ME-SSA: an Advanced Random Access for the Satellite Return Channel

Transcription

1 IEEE ICC 5 SAC - Satellite and Space Counications ME-SSA: an Advanced Rando Access for the Satellite Return Channel G. Gallinaro, N. Alagha, R. De Gaudenzi, K. Kansanen, R. Müller, P. Salvo Rossi Abstract The paper analyzes the perforance of an advanced rando access schee for the return channel of a satellite counication link. The schee is an evolution of the E-SSA schee proposed in [], [] that couples an asynchronous Spread Spectru Aloha access with Successive Interference Cancellation (SIC) at the central Gateway (GW) receiver to increase the channel throughput. Main feature of the proposed schee is the exploitation of an approxiate linear Miniu Mean Square Error (MMSE) detector in place of the conventional Single User Matched Filter (SUMF) detector used in E-SSA. A gain of 5% in ters of spectral efficiency is achieved over E-SSA in ost typical scenarios. Index Ters CDMA, Spread Spectru, Satellite Counications, Up Link, Multiple Access, Rando Access. T I. INTRODUCTION HIS paper investigates the perforances of a new advanced rando access schee specifically designed for the return channel of a satellite counication link. The proposed schee is an evolution of the Enhanced-Spread Spectru Aloha (E-SSA) proposed in [] and []. As the acrony suggests E-SSA iproves the perforance of the conventional Spread Spectru Aloha ([4], [5]) rando access through a process of Successive Interference Cancellation (SIC). It is known that Spread Spectru Aloha (SSA) (assuing a sufficiently high Processing Gain, PG) ay significantly exceed the throughput provided by non-spread Aloha (slotted or unslotted) [5]. An issue with SSA is its sensitivity to signal power unbalance which can significantly reduce its spectral efficiency. The exploitation of interference cancellation in E- SSA actually turns signal power unbalance into an advantage as access efficiency is actually increased by the variation of power in incoing packets. It was actually found [7] that optial perforances of E-SSA are achieved for power distributions of incoing packets which are approxiately unifor when power is easured in logarithic scale (i.e. db). The optial range of packet powers, for a given peak This work was supported by ESA Contract No 48548/3/NL/JK. G. Gallinaro is with Space Engineering S.p.A., Roe, Italy. N. Alagha and R. De Gaudenzi are with the European Space Agency, Noordwijk, The Netherlands. R. Müller is with Friedrich-Alexander-Universität Erlangen-Nürnberg, Gerany. K. Kansanen and P. Salvo Rossi are with NTNU, Trondhei, Norway. E b /N, is also addressed in [7]. The E-SSA access schee is very well suited for supporting a large population of terinals with very bursty traffic (e.g. MM applications, interactive TV, essaging). Deand Assignents Multiple Access (DAMA) strategies are, in fact, not very well suited to cope with such traffic scenario and large size networks due to the large signaling overhead which would result in such cases [5]. The E-SSA rando access has been recently adopted in the S-MIM standard of ETSI for use in the return channel of S-band ultiedia satellites ([8], [9]). The E-SSA schee is particularly attractive since it requires inial processing on the user terinals as all cancellation processing is done at the GW receiver. Also, no network tiing synchronization is required as terinal access to the RF channel is fully asynchronous. Furtherore, no user identification is required prior to user packet detection and decoding. In fact, all users reuse the sae spreading codes. Code collision probability is, in fact, iniized by the use of long spreading codes (not repeating within the packet) and the truly asynchronous nature of user transission. A single spreading code for all users is typically used in E- SSA systes as this reduces the GW receiver coplexity which has only to search for a single preable to detect the presence of packets on air. Given the fact that packet acquisition is likely the ost coputational intensive part of the receiver, the advantage of searching for a single wavefor signature is evident. Current E-SSA rando access adopts a conventional Single User Matched Filter (SUMF) receiver for despreading the received packet. It is well known, however, that a linear MMSE detector can boost the achievable perforances in a CDMA access schee []. Using the MMSE detector in place of the SUMF detector in an E-SSA like rando access has thus the potential of further iproving the spectral efficiency in particular when the packets power unbalance is reduced. On the other hand the MMSE-SIC processing is able to reach the ultiple access channel capacity, as shown in []. Given its use of the MMSE detector we call this new rando access schee ME-SSA (MMSE Enhanced Spread- Spectru Aloha). Incorporating the MMSE detector in an E-SSA like schee is not straightforward. E-SSA is in fact a totally asynchronous syste with packets being randoly transitted. The active transitters are changing continuously. This fact, together /5/$3. 5 IEEE 856

2 IEEE ICC 5 SAC - Satellite and Space Counications with the use of long spreading code sequences and relatively short packets, akes infeasible the use of an adaptive MMSE detector. Since the ipleentation of MMSE through a direct atrix inversion is too cubersoe, the adopted solution is the use of a ultistage detector approxiating the MMSE one ([]-[5]). The reainder of the paper is organized as follows. Section II provides soe detail on the ME-SSA access design as well as soe inforation on the signal design and the trade-offs considered in such design. Section III provides a short description of the considered GW receiver processing. Finally, siulation perforances are addressed in Section IV where perforances of E-SSA and ME-SSA in different conditions are shown and results discussed. II. ME-SSA ACCESS DESIGN A. E-SSA signal design As entioned in the introduction, the ME-SSA is an evolution of the E-SSA rando access. Key design drivers for the E-SSA design were the asynchronous nature of user transission, the lack for user identification (before actual packet decoding), the user terinals low-cost and lowtransitting power requireents and the affordable GW receiver coplexity. The lack of user identification and the need to liit the GW receiver coplexity calls for a signal design based on spread spectru signal with a long spreading code shared by all the users in order to only search for one code in the receiver. The E-SSA wavefor specification was largely based on the 3GPP W-CDMA up-link wavefor ([6]-[8]). The sae FEC schee (turbo code with rate /3) was also retained as well as the presence of a Physical layer Control Channel (PCCH), carrying reference sybols to aid the receiver deodulator and an optional signaling field inforing on the actual carrier forat. The actual traffic data are carried by a Physical Data CHannel (PDCH). Both channels are BPSK odulated in the up-link. Siilarly to 3GPP W-CDMA, the two channels are code ultiplexed (each channel using a different Walsh Hadaard orthogonal code as spreading code) and apped to the I and Q axis of a coplex signal which is in turn scrabled by a coplex long scrabling code (Fig. ). The above approach is justified on the fact that BPSK odulation is actually optiu when a SUMF detector is used at the receiver and ensure robustness to possible channel carrier phase estiation errors. Differently fro the 3GPP W-CDMA the long scrabling code has the sae length as the packet. It never repeats within a packet. The long spreading code approach, in addition to iniizing the collision probability is also optial when used in a FEC coded syste as it randoizes the interference, forcing it to the equivalent of Gaussian noise. B. ME-SSA signal design BPSK odulation is not optial when a linear MMSE detector (or one of its approxiations) is used at the receiver. QPSK odulation appears preferable in such case [9]. In fact, the MMSE detector perforances decrease for higher syste loading []. With QPSK odulation the syste loading is halved with respect to BPSK odulation (assuing the sae FEC code rate for both options) thanks to the signal space diensions doubling. The adoption of QPSK odulation requires a odification of the E-SSA wavefor design. In particular, the code ultiplexing between PCCH and PDCH is not any ore appropriate. Tie doain ultiplexing between the two channels is now adopted as this allows a constant envelope signal (before filtering). A single binary Walsh Hadaard spreading code followed by a coplex long scrabling code is thus adopted (Fig. ) siilarly to the approach chosen for the down-link of the 3GPP W-CDMA. Figure Figure. Spreading and Scrabling strategy in E-SSA S/P I Real Spreading Code Q j Figure. Spreading and Scrabling strategy in ME-SSA III. RECEIVER DESIGN Coplex Scrabling Code I+jQ Different receiver architectures at the GW can be considered for deodulating, decoding and cancelling the incoing packets. In the following we will consider a serial receiver approach as this was ipleented in our software siulator to assess the perforances of both E-SSA and ME- SSA schees (Fig. 3). An alternative receiver architecture is the shared eory ipleentation which is briefly described in []. The two proposed receiver architectures are equivalent as far as perforances are concerned. The serial receiver architecture is actually coposed of a serial chain of all equal odules with each odules coposed of a set of deodulators and decoders for the incoing packets and a set of re-odulators allowing the subsequent cancellation of the S 857

3 IEEE ICC 5 SAC - Satellite and Space Counications 3 Delay Delay Deodulators & Decoders Reodulators - + Deodulators & Decoders Reodulators - + Deodulators & Decoders regenerated signal fro a suitably delayed input signal replica. Such a delayed input signal replica, after cancellation of the decoded packets, is then provided to the next receiver stage for further processing. The nuber of stages in the receiver is thus equivalent to the nuber of SIC iterations in the cancellation process. The above discussed receiver architecture is quite general and applicable for both E-SSA and ME-SSA. Clearly given the different wavefor forats the actual deodulators and re-odulators will differ. However, the ain difference between the receiver for ME-SSA and E-SSA is the spread spectru signal detector used in the deodulator. E-SSA uses a conventional SUMF detector and ME-SSA adopts a detector approxiating MMSE detection. As noted before, a direct MMSE solution coputation is unfeasible for systes with a large nuber of active users. In fact, considering for siplicity of notation a chip synchronous syste (for an asynchronous schee the coplexity would further grow for the need to process interval coprising ultiple sybols) with spreading factor equal to N and with K users on air, the following signal odel can be written: y = Ψ Pb + w () where Ψrepresents the spreading atrix of size N by K and P is K by K a diagonal atrix containing the users powers. The N by N covariance atrix of the theral noise vector w is diagonal with all equal diagonal eleents. We recall that the MMSE atrix M is : H M= R Ψ () with R being the signal plus noise covariance atrix given by: H R = I Ψ PΨ σ w + Stage Decoded Packets Given the previously entioned coplexity of inverting R, a practical solution is instead to approxiate the MMSE detector through a ultistage detector approach ([]-[5]) whose coplexity scales linearly with the nuber of users. The ultistage detector approxiates the inverse of the covariance atrix, R -, by a polynoial expansion in R, i.e.: S k R w k R (3) k= This expansion can be derived applying the Cayley-Hailton theore to the atrix R. The theore states that a square atrix of size K by K whose eigenvalues are Λ={λ, λ, λ K } is a zero of its characteristic polynoial, i.e.: Figure 3. Serial Receiver Architecture K ( R λki) = (4) k= Expanding the above equation we have: S k ckr = (5) k = where the coefficients c k are functions of the eigenvalues Λ. The above equation can be rewritten after ultiplying both sides for R - and solving for R - as: c c K k+ k R R (6) = k= Hence K stages would be sufficient to invert the atrix R. In practice a nuber of stages equal to or 3 can already give a good approxiation of R -. Since the eigenvalues of a large rando atrix only depend on its statistics, the eigenvalue can be coputed off-line thus allowing to derive the coefficients w k also by off-line coputation. For a ultistage detector with S stages (with S< K), optial weighting is discussed in [3]-[5]. It can be shown that the coefficients that iniize the MSE due to the discarded ters in (3), satisfy a set of Yule-Walker equations, i.e.: + σ 3 + σ... = S + σ S 3 + σ + σ + σ w w w + + S where the k are the eigenvalue oents given by: k K k λi K i= = Stage N- Decoded Packets The ultistage detector actually builds an approxiation to R - Ψ H by concatenating S stages with each stage perforing despreading (with a SUMF detector) and then re-spreading of the input signal. These operations are respectively equivalent to ultiply by atrix Ψ H (despreading) and then by atrix Ψ (re-spreading) the input signal. The coefficients w k can be chosen to approxiate the MMSE detector or other detectors (e.g. the decorrelator) Stage N Decoded Packets 858

4 IEEE ICC 5 SAC - Satellite and Space Counications 4 Figure 4. Principle schee of a ultistage detector. However, the optial weighting coefficients ust be calculated. Although the Yule-Walker equations can be solved in quadratic tie (through the Levinson-Durbin recursion) this approach still has excessive coplexity if one considers that eigenvalues have to be estiated for the ethod to be applicable. In general, the coputation of the eigenvalues of a atrix has cubic coplexity. Hence, deriving such optial coefficient can be a proble of siilar coplexity to atrix inversion. Luckily, for sufficiently large Spreading Factor (SF), asyptotic weight values can be coputed off-line. This is because eigenvalue oents do not depend on the actual spreading code on air (at least for sufficiently high spreading factor) but only on syste loading (i.e. nuber of users on air at a given tie), user power distribution and wavefor characteristics (e.g. roll-off). The principle schee of the ultistage detector is suarized in Fig. 4 where it is evident the coposition of each stage by a despread unit (coposed by SUMF detectors) followed by a respreader unit. The output of the various SUMF detectors is then weighted to copute the final despread signals. The resulting ultistage detector coplexity is only arginally higher than that of a SUMF detector. In practice, in an asynchronous rando access environent, coputation of the weights shall be done dynaically as new packets on air are detected or older packets terinate. Still, coplexity of such operations is arginal. Furtherore, E- SSA typically requires a larger nuber of SIC iterations with respect to ME-SSA to achieve the optial perforances particularly in presence of large power unbalance between packets. Finally, we have to stress that the actual coplexity of the receiver in both E-SSA (using SUMF) and ME-SSA (using the ultistage detector) is doinated by the packet acquisition circuitry which is not addressed in this paper (see [] for a discussion of packet acquisition). IV. PERFORMANCE ASSESSMENT Perforances of E-SSA and ME-SSA have been derived using a physical layer siulator also ipleenting the rando access layer by asynchronously generating packets according to soe arbitrary distribution. In this paper we present results where packet distribution is Poisson (exponentially distributed packet inter-arrival tie). For siplicity, in each siulation run, a single packet type (i.e. packet length, spreading factor, and FEC code rate) is considered. Although acquisition has been considered ideal in the following siulation results, a preable of 96 sybols (before spreading) was also included in the siulations and real channel estiation has been considered. A PCCH (control db relative power level was considered for E-SSA. The PCCH was used to carry only pilot sybols for channel estiation purposes. For ME-SSA the tie ultiplexed control channel was also carrying solely pilot sybols. In particular, one sybol out of was associated to the control channel. The overhead for the control channel was thus identical in the perfored siulations for both E-SSA and ME-SSA. However no optiization of the PCCH overhead was done. If not stated otherwise, the nuber of receiver stages used in the siulations is for both E-SSA and ME-SSA, allowing for SIC iterations. For ME-SSA, a three-stage ultistage detector was used instead of the SUMF detector of E-SSA. Fig. 5 shows a perforance coparison between E-SSA and ME-SSA in the case that all packets are received at the sae power. The coparison is done in ters of spectral efficiency and Packet Loss Ratio (). Increasing the syste load results in higher spectral efficiency until a point where the starts to grow very rapidly due to the excessive interference in the syste. Further increase of the syste load will ultiately produce a collapsing of the perforances. As = -3 appears to be a reasonable operating requireent for several application scenarios, we consider as actual spectral efficiency of the syste that one corresponding to such target. The siulations reported in Fig. 5 were done for the sae Es/No after despreading. As ME-SSA uses QPSK odulation instead of the BPSK used in E-SSA whilst the FEC code rate was the sae, the E b /N in ME-SSA was 3 db lower than in E- SSA. Notwithstanding the lower E b /N, fro Fig. 5, it appears that in such unifor packet power case, ME-SSA provide about 5% higher spectral efficiency than E-SSA at the target of -3. Actually, advantage of ME-SSA can be even larger if coparison is done at higher E s /N. In fact, 859

5 IEEE ICC 5 SAC - Satellite and Space Counications E-SSA SF=6 Es/No=6 db Equal Power.E+.E-.E-.E-3.4..E-3...E-4.E Figure 5. Coparison of perforances between E-SSA and db and equal carrier power (SF=6). Packet Length was info bit with 3GPP code rate /3FEC. 96 sybols preable. Ideal acquisition considered. SIC iterations perfored.4. ME-SSA SF=6 Es/No=6 db Equal Power carriers.e+.e-.e- perforance of ME-SSA iproves with the SNR whilst those of E-SSA are alost insensitive (in such unifor packet power case) to the SNR. All incoing packets having the sae power will clearly not happen in practical systes (even on satellite links). Actually [7] has shown that E-SSA perforances are axiized, for a given peak E b /N, if the incoing packet are received at power levels (in db) which are approxiately uniforly distributed. The range of E b /N will thus range fro the peak E b /N down to a iniu value that is larger than the iniu E b /N threshold required for correct packet decoding (for the target ). The argin over the threshold depends on the cancellation efficiency of the SIC process. With ideal cancellation efficiency the optial power range bottos out at the iniu E b /N required to achieve the target. In the S-MIM specification of E-SSA [9] a signaling echanis is foreseen to support an explicit transit packet power randoization aiing at achieving the optial power distribution. In practice, the packet power distribution will never be identical to the optial one. Anyway, coputing the perforances with such optial power distribution will give us an upper liit to the achievable perforances. Fig. 6 shows a further coparison between E- SSA and ME-SSA perforances in presence of packet power randoization (with uniforly distributed packet power). The perforance coparison was done for the sae E b /N (.77 db in both cases). The spreading factor (SF) of ME-SSA was double of that of E-SSA (3 against 6 in the specific case here) as the coparison was done for the sae occupied bandwidth and inforation bit rate. The advantage of ME- SSA is even larger than 5% in such a case. It has to be observed that the required nuber of SIC iterations required to get optial perforances is generally larger in presence of a larger packet power dynaic range. This is particularly true for E-SSA. As stated before in Fig. 6 the continuous curves refer to the perforance achievable with SIC iterations. A single point fro each perforance curve (cross/triangle) has been also re-siulated with a larger nuber of iterations ( for E-SSA and 8 for ME-SSA). Although, the ultistage detector approxiation to an MMSE detector is strictly valid only for very large SF, good perforances are actually achieved with the ultistage detector even for very sall SFs. In this regard, Fig. 7 shows the perforances of and E b /N o =.77 db. For coparison the perforances of E-SSA at the sae E b /N o are also shown. As the coparison is done for the sae bandwidth, a SF= is required for E-SSA. For such extreely sall SF the perforance of E-SSA are particularly degraded (particularly if we consider perforances at a target = -3 ). For ME-SSA there is instead only a very inor decrease of perforance with respect to the case of Fig. 6. Results showed so far assued a-priori knowledge of packets on-air. In practice a preable is used for packet detection. For E-SSA the packet preable is sized in order to have a good detection probability at an E b /(N +I ) such that the decoder can provide, with soe non-negligible probability, a correct decoding. In the S-MIM specs, a 96 sybols preable is defined for such purpose. In ME-SSA, there could be an advantage in also detecting the preable of packets whose E b /(N +I ) is below the threshold for correct decoding, as the ultistage detector requires a good knowledge of the syste load as well as the packet power distribution for optial operation. As a atter of fact, siulation results shown in this paper for E-SSA will not change significantly if also the acquisition process is fully siulated. For ME-SSA, soe ipact is expected when real acquisition is siulated and a longer preable ight be preferable in soe case. V. CONCLUSIONS An evolution of the E-SSA rando access schee exploiting a ultistage detector approxiating the linear MMSE detector has been presented. The new access schee allows a significant iproveent of the achievable throughput aking this rando access schee uch ore appealing for a wide range of services beyond those typically considered as targets for rando access. The new access schee can also be eployed in relatively narrow-band channels as good perforances can be achieved even with very low spreading factor. REFERENCES [] O. del Rio Herrero, R. De Gaudenzi, A high efficiency schee for quasi-real-tie satellite obile essaging systes, in the Proc. of the 86

6 IEEE ICC 5 SAC - Satellite and Space Counications 6 th International Workshop on Signal Processing for Space Counications, SPSC 8, 6-8 Oct. 8, Rhodes Island, Greece. [] O. del Rio Herrero, R. De Gaudenzi, High Efficiency Satellite Multiple Access Schee for Machine-to-Machine Counications, IEEE Trans. on Aerospace Engineering, Oct.. [3] De Gaudenzi, O. Del Rio Herrero, Advances in Rando Access protocols for satellite networks, International Workshop on Satellite and Space Counications, 9, IWSSC 9, Siena, Italy Sept. 9-, 9, pp [4] United States Patent No. 5,537,397, N. Abrason, July 6, 996, Spread Aloha CDMA data counications. [5] O. del Rio Herrero, G. Foti, and G. Gallinaro, Spread-Spectru Techniques for the Provision of Packet Access on the Reverse Link of Next-Generation Broadband Multiedia Satellite Systes, IEEE Journal on Sel. Areas in Co., vol., no. 3, pp , Apr. 4. [6] N. Abrason, The Throughput of Packet Broadcasting Channels, IEEE Trans. Counications, vol. COM-5, no., pp. 7-8, Jan [7] F. Collard and R. De Gaudenzi, On the Optiu Packet Power Distribution for Spread Aloha Packet Detectors with Iterative Successive Interference Cancellation, Subitted to IEEE Trans. on Wireless Counications, Noveber 3. [8] ETSI TS 7- V.. (-), Satellite Earth Stations and Systes; Air Interface for S-band Mobile Interactive Multiedia (S- MIM); Part : General Syste Architecture and Configurations. [9] ETSI TS 7-3 V.. (-), Satellite Earth Stations and Systes; Air Interface for S-band Mobile Interactive Multiedia (S- MIM); Part 3: Physical Layer Specification, Return Link Asynchronous Access. [] S. Verdu, and S. Shaai, Spectral Efficiency of CDMA with Rando Spreading, IEEE Transactions On Inforation Theory, Vol. 45, No., March 999. [] M.K. Varanasi and T. Guess, Optiu Decision Feedback Multiuser Equalization with Successive Decoding Achieves the Total Capacity of the Gaussian Multiple Access Channel, IEEE Proceedings of the 3 st Asiloar Conference on Signals, Systes and Coputers, Vol., pp , -5 Noveber 997, Pacific Grove, USA. [] S. Moshavi, E. G. Kanterakis, and D. L. Schilling, Multistage linear receivers for DS-CDMA systes, Int. J. Wireless Infor. Networks, vol. 3, no., pp. 7, Jan [3] R.R. Müller and S. Verdù, Design and Analysis of Low-Coplexity Interference Mitigation on Vector Channels, IEEE Journal on Selected Areas in Counications, vol. 9, Issue 8, pp , Aug.. [4] L. Cottatelluci, M. Debbah and R.R. Müller, Asyptotic Design and Analysis of Multistage Detectors for Asynchronous CDMA Systes, IEEE International Syposiu on Inforation Theory, 7 June - July 4, Chicago, USA. [5] L. Cottatellucci and R.R. Müller, A Systeatic Approach to Multistage Detectors in Multipath Fading Channels, IEEE Transactions on Inforation Theory, Vol. 5, Issue 9, pp , Sept. 5. [6] ETSI TS 5 V3.. (-9): Universal Mobile Telecounications Syste (UMTS). Physical channels and apping of transport channels onto physical channels (FDD). [7] ETSI TS 5 3 V3.. (-9): Universal Mobile Telecounications Syste (UMTS). Multiplexing and channel coding (FDD). [8] ETSI TS 5 3 V3.9. (3-): Universal Mobile Telecounications Syste (UMTS). Spreading and odulation (FDD). [9] G. Caire, S. Gueghar, A. Rouy, and S. Verdú, Maxiizing the Spectral Efficiency of Coded CDMA Under Successive Decoding, IEEE Transactions on Inforation Theory, vol. 5, no., January 4. [] S. Verdù and S. Shaai (Shitz), Spectral Efficiency of CDMA with Rando Spreading, IEEE Transactions on Inforation Theory, Vol. 45, No., March 999ù [] R. De Gaudenzi, O. del Rio Herrero and G. Gallinaro, Enhanced spread Aloha physical layer design and perforance, Int. J. Satell. Coun. Network. (4), Published online in Wiley Online Library (wileyonlinelibrary.co). DOI:./sat E-SSA SF=6 Peak Eb/No=.77 db 9 db Rando. Range.E+.E-.E-.E- ( it.) (8 it.) ( it.).4.e-3 ( it.).e-3.5 ( it.). ( it.) (8 it.) ( it.).e-4.e Figure 6. Coparison of perforances between E-SSA and E b/n =.77 db and power randoization. Packet Length: info bit. FEC: 3GPP turbo-code rate /3. 96 sybols preable. Ideal acquisition. SIC iterations considered for both E-SSA and ME-SSA although soe perforance iproveent can be obtained increasing the nuber of iterations (as shown fro a single point also plotted for iterations (left) or 8 iterations (right).5.5 ME-SSA SF=3 Peak Eb/No =.77 db 9.5 db power rand. Multistage Det..E+.E E-SSA SF= Peak Eb/No=.77 db 9 db Rando. Range.E+.E-.E-.E-3 Spectral Efficiency (bits/chip).e-4.e Offered Traffic (bits/chip) Figure 7 Coparison of perforances at low SF between E-SSA and b/n =.77 db and power randoization. Packet Length was info bit with 3GPP code rate /3FEC. 96 sybols preable. Ideal acquisition considered. SIC iterations were considered for both E-SSA and ME-SSA ME-SSA, SF=4 Peak Eb/No=.77 db Power Rand. 9.5 db.e+.e-.e-.e-3 86