Network Topology Modulation for Energy and Data Transmission in Internet of Magneto-Inductive Things Burhan Gulbahar Applied Research Center of Technology Products Department of Electrical and Electronics Engineering Ozyegin University, Istanbul, 34794, Turkey Email: [email protected] 7 1 0 2 n Abstract—Internet-of-things (IoT) architectures connecting a transmitting data and energy by using radio frequency (RF) a J massive number of heterogeneous devices need energy efficient, or magneto-inductive (MI) methods [2], [3]. However, the low hardware complexity, low cost, simple and secure mech- 6 existing literature generallyconcentrateson the energyversus anisms to realize communication among devices. One of the 2 capacitytrade-offincludingpowerallocationschemes[2],[4], emergingschemesistorealizesimultaneouswirelessinformation and power transfer (SWIPT) in an energy harvesting network. [5]. The recent studies analyze RF based SWIPT modulation ] T Radio frequency (RF) solutions require special hardware and methods including spatial domain (SD) and intensity based I modulation methods for RF to direct current (DC) conversion energy pattern changes [2], [6]. RF solutions are not mature s. and optimized operation to achieve SWIPT which are currently today to achieve SWIPT and require specialized circuits for in an immature phase. On the other hand, magneto-inductive c RF to direct current (DC) conversion. In this article, novel [ (MI) communication transceivers are intrinsically energy har- vesting with potential for SWIPT in an efficient manner. In and practical network topology modulation and demodulation 2 this article, novel modulation and demodulation mechanisms architectures are presented for MI communications (MIC) v are presented in a combined framework with multiple-access networksbychangingthespatialpatternofpowertransmitting 6 channel(MAC)communicationandwirelesspowertransmission. active coils. The proposed MIC based IoT architecture (MI- 9 The network topology of power transmitting active coils in IoT)providesreliable,simpleandsecuremechanismswithlow 4 a transceiver composed of a grid of coils is changed as a 2 novel method to transmit information. Practical demodulation cost and low latency performances for connecting everyday 0 schemes are formulated and numerically simulated for two- objects with direct power transmission capability. . user MAC topology of small size coils. The transceivers are MIC is an alternative method with the advantage of uni- 1 suitabletoattachtoeverydayobjectstorealizereliablelocalarea 0 formity for varying environments without medium specific network(LAN)communicationperformanceswithtensofmeters 7 attenuation, multi-path and high propagation delay in chal- communication ranges. The designed scheme is promising for 1 future IoT applications requiring SWIPT with energy efficient, lenging environments including underwater, undergroundand : v low cost, low power and low hardware complexity solutions. nanoscale medium with in-body and on-chip applications [7], i Index Terms—Simultaneous wireless information and power [8], [9], [10]. In [4] and [11], a trade-offanalysisis presented X transfer, magneto-inductive communication, network topology for the problem of information and power transfer on a r modulation, internet-of-things a coupled-inductor circuit with power allocation policies. In I. INTRODUCTION [10], a nanoscale communication architecture with graphene coilsispresentedsatisfyingbothpoweranddatatransmissions The Internet of Things (IoT) and 5G architectures re- for in-body and on-chip applications. On the other hand, quire connecting heterogeneous devices including machine- existingstudiesonMICnetworkstreatothercoilsassourcesof to-machine (M2M) and wireless sensor networking (WSN) interferenceincludingmultiple-inputmultiple-output(MIMO) unitswith potentiallylowerdata rate butlowlatency,reliable, and diversity architectures [3], [7], [8], [9], [12], [13]. How- energyefficientandsecuremechanisms[1].Theseapplications ever, SWIPT architectures, MAC schemes utilizing the same require short packets and simple modulation/demodulation time-frequencyresources, and modulationmethodsother than mechanismsforthewidespreadutilizationofIoT.Themassive classical signal waveform approaches are not discussed. number of devices require not only low hardware complexity Inthisarticle,theinformationisembeddedtocoilpositions schemes but also methods of energy harvesting such as si- by modulating the frequency selective MI channelor network multaneouswireless informationand power transfer (SWIPT) topologyinsteadofclassicalsignalwaveformmodulation.The Dr. Burhan Gulbahar is with the Department of Electrical and proposedschemefulfills theidea offullycoupledinformation Electronics Engineering and Applied Research Center of Technol- and power transfer for widespread utilization of MI-IoT ap- ogy Products, Ozyegin University, Istanbul, 34794, Turkey, (e-mail: [email protected]). plications [5]. It eliminates signal modulation and does not TobepublishedinProc.ofIEEEGLOBECOM2016FirstInternationalWorkshopontheInternetofEverything(IoE)(cid:13)c2016IEEE MI BALLOON II. SYSTEMMODEL In this paper,there are K 1 transmitters denotedby Tx j − for j [1, K 1] to transmit data to a single receiver, ∈ − i.e., indexed as the first user and denoted by Rx, in a MAC (a) (b) topology with random positions and orientations as shown in Fig. 1(a). Each user has a grid of closely spaced N N coils Fig.1. (a)MI-IOTforLANsofdailyobjectsformingacoupledMIballoon, × and (b) a grid of N ×N coils at each transceiver with varying topologies where each coil is either in open circuit condition without consisting ofndifferent spatialsymbols. any effect to the network or actively resonating as shown in Fig. 1(b). The number of concurrently active coils in each transceiver with the ability to change its spatial coil waste any power for data transmission. Furthermore, it does distribution is denoted by T. The grid can be realized on a not require transmitter pre-coding, channel state information flexible paper suitable to attach to daily objects and to be andintensitypatternmodulationasinRFcounterpartutilizing recycled considering the billions of devices. Generalised Pre-coding aided Spatial Modulation (GPSM) Circuit theoretical equivalents of transceivers are presented scheme [6]. Moreover, MAC sharing problem is intrinsically in detail in [3], [7], [10]. Each coil has series RLC type solved due to includingwhole network topologyas data sym- circuit with resistance Ri, capacitance Ci and inductance Li bol. The solution requires no receiver feedback and provides for i [1, NTot] where NTot = K T with resonance ∈ × MAC sharing without sacrificing resources for transmitter frequency ω0 = 1/√LiCi. Furthermore, each coil has the synchronization or interference prevention. The contributions samepropertiesto simplifythe analysis,i.e.,Ri =R, Li =L achieved, for the first time, are summarized as follows: and Ci = Cs for i [1, NTot]. The receiver loads denoted ∈ by Z for the coils of Rx are set to R without requiring any L • Network topology modulation mechanism directly mod- calibrationwithatrade-offinthereceivedpowerperformance. ulating the frequency selective MI channel. Transmitter and receiver diversity are provided by N < T R • Practical topology demodulation combined with SWIPT. loaded coils in Rx (the remaining ones as passive coils) and • Topology modulating MACs fully utilizing time- all transmitter coils with the same resonance voltage. frequency bands without synchronization or contention. Voltage sources and coil currents in the network denoted • Reliable, capable of energy harvesting, low cost, low by the column vectors V and I , respectively, satisfy ω ω hardware complexity and low latency IoT networking. V = M I where M is mutual inductance matrix, ω ω ω ω VT = 0 ... 0 V V ... V ... V , ω is the oper- The proposed method supports the widespread adoption of ω 2,1 2,2 i,j K,T MIC networks for IoT satisfying the following requirements: atingfr(cid:2)equency,Vi,j denotesthevoltagelevel(cid:3)injthcoilofith user, i [2, K], j [1, T] and . T denotes transpose. M ω ∈ ∈ {} 1) Low latency: continuous energy and data transmission has the elements M (i,j)=ωM(i,j) for i=j, Z + Z ω i L 6 in a MAC topology without the overhead of resource for i = j [1, N ] and Z for i = j [N + 1,N ] R i R Tot ∈ ∈ sharing, synchronization or receiver feedback. whereZ =R+ωL+1/(ωC ),=√ 1isthecomplex i s − 2) High reliability: the robustness of the MIC channel to unityandM(i,j)=M(j,i)isthemutualinductancebetween fading and low probability of symbol error. ith and jth coil. Next, the network topology modulation is 3) Low hardware complexity: simple and low-cost SWIPT introduced. transceiverwith alreadyavailableMI circuits,and with- out separate structures for data and power transmission. III. NETWORK TOPOLOGYMODULATION 4) Energy harvesting capability: intrinsically by MI coils The spatial grid topologies of each transmitter are changed withoutrequiringseparateRFtoDCconversioncircuits. toformvaryingmutualcouplings.Asignalmodulatingvoltage 5) Security: immunity to RF fields and radiative effects; waveform is not required but only varying spatial patterns of potential detection of intruder coils as the changes in actively resonating coils. Therefore, the network is continu- networktopologysymbolandpowertransmissionlevels. ously transmitting power to the receiver by also embedding 6) Energy Efficiency: no extra energy for signal waveform information into the spatial structure that the energy is trans- based data transmission; the crowded set of MI coils mitted,however,withoutanysignalmodulationcomplexityor forming a waveguide to enhance the communication energyreservationfordata.Thenetworktopologymodulation rangewithoutconsuminganyactivetransmissionpower. is mainly realized by changingM−1 denotedby Γ which is ω ω easily calculated by using matrix identities as follows: The remainder of the paper is organized as follows. In Section II, MI network system model for MAC topologies NR −1 Γ =Q α Λ + I + β ~q ~qHΛ QHG−1 (1) is presented. In Sections III and IV, topologymodulationand ω (cid:18) ω ω i i (cid:19) ω demodulation mechanisms are introduced, respectively. Then, Xi=1 inSectionV,theproposedmethodsaresimulatedfortwo-user where the elements M(i,j) of M = QΛQH are equal to MAC network. Finally, in Sections VI and VII, open issues M(i,j) for i = j and L otherwise, diagonal Λ includes 6 and conclusions are discussed, respectively. eigenvalues, QQH = QHQ = I, G−1 (ω)−1 α I+ ω ≡ ω (cid:0) TABLEI Assume that a set of symbols consisting of n different NETWORKTOPOLOGYMODULATIONTYPESANDPROPERTIES topologies for each transceiver is utilized. The total number of differentsymbolcombinationsexcludingRx is denotedby ID Frequency Time Nc ID Frequency Time Nc Nc and ithnetworksymbolis denotedby si. The possible set 1 ωkforkthcoil Concurrent nK−1 3 Sω (forall) Concurrent nK−1 ofmodulationtypesindexedwithModk fork ∈[1, 4]defined in (2), (3), (4) and (5), respectively, and their properties are 2 ωkforkthcoil TDMA n 4 Sω (forall) TDMA n showninTableI.Mod andMod havetheconstellationsize 2 4 of N = n while it becomes nK−1 for Mod and Mod . c 1 3 β I , I is the identity matrix, ~qH is the ith row of Q, The best time-frequency diversity is achieved with Mod3 ω L i given in (4) which is discussed in the following sections. The IL(i,(cid:1)j) = 1 for i = j [1,NR] and zero otherwise, α = ω2C /ς , β = ∈ω3C2Z / ς2 + ς C ωZ computationalcomplexityoftheproposedmodulationscheme ω s ω ω − s L ω ω s L includes the calculation of total transmit power at the sources and ςω =(CsωR ). The proof is give(cid:0)n in Appendix A.(cid:1) − andnormalizingthepowerresultinginN N multiplications It is observed in simulation studies that different network S ω topologies provide different Λ allowing to realize eigenvalue and T2 additions where NS denotes the number of random modulation.EigenvaluemodulationsymbolsΛ areeitherpre- frequencysets(Sω)toincludemorediversefrequencyeffects. k Next, demodulation mechanisms are discussed. determined for each k [1,N ] by using pilot training phase c ∈ or they are estimated based on the preliminary knowledge of IV. NETWORK TOPOLOGYDEMODULATION spatialmodulationpatternsandphysicalsizeofLANtopology. ItisobservedthatΛ isrobusttoangulardisorientationswhile The demodulation is achieved by finding the spatial mod- the detailed analysis is left as a future work. In the following ulation matrix Γω for different symbols and at different frequencies, and then comparing with the received current at discussions,fourdifferentmodulationmethodsareproposedin varyingfrequencies.Furthermore,the symbolsare definedfor termsof oscillation frequenciesof each coiland time sharing. the whole network by solving MAC sharing problem without A. Modulation with Orthogonal Time-Frequency Sharing requiringsynchronizationor time-frequencyresource sharing. Assume that each coil with the index t in a grid transmits IfweassumethatMod isused,thenetworktopologysymbol 3 vpoolwtaegrealteavedliVffeirsensettfrtoequuneintycywωitthfeoqrutal∈ph[1a,sTe]f.oFruarllthceorimlsofroer, hfraesqtuheenicnydethxekn,fothrekm∈e[a1s,uNrecd]annodisωy cisurthreenptodweenrottreadnsbmyi˜Issi(oin) ω simplicity. Then, the current at ωt in ith coil of the receiver, at the ith coil receiver is found by combining (1) and (4) as i.e., Iωt(i), is given as follows: follows: K−1 ˜I (i)=I (i)+ n =~qH Υ ~q + n (6) I (i)= Γ (i,t +j T) + n (2) ω ω s,i i,k ω,k Σ,k s,i ωt ωt t,i Xj=1 where ~qH is the ith row of Q in (1), ~q = NTot ~q , i,k Σ,k j=T+1 j,k where nt,i is assumed to be independent and identically Υ = α (ω)−1 Θ + β NR Ψ P−1, Ψ = distributed complex valued additive white Gaussian noise ω,k ω ω,k ω i=1 i,k i,k (AWGN) with variance of σ2 and E n n∗ = 0 for ~qi,k~qHi,kΛk and Θω(cid:0),k = αωΛk +PI. The p(cid:1)roof is given t,i { t,k m,l} in Appendix B. It is further simplified with NR = 1, m = t or k = l. Moreover, if time-division multiple access 6 6 NTot = 1+(K 1)T and by using perturbation equality (TDMA) channel sharing is utilized, then the induced current (I +a~x~yH)−1 =−I + b~x~yH where b = a/(1 + a~yH~x) at a time slot is due to some specific Tx for j [1, K 1] − j and the proof is given in Appendix C. Then, Υ becomes ∈ − ω,k and I (i) becomes equal to the following: ωt as follows: Iωt(i)=Γjωt(i,t +j T) + nt,i (3) Υ = αωΘ−ω,1k αωβωΘ−ω,1kΨ1,kΘ−ω,1k(ω)−1 (7) where Γjωt is obtained by deleting the rows and columns of ω,k ω − 1 + βω~qH1,kΛkΘ−ω,1k~q1,k Γ with the indices not in [jT + 1,(j + 1)T]. ωt Then,insertinginto(6)anddroppingi=1,itbecomesI˜ = B. Modulation with Network Topology Diversity ωs ~υT ~q∗ ~q whereω =ω , istheKroneckerproduct If all the coils utilize the same frequency set in parallel, ωs,k 1,k⊗ Σ,k s ⊗ and ~υ(cid:0)ω,k(t) for t (cid:1) [1, NTot] is given by the following: i.e., S = ω : s [1,N ] , where N is the number ∈ ω s ω ω of concurren{t frequen∈cies then,}the transmitter diversity is ~υ (t) = αω ϕ Nj=T1ot̺ω,k,jλk,tϕω,k,t (8) realized and the received current is given as follows: ω,k ω(cid:18) ω,k,t− P1 + NTotλ ̺ (cid:19) j=1 k,j ω,k,j NTot P Iωs(i)= Γωs(i,t) + ns,i (4) where ̺ω,k,j ≡ βω|qk,j|2ϕω,k,j, ϕω,k,t ≡ (αωλk,t + 1)−1, t=XT+1 λk,t Λk(t,t) and qk,t ~q1,k(t). If it is assumed that ≡ ≡ eigenvalues are close to each other then, ~υ (t) is approxi- If TDMA is utilized then, I (i) due to Tx is as follows: ω,k ωs j matedbydroppingtheq2 termsduetotheunitarymagnitude k,j T property of ~q . Then, the received current vector ˜I = I (i)= Γj (i,t+jT) + n (5) 1,k S ωs ωs s,i [I˜ ...I˜ ]T is approximated by ˜I η Υ ~q + n Xt=1 ω1 ωNω S ≈ k k 0,k S (a) Fig.3. Networktopologydemodulation without pilottraining. (b) Pe ≤ (0.5/Nc) Ni=c1 Nk=c1,k6=ierfc(dik/2√2N0) where Fig.2. (a)Networktopologymodulation,and(b)demodulationmechanism dik is the distancPe betwPeen si and sk, and erfc(.) is the forpilotaidedcasewithfullknowledge ofthetransmitted symbolset. complementary error function. B. Pre-computation based Demodulation where ΥT ~υ ... ~υ , n =[n ...n ]T is the k ≡ ω1,k wNω,k S ω1 ωNω complexreceiv(cid:2)ernoisevectorha(cid:3)vingN0/2spectraldensityfor It is assumed that there is no knowledge about the symbol realandimaginarypartsateachfrequency,~q0,k ≡~q∗1,k⊗~qΣ,k set or ηk~q0,k. An average Υk is pre-calculated based on the satisfying NTot~q (t) = 0 and η equalizes the received observation of the small variance of Λ and by considering t=1 0,k k k or transmitPted power for the symbol. the approximation in (8) without q2 terms for the estimated k,j There are two differentdemodulationmethodswith respect range and orientations of the transmitters. The proposed de- totheamountofknowledgeaboutη Υ ~q foreachsymbol modulationschemeisshowninFig.3.Thetransmitpowerfor k k 0,k s such that it is either perfectly known by pilot aided trans- s is equalized with η instead of the receiver power since k m m mission or Υ is estimated based on pre-calculation for the there is no feedback. Then, the estimation of s is achieved k m communicationrangeswithoutanyinformationaboutη ~q . by finding the index k minimizing NTotη q (t)2 It is observed in Section V that the eigenvalues in Λkk h0a,vke where ηkq0,k,m = Υ†k˜IS, ηkq0,k,m |dPento=te1s bpkarba0m,ke,tmer e|s- low variances for varyingpositions and orientations, and they timation regarding symbol s when symbol s is trans- k m are specific for each symbol sk. Demodulation schemes are mitted,b˜I b= η Υ ~q b+ bn and Υ† denotes Moore- S m m 0,m S k realizedbyutilizingthisuniquenessasshowninthefollowing. Penrose pseudoinverse of Υ . Then, NTotη q (t)2 k | t=1 k 0,k,m | A. Pilot Aided Perfect Knowledge of Symbol Set 1eqTuΥals†Υto |zηk,m~q|2 w,hmereT zk,m1T≡Υ†gka,mndP+1TmisTkbtnhSeb,rogwk,mve≡c- ΥkThaendrec~qe0i,vkerfoisr aasnsyumkewdittohhthaveeatihdeoffulpliklontowsylmedbgoelso.fTηhke, tcoerntorafklaclhlmi-osnqmeusa.r0e,χm2kd,imst(r1ikb)u≡t≡ion|zwk,imtkh|2thies naumgebneerraolifzeddegnroenes- modulation schemes are shown in Fig. 2(a) where NS =1 in of freedom of one where zk,m is complex Gaussian random pilot training case. The demodulationmechanism is shown in variable with the expectation E z = g and variance k,m k,m FTisg,.t2ra(bn)s.mEiasscihonsydmeblaoylstkdisanthdenvSectiosrtηhkeΥAkW~qG0,Nk wcoitmhppoenrieondt IVfart{hzek,mpr}ob=abiNlit0ykmPk(kχ222, an(d1)k{.k>2 dχe2}not(e1s))Euiscliddeenaontendormby. m,m j,m with independent and identically distributed elements having P (m,j) then, the probability of error P is bounded by e e variance σnωs =N0 for s∈[1,Nω]. The optimumestimation (1/Nc) Ni=c1 Nk=c1,k6=iPe(i,k).Theprobabilitydistribution is achieved with maximum likelihood (ML) decision using of the difPferencPeoftwo correlatednon-centralchi-squareran- the distance metric, i.e., 0.5sHk sk−˜IHS sk, and choosing the dom variables is complex to formulate explicitly. In addition, minimumbasedon decisionregionsDk foreachsk [14]. The randomsets ofenergytransmissionfrequenciesare utilizedto computational complexity includes finding the maximum of improvedetectionperformancebyaveragingthetruedecisions tNheωcmorurletliaptliiocnatsioannsdathnedtaedrmditOion(Ns ωinn˜KIHS−1s)kfworitMh NodS3 d=ue1t.o gasivesnhofworneainchFsiegt.S3ω.,iF=urt{hωesr,mi o:rse,∈t[h1e,Nnωoi]s}efoisr ism∈a[l1le,rNSin] Transmitpowersareadaptedwithηk toequalizetotalreceived each detection procedure where Nω different frequencies are power for each k for simplicity. Then, with equally probable utilized instead of N N . The computational complexity ω S symbols, the probability of error per transfer is given by includes finding the m×inimums of N different vectors of Pe = 1− N1c Nmc=1 Dmf˜IS(˜IS|sm)d˜IS where f˜IS(˜IS|sm) length Nc, finding the maximum of aSvector of size Nc and isapproximatePdby(πRN0)−N2ωexp −N10 Nj=ω1|I˜ωj−smj|2 the term O(NSNωNc)=O(NSNωnK−1) for Mod3 due to andsmj isthejthelementofsm.T(cid:0)heuppePrboundisgivenby(cid:1) NSNc(Nω + 1) multiplications and NSNcNω additions. (a) (b) Fig.4. (a)Two-userMACtopologywithmirrorsymmetricpositionsofthe parallel transmitters and(b)threedifferent network topologysymbols. TABLEII (a) SIMULATIONPARAMETERS PARAM. MEANING VALUE r,∆c Thecoilradiusandinter-coildis- 1cm,1mm tanceinthegrids R,L,Cs The resistance, inductance and 65mΩ,84 nH,3-300nF capacitanceofasinglecoilunit f0 Operatingresonancefrequency 1MHz,10MHz ωl,ωh(×ω0−1) Transmissionfrequencyinterval {0.92-1.08},{0.99-1.01} NS,Nω,Nc Thenumberofrandomfreq.sets, {1,24},{24,32},3 freqs.inaset,andthesymbols N0 Noisepowerspectraldensity 10−21,10−16,10−12W/Hz PT Averagetrans.powerpersymbol 1/Nc=1/3mW (b) Fig.5. (a)Eigenvalues ofQfordifferentsymbolsand(b)theupperbound V. NUMERICAL SIMULATIONS onthesymbolerrorprobability forvarying D andnoiselevels. The proposed system is simulated for two-user MAC scheme as a proof of concept. It is quite easy to extend to crowdedIoTnetworks.Thereceiverischosentohaveasingle where lc ≡ 2πr [15]. The capacitance is Cs =1/(ω02L)≈ coiltosimplifytheanalysis.Theorientationsofthetransmitter 300nFor3nFforf0 =1MHzor10MHz,respectively,with coils are chosen in parallel with Rx with equal distances to f 10MHztoreduceparasiticeffects[3],[7].Noisespectral ≤ simplify the analysis as shown in Fig. 4(a). The positions of density is simulated for N0 = 1 105 109 NTh to analyze × the coils are represented in Cartesian coordinate system as worsecasesthanthethermalno(cid:2)isewhereN(cid:3)Th =4κBTrR ≈ p = 0 0 D T and p = 0 0 D T. Their equal normal 10−21 W/Hz, κB = 1.38 10−23 J/K and Tr = 300 K [3], 1 2 × vector(cid:2)is n˜ −= [0(cid:3) 0 1]. N and(cid:2) T are(cid:3)set to three and five, [7].TotaltransmitpowerforNc symbolsissetto1mWinlow powerregimewithanaverageP =1/N (mW)formassive respectively.Modulationtopologyforasinglecoilisassumed T c utilization in everyday objects. Mutual inductance calculation tobeoneoftwoschemes,i.e.,eitherthecoilswiththeindices is based on the general model in [3] and [16] including 1,2,3,4,5 or 1,3,7,8,9 areactiveatanytime.Therefore, { } { } position, orientation and size dependency. The demodulation thereis a totalof threedifferentnetworktopologysymbolsas mechanismswithandwithoutanytrainingaresimulatednext. showninFig.4(b).Singlechannelusetransmitslog 3bitsof 2 total MAC data. N can be easily increased, however, three c A. Pilot Training symbols better clarify the proposed system. The simulation parameters are shown in Table II. The coil The case with pilot training is simulated for f = 1 MHz, radius and the boundary distance between the coils are set to (ω ,ω ) = (0.99,1.01) ω , N = 32 and N = 1 due to l h 0 ω S × r =1 cm and ∆ =1 mm, respectively, so that a transceiver perfectlyknownsymbolfamily allowingto decrease N . The c S with a planar grid of coils has the area of 6.2 6.2 cm2 received power for each symbol is equalized to simplify the × to be easily attached to daily objects as shown in Fig. 1(a). system and the analysis where unequal transmit powers are The number of turns is one for simplicity. R is calculated used for each symbol with the average P = 1/3 mW and T by using πrρ/(2w δ) for square cross-section copper wire distributed equally among N frequencies. c ω of width and height w = 2 mm compatible with [3] and Eigenvalues are plotted for varying distances reaching to c [7] where the resistivity is ρ = 1.7210−8Ωm, skin depth is 50 meters as shown in Fig. 5(a). They are robust with ≈ δ = ρ/(πµf ) and µ = 4π10−7 H/m. L is equal to respect to the distance which is utilized in the next section 0 µl /p(2π) ln l /w + 0.5 + 0.447w /l 84 nH to demodulate by directly using the eigenvalues. The upper c c c c c ≈ (cid:0) (cid:1)(cid:0) (cid:0) (cid:1) (cid:1) VI. OPENISSUESAND FUTURE WORK Thefollowingtopicspromisetofurtherimprovethesystem: • The computational complexity exponentially dependent onK duetomulti-userMLdetectionneedstobereduced withmodernmethods,e.g.,spheredecoding,byanalyzing † the structures of η , Υ , ~q and Υ . k k 0,k k • Analysis and experimentation of performance for vary- ing the following: types and numbers of constellations; numbers, positions and orientations of coil grids; coil and grid dimensions;frequencysets, voltage sourcesand synchronization;passive relaying and waveguides. • Theeffectsofintrudercoilsduetochangeinthenetwork topology need to be analyzed in more detail and a Fig. 6. Monte Carlo simulation of the probability of symbol error for protocol needs to be developed to detect the changes in demodulating withoutanypilottraining forvaryingD andnoiselevels. the received data and the consumed powers in the coils. boundontheprobabilityofsymbolerrorisshowninFig.5(b) VII. CONCLUSION where it drops to 10−8 at several meter distances including high-noiseregimesallowing for reliable IoT communications. In this paper, novel modulation and demodulation mech- It is possible to realize a LAN having the range of tens of anisms are proposed with energy efficient, low cost and low meterswithlowcost,planarandsmallcoilswiththepotential hardwarecomplexitymechanismsforSWIPTinawirelessMI to tune system performance by changing system parameters. MAC network. The network topology of power transmitting coilsischangedtomodulateinformationsymbolsandpractical demodulation schemes are presented. The designed scheme B. Without Pilot Training is numerically simulated for two-user MAC topology with The small variations of eigenvalues shown in Fig. 5(a) small size coils suitable to attach to daily objects. The spatial for varying distances are utilized to practically demodulate modulationmethod providesreliable LAN performanceswith without any training or synchronization among the transmit- tens of meters of ranges and low power operation of mW ters. The case is simulated for f = 10 MHz, (ω ,ω ) = leveltransmitpowers.SWIPT obtainedwith MICnetworksis l h (0.92,1.08) ω and N = N = 24. P is distributed future promising for IoT applications and MACs with energy 0 S ω T × amongN N frequenciesequallywhileequalizingonlythe efficient, simple and low cost solutions. S ω × transmitpowerforeachsymbolwithoutanyreceiverfeedback. The calculation of error probability is realized through APPENDIX A MonteCarlosimulationswith4 106 trialsduetocomplexity × of obtaining explicit expressions. Therefore, the only errors M is equal to ωM + RI + Z I +I/(ωC). It is ω L L higher than 2.5 10−6 are visible in the simulation results. It converted to (ω)(M + E ) where E = RI/(ω) + × ω ω isobservedinFig.6thattheproposedsystemachievesreliable Z I /(ω) I/(ω2C). Then, M−1 is given by (M + communication in several meters without any pilot training EL)L−1/(ω)−= (I + E−1M)−1E−ω1/(ω). Furthermore, ω ω ω or receiver feedback. In comparison with far field RF based E−1 is simplified as α I+β I . The inverse Γ =M−1 is ω ω ω L ω ω SWIPT systems, far field directive power beaming achieves found by inserting the simplification of E−1 as follows: ω several meters of ranges for indoor and outdoor applications achieving to transmit several mWs of power [17]. However, Γ =1 (α M +I + β I M)−1G−1 ω ω ω L ω high efficiency rectifiers require input powers between 0.5 - =2 (α QΛQH +I + β I QΛQH)−1G−1 5 mW with much higher transmission powers although the ω ω L ω −1 radiative RF range is not limited. On the other hand, it is =3 Q α Λ + I + β Q−1I QΛ QHG−1 possibletoachieveMISWIPTwithseveralmWstransmission (cid:18) ω ω L (cid:19) ω power with the possibility to tune and further improve the NR −1 range performance by optimizing coil dimensions, the num- =4 Q α Λ + I + β ~q ~qHΛ QHG−1 (9) (cid:18) ω ω i i (cid:19) ω ber of coils, constellations and operating frequency, and by Xi=1 utilizing passive waveguides or relays, e.g., enabling ranges reachingkilometerswithlargerradiuscoils[7].Theproblems where G−1 (ω)−1E−1 and ~qH is the ith row of Q. The ω ≡ ω i 3 4 of synchronization, channel access and resource sharing are simplification while passing from = to = is realized by using solved by utilizing time-frequency resources together by all the equality I Q = Q NR ~q ~qH which can be obtained L i=1 i i the transmitters and by only changing the network topology. by using simple matrix cPalculations. APPENDIX B [6] R. Zhang, L. L. Yang, L. Hanzo, “Energy Pattern Aided Simultaneous Wireless Information and Power Transfer,” IEEE Journal on Selected The expression Γ V to obtain the received current is ω ω AreasinCommunications, vol.33,no.8,pp.1492–1504,2015. simplified by using (9) as follows: [7] B.Gulbahar and O.B. Akan, “A Communication Theoretical Modeling andAnalysisofUnderwaterMagneto-inductiveWirelessChannels,”IEEE NR Trans.onWirelessCommunications,vol.11,no.9,pp.3326–3334,2012. Γ V =1 Q Θ + β Ψ −1QHG−1V [8] S. Kisseleff, I. F. Akyildiz and W. Gerstacker, “Beamforming for Mag- ω ω ω,k ω i,k ω ω netic Induction based Wireless Power Transfer Systems with Multiple (cid:0) Xi=1 (cid:1) Receivers,” arXivpreprintarXiv:1508.02566,2015. =2 1 Q Θω,k + βω NR Ψi,k −1QH [9] mS.uLnii,caYt.ioSnunfoarndWWire.lSeshsi,U“CnadpeargcritoyunodfMSeangsnoerticN-Ientwduocrtkios,n”MInItMernOatCioonma-l ω (cid:0) Xi=1 (cid:1) JournalofDistributedSensorNetworks, Article ID426324,2015. [10] B.Gulbahar,“TheoreticalAnalysisofMagneto-Inductive THzWireless (α I+β I )V ω ω L ω Communications and Power Transfer with Multi-layer Graphene Nano- =3 αωQ Θ + β NR Ψ −1QHV cMouillsti,”-SctoaleapCpoeamrmiunniIcEaEtiEonsT,rapnps.a1c–ti1o,n2s0o1n7.Molecular, Biological, and ω,k ω i,k ω ω [11] L.Kisong,C.Dong-Ho,“SimultaneousInformationandPowerTransfer (cid:0) Xi=1 (cid:1) UsingMagneticResonance,” ETRIJournal,vol.36,no.5,pp.808–818, =4 αωQ Θ + β NR Ψ −1 [12]2H01.4N. guyen, J. I. Agbinya and J. Devlin, “FPGA-Based Implementa- ω,k ω i,k ω (cid:0) Xi=1 (cid:1) tion of Multiple Modes in Near Field Inductive Communication Using Frequency Splitting and MIMO Configuration,” IEEE Transactions on NTot Circuits andSystems I,vol.62,no.1,pp.302–310, 2015. ~qj,kV(j,ω) (10) [13] N.Ahmedetal.,“AMulti-coilMagneto-InductiveTransceiverforLow- j=XT+1 costWirelessSensorNetworks,”IEEEUnderwaterCommunicationsand Networking (UComms),pp.1–5,2014. where it is assumed that kth symbol is transmitted, V (t)= [14] J.G.Proakis,“Digitalcommunications,”McGraw-Hill,NewYork,1995. ω V(t,ω) is the oscillating voltage for the tth coil at ω for t [15] E.B.RosaandF.W.Grover,“FormulasandTablesfortheCalculation ∈ ofMutualandSelf-Inductance,”USDepartmentofCommerceandLabor, 3 [T +1,N ] and zero otherwise, the equality = is obtained BureauofStandards, 1912. Tot by using I V = 0, and QHV = NTot ~q V(j,ω) [16] S.Babicetal.,“MutualInductance Calculation Between Circular Fila- L ω ω j=T+1 j,k ≡ mentsArbitrarilyPositionedinSpace:Alternative toGrover’sFormula,” ~q is utilized in the equality =4. TPhen, the current at ith IEEETransactions onMagnetics, vol.46,no.9,pp.3591-3600, 2010. Σ,k receiver coil is found by replacing Q with ~qH . [17] I. Krikidis et al., “Simultaneous Wireless Information and Power i,k Transfer in Modern Communication Systems”, IEEE Communications Magazine, vol.5,no.11,pp.104–110,2014. APPENDIX C I (i) is calculated by using (10) as follows: ω I (i) =1 αω ~qH Θ + β NR Ψ −1~q ω ω i,k ω,k ω i,k Σ,k (cid:0) Xi=1 (cid:1) =2 αω ~qH I + β Θ−1 A −1Θ−1 ~q (11) ω i,k ω ω,k k ω,k Σ,k (cid:0) (cid:1) where A = NR Ψ = NR ~q ~qH Λ . If N = 1, k i=1 i,k i=1 i,k i,k k R then the pertuPrbation equalitPy, i.e., (I + a~x~yH)−1 = I + b~x~yH with b = a/(1 + a~yH~x), is utilized where ~x Θ−1~q , ~y Λ−H~q and a β . Then, (7) is easil≡y ω,k 1,k ≡ k 1,k ≡ ω obtained. ACKNOWLEDGMENT This work is supported by Vestel Electronics Inc., Manisa, 45030 Turkey. REFERENCES [1] N.A.Johansson,etal.,“RadioAccessforUltra-reliableandLow-latency 5GCommunications,”IEEEInternationalConferenceonCommunication Workshop(ICCW), pp.1184–1189, 2015. [2] S.Bi,C.K.Ho,R.Zhang,,“Wireless PoweredCommunication: Oppor- tunitiesandChallenges,”IEEECommunicationsMagazine,vol.53,no.4, pp.117–125, 2015. [3] B.Gulbahar,“ACommunicationTheoreticalAnalysisofMultiple-access Channel Capacity in Magneto-inductive Wireless Networks,” submitted forpublication, 2016. [4] P. Grover, A. Sahai, “Shannon meets Tesla: Wireless Information and PowerTransfer,”ISIT,pp.2363–2367, 2010. [5] L.R.Varshney, “Transporting Information and EnergySimultaneously,” IEEEInt.Symposium onInformation Theory,pp.1612–1616, 2008.