Nesimoglu, T., Beach, MA., MacLeod, JR., & Wa, PA. (00). Mixe lineaisation fo softwae defined adio applications. In ehicula Technology Confeence 00 (TC 00-Fall) (ol. 1, pp. 534-538). Institute of Electical and Electonics Enginees (IEEE). https://doi.og/10.1109/etecf.00.1040401 Pee eviewed vesion Link to published vesion (if available): 10.1109/ETECF.00.1040401 Link to publication ecod in Exploe Bistol each PF-document Univesity of Bistol - Exploe Bistol each Geneal ights This document is made available in accodance with publishe policies. Please cite only the published vesion using the efeence above. Full tems of use ae available: http://www.bistol.ac.uk/pue/about/eb-tems
Mixe Lineaisation fo Softwae efined Radio Applications T. Nesimoglu, M. A. Beach, J. R. MacLeod and P. A. Wa Cente fo Communications each, Univesity of Bistol, Univesity Walk, Bistol BS8 1TR, United Kingdom Email: T.Nesimoglu@Bistol.ac.uk Abstact The inheent non-lineaity of mixes is paticulaly acute in boadband eceive design fo softwae defined adio (SR) applications. Hee, the eceive fontend sees not only the wanted channel, but also a numbe of neaby signals. A conventional mixe will downconvet all of these eceived channels to, thus adding inband intefeence to the wanted channel. In this pape, known mixe lineaisation schemes ae explained and a new technique using fequency etanslation within a lineaised mixe achitectue is pesented. Two-tone-test esults fom a pototype offeed 33dB eduction in the distotion poducts and db suppession of adjacent channel intefeence (ACI) fo a π/4-qpsk modulated caie. A theoetical analysis is also caied out to demonstate the amplitude and phase matching equiements of the technique. Keywods mixe lineaisation, SR eceive fontend, feedfowad, signal cancellation, fequency etanslation. I. INTROUCTION Mixes ae key components in communication systems fo fequency tanslating signals. In eceives, mixes ae used fo downconveting the eceived adio fequency () signal to baseband o to an intemediate fequency () fo futhe pocessing. In tansmittes, mixes upconvet the baseband signal to o to fo tansmitting via an antenna. Howeve, the inheent non-lineaity of mixes in communication systems ceates numeous undesied effects like hamonic (H) and intemodulation distotion (IM), speading the spectum to a wide bandwidth. The H can be filteed out since it appeas at one octave highe fequency than the fundamentals, but this equies appopiate filteing, wheeas IM cannot be emoved by this means and ceates ACI to othe neaby channels as well as co-channel intefeence within the same channel. This is paticulaly acute fo a SR eceive fontend [1-3] which eceives not only the wanted channel, but also a numbe of neaby signals. A non-linea mixe will downconvet all of these eceived channels togethe with the wanted channel to. uing this fequency tanslation pocess inband intefeence will be added to the wanted channel, making it moe difficult o even impossible fo the eceive to coectly detect the infomation. This places demanding filteing equiements [4, 5] on a boadband eceive fontend to eject the out-of-band unwanted channels (blockes) enteing the mixe, hence peventing the geneation of sufficient inband powe to intefee with the wanted signal. Howeve, stong intefeing neaby signals may not be ejected. Also, in a taditional adio application the fequency of tansmission and eception will be fixed and the filte paametes will be set only fo these known fequencies. This is incompatible with the SR concept and filteing out the blockes of multiple standads will be difficult, thus a linea mixe is highly desiable. II. MIXER LINEARISATION SCHEMES Feedfowad has been peviously applied to amplifies [6] yielding significant eduction in IM poducts. Applying feedfowad to mixes necessitates a diffeent appoach, since fequency tanslation occus making the geneation of the efeence and eo signals difficult. Consideing a eceive, the efeence (undistoted clean signal at input) and the output signals whee the IM poducts exist (at ) ae at diffeent fequencies, and thus a diect compaison is not possible. Two feedfowad lineaisation achitectues have been poposed fo mixes within adio eceive applications, whee the efeence signal was fequency tanslated by a backed-off o a satuated seconday mixe. A eview of these achitectues is given below. In ode to show the actual capabilities of each technique, expeimental esults and poblems associated with thei application to a eceive fontend ae also included. A. Feedfowad Mixe In [7] the seconday mixe is backed-off to opeate in its linea egion as shown in Fig. 1. This mixe downconvets the efeence signal to the same as the output of the main mixe ideally undistoted, but if such a mixe wee available, it would no longe be necessay to lineaise mixes. This signal when used as a efeence is only an appoximation to the equied efeence signal. The output of the main mixe, which includes IM is coupled and added in anti-phase to the output of the seconday mixe, thus cancelling the fundamental signals. This eo signal is also an appoximation to the equied eo signal, which is then ecombined at the output This pape is based upon wok pefomed within the famewok of the IST poject IST-1999-1070 TRUST, patly funded by the Euopean Union. 0-7803-7467-3/0/$17.00 00 IEEE. 534
combine to suppess the IM at the output. Accoding to measued esults fom a simila pototype at Univesity of Bistol [8], the disadvantage of this achitectue is that the signal-to-noise atio (SNR) of the efeence path is significantly educed since it is opeating at a much lowe powe. This adds noise to the main path when the eo signal is combined at the output combine to suppess the IM, which would make the eceive less sensitive to the eceived signals and educe the dynamic ange of the eceive. Pactical esults indicated a 5dB eduction in the thid-ode IM (IM3) at the output when the pototype was used as a downconvete. Main Mixe III. FREQUENCY RETRANSLATION Cuent mixe lineaisation techniques ae unable to simultaneously offe a lage dynamic ange, low noise pefomance and suppess IM. Fig. 3 shows a novel eceive achitectue in ode to ovecome these shotcomings [10, 11]. The system will be explained consideing a eceive application downconveting to, but it can also be applied to a tansmitte. e e in ownconveting Mixe A α K e LO LO ~ e _ + B Upconveting Mixe Figue 3: Fequency etanslation technique applied as a eceive. Seconday Mixe (Backed-off) + _ Figue 1: Feedfowad eo coection. B. Single-Loop Feedfowad Mixe The addition of noise to the final output of the pevious configuation (see Fig. 1) was eliminated in the achitectue shown in Fig. [9]. Hee, the seconday mixe is diven with a much highe signal than the main mixe to povide a high level of IM which is an appoximation to the equied eo signal, also poviding a high SNR. This eo signal is amplitude and phase adjusted befoe being added to the final output fo suppessing the IM. High levels of IM3 eduction can be obtained, about 30dB at a single opeating fequency and signal level. This technique offes a low dynamic ange, since pefomance is citically dependent on the amplitude matching of the IM poducts and the mismatching chaacteistics of the two mixes. Main Mixe The distoted output of the downconveting mixe at is coupled. This output is amplified, fequency etanslated back to by the upconveting mixe and filteed to emove unwanted image signals. The clean (efeence) signal at the eceive fontend is also coupled and added in anti-phase to the fequency etanslated sample of the output (which is now at ) with amplitude coection. This pocess cancels the fundamental signals and poduces an eo signal including only the IM poducts. This eo signal is then combined with the eceived input signal with coect amplitude and phase elation to pedistot the satuated downconveting mixe. This povides suppession of the IM without affecting the fundamental signal level, if the signal cancellation is also coectly optimised. The lineaity of the second (upconveting) mixe is not so citical since it is not fequency tanslating the efeence signal, but the aleady distoted output. Hee, signal cancellation is the vecto addition of the efeence and fequency etanslated output, with system pefomance citical on the optimisation of this paamete, in common with othe feedfowad lineaisation achitectues. Also, afte achieving efeence signal cancellation, suppessing the IM poducts equies pedistoting the downconveting mixe with an eo signal. These mechanisms will be descibed in moe detail below. β LO ~ Seconday Mixe (Satuated) Figue : Single-loop feedfowad. A. Signal Cancellation The fist function within fequency etanslation system is the suppession of the efeence signal vecto ( ) at by the anti-phase and equal magnitude combination of fequency etanslated output ( ). This is a vecto addition and it can be mathematically defined. In Fig. 4, two phasos of unequal amplitude and abitay offset fom a pefect anti-phase ae illustated. 0-7803-7467-3/0/$17.00 00 IEEE. 535
cos( α) α sin( α) Pefect magnitude and phase match Figue 4: ecto addition of efeence and fequency etanslated output. Hee, the phase adjustment is caied out on the vecto, which can be esolved into its othogonal components. One of these components is anti-phase and the othe is othogonal to. The anti-phase component, cos(α ) will add to destuctively and achieve the cancellation of efeence signals as shown in (1), whee the is the esulting vecto. = cos( ) (1) α The emaining othogonal vectos and sin(α ) can be expessed using Pythagoas theoem, yielding the final esultant vecto as: = = = + ( sin( α) ) ( cos( α) ) + ( sin( α) ) ( cos( α) ) + ( sin( α) ) Fom Fig. 3 it can be seen that the amplitude adjustment is also caied on the efeence signal, theefoe it can be defined in tems of with a voltage offset of, as = + giving: ( ( + )cos( α) ) + (( + )sin( α ) ) () = (3) Nomalising to yields: 1 1 cos( ) = + 1 sin( ) + + α α (4) and the suppession ( Sup ) efeed to is: = 1 (5) Sup A suppession of unity epesents thee is no output efeence signal magnitude. Equation 5 can be used to calculate the signal suppession atio fo vaying magnitude and phase offsets. Fig. 5 shows the magnitude and phase matching equiements fo diffeent values of suppession fom 50dB to 5dB in incements of 5dB. As the suppession inceases, the pefomance is highly dependent on the magnitude/phase accuacy. Howeve, at lowe degees of signal suppession, vaiation in magnitude/phase mismatch esults in smalle degadation in the signal cancellation pefomance indicated by the widening of the taces. In ode to achieve a signal suppession of 45dB, magnitude and phase match should be bette than 0.05dB and 0.5 espectively. These matching equiements ae simila to a signal cancellation loop within a feedfowad amplifie [1, 13]. Magnitude Mismatch (db) 0.30 0.5 0.0 0.15 0.10 0.05 40dB 35dB 30dB 5dB 45dB 50dB 0 0 0.5 1 1.5.5 3 3.5 Phase Mismatch (egees) Figue 5: Magnitude and phase match equiement fo vaious levels of signal suppession. 1)Pedistotion Afte the signal cancellation of the fundamental signals, suppessing the IM poducts will be achieved by pedistoting the downconveting mixe with the eo signal. Assuming that signal cancellation was successful, the gain and phase accuacy of the eo signal will detemine the IM3 suppession pefomance, which is simila to the matching equiements of the signal cancellation loop, thus it is not epeated. Hee, the system pefomance is citical on the optimisation of the both signal cancellation and eo signal pedistotion loop. Theefoe, in common with othe feedfowad cicuits, this system is equally sensitive to impefections. Fo maintaining the IM suppession in a pactical system, an adaptive contol scheme is necessay in ode to maintain system pefomance with changing cicuit paametes and input signal conditions. B. Implementation and Pactical ults A pototype demonstation system is shown in Fig. 6, whee two passive double-balanced SRA-000 Mini-Cicuits 0-7803-7467-3/0/$17.00 00 IEEE. 536
mixes wee used [14]. The amplifie peceding the downconveting mixe is offeing a high 1dB gaincompession point, thus diving the downconveting mixe to satuation without adding any additional distotion itself. This is to ensue that the technique is coecting the non-lineaity of the mixe and not othe cicuit elements. The eo and amplifies povide sufficient gain at thei opeating fequencies to compensate fo the losses such as coupling, powe splitting/combining, filteing and the convesion losses of the mixes. Futhe, the eo amplifie is also opeating in its linea ange, thus not distoting the eo signal. The amplitude and phase adjustment of signal cancellation and pedistotion loops has been pefomed using voltage vaiable components. intecept (TOI) point of the mixe fom 0.17dBm to 1.16dBm. The signal cancellation loop povides moe than 40dB suppession of fundamentals as shown in Fig. 8. In ode to obtain this eo signal, the amplitude and phase match should be within 0.1dB and 1 (see Fig. 5). Noise powe measuements at the output indicate only 0.dB incease in the noise figue when the lineaisation is applied, which is consideed to be negligible. This illustates that the technique does not degade the noise pefomance of the eceive and by coect choice of components it can be futhe minimised. The same pototype was also tested with a TETRA π/4-qpsk modulated caie again downconveted fom 90MHz to 160MHz, with Fig. 9 showing a db impovement in ACI. Fig. 10 shows the eo signal used fo obtaining this ACI impovement. -10-0 Fequency etanslated output Eo signal Powe (dbm) -10 Figue 6: Plan view of the fequency etanslation pototype. 919.55 919.63 919.7 919.80 919.88 919.97 90.05 90.13 90. 90.30 90.38 90.47 90.55 Fequency (MHz) Figue 8: The signal cancellation pocess with f=100khz. -0 Without technique With technique -0 Without technique With technique Powe (dbm) Powe (dbm) 159.55 159.63 159.7 159.80 159.88 159.97 160.05 160.13 160. 160.30 160.38 160.47 160.55 Fequency (MHz) Figue 7: Measued two-tone-test showing a maximum 33dB IM3 impovement with f=100khz. A two-tone-test was applied at 90MHz with a tone sepaation ( f) of 100kHz to povide a downconveted signal at an of 160MHz. The output of the downconvete with and without the technique applied is given in Fig. 7, indicating an impessive 33dB suppession of IM3. Futhe amplitude and phase adjustment was pefomed to suppess IM3 to the same level of fifth-ode IM, whee the IM3 impovement is 5dB, i.e. the technique has inceased the output thid-ode 159.95 160.00 160.05 Fequency (MHz) Figue 9: Measued π/4-qpsk output spectum showing db ACI suppession. A two-tone-test was also applied at 90MHz with f=500khz and downconveted to an of 160MHz. Afte applying the poposed technique, 19dB IM3 suppession was obtained. Inceasing the fequency sepaation degades the signal cancellation and hence the IM3 suppession. ue to the delay mismatch between the two paths in signal cancellation 0-7803-7467-3/0/$17.00 00 IEEE. 537
loop, it is not possible to maintain the equied 180 phase diffeence fo ideal cancellation at all fequencies. Howeve, it is possible to match the phase at one fequency (in ou pototype this is 90MHz), whee the pefect cancellation will occu. As the signal fequency deviates fom the cente fequency, the cancellation will degade and eoptimisation will be equied to maintain the pefect cancellation. This elationship is measued and shown in Fig. 11. At 90MHz the signal cancellation is about 88dB (Make 1) and at.5mhz offset it educes to 7.8dB (Make ). IM3 by up to 33dB, with aveage suppession of 5dB can be obtained. At this opeating point, the calculations show that the output TOI point of the mixe has been inceased fom -0.17dBm to 1.16dBm. The tests with π/4-qpsk modulated caie has shown db ACI impovement at the output. Theoetical analysis demonstates that fo a high level IM suppession, accuate amplitude/phase match is equied. The futue wok will focus on impoving the lineaisation bandwidth and dynamic ange as well as an adaptive contol scheme fo a pactical application of this technique. Powe (dbm) -0 Fequency etanslated output Eo signal ACKNOWLEGEMENT The authos would like to acknowledge the contibutions of thei colleagues fom Siemens AG, Fance Télécom - CNET, Cente Suisse d Electonique et de Micotechnique S.A., King s College London, Motoola Ltd., Panasonic Euopean Laboatoies GmbH, Telefonica Investigacion Y esaollo S.A. Unipesonal, Toshiba each Euope Ltd., TTI Note S.L., Univesity of Bistol, Univesity of Southampton. 919.95 90.00 90.05 Fequency (MHz) 10 db/ db Figue 10: The signal cancellation pocess with π/4-qpsk modulated caie. CH1 S1 db MAG 10 db/ REF 0 db 10 db 1: -88.30 db 94 90 MHz MHz : -7.81 db 0 db 9.5 MHz START 915 MHz 1 MHz/ STOP 95 MHz Figue 11: The signal cancellation degading due to the delay mismatch as the signal fequency deviates fom the cente fequency. I. 1 CONCLUSION In this pape, peviously investigated mixe lineaisation techniques ae summaised and a new technique is pesented. A hadwae pototype was constucted and the technique was evaluated by pactical means. It povides consideable impovement of mixe non-lineaity without compomising the noise pefomance. Measuements indicate suppession of A F 1 REFERENCES [1] J. Mitola, The Softwae Radio Achitectue, IEEE Communications Magazine, May 95, Pages: 6-38. [] J. M. Peeia, Beyond softwae adio, towads e-configuability acoss the whole system and acoss netwoks, IEEE ehicula Technology Confeence, v 5, 1999, Pages: 815-818. [3] J. R. MacLeod, M. A. Beach, P. A. Wa and T. Nesimoglu, A Softwae efined Radio Receive Test-Bed, IEEE ehicula Technology Confeence, TC 001, Atlantic City, olume: 3, Pages: 1565-1569. [4] J. R. MacLeod, T. Nesimoglu, M. A. Beach and P. A. Wa, Miniatue istibuted Filtes fo Softwae Re-configuable Radio Applications, IST Mobile Summit, June 00, Thessaloniki, Geece, Pages: 159-163. [5] J. R. MacLeod, M. A. Beach, P. A. Wa and T. Nesimoglu, Filte Consideations in the esign of a Softwae-efined Radio, IST Mobile Summit, Septembe 001, Bacelona, Spain, Pages: 363-368. [6] N. M. Pothecay, Feedfowad linea powe amplifies, London, Atech House,1999. [7] T. J. Ellis, Modified feed-fowad technique fo mixe lineaization, IEEE MTT-S Intenational Micowave Symposium igest, v 3, 1998, Pages: 143-146. [8] J. P. McGeehan, M. A. Beach, A. T. Muno, N. Kingswood, R. J. Wilkinson, K. Mois and W. Peasgood, Pogammable Achitectues fo Communications Test, FINAL REPORT: EPSRC Gant Refeence GR/K39707. [9] M. Chongcheawchamnan and I.. Robetson, Lineaized micowave mixe using simplified feedfowad technique, Electonics Lettes, v 35, n 9, 1999, Pages: 74-75. [10] T. Nesimoglu and M. A. Beach, Lineaised mixe using fequency etanslation, UK Patent Application No. 0117801.1, 0th July 001. [11] T. Nesimoglu, M. A. Beach, P. A. Wa and J. R. MacLeod, Lineaised Mixe using Fequency Retanslation, IEE Electonics Lettes, ecembe 001, ol. 37, No. 5, Pages: 1493-1494. [1] R. J. Wilkinson and P. B. Kenington, Specification of eo amplifies fo use in feedfowad tansmittes, Cicuits, evices and Systems, IEE Poceedings-G, olume: 139 Issue: 4, Aug. 199, Pages: 477-480. [13] K. J. Pasons and P. B. Kenington, Effect of delay mismatch on a feedfowad amplifie, IEE Poceedings on Cicuits, evices and Systems, olume: 141 Issue:, Apil 1994, Pages: 140-144. [14] http://www.mini-cicuits.com 0-7803-7467-3/0/$17.00 00 IEEE. 538