Aalbog Univesitet MIMO OTA Testing in Small Multi-Pobe Anechoic Chambe Setups Lloente, Ines Caton; Fan, Wei; Pedesen, Get F. Published in: I E E E Antennas and Wieless Popagation Lettes DOI (link to publication fom Publishe):.9/LAWP.25.2497542 Publication date: 26 Document Vesion Accepted autho manuscipt, pee eviewed vesion Link to publication fom Aalbog Univesity Citation fo published vesion (APA): Lloente, I. C., Fan, W., & Pedesen, G. F. (26). MIMO OTA Testing in Small Multi-Pobe Anechoic Chambe Setups. I E E E Antennas and Wieless Popagation Lettes, 5, 67-7. DOI:.9/LAWP.25.2497542 Geneal ights Copyight and moal ights fo the publications made accessible in the public potal ae etained by the authos and/o othe copyight ownes and it is a condition of accessing publications that uses ecognise and abide by the legal equiements associated with these ights.? Uses may download and pint one copy of any publication fom the public potal fo the pupose of pivate study o eseach.? You may not futhe distibute the mateial o use it fo any pofit-making activity o commecial gain? You may feely distibute the URL identifying the publication in the public potal? Take down policy If you believe that this document beaches copyight please contact us at vbn@aub.aau.dk poviding details, and we will emove access to the wok immediately and investigate you claim. Downloaded fom vbn.aau.dk on: septembe 6, 28
MIMO OTA Testing in Small Multi-Pobe Anechoic Chambe Setups Inés Catón Lloente, Wei Fan, Get F. Pedesen Abstact Ove the Ai (OTA) testing of MIMO capable teminals is often pefomed in lage anechoic chambes, whee plana waves impinging the test aea ae assumed. Futhemoe, eflections fom the chambe, and pobe coupling ae often consideed negligible due to the lage dimensions of the chambe. This pape investigates the feasibility of educing the physical dimension of 2D multi-pobe anechoic chambe setups fo MIMO OTA testing, with the pupose of educing the cost and space of the setup. In the pape, a channel emulation algoithm and chambe compensation technique ae poposed fo MIMO OTA testing in small anechoic chambes. The pefomance deteioation in a small anechoic chambe, i.e., with a ing adius of.5 m, is demonstated via simulations. Index Tems Channel Emulation, MIMO OTA testing, small multi-pobe anechoic chambe setup, test aea sampling. I. INTRODUCTION THE pupose of Ove the Ai (OTA) testing is to assess the pefomance of wieless devices including antenna pefomance. MIMO capable devices should be tested unde ealistic channel conditions, so that its tue pefomance is assessed. The multi-pobe anechoic chambe (MPAC) method is a pomising solution to achieve this. In this method, a numbe of souce antennas, efeed to as pobe antennas, ae placed inside an anechoic chambe suounding the Device Unde Test (DUT). Though the use of a channel emulato connected to the pobe antennas, vaious spatial channels can be ceated []. The main disadvantage of the MPAC method is its cost. A sufficient numbe of pobe antennas is equied to accuately emulate the channel, which leads to costly designs [2]. In addition, the adius of the ing whee the pobes ae placed is typically assumed to be sufficiently lage so that waves adiated fom the pobe antennas ae plana in the test aea. Typical ing adius in pactical OTA setups epoted in the liteatue is, fo example, 2 m in Aalbog Univesity and ETS-Lindgen setups, to name a few [3]. A lage multi-pobe setup and consequently a lage anechoic chambe ae often cost-pohibitive. Thee is an inceasing inteest fom the industy to have a cost-effective MPAC setup by building a space-saving, flexible and potable system. An attempt in this diection is the commecial solution in [4] with a ing adius of.2 m. A compact and potable system is advantageous because laboatoy space is often valuable and limited. Methods to educe the physical dimensions of MPAC setups used fo MIMO OTA testing while still accuately emulating desied adio channels ae highly desiable. As the size of the setup is educed, the waves adiated fom the pobe antennas become spheical. Futhemoe, eflections and pobe coupling might become a concen. The measuement uncetainty inceases with pobe coupling as demonstated in [5]. Results in [6] indicate that eflections fom the chambe significantly affect the quality of the quiet zone in small setups. This pape investigates such effects on the channel emulation accuacy in small MPAC setups. We efe to small MPAC setups as those whee the adiated waves cannot be consideed plana. One of the majo challenges in OTA testing is to emulate the desied envionment by contolling the signals adiated fom the pobes. The Pefaded Signals Synthesis (PFS) technique is one of the channel emulation techniques that has gained populaity in commecial poducts []. The PFS technique aims at epoducing Geometically Based Stochastic Channel (GBSC) models, by tansmitting pefaded signals with specific powe weights fom the pobes []. Fo the PFS technique, it is typically assumed that the signals have plana wavefont. Howeve, this assumption might not be valid in small MPAC setups. Pape [7] povides theoetical eos due to the poximity between the DUT and the pobes, yet channel emulation was not addessed. In this pape, we evise the PFS technique and popose a new way to sample the test zone appopiate fo chambes of any size. The main contibution of this pape is to study the feasibility of emulating GBSC models using the PFS method in small MPAC setups fo MIMO OTA testing. To enable this, we have investigated diffeent aspects: A method fo channel emulation in small MPAC setups using the PFS technique is poposed in Section II. This includes novel methods to sample the test aea that take into account the spheical wave effect and ae suitable fo chambes of abitay sizes. A chambe compensation technique is poposed to minimize eflections and pobe coupling in MPAC setups. Channel emulation accuacy deteioation due to spheical waves, eflections and coupling in small MPACs ae shown in Section IV, along with simulations validating the pevious contibutions fo diffeent channel models. II. CHANNEL EMULATION IN SMALL MPAC SETUPS The main contibutions of this pat lies in two aspects: an extension of the PFS technique and a novel test aea sampling method to make the PFS technique suitable fo chambes of abitay size. GBSC models specify a continuous Powe Angula Spectum (PAS) at the eceive side. The focus of the PFS technique is to econstuct the taget PAS with a limited numbe of pobes [8]. A. Taget Spatial Coelation The taget spatial coelation ρ between a pai of antennas depends on the taget continuous PAS and complex pattens of
2 Opposite points Deteministic gid Random paiing Fig.. Wave geneated by the k-th pobe antenna impinging antennas u and v, fo plana wave assumption (left), and spheical wave (ight). The phase diffeence between u and v depends on the path length diffeence d. the antennas []. Since the pattens of the DUT antennas ae unknown, they ae often assumed to be omnidiectional with a phase diffeence dependent on the antenna sepaation [], [8]. The taget coelation is: ρ = π π exp (j2πd cos(α)/) P (φ) dφ () whee is the caie wavelength, d, α ae shown in Fig., and P (φ) is the taget PAS, composed of ideal plane waves []. B. Emulated Spatial Coelation The emulated spatial coelation ˆρ geneated by a limited numbe of pobe antennas K, fo ideal MPAC setups is []: ˆρ ideal = K g k exp (j2πd cos(α)/) (2) k= whee g k is the powe weight fo the k-th pobe antenna. The emulated coelation fo small chambes should include the phase and gain eos poduced by spheical waves [7]: K k= ˆρ small = g k F k,u F k,v exp(j2π(d k,u d k,v )/) K k= F k,u 2 g k K (3) k= F k,v 2 g k whee F k,u, F k,v ae the path loss tems fom pobe k to antennas u, v, and d k,u, d k,v ae shown in Fig.. Note that (2) is a special case of (3) with the pobes placed sufficiently fa. C. Objective Function The objective is to obtain the pobe weights g = {g k }, g k [, ] with k =,... K, to minimize the eo between the taget spatial coelation ρ = {ρ,, ρ M } and the emulated spatial coelation ˆρ = {ˆρ,, ˆρ M } ove M antenna pais [8]: min g ρ ˆρ small (g) 2 (4) whee g is obtained by applying convex optimization to (4) [8]. D. Test Aea Sampling In ideal MPAC setups, M antenna pais ae selected to sample the test aea as opposite points on cicles of diffeent adius [8] as shown in Fig. 2 (left). Howeve, the selected antenna pais ae always symmetic with espect to the cente, which is not epesentative of the test aea samples in small setups, as demonstated in Section IV. Instead, the test aea should be sampled by pais of antennas that epesent all - - - - - - Fig. 2. Test aea sampling methods. The cicles epesent the position of the antennas, and the lines show how these antennas would be paied. Note that only few antenna pais ae shown fo illustation pupose. possible locations of the DUT antennas, so that spheical wave effects ae taken into account. Two sampling methods ae poposed in this pape, namely andom paiing and deteministic gid paiing, illustated in Fig. 2. The fome consists on paiing antennas andomly placed inside the test aea. In the late, the antennas ae located on a squae gid coveing the test aea and combinations of any two samples ae paied. III. CHAMBER COMPENSATION TECHNIQUE In this section, we conside the case whee the system is no longe ideal, i.e., eflections and coupling between the pobes affect channel emulation. In this case, the field within the test aea includes both the desied signals fom the pobes and the undesied signals caused by eflections and coupling between the pobes. By identifying the stength and diection of the undesied paths, we can build a chambe model that can be pe-compensated to educe undesied signals in the channel emulation stage. We assume that the dominant souces of eo ae eflections and coupling between the pobes, wheeas coupling between the pobes and the DUT is consideed negligible. The poposed technique is inspied by the wok in [9], whee the goal was to compensate nea-field effects and scatteing contibutions fom neighboing pobes. A. Chambe Model An antenna with an omnidiectional patten, e.g., a calibation dipole, can be used to ceate the chambe model. The calibation antenna is swept ove P positions on a cicle of adius equal to the taget test aea adius, as shown in Fig. 3, to identify the incoming diection of undesied and desied paths. The minimum numbe of calibating positions equied is P = K. The S 2 is measued using a netwok analyze fo each antenna position with each pobe antenna active at a time, leading to the chambe model S which is a complex matix of size P K. Note that the eflection coefficient is fequency dependent. Theefoe, the chambe model needs to be measued fo each fequency band to be tested. B. Compensation Technique Knowing the geometic position of the pobe antennas and the calibation antenna, the taget field T C P K fo each active pobe antenna can be calculated. The goal is to calculate the weights D C K K that compensate the undesied paths in S, while maintaining the desied (taget) ones: S D = T (5)
3 Channel Emulato h(t) h2(t) hk(t) Base Station Emulato Chambe compensation D h(t) h2(t) hk(t) Anechoic Chambe y -.5 DUT position Pobe antennas Calibation antenna positions Fig. 3. MPAC setup including chambe compensation. K pobe antennas ae placed on a ing of adius R, suounding the test aea of adius. The DUT would be placed in the cente of the chambe since its antennas must be within the test aea duing pefomance testing. A eflection poduced by the ing, incoming fom in between two pobes due to pobe being active is shown. Using the Mooe-Penose pseudoinvese, (5) is solved as: D = (S H S) S H T (6) whee ( ) H is the hemitian tanspose. Note that both the taget field and the measued field include the spheical wave effect. The pocedue would minimize eflections and coupling on the azimuth plane only. Reflections fom othe elevation angles ae assumed to be negligible, as absobes would be placed on the walls in pactical setups. Fig. 3 shows the idea of the chambe compensation in a MPAC setup. Fo a single cluste, the spatial channel chaacteistics ae achieved by weighting the coefficients with the optimized weights g []. To compensate fo undesied effects in the chambe, we multiply the channel coefficients h k (t) by matix D, obtaining the compensated coefficients ĥk(t): ĥ k (t) = K D k,n h n (t) (7) n= whee D k,n coesponds to the element in ow k and column n of matix D. The compensated channel coefficients ae a linea combination of the oiginal ones. Fo channel models composed by multiple clustes, each cluste is emulated individually and independently [8]. The chambe compensation technique is applied to each cluste independently as well. IV. SIMULATION RESULTS In the following esults, a small MPAC setup with ing adius.5 m and a test aea of diamete.2 m is consideed. A. Validation of Test Aea Sampling Methods We use the deviation between the taget spatial coelation calculated fom an ideal PAS, i.e., using (), and the simulated coelation fo the same PAS in a small chambe. The deviation depends on the location of the pai antennas and the taget PAS [7]. Table I shows the deviation fo fou epesentative PASs. Note that the eo is exclusively caused by spheical wave effect. The method poposed in [8] fails to epesent the actual eo ove the test aea. On the othe hand, the othe two methods poposed in this pape ae moe suitable. Both 3 2 x K RMS coelaton eo TABLE I RMS CORRELATION ERROR FOR DIFFERENT SAMPLING METHODS. Opposite points Random on a cicle paiing Gid sampling Unifom PAS.7.6.56 Laplacian (AS = 35 ).7.7.7 SCME-Umi.9.72.65 SCEM-Uma.8.66.6.2.5..5 RMS coelation eo vs ing adius R.85.64 K = @ 75MHz K = @ 2655MHz.7.3.5.5 2 2.5 R Fig. 4. RMS eo fo the SCME Uban mico model as a function of R. Deviation values fo R=.5 m and R=2.5 m ae maked. methods achieve simila esults, yet the gid paiing has the advantage of being deteministic. B. Spheical Wave Effect Fig. 4 shows the RMS deviation as a function of the ing adius R fo two epesentative fequencies: 75 MHz and 2655 MHz, cental fequencies of LTE bands 3 and 7. The eo is caused only by the spheical wave effect, i.e., K =. The emulation eo due to a limited numbe of pobes is not pesent. The eo poduced by the spheical wave effect deceases as R inceases, as expected. The eo fo R =.5 m is below. fo both fequencies. Fig. 5 demonstates the impact of using a small anechoic chambe and limited numbe of pobes on channel emulation accuacy. The cuve labeled as K =, R =.5 m epesents the eo due to the spheical wave effect. On the othe hand, the cuves labeled with R = show the emulation eo due to the limited numbe of pobes in an ideal setup. As we can see, the impovement of using a lage anechoic chambe geneally diminishes as fequency inceases fo K = 8 and K = 6. This is because the test aea inceases with espect to the wavelength, causing the emulation accuacy to deteioate egadless of the chambe size [2]. With K = 8, the emulation accuacy in tems of RMS eo in an ideal setup (R = ) impoves aound.64 at 75 MHz and. at 2655 MHz, compaed to esults in a setup with R =.5 m. With K = 6, the impovement is.64 at 75 MHz and.8 at 2655 MHz. C. Validation of the Chambe Compensation Technique We will use the example shown in Fig. 3, whee an undesied eflection exists when pobe is active. Fig. 6 shows the amplitude of the taget field T, distoted field S, and compensated field fo only pobe active. The taget field has spheical wave font due to the small size of the chambe. Due to the eflection, the field within the test aea is distoted,
4 RMS coelation eo.4.3.2. RMS coelation eo vs fequency f c K =, R =.5 m K = 6, R =.5 m K = 8, R =.5 m K = 6, R = K = 8, R =.64..8.5 2 2.5 3 3.5 4 f c [GHz] Fig. 5. RMS eo fo the SCME Uban mico channel model as a function of caie fequency. Deviation between ideal and small setups ae maked at 75 MHz and 2655 MHz. - Taget - - Distoted - - Compensated - Fig. 6. Compensation of a static wave. White cicle epesents the test aea. R =.5 m, =. m, K = 6, f = 2 GHz. SCF.5 Spatial coelation. f c = 2 GHz Taget, K =, R = Small chambe, K =, R =.5 m Compensated. K = 6, R =.5 m Not compensated, K = 6, R =.5 m.5.5 d [] Fig. 7. Spatial Coelation Function (SCF) fo the SCME Uban mico model, DUT placed along the y axis. i.e., thee is a fading patten ove space caused by the coheent summation of the taget field and the eflection. The distotion caused by the eflection is compensated inside the test aea using the poposed technique as shown in Fig. 6. Once poven that the field within the test aea can be compensated fo a simple case, we calculate the channel coefficients fo the SCME uban mico model. The spatial coelation emulated using the oiginal channel coefficients and the compensated ones ae calculated, and shown in Fig. 7. The taget spatial coelation is calculated as in () using plana wave assumption. If the compensation technique is not applied, the eflection distots the field in the test aea and deteioates the emulated spatial coelation. On the othe hand, applying the compensation technique minimizes these effects and educes the deviation consideably..5 V. CONCLUSION AND FUTURE WORK In this pape, we have investigated the feasibility of channel emulation in a small MPAC setup with the pupose of deceasing the cost of such setups and saving space. A OTA ing with adius R =.5 m has been consideed in the simulations, which is less than half the size of othe epoted setups in the liteatue. Simulation esults demonstated that the eo caused by the spheical wave effect is slightly wose fo a chambe with R =.5 m compaed to that with R = 2.5 m. Moeove, the emulation eo caused by the use of a limited numbe of pobes becomes dominant as the fequency inceases, egadless of the ing size. Finally, we demonstated that the poposed compensation technique can effectively minimize undesied effects such as eflections o coupling that might exist in small anechoic chambes. A spatial coelation eo values below. is achieved using K = 6 pobes when a stong eflection exists. Some logic extensions could follow the wok pesented in this pape. The compensation technique was poposed fo a 2D setup, yet in pinciple it would be applicable to a 3D setup as well. Thoughout this pape, we have consideed a RMS deviation up to. in the spatial coelation as a measue. Howeve, it would be of inteest to see the effect of the chambe size on othe paametes, e.g., channel capacity. Futhemoe, many pactical aspects have been left fo futhe study, e.g., design of the pobe antennas, chambe design, measuement uncetainty levels in small MPAC setups, etc. ACKNOWLEDGMENT This wok has been suppoted by the Danish High Technology Foundation via the VIRTUOSO poject. REFERENCES [] P. Kyösti, T. Jämsä, and J.-P. Nuutinen, Channel modelling fo multipobe ove-the-ai MIMO testing, Intenational Jounal of Antennas and Popagation, vol. 22, 22. [2] A. Khatun, T. Laitinen, V.-M. Kolmonen, and P. Vainikainen, Dependence of eo level on the numbe of pobes in ove-the-ai multipobe test systems, Intenational Jounal of Antennas and Popagation. [3] W. Fan, I. J. Szini, M. D. Foegelle, J. Ø. Nielsen, and G. F. Pedesen, Measuement uncetainty investigation in the multi-pobe OTA setups, in Antennas and Popagation (EuCAP), 24 8th Euopean Confeence on, Apil 24, pp. 68 72. [4] SATIMO, StaMIMO. [Online]. Available: http://www.satimo.com/content/poducts/stamimo-hu [5] A. Alayon Glazunov, S. Pasad, P. Handel, T. Bolin, and K. Pytz, Impact of scatteing within a multipath simulato antenna aay on the Ricean fading distibution paametes in OTA testing, IEEE Tansactions on Antennas and Popagation, vol. 62, no. 6, pp. 3257 3269, Mach 24. [6] T. Hansen, Plane wave geneation within a small volume of space fo evaluation of wieless devices, U.S. Patent 23/ 52 962, Aug. 8, 22. [7] P. Kyösti and L. Hentilä, Citeia fo physical dimensions of MIMO OTA multi-pobe test setup, in 6th Euopean Confeence on Antennas and Popagation (EuCAP), Mach 22, pp. 255 259. [8] W. Fan, X. Caeño, F. Sun, J. Ø. Nielsen, M. B. Knudsen, and G. F. Pedesen, Emulating spatial chaacteistics of MIMO channel fo OTA testing, IEEE Tansactions on Antennas and Popagation, vol. 6, no. 8, pp. 436 434, May 23. [9] D. Pavegi, T. Laitinen, A. Khatun, V.-M. Kolmonen, and P. Vainikainen, Calibation pocedue fo 2-D MIMO ove-the-ai multi-pobe test system, in 6th Euopean Confeence on Antennas and Popagation (EuCAP), Mach 22, pp. 594 598.