channel estimation and peak-to-average power ratio...

Research ArticleChannel Estimation and Peak-to-Average Power Ratio Analysisof Narrowband Internet of Things Uplink Systems

Md Sadek Ali Yu Li Md Khalid Hossain JewelOluwole John Famoriji and Fujiang Lin

Micro-Nano-Electronic System Integration RampD Center (MESIC) Department of Electronic Science and TechnologyUniversity of Science and Technology of China (USTC) Hefei Anhui 230026 China

Correspondence should be addressed to Md Sadek Ali sadekmailustceducn

Received 27 November 2017 Accepted 21 May 2018 Published 5 July 2018

Academic Editor Pavlos I Lazaridis

Copyright copy 2018 Md Sadek Ali et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Narrowband Internet of Things (NB-IoT) is a cellular based promising low-power wide-area network (LPWN) technologystandardized by the 3rd Generation Partnership Project (3GPP) in release-13 as a part of the future 5th Generation (5G) wirelesscommunication systems The main design target of NB-IoT was to enhance radio coverage by repeating signal over an additionalperiod of time for the ultralow-end IoT devices that would be operated in extreme coverage environments But the power efficiencyof the low-cost NB-IoT user equipment (NB-IoT UE) in the uplink is the major concern Coverage improvement from signalrepetitions depends on the channel estimation quality at extremely bad radio conditions The typical operating signal-to-noiseratio (SNR) for NB-IoT is expected to be much lower than the zero In this paper we have proposed two efficient narrowbanddemodulation reference signal (NDMRS)-assisted channel estimation algorithms based on the conventional least squares (LS)and minimum mean square error (MMSE) estimation methods The theoretical analysis and the link-level performance of ourproposed estimation methods are presented Simulation results exhibit that the proposed methods provide better estimationprecision compared to the traditional LS andMMSEmethods at the low SNR situations Furthermore we have analyzed the raised-cosine (RC) and square-root-raised cosine (RRC) pulse shaping to reduce peak-to-average power ratio (PAPR) as an uplink transmitfilter The PAPR values are evaluated through extensive computer simulations for both single-tone and multi-tone transmissionsOur evaluation results vindicate that the RRC pulse shaping with lower PAPR values is feasible to design of practical NB-IoT uplinktransmitter and increases power efficiency

1 Introduction

The Internet ofThings (IoT) is a novel prototype which offersmassive connectivity to physical objects radio-frequencyidentification (RFID) tags vehicles sensors actuators andother things embedded with electronics to the InternetIoT allows things to be connected across existing networkinfrastructure interacting with each other through uniqueaddressing schemes thus reducing extra deployment cost andimproving accuracy and efficiency According to [1] therewill be more than 30 billion devices connected wirelesslyto the IoT by 2020 Nokia [2] based on Machina research2015 predicted that about 30 billion connected IoT deviceswill be deployed by 2025 of which cellular IoT (CIoT)and low-power wide-area (LPWA) modules are about 23

percent Traditionally mobile broadband networks need highthroughput and low latency whereas LPWA applicationsrequire low-throughput extended coverage low-cost lowcomplexity scalability low delay sensitivity and high powerefficiency [3ndash6]

There are many short-range wireless communicationtechnologies [7 8] like Bluetooth low energy (BLE) Wi-FiLi-Fi ZigBee and Z-wave to enable the IoT Some of theIoT enabling technologies [9 10] such as SigFox and LoRathat are operating in license-exempted band (ie industrialscientific and medical (ISM) band) On the other handGlobal System for Mobile Communications (GSM) and the3rd Generation Partnership Project (3GPP) standard Long-Term Evolution (LTE) are operating in licensed spectrum toenable the IoT A new cellular based IoT enabling technology

HindawiWireless Communications and Mobile ComputingVolume 2018 Article ID 2570165 15 pageshttpsdoiorg10115520182570165

2 Wireless Communications and Mobile Computing

named narrowband IoT (NB-IoT) was specifically designedfor ultralow-end IoT applications The 3GPP finalized thespecifications ofNB-IoT in LTE release-13 [11] It is secure andreliable for data transmission because of the deployment inlicensed spectrum of GSM or LTE [12] NB-IoT enables IoTwhose applications are very diverse including smart meter-ing smart cities smart water smart environment smartagriculture smart animal firming retail logistics securityand emergencies industrial control and domestic and homeautomation Therefore the channel condition of NB-IoT isvery complicated due to its diverse application scenarios

Since NB-IoT is still in its infancy the proper guide-lines for efficient channel estimation and equalization arestill missing in the current literature Channel estimationbased on the pilot signal has been studied well for tradi-tional orthogonal frequency division multiplexing (OFDM)and single-carrier frequency division multiple access (SC-FDMA) systems eg [13 14] In the literature most ofthe NB-IoT research focused on frame structure design[15] scheduling and link adaptation [16] random accessprocedure [17] and system acquisition [18] Positioningperformance of NB-IoT has been studied in [19 20] YD Beyene et al in [21] have investigated the performanceof three traditional channel estimators only for 15 kHzsubcarrier spacing L Zhang et al in [22] have analyzed thechannel equalization and coexistence problem ofNB-IoT andLTE signals using only traditional LTE supported subcarrierspacing To the best of our knowledge the channel estimationof NB-IoT uplink with 375 kHz subcarrier spacing hasnot been studied yet Thus efficient channel estimation isthe prerequisite of coverage improvement equalization andsignal decoding at the receiver

In NB-IoT uplink low peak-to-average power ratio(PAPR) enhances the efficiency of the low-cost power ampli-fier Low out-of-band radiation is desired for the uplinktransmitter due to its very narrow bandwidth The PAPRproblem is more of a concern in the NB-IoT uplink (ielow-cost transmitter)The efficiency of the power amplifier iscrucial for the low-costNB-IoT user equipment (NB-IoTUE)with limited battery power Thus low PAPR in the NB-IoTuplink is ultimate desire owing to low-cost power amplifierPAPR reduction techniques such as scrambling discreteFourier transform (DFT) spreading and cyclic prefix (CP)insertion can be applied at the uplink transmitter NB-IoTsupports modulation schemes like 1205872-BPSK (binary phaseshift keying) and 1205874-QPSK (quadrature phase shift keying)which are also robust against PAPR by applying constellationrotation to make smooth transition between constellationpoints However these techniques are not fully preservedFurther reduction of PAPR would be great demand at theuplink because transmitter is the low-cost and low-power IoTdevice In [23] the authors have been evaluated the PAPRvalues employing root-raised cosine (RRC) pulse shaping(PS) filter only for the single-tone transmission This is thefirst time that the partial analysis of PAPR was taken intoaccount in the NB-IoT uplink transmitter

In this paper we have developed an uplink NB-IoTsystemmodel according to the 3GPP specifications in release-13 [24ndash26] The motivation was that the uplink transmission

of NB-IoT systems is more complicated compared to thedownlink transmission We have considered both types oftransmission schemes and subcarrier spacing for channelestimation and PAPR analysis The major contributions ofthis paper can be summarized as follows

(1) We have provided a brief overview of NB-IoT tech-nology including deployment options physical chan-nels and signals uplink frame structure and resourceunit (RU) definition An analytic NB-IoT uplinkreceived signalmodel is derived as a function of trans-mitted signal and channel impairments Narrowbanddemodulation reference signal (NDMRS) generationandmapping to time-frequency grid is also presented

(2) We have proposed two NDMRS-aided channel esti-mation algorithms based on the traditional leastsquares (LS) and minimum mean square error(MMSE) estimators that can be coped with the com-plicated channel conditions of NB-IoT systemsThrough simulations we have investigated and ver-ified the effectiveness of our proposed algorithmscompared with the conventional LS and MMSE algo-rithms in terms of bit error rate (BER) related tosignal-to-noise ratio (SNR) Simulation results showthat our proposed channel estimation algorithmsoutperform the others

(3) In addition we have provided theoretical analysis ofPAPR for the NB-IoT uplink employing raised-cosine(RC) and square-root-raised cosine (RRC) pulseshaping filters We have also showed the compari-son of PAPRvalues that are obtainedwith andwithoutPS through computer simulations for both single-tone andmulti-tone transmissions Numerical resultselucidate that RRC PAPR reduction technique is fea-sible for the implementation of NB-IoT uplink trans-mitter

The rest of the paper is organized as follows in Section 2 abrief overview of theNB-IoT technology is provided NB-IoTuplink signal model and NDMRS generation and mappingare presented in Section 3 Theoretical analysis of channelestimation and its performance analysis by simulations aregiven in Section 4 In Section 5 PAPR reduction techniquesand its numerical results are presented Finally we concludethe paper in Section 6

Notations Bold face lowercase letters are used to rep-resent time-domain vectors (or matrices) while frequency-domain vectorsmatrices are denoted by uppercase boldcharacters Superscripts (sdot)119879 and (sdot)119867 denote the transposeand Hermitian of a vector a scalar or a matrix respectivelyand (sdot)minus1 denotes matrix inversion The circular convolutionoperation is denoted by otimes and the operators Ε[sdot] | sdot | and sdot represent the expectation absolute value and Euclideannorm respectively 119868119871 denotes the 119871 times 119871 identity matrix

2 Overview of NB-IoT Technology

NB-IoT was designed by 3GPP as a key technology to meetthe demands of massive low-power IoT connectivity for the

Wireless Communications and Mobile Computing 3

LTE guard-band

LTEguard-band

LTE Carrier

NB-IoT (180 kHz) (In-band)

(a)

LTEguard-band

LTEguard-band

LTE Carrier

NB-IoT (180 kHz) (Guard-band)

(b)

NB-IoT (200 kHz)(Stand-alone)

GSM Carriers

(c)

Figure 1 NB-IoT modes of operation (a) in-band (b) guard-band and (c) standalone

evolution of future 5G wireless communication systems Itcan be implemented in three different operationmodes spec-ified by 3GPP in release-13 [27 28] standalone in-band andguard-bandThe illustrations of three deployment options aregiven in Figure 1 NB-IoT can be deployed by replacing one ormore efficiently reframed 200 kHz GSM carriers a so-calledstandalone mode of operation Radio coverage for NB-IoTcan be enhanced significantly by using all the transmit powerat the evolved node B (eNB) also known as base station Inin-band operation it can be implemented inside the LTEcarrier using one or more physical resource blocks (PRBs)a PRB corresponds to 180 kHz bandwidth LTE and NB-IoTshare the total transmit power at the eNBWide-area coveragecan also be achieved by boosting power on the NB-IoT PRBThe spectrum efficiency can also be increased by the sharingof PRB between LTE and NB-IoT The third option can bedeployed within the LTE carrierrsquos unused guard-band Thisis allowed only for 5 MHz or higher LTE system bandwidthIn-band and guard-band deployments of NB-IoT reuse theexisting LTE base stationrsquos radio-frequency (RF) front-endand the baseband numerologies with some modificationsto fit into the narrow bandwidth [29] The coexistence ofLTE and NB-IoT has been investigated through rigoroussimulations in [27] Thus it will not incur extra deploymentcost and time to come in operation The modes of operationshould be known to the NB-IoT UE when it is turned onand searches for anNB-IoT carrier NB-IoT supports 100 kHzchannel raster for all types of operation modes

NB-IoT channels and signals are designed based on theexisting LTE channels and signals with required modifi-cations and simplifications to fit the 3GPP specified 180kHz narrow bandwidth The NB-IoT downlink and uplinkchannels and signals with their functions according torelease-13 are given in Table 1 The 3GPP is specified asnarrow as system bandwidth of 180 kHz for both downlinkand uplink transmissions NB-IoT supports only frequencydivision duplexing (FDD) with half duplex transmission Inthe downlink NB-IoT inherits the downlink numerologyfrom existing LTE although with further restricted supportIt uses OFDM with 15 kHz subcarrier spacing as in LTE The

basic time unit for NB-IoT is specified by a factor of 119879s =1(15000 times 2048) seconds The slot duration 119879slot = 15360 times119879s = 05ms A pair of consecutive slots constitute a subframewith duration 1 ms The NB-IoT radio frame for downlinkconsists of 10 subframes with duration 119879f = 307200 times119879s = 10ms Thus a radio frame contains 20 slots the slot numberwithin a radio frame is denoted here as 119899s where 119899s isin0 1 sdot sdot sdot 19 In NB-IoT uplink two schemes are supportedfor the basebandmodulation single-tone transmission basedon frequency division multiple access (FDMA) and multi-tone (ie 3 6 or 12 subcarriers) transmission accordingto SC-FDMA Two different subcarrier spacing types areallowed for single-tone transmission 15 kHz and 375 kHz[25] NB-IoT with 375 kHz subcarrier spacing is designedto provide more capacity in power limited scenarios [27]The first one is the same as in LTE with 05 ms slot 1ms subframe and 10 ms frame in time-domain whereasthere are 12 subcarriers within 180 kHz system bandwidthin frequency-domain A time-frequency grid structure for15 kHz subcarrier spacing within a frame is illustrated inFigure 2 On the other hand the second one is different fromLTE with slot duration 119879slot = 61440 times 119879s = 2 ms Thus fiveconsecutive slots constitute a radio framewith the duration of61440times119879stimes5 = 10mswhere the slot number 119899swithin a radioframe can be selected from the set 119899s isin 0 1 sdot sdot sdot 4 In thefrequency-domain 180 kHz system bandwidth consists of 48subcarriers Figure 3 shows the NB-IoT uplink resource gridstructure within a frame for 375 kHz subcarrier spacing Forsingle-tone transmission both subcarrier spacing types can beused whereas only 15 kHz subcarrier spacing is specified formulti-tone transmission

The 3GPP in [24] defined a new feature forNB-IoT uplinkcalled resource unit (RU) which is the basic unit for narrow-band physical uplink shared channel (NPUSCH) allocationIn the time-domain one transport block size (TBS) can bemapped to multiple RUs from the set 1 2 3 4 5 6 8 10In 3GPP [25] TBSs are defined as a function of the num-ber of RUs and modulation and coding scheme (MCS)level The maximum TBS is 1000 bits for the uplink Thecharacterization of a RU is given in Table 2 In NB-IoT


Table 1 Physical channels and signals for NB-IoT systems

ChannelsSignals Functions

Downlink

Narrowband Physical Downlink Control Channel(NPDCCH)

Scheduling information for both downlink and uplinkdata channels

Narrowband Physical Downlink Shared Channel(NPDSCH) Downlink dedicated and common data

Narrowband Physical Broadcast Channel (NPBCH) Master information for system access

Narrowband Synchronization Signal (NPSSNSSS) Cell search including time and frequencysynchronization and cell identity detection

Uplink

Narrowband Physical Uplink Shared Channel(NPUSCH) Uplink dedicated data and control information

Narrowband Physical Random AccessChannel(NPRACH) Random access procedure

Narrowband Demodulation Reference Signal(NDMRS) Uplink channel estimation

Table 2 Characterization of resource unit (RU)

Subcarrier spacing No of subcarriers No of slots No of SC-FDMA symbols Tx time interval (TTI)375 kHz 1 16 112 32 ms

15 kHz

1 16 112 8 ms3 8 56 4 ms6 4 28 2 ms12 2 14 1 ms

systems repetition of user data and associated control signaltransmission has been taken in the 3GPP as a key techniqueto achieve wide-area coverage Repetitions of same signalcan be increased transmission reliability but reduced spectralefficiencyMaximum 2048 and 128 repetitions are allowed fordownlink and uplink transmissions respectively suggestingthat the received data would be decoded even when thenoise power is far greater than the signal power In otherwords the eNB and NB-IoT UE transmits the same TBSrepeatedly as many times as indicated in the downlink anduplink respectively The NB-IoT UE and eNB at the receivercombines the repetitions before decoding the transmitteddataThenumber of repetitions is determined by the eNB andNB-IoTUE to achieve the desired SNR at the NB-IoTUE andeNB receiver respectively

3 NB-IoT Uplink System Model

The system model of uplink NB-IoT systems with NDMRSsequence channels and the associated estimation and equal-ization blocks is shown in Figure 4 The uplink transmittercomprises transport channel also known as uplink sharedchannel (UL-SCH) and data channel (NPUSCH) processingBinary input data arrives to the channel coding unit in theform of one transport block over a number of RUs per uplinkcell The number of RUs is scheduled according to [25] In3GPP [30] UL-SCH processing consists of transport blockcyclic redundancy check (CRC) attachment (eg 24 bitswith generator polynomial gCRC24A(D)) 13-rate based turbocoding and rate matching to yield a codeword input to theNPUSCH

31 NDMRS Sequence Generation and Mapping A NDMRSsequence 119903119906(119899) can be generated for the case when thenumber of subcarriers119873RU

sc = 1 in a RU is as

119903119906 (119899) = 1radic2 (1 + 119895) (1 minus 2119888 (119899)) 119908 (119899mod 16) 0 le 119899 lt 119872119873RU

slots119873RU

(1)

where 119888(119899) is the binary sequence defined by a length-31 Goldsequence 119872 denotes the repetition number of same signaltransmissions119873RU

slots represents number of slots in a RU and119873RU is the number of RUs The initialization value of thefirst sequence is specified with a unit impulse function oflength-31 The second sequence is initialized with the seed119888init = 35 at the start of the NPUSCH transmission [24]The variable119908(119899) is defined in [24] where the base sequenceindex is 119906 = 119873Ncell

ID mod16 for NPUSCH format-1 withoutenabling group hopping Thus the NDMRS sequence 119903119906(119899)for NPUSCH format-1 can then be represented as

119903119906 (119899) = 119903119906 (119899) (2)

The NDMRS sequence 119903119906(119899) for the number of subcarriersgreater than one in a RU is defined by a cyclic shift 120572 of a basesequence as

119903119906 (119899) = 119890119895120572119899119890119895120593(119899)1205874 0 le 119899 lt 119873RUsc (3)

where 120593(119899) is defined in [24] for the scheduled number ofsubcarriers in a RU Without loss of generality we assume


1 Frame = 10 ms

1 slot = 05 ms

012

12 S

ubca

rrie

rs =

180

kH

z

3456789

1011

ns = 0 ns = 1 ns = 19

Figure 2 NB-IoT uplink resource grid structure for 15 kHz sub-channel bandwidth

48 S

ubca

rrie

rs =

180

kH

z

1 Frame = 10 ms

1 slot = 2 ms

0123456

424344454647

ns = 0 ns = 1 ns = 4


that there is no higher layer signaling then the base sequenceindex 119906 can be obtained as

119906 =

119873NcellID mod12 for 119873RU

sc = 3119873Ncell

ID mod14 for 119873RUsc = 6


sc = 12(4)

The cyclic shift120572 for119873RUsc = 3 and 6 is defined in [24] whereas120572 = 0 for119873RU

sc = 12The NDMRS sequence is also known as pilot symbol

which is transmitted together with the user data symbolsto estimate channel response in uplink NB-IoT systems IneachNB-IoT uplink slot NDMRS symbols aremapped to theallocated number of subcarriers in a RU of the fourth SC-FDMA symbol for 15 kHz subcarrier spacing whereas the

fifth symbol is for 375 kHz subcarrier spacing An inversediscrete Fourier transform (IDFT) operation is performedon the contents of resource grid that contains the NDMRSsymbols to convert time-domain reference sequence followedby CP addition

32 Analytical Uplink Signal Model Let us consider that anNB-IoT UE transmits a block of bits 119887 = [119887(0) 119887(1) 119887(119873bit minus 1)] where119873bit is the number of transmitted bits in acodeword on the NPUSCH in one subframe The codewordbits 119887 are scrambled using NB-IoT UE specific scramblingsequence in neighboring cells to ensure that the interferenceis randomized and the transmission from different cells isseparated prior to decoding at the eNB receiver Thus weobtain a block of scrambled bits (119894) as

(119894) = (119887 (119894) + 119888 (119894))mod 2 (5)


DFT

Para

llel-t

o-Seri

al (P

S)

Dem

odul

atio

n

Des

cram

blingEst Data

Sequence

Rem

ove C

P

Phys

ical

Re

sour

ce

Dem

appi

ng an

d Eq

ualiz

atio

n

Channel Estimation

IDFT

CP Cyclic PrefixPS Pulse Shaping

RF Fro

nt-E

nd

(Rx)

Seri

al-to

-Pa

ralle

l (S

P)

Scra

mbl

ing

Mod

ulat

ion

DFT

Phys

ical

Re

sour

ce

Map

ping

IDFT

Add

CPPS

Para

llel-t

o-Seri

al (P

S)

RF Fro

nt-E

nd

(Tx)

Seri

al-to

-Pa

ralle

l (S

P)Data Sequence

Multipath Fading ChannelNDMRS

Sequence

Physical Resource Mapping

IDFT Add CP

Figure 4 Block diagram of uplink NB-IoT systems

where 119894 = 0 1 119873bitminus1 and 119888(119894) is the scrambling sequencedefined by a length-31 Gold sequence [24] The initializationvalue of the first sequence is specified with a unit impulsefunction of length-31 The second scrambling sequence willbe initialized with the seed according to

119888init = 119899RNTI sdot 214 + 119899fmod 2 sdot 213 + lfloor119899s2 rfloor sdot 29 + 119873NcellID (6)

where 119899RNTI denotes the index of the radio network tempo-rary identifier (RNTI) 119899s is the first slot of the transmissionof the codeword and the narrowband cell identity numbercan be selected from the set 119873Ncell

ID isin 0 1 sdot sdot sdot 503 Thescrambling sequence will be reinitialized for the repetitionsof NPUSCH according to (6) after every119873NPUSCH

identical transmis-sion with 119899s and 119899f set to the first slot and the framerespectively In constellation mapping of NPUSCH trans-mission the block of bits (119894) is modulated by employinglow PAPR modulation schemes (eg 1205872-BPSK and 1205874-QPSK) which are specified for NB-IoT systems to improvethe power efficiency at the transmitter (ie NB-IoT UE)Thuswe have a block of complex-valuedmodulation symbols119904 = [119904(0) 119904(1) 119904(119873symb minus 1)]119879 where 119873symb denotes thenumber of modulated symbols

The block of modulation symbols 119904 is divided into119873symb119872NPUSCHsc sets each corresponding to one SC-FDMA symbol

The parameter 119872NPUSCHsc = 119873NPUSCH

RB sdot 119873RBsc indicates the

number of subcarriers allocated for NPUSCH transmission

where119873NPUSCHRB (eg119873NPUSCH

RB = 1 for NB-IoT) correspond-ing to the bandwidth of NPUSCH in terms of PRB and 119873RB

scis the number of subcarriers in a PRBThe frequency-domainsymbols after performing DFT operation can be representedas

119878 (119897 sdot 119872NPUSCHsc + 119896) = 1

radic119872NPUSCHsc

sdot 119872NPUSCHsc minus1sum119894=0

119904 (119897 sdot 119872NPUSCHsc + 119894) 119890minusj2120587119894119896119872NPUSCH

sc 0 le 119896 le 119872NPUSCH

sc minus 1 0 le 119897 le 119873symb119872NPUSCHsc minus 1

(7)

The physical resource element mapping is accomplished byplacing frequency-domain user data symbols and knownNDMRS symbols within the uplink time-frequency gridNPUSCH can be mapped to one or more than one RUaccording to [25] each of which can be transmitted119872 timesThe block of frequency-domain symbols is mapped in asequential manner (ie localized mapping) to subcarriersassigned for transmission [24 26 31 32] The mapping toresource elements (119896 119897) corresponding to subcarriers allo-cated for transmission within a RU will be in increasingorder of the first subcarrier index 119896 then the symbol index119897 and finally the slot number After mapping to 119873slots slots119873slots repeats 119873NPUSCH

identical additional times before continuing


NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs

15 kHz Subcarrier spacing

(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)

Figure 5 Resource grid mapping for (a) multi-tone (eg 12 tone) with 15 kHz subcarrier spacing and (b) single-tone with both 15 kHz and375 kHz subcarrier spacing

themapping of 119878(sdot) to the following slot where the quantities119873NPUSCHidentical and119873slots can be defined as follows

119873NPUSCHidentical =

min(lceil1198722 rceil 4) for 119873RUsc gt 1

1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)

where Δ119891 denotes the subcarrier spacing The mappingof 119878(sdot) is then repeated until 119872119873RU119873RU

slots slots have beentransmitted Figure 5 shows the mapping pattern of userdata (NPUSCH) symbols and NDMRS symbols within aresource grid for NPUSCH format-1 for example a RUcontains 12 subcarriers for multi-tone transmission and onlyone subcarrier for single-tone transmission

The physical resource element mapping is followed byan inverse DFT (IDFT) operation to convert the data intotime-domain signal For single-tone transmission the time-domain baseband signal 119909119896119897(119905) after the CP insertion withlength 119873CP119897 and PS operation for the 119896-th subcarrier in SC-FDMA symbol 119897 in an uplink slot can be expressed as

119909119896119897 (119905) = 119860119896(minus) 119897 sdot 119890119895120593119896119897 sdot 1198901198952120587(119896+12)Δ119891(119905minus119873CP119897119879s)119896(minus) = 119896 + lfloor119873RU

sc2 rfloor (10)

for 0 le 119905 lt (119873CP119897+119873)119879s where parameters forΔ119891 = 15 kHzand Δ119891 = 375 kHz are specified in Table 3 119860119896(minus) 119897 is the

frequency-domain modulation value of symbol 119897 and thephase rotation 120593119896119897 is defined as [24]

120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =

1205872 for BPSK1205874 for QPSK

120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU

symb minus 1 119897 = mod119873RUsymb

(11)

where is the symbol counter that is reset at the start of atransmission and incremented for each symbol during thetime of transmission

The time-domain signal 119909119897(119905) in SC-FDMA symbol 119897 in anuplink slot for multi-tone transmission can be modelled as

119909119897 (119905) = lceil119873RUsc 2rceilminus1sum119896=minuslfloor119873RU

sc 2rfloor

119860119896(minus) 119897 sdot 1198901198952120587(119896+12)Δ119891(119905minus119873CP119897119879s) (12)

for 0 le 119905 lt (119873CP119897 + 119873) times 119879s where 119896(minus) = 119896 + lfloor119873RUsc 2rfloor119873 = 2048 Δ119891 = 15 kHz and 119860119896(minus) 119897 is the content of

resource element (119896 119897) Note that only normal CP length119873CP119897 of existing LTE is supported in release-13 of the NB-IoTspecification

The time-domain baseband signal is upconverted bya RF front-end and then transmits through a multipathfading channel whose delay speared is assumed to be smallerthan the CP length The received signal is composed of


Table 3 SC-FDMA parameters for119873RUsc = 1

Parameter Subcarrier spacing375 kHz 15 kHz119873 8192 2048

Cyclic prefix length119873CP119897 256 160 for 119897 = 0144 for 119897 = 1 2 6Set of values for 119896 -24-23 23 -6-5 5the signals from different channel paths and additive noisethen resultant signal for both single-tone and multi-tonetransmissions can be represented as the circular convolutionof transmitted signal and channel impulse response (CIR)Thus we have

119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)

where 119899(119905) is the additive white Gaussian noise (AWGN)withzero mean and variance 1205901198992 119910single(119905) and 119910multi(119905) are thereceived signal for single-tone and multi-tone transmissionsrespectively and ℎ(119905) denotes the CIR of themultipath fadingchannel with 119871 distinct complex-taps which can be expressedas

ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)

where 120573119894 and 120591119894 represent the attenuation and the delay ofthe 119894-th path respectively Therefore the noisy and delayedversion of the signals at the receiver can be written as

119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)

After removing CP the receiver performs inverse opera-tions of the NPUSCH and UL-SCH processing In additionNDMRS-assisted frequency-domain channel estimation andequalization are performed

4 Channel Estimation in NB-IoT Uplink

41 Theoretical Analysis We first compute the channel esti-mates for all the allocated subcarriers in a RU of the symbols(ie 119897 = 3 10 or 4 11 depending on the subcarrier spacing)within a subframe that contain NDMRS sequences Thenwe obtain the channel estimates for the rest of the symbolsemploying one dimensional (1D) time-domain interpolationof the channel estimates within one subframe of a RUNPUSCH and NDMRS hopping are not considered in thiswork to make out derivations generally applicable to anymulticarrier communication systems The NDMRS-aidedchannel estimation can be done by using widely used esti-mation algorithms like LS [33] estimator and MMSE [34]

estimator We assume that all the scheduled number ofsubcarriers 119873RU

sc in a RU are occupied by NDMRS symbols(ie pilots) 119903119906(119899) generated in Section 31 within the specifiedsymbol locations Then the group of received pilot symbols119877 in the frequency-domain can be represented as

119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)

For the pilot symbol 119877 119867119877 is the true channel frequencyresponse (CFR) at the pilot locations and 119877 represents 119877times1Gaussian white noise vector and its noise variance 1205902

119877 Then

CFR estimates 119877 can be written as

119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)

where119867 is the 119871times 1 channel coefficient matrix in frequency-domain 119871 denotes the maximum channel delay spearedwhich is assumed to be shorter than the NB-IoT supportedCP length119873CP119897 and119865R represents119877times119871matrixTherefore thechannel estimates con

LS based on the conventional LSmethodof the whole channel response can be obtained as

conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)

where 119865119871 is the 119873RUsc times 119871 matrix which has the lines where

NDMRS symbols are located and the previous column of119873RUsc times 119873RU

sc DFT matrixThe LS algorithm is computationally less complex but the

problem is that the quantity (119865119867119877119865119877)minus1 in (20) which turnsout to be an ill-conditionedmatrixThus the conventional LSestimator cannot be a practical estimator to NB-IoT uplinksystems due to the presence of some subcarriers withoutSC-FDMA modulation The problem of conventional LSestimator can be mitigated to fit in the low complexity NB-IoT systems by adding a normalization matrix 120578119868119871 where 120578is a regularization parameter and its value has to be chosenfrom the range 0sim1 such that the resulting eigenvalues areall defined and the inverse matrix is least perturbed and 119868119871denotes the identity matrix Therefore the channel estimates

propLS of the proposed LS estimator in frequency-domain can

be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)

The mean square error (MSE) 120576propLS of the proposed LSestimator can be computed as

120576propLS = Ε [10038171003817100381710038171003817propLS minus119867100381710038171003817100381710038172] (22)


Consequently after simplification of (22) we have

120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)

The MMSE is an optimal estimation technique that exploitsthe knowledge of the channel statistics and channel covari-ance matrix For the conventional MMSE estimator we have

conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)

where Λ = Ε[119867119867119867] represents the autocovariance matrix of119867 MMSE is a modified form of conventional LS estimator in(20) but it is very intricate to obtain the precise knowledgeof the channel covariance matrix in very low SNR regimeFor the application of MMSE in NB-IoT uplink systems weassume that the delay spectrumof the channel power is evenlydistributed then the channel covariance matrix Λ turns outto be an identity matrix 119868119871 resulting in the elimination of realtime matrix inversion Furthermore the noise power is alsonormalized by dividing the average power 1205902119877 of the NDMRSsymbols Thus channel estimates prop

MMSE for the proposedMMSE estimator can be estimated as

propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)

TheMSE of the proposed method 120576propMMSE can be computed as

120576propMMSE = Ε [10038171003817100381710038171003817propMMSE minus119867100381710038171003817100381710038172] (26)

Subsequently the simplified form of (26) can be representedas the following form

120576propMMSE = 100381710038171003817100381710038171003817100381710038171003817Λ minus Λ(1 +ΓΥ (Λminus1))minus1100381710038171003817100381710038171003817100381710038171003817 (27)

where Υ represents the average SNR which is defined as

Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)

where Γ is the modulation scheme dependent constant forexample Γ = 1 for QPSK modulation

42 Simulation Results and Analysis We have consideredLTE-based NB-IoT uplink systems whose parameters areselected based on the specifications of 3GPP NB-IoT inrelease-13 We have investigated and compared the perfor-mance of our proposed NDMRS-assisted channel estimationalgorithms with conventional LS and MMSE algorithms interms of BER in contrast to SNR In this paper we haveconsidered a simple single-input single-output (SISO) system

Table 4 Simulation parameters

Parameter ValueSystem bandwidth 180 kHzCarrier bandwidth 900 MHzSubcarrier spacing 15 kHz and 375 kHzTransmission mode Singe-tone and multi-tone (3 6 or 12)Channel coding Turbo (13-coding rate)Modulation schemes BPSK and QPSKCRC 24 bitsAntenna configuration SISO (1Txtimes1Rx)Propagation channel Typical urban (TU) 119891d = 1HzChannel estimation Modified LS and MMSEChannel equalization Zero forcing (ZF)Number of iterations 105

for both single-tone transmission with 15 kHz and 375 kHzsubcarrier spacing and multi-tone transmission with 15 kHzsubcarrier spacing We have set the repetition number toguarantee the transmission reliability (ie BERlt10minus1) at lowSNR Transmission time and resource utilization are alsoour concern because low transmission time and high rateof resource utilization can improve the data rate of NB-IoT systems Low complexity zero forcing (ZF) equalizer isemployed In this simulation we have considered identicaltransmission time and resource utilization The fundamentalparameters are used to carry out simulations as listed inTable 4 and referred to figure captions for better readability

Simulation results of the performance of single-tone transmission for different channel estimators using1205872ndashBPSK modulation are shown in Figure 6 It is observedthat the channel estimation accuracy cannot be improvedwhen SNR is extremely low but estimation precision risesas the receive SNR increases (ie better channel condition)For 15 kHz subcarrier spacing as shown in Figure 6(a)our proposed LS and MMSE estimators perform betterthan the traditional LS and MMSE estimators As shown inFigure 6(b) the system performance of 375 kHz subcarrierspacing employing 1205872ndashBPSK for all estimation methods isslightly lower compared to 15 kHz subcarrier spacing

The BER performance curves of different channel esti-mators employing 1205874-QPSK constellation for single-tonetransmission are shown in Figure 7 The simulation resultselucidate that the system performance with 1205874-QPSK mod-ulation is little bit lower than 1205872ndashBPSK modulation due toextremely low SNR values However the system performanceimproves with our proposed algorithms compared to theconventional LS and MMSE algorithms regardless of themodulation scheme and subcarrier spacing

The BER performance curves of NPUSCH format-1 formulti-tone (eg 12-tone) transmission for different channelestimation techniques are shown in Figure 8 It is also seenthat the systemperforms better with our proposed algorithmsthan the traditional LS and MMSE algorithms Since NB-IoT supports only phase-shift-keying (PSK) modulation thereceiverrsquos performance of such two algorithms has linearchange and no significant variation when SNR is extremely


NPUSCH Single-tone (15 kHz) SISO (1Tx-1Rx) TU

Conventional LS estProposed LS est

Conventional MMSE estProposed MMSE est

0minus6minus8 minus4 minus2minus12minus14minus16 minus10minus18minus20

SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER

minus18 minus16 minus14 minus12 minus10 minus8 minus6 minus4 minus2 0minus20

SNR (dB)

(b)

Figure 6 BER performance of NPUSCH for single-tone transmission with 1205872 ndashBPSK modulation when MCS = 0 RU = 1 TBS = 16 andthe transmission time is 512 ms (a) 15 kHz subcarrier spacing with repetitions119872 = 64 and (b) 375 kHz subcarrier spacing with repetitions119872 = 16





SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)

Figure 7 BER performance of NPUSCH for single-tone transmission with 1205874 ndashQPSK modulation when MCS = 4 RU = 1 TBS = 56 andthe transmission time is 512 ms (a) 15 kHz subcarrier spacing with repetitions119872 = 64 and (b) 375 kHz subcarrier spacing with repetitions119872 = 16


NPUSCH 12-tone (15 kHz) SISO (1Tx-1Rx) TU




SNR (dB)

10minus2

10minus1

100

BER

Figure 8 BER performance of NPUSCH for multi-tone (eg 12-tone) transmission with 15 kHz subcarrier spacing using QPSKmodulation when MCS = 4 RU = 8 TBS = 552 and repetitions119872 = 64 The transmission time is 512 ms

lower Finally we conclude that our proposed MMSE algo-rithm can be coped with the practical implementation of NB-IoT uplink systems to ensure successful transmission of userdata for both single-tone and multi-tone transmissions

5 PAPR Analysis of NB-IoT Uplink

51 Theoretical Analysis The baseband time-domain trans-mit signals 119909119896119897(119905) and 119909119897(119905) are derived in (10) and (12) for

single-tone and multi-tone transmissions respectively Tomake our derivations generally applicable to any multicarriercommunication systems we assume that 119909(119905) is the con-tinuous time baseband SC-FDMA signal for both types oftransmission The PAPR of the time-domain baseband SC-FDMA signal119909(119905) can be defined as the ratio of themaximuminstantaneous power 119875max (ie peak power) to the averagepower 119875avg of the signal Thus we have

PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)

where 119879119904 is the symbol duration In NB-IoT uplink transmit-ter (ie NB-IoT UE) the PAPR can be reduced by exploitinglinear filtering operation referred to as pulse shaping tolimit the out-of-band radiation which decreases the spectralefficiency In this paper RC and RRC filters are employedto pulse shape the SC-FDMA signals The RC filter can becharacterized by the roll-off factor 120575 and the symbol duration119879119904 Then the impulse response of the RC filter in time-domain can be expressed as

ℎRC (119905) = sin (120587119905119879119904) sdot cos (120587120575119905119879119904)(120587119905119879119904) (1 minus 4120575211990521198792119904 ) (33)

Equation (33) can also be expressed in frequency-domain as

119867RC (119891) =

119879119904 0 le 10038161003816100381610038161198911003816100381610038161003816 le 1 minus 12057521198791199041198791199042 1 + cos [120587119879119904120575 (10038161003816100381610038161198911003816100381610038161003816 minus 1 minus 1205752119879119904 )] 1 minus 1205752119879119904 le 10038161003816100381610038161198911003816100381610038161003816 le 1 + 12057521198791199040 10038161003816100381610038161198911003816100381610038161003816 ge 1 + 1205752119879119904

(34)

The square-root of the RC filter output characterizes theimpulse response of the RRC filter Therefore the impulseresponse of the RRC filter in frequency-domain can bewritten as

119867RRC (119891) = radic119867RC (119891) (35)

Consequently the channel impulse response of RRC filter intime-domain can be represented asℎRRC (119905)= sin (120587119905119879119904) (1 minus 120575) + (4120575119905119879119904) cos (120587119905119879119904) (1 + 120575)(120587119905119879119904) (1 minus 16120575211990521198792119904 ) (36)

Finally the distribution of PAPR of the baseband SC-FDMAsignal 119909(119905) is the most practical performance indicator DWulich et al in [35] have investigated the amplitude of asingle-carriermodulated signal that does not have a Gaussiandistribution and it is also hard to deduce analytically theprecise form of the distribution In this paper we performnumerical analysis to investigate the PAPR properties of SC-FDMA signals For a given threshold value of PAPR 1205950the cumulative distribution function (CDF) can be definedas

119865120595 (1205950) = Pr (120595 le 1205950) (37)


Table 5 999 percentile PAPR for single-tone transmission

Modulation Subcarrier spacing (kHz) CCDF of PAPR (dB)No PS RC RRC

1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270

Single-tone (15 kHz) CCDF of PAPR

No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)

Figure 9 Comparison of CCDF of PAPR for NB-IoT uplink single-tone transmission with and without pulse shaping using 1205872-BPSKmodulation when TBS = 16 and roll-off factor 120575 = 02 (a) 15 kHz subcarrier spacing and (b) 375 kHz subcarrier spacing

where 120595 = PAPR[119909(119905)] The complementary CDF (CCDF) ofPAPR is the probability that the PAPRof the SC-FDMAsignalexceeds a given threshold 1205950 which can then be expressed as

Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)

52 Simulation Results and Analysis The CCDF is takento represent the statistical probability that the PAPR valueof a TBS exceeds a predefined threshold PAPR0 We haveconsidered an NB-IoT uplink transmission system for bothsingle-tone and multi-tone transmissions with 180 kHz sys-tem bandwidth Low PAPR modulation schemes like 1205872 -BPSK and 1205874 -QPSK for single-tone and only QPSK formulti-tone transmissions are employed Total 105 repetitionsare employed to calculate the CCDF of PAPR In addition theRC and RRC pulse shaping filters with roll-off factor 120575 = 02and oversampling factor of 4 are used as transmit filter to limitthe out-of-band radiationWehave compared the PAPRvaluethat is exceeded with the probability less than 01 percent (iePrPAPR gt PAPR0 = 10minus3) PAPR

Figure 9 shows the comparison of CCDF of PAPR amongno pulse shaping RC and RRC pulse shaping for single-tonetransmission with 1205872-BPSK modulation In this case both

15 kHz and 375 kHz subcarrier spacing types are consideredAs shown in Figure 9(a) it is observed that the 01 percentor 999 percentile PAPR of 15 kHz subcarrier spacing usingRRC filter are approximately 13 and 04 dB less compared tothe no pulse shaping and the RC filter respectively On theother hand 375 kHz subcarrier spacing with RRC filter asdepicted in Figure 9(b) shows about 13 and 021 dB less PAPRvalue at 01 percent of CCDF than without pulse shaping andRC filter respectively Figure 10 shows the comparison ofCCDF of PAPR with and without pulse shaping for single-tone transmission employing 1205874-QPSK modulation It canbe seen that the PAPR values for 1205874-QPSK modulationare higher than the PAPR values evaluated with 1205872-BPSKmodulation in Figure 9 regardless of the subcarrier spacingThe PAPR evaluation results for single-tone transmission canbe summarized in Table 5

The CCDF of PAPR curves with and without pulseshaping for multi-tone (eg 3 6 and 12-tone) transmissionemploying 1205874 -QPSK modulation are shown in Figure 11As shown in Figure 11 the PAPR value is increasing asthe number of tones increases at the 999 percentile ofCCDF Table 6 shows the summery of our evaluations formulti-tone transmission Finally we conclude that the lower


Table 6 999 percentile PAPR for multi-tone transmission

Modulation No of subcarriers CCDF of PAPR (dB)No PS RC RRC

QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)

Figure 10 Comparison of CCDF of PAPR for NB-IoT uplink single-tone transmission with and without pulse shaping using 1205874-QPSKmodulation when TBS = 56 and roll-off factor 120575 = 02 (a) 15 kHz subcarrier spacing and (b) 375 kHz subcarrier spacing

Multi-tone (15 kHz) CCDF of PAPR

Nsc=3 No PSNsc=3 RC PSNsc=3 RRC PSNsc=6 No PSNsc=6 RC PS

Nsc=6 RRC PSNsc=12 No PSNsc=12 RC PSNsc=12 RRC PS

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)

Figure 11 Comparison of CCDF of PAPR for NB-IoT uplink multi-tone transmission with and without pulse shaping transmit filterusing QPSK modulation when TBS = 56 and roll-off factor 120575 = 02

values of PAPR by using RRC filter is feasible for NB-IoTuplink transmitter thus requiring very little power back-offto maintain the linearity of the power amplifier

6 Conclusion

In this paper we have provided a brief survey of NB-IoTtechnology including deployment options physical channelsand signals uplink resource grid structure and resourceunit configuration We have developed a system model foruplink NB-IoT based on the 3GPP specifications in release-13 An analytical signal model and NDMRS generation andmapping are presented To guarantee the successful detectionof user data (ie BERlt10minus1) in extremely low SNR regimewe have proposed two channel estimation algorithms as amodified form of traditional LS and MMSE estimators Wehave investigated the effectiveness of our proposed NDMRS-assisted channel estimators compared with others throughextensive link-level computer simulations The simulationresults vindicate that our proposed estimation techniquesperform better at the SNRlt0 dB compared to the con-ventional LS and MMSE algorithms and suggesting thatthe proposed algorithms can be adopted to NB-IoT uplinkreceiver The improved channel estimation techniques can


be applied to not only NB-IoT systems but also in anymulticarrier communication systems Furthermore we haveanalyzed and evaluated the PAPR by employing RC andRRC pulse shaping at the transmitter Through numericalsimulations the PAPR values are evaluated for both single-tone and multi-tone transmissions Our evaluation resultsshow that the RRC pulse shaping with lower PAPR values isfeasible to the actual hardware design of low-costNB-IoTUEIn the future we will consider carrier frequency offset (CFO)and receiver diversity to improve the system performance inuplink NB-IoT systems

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

The authors would like to acknowledge the CAS-TWASPresidentrsquos Fellowship ProgramTheywould also like to thankthe Information Science Laboratory Center of University ofScience and Technology of China (USTC) for hardware andsoftware services

References

[1] A Nordrum ldquoPopular Internet of Things forecast of 50 billiondevices by 2020 is outdatedrdquo IEEE Spectrum 2016

[2] ldquoCellular networks for massive IoT-enabling low power widearea applications Ericsson White paper 2016rdquo httpswwwericssoncomresdocswhitepaperswp iotpdf

[3] A Diaz-Zayas C A Garcia-Perez A M Recio-Perez and PMerino ldquo3GPP Standards to Deliver LTE Connectivity for IoTrdquoin Proceedings of the 2016 IEEE First International Conference onInternet-of-Things Design and Implementation (IoTDI) pp 283ndash288 Berlin Germany April 2016

[4] F Liu C Tan E T Lim and B Choi ldquoTraversing knowledgenetworks an algorithmic historiography of extant literature onthe Internet of Things (IoT)rdquo Journal of Management Analyticsvol 4 no 1 pp 3ndash34 2017

[5] L D Xu W He and S Li ldquoInternet of things in industries asurveyrdquo IEEE Transactions on Industrial Informatics vol 10 no4 pp 2233ndash2243 2014

[6] S Li L D Xu and S Zhao ldquoThe internet of things a surveyrdquoInformation Systems Frontiers vol 17 no 2 pp 243ndash259 2015

[7] R Want B N Schilit and S Jenson ldquoEnabling the internet ofthingsrdquoThe Computer Journal vol 48 no 1 pp 28ndash35 2015

[8] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of things a survey on enabling tech-nologies protocols and applicationsrdquo IEEE CommunicationsSurveys amp Tutorials vol 17 no 4 pp 2347ndash2376 2015

[9] KMekki E Bajic F Chaxel and FMeyer ldquoA comparative studyof LPWAN technologies for large-scale IoT deploymentrdquo ICTExpress 2018

[10] J Petajajarvi K Mikhaylov M Hamalainen and J IinattildquoEvaluation of LoRa LPWAN technology for remote health andwellbeing monitoringrdquo in Proceedings of the 10th InternationalSymposium on Medical Information and Communication Tech-nology ISMICT 2016 USA March 2016

[11] Introduction of NB-IoT in 36331 3GPP RP-161248 3GPP TSG-RANMeeting 72 Ericsson Nokia ZTE NTT DOCOMO IncBusan South Korea Jun 2016

[12] N Mangalvedhe R Ratasuk and A Ghosh ldquoNB-IoT deploy-ment study for low power wide area cellular IoTrdquo in Proceedingsof the 27th IEEE Annual International Symposium on PersonalIndoor and Mobile Radio Communications PIMRC 2016 espSeptember 2016

[13] A Kiayani L Anttila Y Zou and M Valkama ldquoChannelEstimation and Equalization in Multiuser Uplink OFDMA andSC-FDMA Systems Under Transmitter RF Impairmentsrdquo IEEETransactions on Vehicular Technology vol 65 no 1 pp 82ndash992016

[14] J Xue and S Li ldquoAn SC-FDMA Channel Estimation AlgorithmResearch Based on Pilot Signalsrdquo in Proceedings of the 2nd Inter-national Symposium on Computer Communication Control andAutomation China Feburary 2013

[15] Y-P E Wang X Lin A Adhikary et al ldquoA premier on 3GPPnarrowband Internet ofThings (NB-IoT)rdquo IEEE Com Mag pp117ndash123 2017

[16] C Yu L Yu Y Wu Y He and Q Lu ldquoUplink schedulingand link adaptation for narrowband internet of things systemsrdquoIEEE Access vol 5 pp 1724ndash1734 2017

[17] J Zou H Yu W Miao and C Jiang ldquoPacket-Based PreambleDesign for Random Access in Massive IoT CommunicationSystemsrdquo IEEE Access vol 5 pp 11759ndash11767 2017

[18] W Yang M Hua J Zhang et al ldquoEnhanced SystemAcquisitionfor NB-IoTrdquo IEEE Access vol 5 pp 13179ndash13191 2017

[19] X Lin J Bergman F Gunnarsson et al ldquoPositioning for theInternet ofThings A 3GPP Perspectiverdquo IEEE CommunicationsMagazine vol 55 no 12 pp 179ndash185 2017

[20] S Hu A Berg X Li and F Rusek ldquoImproving the Perfor-mance of OTDOA Based Positioning in NB-IoT Systemsrdquo inProceedings of the 2017 IEEEGlobal Communications Conference(GLOBECOM 2017) pp 1ndash7 Singapore December 2017

[21] Y D Beyene R Jantti K Ruttik and S Iraji ldquoOn the perform-ance of narrow-band internet of things (NB-IoT)rdquo in Proceed-ings of the 2017 IEEE Wireless Communications and NetworkingConference WCNC 2017 USA March 2017

[22] L Zhang A Ijaz P Xiao and R Tafazolli ldquoChannel Equaliza-tion and Interference Analysis for Uplink Narrowband Internetof Things (NB-IoT)rdquo IEEE Communications Letters vol 21 no10 pp 2206ndash2209 2017

[23] R Ratasuk N Mangalvedhe J Kaikkonen and M RobertldquoData Channel Design and Performance for LTE NarrowbandIoTrdquo in Proceedings of the 2016 IEEE 84th Vehicular TechnologyConference (VTC-Fall) pp 1ndash5Montreal QC Canada Septem-ber 2016

[24] Evolved Universal Terrestrial Radio Access (E-UTRA) ldquoPhy-sical channels andmodulationrdquo 3GPP Tech Spec Group RadioAccess Network V 1340 Rel 13 Tech Spec TS 36211 2016httpwww3gpporgftpSpecsarchive36 series3621136211-d40zip

[25] Evolved Universal Terrestrial Radio Access (E-UTRA) ldquoPhysi-cal layer proceduresrdquo 3GPP Tech Spec Group Radio AccessNetwork V 1340 Rel 13 Tech Spec TS 36213 2016 httpwww3gpporgftpSpecsarchive36 series3621336213-d40zip

[26] Evolved Universal Terrestrial Radio Access (E-UTRA) ldquoUserEquipment (UE) Conformance Specificationrdquo Radio Transmis-sion and Reception 3GPP Tech Spec V1330 Rel 13 TechSpec TS 36521-1 2016


[27] Evolved Universal Terrestrial Radio Access (E-UTRA) ldquoNB-IoT Technical Report for BS and UE radio transmission andreceptionrdquo 3GPP Tech Rep V 1300 Rel 13TR 36802 2016

[28] GSMA ldquo3GPP Low Power Wide Area Technologiesrdquo GSMAWhite Paper 2016

[29] R Ratasuk B Vejlgaard N Mangalvedhe and A Ghosh ldquoNB-IoT system for M2M communicationrdquo in Proceedings of the2016 IEEEWireless Communications andNetworking ConferenceWorkshops WCNCW 2016 pp 428ndash432 qat April 2016

[30] Evolved Universal Terrestrial Radio Access (E-UTRA) ldquoMulti-plexing and channel codingrdquo 3GPP Tech Spec Group RadioAccess Network V 1340 Rel 13 Tech Spec TS 36212 2016httpwww3gpporgftpSpecsarchive36 series3621236212-d40zip

[31] F E Abd El-Samie F S Al-kamali A Y Al-Nahari and M IDessouky SC-FDMA for Mobile Communications CRC PressBoca Raton FL USA 2013

[32] Evolved Universal Terrestrial Radio Access (E-UTRA) ldquoUserEquipment (UE) Radio Transmission and Receptionrdquo 3GPPTech Spec V131 Rel 13 Tech Spec TS36101 2017

[33] J-J van de Beek O Edfors M Sandell S K Wilson and PO Borjesson ldquoOn channel estimation in OFDM systemsrdquo inProceedings of the 1995 IEEE 45th Vehicular Technology Con-ference Part 2 (of 2) pp 815ndash819 July 1995

[34] M Morelli and U Mengali ldquoA comparison of pilot-aided chan-nel estimation methods for OFDM systemsrdquo IEEE Transactionson Signal Processing vol 49 no 12 pp 3065ndash3073 2001

[35] DWulich and L Goldfeld ldquoBound of the distribution of instan-taneous power in single carrier modulationrdquo IEEE Transactionson Wireless Communications vol 4 no 4 pp 1773ndash1778 2005

International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018


Active and Passive Electronic Components

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of



Journal ofEngineeringVolume 2018

SensorsJournal of



RotatingMachinery


Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom


named narrowband IoT (NB-IoT) was specifically designedfor ultralow-end IoT applications The 3GPP finalized thespecifications ofNB-IoT in LTE release-13 [11] It is secure andreliable for data transmission because of the deployment inlicensed spectrum of GSM or LTE [12] NB-IoT enables IoTwhose applications are very diverse including smart meter-ing smart cities smart water smart environment smartagriculture smart animal firming retail logistics securityand emergencies industrial control and domestic and homeautomation Therefore the channel condition of NB-IoT isvery complicated due to its diverse application scenarios

Since NB-IoT is still in its infancy the proper guide-lines for efficient channel estimation and equalization arestill missing in the current literature Channel estimationbased on the pilot signal has been studied well for tradi-tional orthogonal frequency division multiplexing (OFDM)and single-carrier frequency division multiple access (SC-FDMA) systems eg [13 14] In the literature most ofthe NB-IoT research focused on frame structure design[15] scheduling and link adaptation [16] random accessprocedure [17] and system acquisition [18] Positioningperformance of NB-IoT has been studied in [19 20] YD Beyene et al in [21] have investigated the performanceof three traditional channel estimators only for 15 kHzsubcarrier spacing L Zhang et al in [22] have analyzed thechannel equalization and coexistence problem ofNB-IoT andLTE signals using only traditional LTE supported subcarrierspacing To the best of our knowledge the channel estimationof NB-IoT uplink with 375 kHz subcarrier spacing hasnot been studied yet Thus efficient channel estimation isthe prerequisite of coverage improvement equalization andsignal decoding at the receiver

In NB-IoT uplink low peak-to-average power ratio(PAPR) enhances the efficiency of the low-cost power ampli-fier Low out-of-band radiation is desired for the uplinktransmitter due to its very narrow bandwidth The PAPRproblem is more of a concern in the NB-IoT uplink (ielow-cost transmitter)The efficiency of the power amplifier iscrucial for the low-costNB-IoT user equipment (NB-IoTUE)with limited battery power Thus low PAPR in the NB-IoTuplink is ultimate desire owing to low-cost power amplifierPAPR reduction techniques such as scrambling discreteFourier transform (DFT) spreading and cyclic prefix (CP)insertion can be applied at the uplink transmitter NB-IoTsupports modulation schemes like 1205872-BPSK (binary phaseshift keying) and 1205874-QPSK (quadrature phase shift keying)which are also robust against PAPR by applying constellationrotation to make smooth transition between constellationpoints However these techniques are not fully preservedFurther reduction of PAPR would be great demand at theuplink because transmitter is the low-cost and low-power IoTdevice In [23] the authors have been evaluated the PAPRvalues employing root-raised cosine (RRC) pulse shaping(PS) filter only for the single-tone transmission This is thefirst time that the partial analysis of PAPR was taken intoaccount in the NB-IoT uplink transmitter

In this paper we have developed an uplink NB-IoTsystemmodel according to the 3GPP specifications in release-13 [24ndash26] The motivation was that the uplink transmission

of NB-IoT systems is more complicated compared to thedownlink transmission We have considered both types oftransmission schemes and subcarrier spacing for channelestimation and PAPR analysis The major contributions ofthis paper can be summarized as follows

(1) We have provided a brief overview of NB-IoT tech-nology including deployment options physical chan-nels and signals uplink frame structure and resourceunit (RU) definition An analytic NB-IoT uplinkreceived signalmodel is derived as a function of trans-mitted signal and channel impairments Narrowbanddemodulation reference signal (NDMRS) generationandmapping to time-frequency grid is also presented

(2) We have proposed two NDMRS-aided channel esti-mation algorithms based on the traditional leastsquares (LS) and minimum mean square error(MMSE) estimators that can be coped with the com-plicated channel conditions of NB-IoT systemsThrough simulations we have investigated and ver-ified the effectiveness of our proposed algorithmscompared with the conventional LS and MMSE algo-rithms in terms of bit error rate (BER) related tosignal-to-noise ratio (SNR) Simulation results showthat our proposed channel estimation algorithmsoutperform the others

(3) In addition we have provided theoretical analysis ofPAPR for the NB-IoT uplink employing raised-cosine(RC) and square-root-raised cosine (RRC) pulseshaping filters We have also showed the compari-son of PAPRvalues that are obtainedwith andwithoutPS through computer simulations for both single-tone andmulti-tone transmissions Numerical resultselucidate that RRC PAPR reduction technique is fea-sible for the implementation of NB-IoT uplink trans-mitter

The rest of the paper is organized as follows in Section 2 abrief overview of theNB-IoT technology is provided NB-IoTuplink signal model and NDMRS generation and mappingare presented in Section 3 Theoretical analysis of channelestimation and its performance analysis by simulations aregiven in Section 4 In Section 5 PAPR reduction techniquesand its numerical results are presented Finally we concludethe paper in Section 6

Notations Bold face lowercase letters are used to rep-resent time-domain vectors (or matrices) while frequency-domain vectorsmatrices are denoted by uppercase boldcharacters Superscripts (sdot)119879 and (sdot)119867 denote the transposeand Hermitian of a vector a scalar or a matrix respectivelyand (sdot)minus1 denotes matrix inversion The circular convolutionoperation is denoted by otimes and the operators Ε[sdot] | sdot | and sdot represent the expectation absolute value and Euclideannorm respectively 119868119871 denotes the 119871 times 119871 identity matrix

2 Overview of NB-IoT Technology

NB-IoT was designed by 3GPP as a key technology to meetthe demands of massive low-power IoT connectivity for the


LTE guard-band

LTEguard-band

LTE Carrier


(a)

LTEguard-band

LTEguard-band

LTE Carrier


(b)


GSM Carriers

(c)









Downlink






Uplink






15 kHz

1 16 112 8 ms3 8 56 4 ms6 4 28 2 ms12 2 14 1 ms







slots119873RU

(1)




119903119906 (119899) = 119903119906 (119899) (2)


119903119906 (119899) = 119890119895120572119899119890119895120593(119899)1205874 0 le 119899 lt 119873RUsc (3)



1 Frame = 10 ms

1 slot = 05 ms

012

12 S

ubca

rrie

rs =

180

kH

z

3456789

1011

ns = 0 ns = 1 ns = 19


48 S

ubca

rrie

rs =

180

kH

z

1 Frame = 10 ms

1 slot = 2 ms

0123456

424344454647

ns = 0 ns = 1 ns = 4



119906 =


sc = 3119873Ncell



sc = 12(4)






(119894) = (119887 (119894) + 119888 (119894))mod 2 (5)


DFT

Para

llel-t

o-Seri

al (P

S)

Dem

odul

atio

n

Des

cram

blingEst Data

Sequence

Rem

ove C

P

Phys

ical

Re

sour

ce

Dem

appi

ng an

d Eq

ualiz

atio

n

Channel Estimation

IDFT


RF Fro

nt-E

nd

(Rx)

Seri

al-to

-Pa

ralle

l (S

P)

Scra

mbl

ing

Mod

ulat

ion

DFT

Phys

ical

Re

sour

ce

Map

ping

IDFT

Add

CPPS

Para

llel-t

o-Seri

al (P

S)

RF Fro

nt-E

nd

(Tx)

Seri

al-to

-Pa

ralle

l (S

P)Data Sequence


Sequence


IDFT Add CP














119878 (119897 sdot 119872NPUSCHsc + 119896) = 1

radic119872NPUSCHsc





(7)




NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs


(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)





1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)





sc2 rfloor (10)



120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =


120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU


(11)




sc 2rfloor









119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



LTE guard-band

LTEguard-band

LTE Carrier


(a)

LTEguard-band

LTEguard-band

LTE Carrier


(b)


GSM Carriers

(c)









Downlink






Uplink






15 kHz

1 16 112 8 ms3 8 56 4 ms6 4 28 2 ms12 2 14 1 ms







slots119873RU

(1)




119903119906 (119899) = 119903119906 (119899) (2)


119903119906 (119899) = 119890119895120572119899119890119895120593(119899)1205874 0 le 119899 lt 119873RUsc (3)



1 Frame = 10 ms

1 slot = 05 ms

012

12 S

ubca

rrie

rs =

180

kH

z

3456789

1011

ns = 0 ns = 1 ns = 19


48 S

ubca

rrie

rs =

180

kH

z

1 Frame = 10 ms

1 slot = 2 ms

0123456

424344454647

ns = 0 ns = 1 ns = 4



119906 =


sc = 3119873Ncell



sc = 12(4)






(119894) = (119887 (119894) + 119888 (119894))mod 2 (5)


DFT

Para

llel-t

o-Seri

al (P

S)

Dem

odul

atio

n

Des

cram

blingEst Data

Sequence

Rem

ove C

P

Phys

ical

Re

sour

ce

Dem

appi

ng an

d Eq

ualiz

atio

n

Channel Estimation

IDFT


RF Fro

nt-E

nd

(Rx)

Seri

al-to

-Pa

ralle

l (S

P)

Scra

mbl

ing

Mod

ulat

ion

DFT

Phys

ical

Re

sour

ce

Map

ping

IDFT

Add

CPPS

Para

llel-t

o-Seri

al (P

S)

RF Fro

nt-E

nd

(Tx)

Seri

al-to

-Pa

ralle

l (S

P)Data Sequence


Sequence


IDFT Add CP














119878 (119897 sdot 119872NPUSCHsc + 119896) = 1

radic119872NPUSCHsc





(7)




NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs


(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)





1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)





sc2 rfloor (10)



120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =


120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU


(11)




sc 2rfloor









119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





Downlink






Uplink






15 kHz

1 16 112 8 ms3 8 56 4 ms6 4 28 2 ms12 2 14 1 ms







slots119873RU

(1)




119903119906 (119899) = 119903119906 (119899) (2)


119903119906 (119899) = 119890119895120572119899119890119895120593(119899)1205874 0 le 119899 lt 119873RUsc (3)



1 Frame = 10 ms

1 slot = 05 ms

012

12 S

ubca

rrie

rs =

180

kH

z

3456789

1011

ns = 0 ns = 1 ns = 19


48 S

ubca

rrie

rs =

180

kH

z

1 Frame = 10 ms

1 slot = 2 ms

0123456

424344454647

ns = 0 ns = 1 ns = 4



119906 =


sc = 3119873Ncell



sc = 12(4)






(119894) = (119887 (119894) + 119888 (119894))mod 2 (5)


DFT

Para

llel-t

o-Seri

al (P

S)

Dem

odul

atio

n

Des

cram

blingEst Data

Sequence

Rem

ove C

P

Phys

ical

Re

sour

ce

Dem

appi

ng an

d Eq

ualiz

atio

n

Channel Estimation

IDFT


RF Fro

nt-E

nd

(Rx)

Seri

al-to

-Pa

ralle

l (S

P)

Scra

mbl

ing

Mod

ulat

ion

DFT

Phys

ical

Re

sour

ce

Map

ping

IDFT

Add

CPPS

Para

llel-t

o-Seri

al (P

S)

RF Fro

nt-E

nd

(Tx)

Seri

al-to

-Pa

ralle

l (S

P)Data Sequence


Sequence


IDFT Add CP














119878 (119897 sdot 119872NPUSCHsc + 119896) = 1

radic119872NPUSCHsc





(7)




NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs


(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)





1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)





sc2 rfloor (10)



120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =


120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU


(11)




sc 2rfloor









119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



1 Frame = 10 ms

1 slot = 05 ms

012

12 S

ubca

rrie

rs =

180

kH

z

3456789

1011

ns = 0 ns = 1 ns = 19


48 S

ubca

rrie

rs =

180

kH

z

1 Frame = 10 ms

1 slot = 2 ms

0123456

424344454647

ns = 0 ns = 1 ns = 4



119906 =


sc = 3119873Ncell



sc = 12(4)






(119894) = (119887 (119894) + 119888 (119894))mod 2 (5)


DFT

Para

llel-t

o-Seri

al (P

S)

Dem

odul

atio

n

Des

cram

blingEst Data

Sequence

Rem

ove C

P

Phys

ical

Re

sour

ce

Dem

appi

ng an

d Eq

ualiz

atio

n

Channel Estimation

IDFT


RF Fro

nt-E

nd

(Rx)

Seri

al-to

-Pa

ralle

l (S

P)

Scra

mbl

ing

Mod

ulat

ion

DFT

Phys

ical

Re

sour

ce

Map

ping

IDFT

Add

CPPS

Para

llel-t

o-Seri

al (P

S)

RF Fro

nt-E

nd

(Tx)

Seri

al-to

-Pa

ralle

l (S

P)Data Sequence


Sequence


IDFT Add CP














119878 (119897 sdot 119872NPUSCHsc + 119896) = 1

radic119872NPUSCHsc





(7)




NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs


(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)





1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)





sc2 rfloor (10)



120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =


120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU


(11)




sc 2rfloor









119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



DFT

Para

llel-t

o-Seri

al (P

S)

Dem

odul

atio

n

Des

cram

blingEst Data

Sequence

Rem

ove C

P

Phys

ical

Re

sour

ce

Dem

appi

ng an

d Eq

ualiz

atio

n

Channel Estimation

IDFT


RF Fro

nt-E

nd

(Rx)

Seri

al-to

-Pa

ralle

l (S

P)

Scra

mbl

ing

Mod

ulat

ion

DFT

Phys

ical

Re

sour

ce

Map

ping

IDFT

Add

CPPS

Para

llel-t

o-Seri

al (P

S)

RF Fro

nt-E

nd

(Tx)

Seri

al-to

-Pa

ralle

l (S

P)Data Sequence


Sequence


IDFT Add CP














119878 (119897 sdot 119872NPUSCHsc + 119896) = 1

radic119872NPUSCHsc





(7)




NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs


(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)





1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)





sc2 rfloor (10)



120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =


120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU


(11)




sc 2rfloor









119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



NDMRS

User data

1 slot = 05 ms

12 S

ubca

rrie

rs


(a)



User data

NDMRS1 slot = 05 ms

1 slot = 2 ms

(b)





1 for 119873RUsc = 1 (8)

and

119873slots = 1 Δ119891 = 375 kHz

2 Δ119891 = 15 kHz(9)





sc2 rfloor (10)



120593119896119897 = 120588 ( mod 2) + 120593119896 ()120588 =


120593119896 ()=

0 = 0120593119896 ( minus 1) + 2120587Δ119891(119896 + 12) (119873 + 119873CP119897) 119879s gt 0

= 0 1 119872119873RU119873RUslots119873RU


(11)




sc 2rfloor









119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia






119910single (119905) = 119909119896119897 (119905) otimes ℎ (119905) + 119899 (119905) (13)

119910multi (119905) = 119909119897 (119905) otimes ℎ (119905) + 119899 (119905) (14)


ℎ (119905) = 119871minus1sum119894=0

120573119894120575 (119905 minus 120591119894) (15)


119910single (119905) = 119871minus1sum119894=0

120573119894119909119896119897 (119905 minus 120591119894) + 119899 (119905) (16)

119910multi (119905) = 119871minus1sum119894=0

120573119894119909119897 (119905 minus 120591119894) + 119899 (119905) (17)






119877 = [119877 (0) 119877 (1) 119877 (119873RUsc minus 1)]119879 (18)


119877 Then


119877 = 119867119877 + 119877 = 119865119877119867 + 119877 (19)



conLS = 119865119871 (119865119867119877119865119877)minus1 119865119867119877 119877 (20)






be estimated as

propLS = 119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 119877 (21)





120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




120576propLS = 1205902119877119865119871 (119865119867119877119865119877 + 120578119868119871)minus1 119865119867119877 (23)


conMMSE = 119865119871 (119865119867119877119865119877 + 1205902

119877Λminus1)minus1 119865119867119877 119877 (24)



propMMSE = 119865119871[[119865

119867119877119865119877 + (1205902

1198771205902119877 ) 119868119871]]minus1

119865119867119877 119877 (25)






Υ = 12059021198771205902119877

(28)

and

Γ = Ε [10038161003816100381610038161003816119877 (119873RUsc )100381610038161003816100381610038162] Ε[

1003816100381610038161003816100381610038161003816100381610038161119877 (119873RUsc )

1003816100381610038161003816100381610038161003816100381610038162] (29)














SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia







SNR (dB)

10minus2

10minus1

100

BER

(a)




10minus2

10minus1

100

BER


SNR (dB)

(b)






SNR (dB)

10minus2

10minus1

100

BER

(a)





SNR (dB)

10minus2

10minus1

100

BER

(b)







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia







SNR (dB)

10minus2

10minus1

100

BER






PAPR [119909 (119905)] = 119875max119875avg (30)

where

119875max = max0le119905le119873RU

sc 119879119904

[|119909 (119905)|2] (31)

and

119875avg = 1119873RUsc

int119873RUsc 119879119904

0Ε [|119909 (119905)|2] 119889119905 (32)




119867RC (119891) =


(34)


119867RRC (119891) = radic119867RC (119891) (35)



119865120595 (1205950) = Pr (120595 le 1205950) (37)




1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





1205872-BPSK 15 364 274 234375 355 246 225

1205874-QPSK 15 440 350 275375 370 345 270


No PSRC PSRRC PS

15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)



Pr (120595 ge 1205950) = 1 minus 119865120595 (1205950) (38)








QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia





QPSK3 44 370 2806 545 380 3012 640 390 340


No PSRC PSRRC PS

05 1 15 2 25 3 35 454

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(a)


No PSRC PSRRC PS

05 1 15 2 25 3 35 4

PAPRＩ(dB)

10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

(b)





10minus5

10minus4

10minus3

10minus2

10minus1

100

Pr(P

APR

gtPA

PRＩ)

1 2 3 4 5 6 70PAPRＩ(dB)



6 Conclusion






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia






Acknowledgments


References







































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia














RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


channel estimation and peak-to-average power ratio...

Documents