R Fujtsu R Average gan G = 10 db Ercsson R Fujtsu R Average gan

42 41 TR V7.0.0 ( ) Table 8.A6: Type 2 / 2, HSDPA + R99, PA3, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference Fujtsu R G= 5 db Fujtsu R G = 10 db Fujtsu R Table 8.A7: Type 3 / 3, HSDPA + R99, PB3, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference Ercsson R Fujtsu R Average gan G= 5 db Ercsson R Fujtsu R Average gan G = 10 db Ercsson R Fujtsu R Average gan

43 42 TR V7.0.0 ( ) Table 8.A8: Type 3 / 3, HSDPA + R99, VA30, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference Ercsson R Fujtsu R Average gan G= 5 db Ercsson R Fujtsu R Average gan G = 10 db Ercsson R Fujtsu R Average gan Table 8.A9: Type 3 / 3, HSDPA + R99, PA3, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference Fujtsu R G= 5 db Fujtsu R G = 10 db Fujtsu R

44 43 TR V7.0.0 ( ) Table 8.A10: Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Medan DIP Rx Type Gan Gan Gan Gan Reference Motorola R Average gan Fujtsu R Ercsson R Intel R Motorola R oka R Panasonc Average gan G= 5 db Fujtsu R Ercsson R Intel R Motorola R oka R Panasonc Average gan G = 10 db Fujtsu R Ercsson R Intel R Motorola R oka R Panasonc Average gan

45 44 TR V7.0.0 ( ) Table 8.A11: Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Medan DIP Rx Type Gan Gan Gan Gan Reference Motorola A 2.33 R Average gan A 2.33 Fujtsu R Ercsson R Intel R Motorola R oka R Panasonc Average gan G= 5 db Fujtsu R Ercsson R Intel R Motorola R oka R Panasonc Average gan G = 10 db Fujtsu R Ercsson R Intel R Motorola R oka R Panasonc Average gan Table 8.A12: Type 3 / 3, HSDPA + R99, PA3, H-Set 6, Medan DIP Rx Type Gan Gan Gan Gan Reference Fujtsu R G= 5 db Fujtsu R G = 10 db Fujtsu R

46 45 TR V7.0.0 ( ) Table 8.A13: Type 3 / 3, HSDPA only, PB3, H-Set 6, Medan DIP Rx Type: 3:3:3:3:3:3:3:3:Gan:Gan:Gan:Gan:Reference AT&T R Motorola R Average gan AT&T R Motorola R Average gan G= 5 db Motorola R Average gan G = 10 db Motorola R Average gan Table 8.A14: Type 3 / 3, HSDPA only, VA30, H-Set 6, Medan DIP Rx Type Gan Gan Gan Gan Reference Motorola A 2.29 R Average gan A 2.29 G = - 2dB InterDgtal Average gan Motorola R InterDgtal Average gan G= 5 db Motorola R Average gan G = 10 db Motorola R Average gan

47 46 TR V7.0.0 ( ) Table 8.A15: Type 3 / 3, HSDPA only, PB3, H-Set 3, Medan DIP Rx Type Gan Gan Gan Gan Reference AT&T R AT&T R Table 8.A16: Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Weghted DIPS Rx Type Gan Gan Gan Gan Reference Agere A 3.77 R Intel R Average gan Agere R Fujtsu R Intel R oka R Average gan Table 8.A17: Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Weghted DIPS Rx Type Gan Gan Gan Gan Reference Agere A 6.70 R Intel R Average gan Agere R Fujtsu R Intel R oka R Average gan

48 47 TR V7.0.0 ( ) Table 8.A18: Type 3 / 3, HSDPA, PB3, H-Set 6, Weghted DIPS Rx Type Gan Gan Gan Gan Reference Agere A 3.70 R Ercsson A 2.28 R InterDgtal R Motorola R Average gan Agere R Ercsson A 1.34 R InterDgtal R Motorola R oka R Panasonc R Average gan Table 8.A19: Type 3 / 3, HSDPA, VA30, H-Set 6, Weghted DIPS Rx Type Gan Gan Gan Gan Reference Agere A 6.71 R Ercsson A 3.22 R InterDgtal #DIV/0! 4.05 R Motorola R Average gan #DIV/0! 4.38 Agere R Ercsson A 1.34 R InterDgtal R Motorola R oka R Panasonc A 1.37 R Average gan

49 48 TR V7.0.0 ( ) Table 8.A20: Type 3 / 3, Combned HSDPA, & HSDPA+R99 PB3, H-Set 6, Weghted DIPS Rx Type Scenaro Gan Gan Gan Gan Reference Modulaton QPSK QPSK QPSK QPSK 16QAM 16QAM 16QAM 16QAM QPSK QPSK 16QAM 16QAM Agere H A 3.70 R Ercsson H A 2.28 R Intel R R InterDgtal H R Motorola H R Average gan Agere H R Ercsson H A 1.34 R Fujtsu R R Intel R R InterDgtal H R Motorola H R oka H R Panasonc H R Average gan Table 8.A21: Type 3 / 3, Combned HSDPA & HSDPA+R99, VA30, H-Set 6, Weghted DIPS Rx Type Gan Gan Gan Gan Reference Agere H A 6.71 R Ercsson H A 3.22 R Intel R R InterDgtal H A 4.05 R Motorola H R Average gan Agere H R Ercsson H A 1.34 R Fujtsu R R Intel R R InterDgtal H R Motorola H R oka H R Panasonc H A 1.37 R Average gan

50 49 TR V7.0.0 ( ) Table 8.A22: Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference LG Electroncs R Marvell R Motorola R Average gan AT&T R Motorola R Average gan Table 8.A23: Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference LG Electroncs A 3.43 R Marvell R Motorola A 2.82 R Average gan Motorola R Average gan Table 8.A24: Type 3 / 3, HSDPA + R99, PB3, H-Set 3, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference LG Electroncs R Motorola R Average gan Motorola R Average gan

51 50 TR V7.0.0 ( ) Table 8.A25: Type 3 / 3, HSDPA + R99, VA30, H-Set 3, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference LG Electroncs R Motorola R Average gan Motorola R Average gan Table 8.A26: Type 3 / 3, HSDPA only, PB3, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference AT&T R Fujtsu R InterDgtal R LG Electroncs R Motorola R oka R Average gan AT&T R Motorola R oka R Average gan

52 51 TR V7.0.0 ( ) Table 8.A27: Type 3 / 3, HSDPA only, VA30, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Fujtsu R InterDgtal R LG Electroncs A 3.39 R Motorola R oka A 4.11 R Average gan Motorola R oka R Average gan Table 8.A28: Type 3 / 3, HSDPA only, PB3, H-Set 3, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference AT&T R Fujtsu R LG Electroncs R Motorola R Average gan AT&T R Motorola R Average gan

53 52 TR V7.0.0 ( ) Table 8.A29: Type 3 / 3, HSDPA only, VA30, H-Set 3, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Fujtsu R LG Electroncs R Motorola R Average gan Motorola R Average gan Table 8.A30: Type 3 / 3, Combned HSDPA & HSDPA + R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference AT&T H R Fujtsu H R Marvell R R InterDgtal H R LG Electroncs H R Motorola H R oka H R Average gan AT&T H R Motorola H R oka H R Average gan

54 53 TR V7.0.0 ( ) Table 8.A31: Type 3 / 3, Combned HSDPA & HSDPA + R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Fujtsu H R Marvell R R InterDgtal H R LG Electroncs H A 3.39 R Motorola H R oka H A 4.11 R Average gan Motorola H R oka H R Average gan Table 8.A32: Type 3 / 3, HSDPA + R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) R LG Elec. (Un PC) A 2.33 R LG Elec. (m PC) A 2.18 R Motorola (o PC) R Motorola (Un PC) R Motorola (m PC) R oka (Un PC) R oka (m PC) R AT&T (o PC) R AT&T (Un PC) R AT&T (m PC) R Motorola (o PC) R Motorola (Un PC) R Motorola (m PC) R oka (Un PC) R oka (m PC) R

55 54 TR V7.0.0 ( ) Table 8.A33: Type 3 / 3, HSDPA + R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) A 3.43 R LG Elec. (Un PC) A 3.90 R LG Elec. (m PC) A 3.34 R Motorola (o PC) A 2.82 R Motorola (Un PC) A 3.04 R Motorola (m PC) A 2.79 R oka (Un PC) A 4.75 R oka(m PC) A 4.23 R Motorola (o PC) R Motorola (Un PC) R Motorola (m PC) R oka (Un PC) R oka (m PC) R Table 8.A34: Type 3 / 3, HSDPA + R99, PB3, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) R LG Elec. (Un PC) R LG Elec. (m PC) R Motorola (o PC) R Motorola (Un PC) R Motorola (m PC) R Motorola (o PC) R Motorola (Un PC) R Motorola (m PC) R

56 55 TR V7.0.0 ( ) Table 8.A35: Type 3 / 3, HSDPA + R99, VA30, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) R LG Elec. (Un PC) R LG Elec. (m PC) R Motorola (o PC) R Motorola (Un PC) R Motorola (Un PC) R Motorola (o PC) R Motorola (Un PC) R Motorola (m PC) R Table 8.A36: Type 3 / 3, HSDPA only, PB3, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) R LG Elec. (Un PC) A 2.23 R LG Elec. (m PC) A 2.25 R AT&T (o PC) R AT&T (Un PC) R AT&T (m PC) R Table 8.A37: Type 3 / 3, HSDPA only, VA30, H-Set 6, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) A 3.39 R LG Elec. (Un PC) A 3.45 R LG Elec. (m PC) A 3.32 R

57 56 TR V7.0.0 ( ) Table 8.A38: Type 3 / 3, HSDPA only, PB3, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan Reference LG Elec. (o PC) R LG Elec. (Un PC) R LG Elec. (m PC) R Table 8.A39: Type 3 / 3, HSDPA only, VA30, H-Set 3, Revsed G= -3 DIP & Codes wth Power Control Rx Type Gan Gan Gan Gan LG Elec. (o PC) R LG Elec. (Un PC) R LG Elec. (m PC) R Table 8.A40: Type 3 / 3, HSDPA only, PB3, H-Set 6, Feld Derved DIP Values Rx Type Gan Gan Gan Gan Reference AT&T R AT&T R Table 8.A41: Type 3 / 3, HSDPA only, PB3, H-Set 3, Feld Derved DIP Values Rx Type Gan Gan Gan Gan Reference AT&T R AT&T R

58 57 TR V7.0.0 ( ) Table 8.A42: Type 2 / 2, HSDPA+R99, PB3, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Tensorcomm A 1.06 R Tensorcomm R Table 8.A43: Type 2 / 2, HSDPA+R99, VA30, H-Set 6, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Tensorcomm A A R Tensorcomm A 2.98 R Table 8.A44: Type 2 / 2, HSDPA+R99, PB3, H-Set 3, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Tensorcomm R Tensorcomm R Table 8.A45: Type 2 / 2, HSDPA+R99, VA30, H-Set 3, Revsed G= -3 DIP & Codes Rx Type Gan Gan Gan Gan Reference Tensorcomm A 1.64 R Tensorcomm R

59 58 TR V7.0.0 ( ) 9 System performance characterzaton 9.0 General Ths chapter dscusses the benefts of Type 3 recevers from a system performance perspectve. Two dfferent smulaton studes were made wthn the Study Item, both showng sgnfcant beneft when usng nterference cancellaton for users at cell borders [66]. Detals concernng the two studes that were made can be found n clauses 9.1 and 9.2. Conclusons can be found n clause Frst system-level study (Ercsson) Smulaton setup We model a macro-cell envronment, where the ste deployment conssts of a unform hexagonal pattern contanng 19 base staton stes, each servng 3 cells. The ste-to-ste dstance s 3000 m. We use a 2-D sectorzaton antenna model whch has antenna gans as shown n Fg.1. Antenna tltng s not consdered n our smulatons. The transmt power of the base staton s 20 watt per carrer per cell. The path loss model s *log(r) n db, where r s the dstance n km from the moble to the base staton. The shadowng loss s log-normal wth a standard devaton of 8 db. The recever s assumed to operate at 9 db nose fgure. To smplfy our analyss, we assume that all the rado lnks have the same power delay profle. All the mobles n the system are movng at 3 km/h. Two multpath models are consdered, a heavly dspersve model and a mldly dspersve model. The heavly dspersve model conssts of four chp-spaced rays wth exponental power delay profle. The average relatve powers for the four paths are 0, -3, -6 and -9 db, respectvely. Ths power delay profle s dentcal to the power delay profle of the Case3 channel specfed n [67]. Hence, we wll refer to ths channel as smply the Case3 channel. The mldly dspersve model has three chp-spaced paths wth average relatve power of 0, -12.5, and db. Ths channel model resembles the Pedestran A channel model n [67]. Fgure 9.1: 2-D sectorzaton antenna pattern used n our smulatons For lnk adaptaton, we use a MCS table based on lnk smulaton results of an deal recever n AWG. The MCS table s shown n Table 1. The SIR n Table 1 s for every HS-PDSCH symbol (16 chps) per code. The SIR range s determned to acheve less than 10% block error rate for the 1 st transmsson. In system-level smulatons however, we nclude a 2 db mplementaton loss for both Type 3 and Type 3 recevers. We use the same fnger postons for Type 3 and Type 3 recevers.

60 59 TR V7.0.0 ( ) In our smulatons, we further assume that 15 codes and 75% of base staton power are avalable for servng the desred user"s HS-DPDCH. Code and power allocatons however do not mpact the relatve performance between Type 3 and Type 3 recevers. Table 9.1: MCS table used for lnk adaptaton SIR (db) range bts/hs-pdsch symbol/code [-11.5, -10.5] [-10.5, -9.5] [-9.5, -8.5] [-8.5, -7.5] [-7.5, -6.5] [-6.5, -5.5] [-5.5, -4.5] [-4.5, -3.5] [-3.5, -2.5] [-2.5, -1.5] [-1.5, -0.5] [-0.5, 0.5] [0.5, 1.5] [1.5, 2.5] [2.5, 3.5] [3.5, 4.5] [4.5, 5.5] [5.5, 6.5] [6.5, 7.5] [7.5, 8.5] [8.5, 9.5] [9.5, 10.5] [10.5, 11.5] [11.5, 12.5] [12.5, 13.5] [13.5, 14.5] and above Smulaton results We evaluate dstrbutons of achevable data rates over fadng realzatons for users at a certan dstance from the servng base staton. Each of these dstrbutons s equvalent to the dstrbuton of CQI reports collected from users at the same dstance away from the base staton. Smulaton results for the Case 3 (heavly dspersve) channel are shown n Fg. 9.2 and Fg. 9.3, for 10 th percentle and medan data rate, respectvely. The 10 th percentle data rate s acheved by 90% of the users, and t s an mportant ndcator for coverage. From Fg. 9.2, we see that Type 3 recever mproves coverage sgnfcantly. The mprovement s around 25-35% n data rate dependng on the user locaton. It s nterestng to see that Type 3 also mproves the 10 th percentle data rate when users are close to the base staton. In fact, the gans of Type 3 are hgher for users close to the base staton. Ths s because those users close to the base staton experence other-cell nterference manly due to nter-sector nterference, and Type 3 s effectve n suppressng few (most lkely one) nter-sector nterference. From Fg. 9.3, we observe the gans for medan data rates are moderate.

61 60 TR V7.0.0 ( ) Fgure 9.2: 10th percentle data rate for users n a hghly dspersve channel Fgure 9.3: Medan data rate for users n a hghly dspersve channel Smulaton results for the mldly dspersve channel are shown n Fg. 9.4 and Fg. 9.5, for 10 th percentle and medan data rate, respectvely. The mprovement for the 10 th percentle data rate s n the range of 20-55% throughout the cell coverage area. On the other hand, we observe that the gans for medan data rates are moderate.

62 61 TR V7.0.0 ( ) Fgure 9.4: 10th percentle data rate for users n a mldly dspersve channel Fgure 9.5: Medan data rate for users n a mldly dspersve channel 9.2 Second system-level study (oka) Smulaton setup for second study The smulatons were performed n a macro cell scenaro, whch conssts of 7 ode Bs and 21 hexagonal cells (sectors) of radus of 933 meters. Thus the ste-to-ste dstance was 2800 m, whch dffers from the 1000 m, used n [4]. Propagaton model was based on [7] and log-normally dstrbuted slow fadng wth an 8 db standard devaton and a spatal correlaton dstance of 50 meters were assumed. The evaluated channel profles was modfed Vehcular A. The

63 62 TR V7.0.0 ( ) power delay profles were modfed from the orgnal ITU power delay profles so that the tap delays are nteger chps. Average path powers were [-3.1, -5.0, -10.4, -13.4, -13.9, -20.4] db n Vehcular A channel. MAC-hs packet schedulng based on Proportonal Far schedulng algorthms was used wthout code-multplexng,.e. only one UE s scheduled per TTI. The maxmum numbers of HS-DSCH codes was 10 wth spreadng factor 16. HS- DSCH power allocaton was 14 W, whch s 70% of the total base staton transmsson power. One code was allocated for HS-SCCH wth spreadng factor of 128. HS-SCCH was power controlled so that the power follows the average power over the last TTI of the assocated DCH wth an offset. Realstc recepton of HS-SCCH was consdered. Sx parallel stop-and-wat (SAW) channels were used for the Hybrd ARQ. At the maxmum 4 retransmssons were allowed per transport block. Chase Combnng was used for the retransmssons [68]. HS-DSCH lnk adaptaton was based on the UE reported channel qualty ndcators (CQI's) (nner loop) and UE reported Ack/acks from past retransmssons (outer loop). Amed resdual block error rate (BLER) after the second transmsson was 1% and lnk adaptaton outer loop was used to control the BLER target. The MCS tables used n ode B were throughput optmzed. CQI reportng granularty of 1dB was accounted. CQI reportng error, whch was modeled as log-normally dstrbuted wth standard devaton of 1 db, was ncluded n the smulatons. The CQI"s reported by UE"s were always based on normal (or nonnterference aware) LMMSE chp level equalzer. The lnk adaptaton outer loop was set to account the dfference between normal LMMSE and nterference aware LMMSE equalzer n SIR calculaton. Moblty and traffc models were based on UMTS [69]. UE velocty was 3km/h. Modfed web browsng traffc model, n whch the users do not have a readng tme durng a download sesson.e. they only have one packet call per sesson, was used. The total smulaton tme was 6 mnutes. The call arrval rate n the network was 140 calls per second and the average packet call sze was 112 klobytes. Thus, the total average offered load per cell can be calculated as A * B / C, where A s the call arrval rate, B s the average packet call sze and C s the number of cells n the network. In these smulatons the average offered load per cell was approxmately 6 Mbps. ew calls were generated accordng to homogeneous Posson process. The offered traffc was hgh enough to have almost 100 % utlzaton of the HS-DSCH. Admsson control allowed up to 16 HSDPA users per cell. The LMMSE equalzer and nterference aware LMMSE equalzer were used for HS-DSCH wth and wthout Rx dversty. For determnng the SIR used wth the nterference aware LMMSE equalzer under study (.e. ether Type 2 or Type 3) the nterference seen from strongest nterferng cells was explctly accounted by modellng the actual channel matrces of the cells [2]. The calculaton of nose covarance matrx n SIR calculaton was thus done n the assumpton that the channel matrces of the strongest nterferng cells are deally known at the recever. Three strongest nterferng other cells were accounted n the calculaton as t was notced that consderng fourth strongest nterferer or lower dd not affect the results sgnfcantly. The man smulaton parameters are also lsted n Table 9.2 below.

64 63 TR V7.0.0 ( ) Table 9.2: System Smulaton Parameters Parameter Explanaton/Assumpton Comments Cellular layout Hexagonal cell grd, wrap-around 7 ode-bs and 21 sectors Cell radus 933 m Corresponds to a ode-b to ode-b dstance of 2800 m. Propagaton Model L= Log 10(R km) Rado propagaton condton Vehcular A wth 3 km/h Std. devaton of slow fadng 8 db Correlaton between sectors 1.0 The correlaton n the slow fadng between the sectors. The UE experences the same knd of slow fadng n the area of the correlatng sectors,.e. the fadng s not entrely random. Correlaton between ode-bs 0.5 The correlaton n the slow fadng between the ode-bs. Correlaton dstance of slow fadng 50 m Ths parameter defnes the maxmum dstance wthn whch the UE experences correlated slow fadng to a sector. Mnmum path loss 70 db BS antenna gan 18 db Antenna front to back rato -20 db ode-b total Tx power 43 dbm Corresponds to 20 W. Power resource for HS-DSCH 14 W HSDPA packet schedulng algorthm Proportonal far Used Redundancy Verson Chase Combnng Maxmum number of retransmssons 4 Maxmum number of retransmsson before the correspondng HARQ channel s cleared Traffc model Web browsng wthout readng tme Average packet call sze was 112 kbytes HSDPA RLC PDU sze 320 bts Code resource for HS-DSCH 10 SF=16 UE HS-DSCH recever LMMSE equalzer or nterference Type2/3 and Type2/3 aware LMMSE equalzer wth and wthout recever dversty. umber Of HARQ channels n UE Smulaton results for second study In Fgure 9.6 the CDF of cell throughput obtaned wth dfferent recevers s presented. In Fgure 9.7 the scheduled user E s / 0 s depcted.

65 64 TR V7.0.0 ( ) Fgure 9.6: Cell throughput Fgure 9.7: HS-DSCH E s / 0 dstrbuton of scheduled user In order to more accurately evaluate the recever gans and the effect of dfferent network stuatons to them, more specfc throughput statstcs were gathered. As nterference aware LMMSE equalzer s assumed to provde gan specfcally when a strong nterferer s present, ths effect was attempted to be captured by collectng statstcs from UEs wth cells of dfferent strength n ther vcnty. As the exstence of a cell n UEs actve set s a good measure of the strength of the cell, the throughput statstcs were gathered from UEs n dfferent DCH soft handover states. Statstcs for two dfferent handover states were consdered. Frst, the statstcs were collected separately for users n DCH soft handover e.g. UEs that have more than one cell n actve set and all the cells do not belong to the same ode B. Second state conssted of users that were n softer handover e.g. UEs that have exactly two cells (sectors) n ther actve set and both are from the same ode B. Fgure 9.8 presents the spatal dstrbuton of users n DCH soft handover and n Fgure 9.9 the dstrbuton of users n softer handover s depcted. The terms "soft handover" and "softer handover" refer to DCH handover states and they are only used to refer to the area of nterest n the cell.

66 65 TR V7.0.0 ( ) Fgure 9.8: Spatal dstrbuton of users n DCH soft handover n respect to the servng ode B Fgure 9.9: Spatal dstrbuton of users n DCH softer handover n respect to the servng ode B In Table 9.3 the average call throughputs of users n dfferent DCH soft handover states s presented. It can be observed that the beneft of Type 3 recevers s largest at the border regons, the largest gans observed for the soft and softer handover rangng from 22% to 21%. Thus the Type 3 nterference aware recevers seems to provde some benefts for the cell edge users, roughly ncreasng the obtaned user throughput by 50kbps. For Type 2 recever some gan can be seen also for the cell border regons, but for all users a slght loss s seen. As the performance of the cell border users s mproved, leadng to ncreased schedulng probablty, resultng slght decrease n overall user throughput. The DCH soft handover state of the user used n statstcs collectng s determned at the end of the call to be the one n whch UE has been longest tme durng a whole call, thus there may be some varance n the observed call throughputs. In Table 9.4 the average nstantaneous HS-DSCH TTI throughputs of users n the aforementoned states are presented for dfferent recevers evaluated. It can be seen that smlarly as n case of the call throughputs, the gan of Type 2 and Type 3 recevers s the hghest n the border regons between two cells of a three sector ode B. As the overall gan, consderng all users, s 3 % wth Type 2 and 6 % wth Type 3, the correspondng gans n the border regons between two sectors s 4 % and 19 %. The small effect of the hgher gans to the total average gans s due to low percentage of the users n the gven regons. Only 3-4 % of the scheduled users are located between sector borders, as can be seen n Table 9.5. It should be noted