Dual Mode GSM/UMTS Terminal Complexity

Transcription

1 ETSI SMG Tdoc SMG 1069/97 Meeting no 24 Madrid, Spain December 1997 Source: Nokia, Ericsson Dual Mode GSM/UMTS Terminal Complexity Dual Mode GSM/UMTS Terminal Complexity Introduction This contribution discusse the dual mode terminal implementation between GSM and UMTS, concentrating mainly on the WCDMA and WB-TDMA/CDMA (For hybrid solution names TD/CDMA or TDMA/CDMA also used) issues. However, issues valid for other proposals are presented as well. The focus on this discussion is on the terminals reaching beyond GSM capability, up to 384 kbits/s or up to 2 Mbits/s services. As a technology background for baseband processing, the following assumption for processing power from Tdoc SMG2 351/97 could be used. There the stated expectation is that 400 MIPS dual ALU general purpose signal processor with 200 MHz and 0.1 mw/mhz power consumption to be available for 2002 for UMTS terminals.

2 Dual mode terminal radio parts The RF parts needed for a UMTS/GSM multimode terminal are shown in Figure 1. The terminal is capable of high medium/high rate applications, and thus the layout can be considered valid for both WCDMA and WB-TDMA/CDMA terminals. With the low rate services the main exception is the absence of duplex filter which is replaced with a TX/RX switch for low rate terminals for TDMA based solutions. Introduction may be necessary even earlier for TDMA solutions if there is need to get rid of the limitations to have the TX and RX s in certain time relation to each other. The complexity of the duplexer is put to right perceptive by understanding that it s used on several second generation phones today, including some of the GSM handportables in the market. Also based on the SMG2 results on the link level Eb/No performance, as the difference in Eb/No values is in the benefit of WCDMA of several dbs, the differences in insertion loss between duplexer and a TX/RX switch become insignificant. The synthesisers indicated in Figure 1. have the following reasoning. For GSM and DCS different frequency bands are used thus separate synthesisers are given. Naturally they all need not to be in use at the same time. For the UMTS part there are double synthesisers as the transmit and receive part operate independently and at the same time. This again is needed for TDMA based solutions when more than one burst is used and frequency hopping between consecutive burst is applied, like assumed in SMG2 performance studies for WB-TDMA/CDMA results with few exceptions. The different filters for different frequency bands are obviously needed and again do not depend on the selected multiple access mechanism, rather on the desired bands for operation. Whether some of the interfrequency (IF) parts could be shared is rather a question on the architecture, the main point is that bandwidths are anyway different and different filtering functionality is needed. The converters for GSM operate with lower bit stream for the output and UMTS solutions need different rates from their converters, for low rate terminals some reuse expected to be possible but that depends very much on the converter type. The WCDMA converters will cover a larger bandwidth, but will on the other hand need to operate with much lower dynamic range (half or less than that of WB-TDMA/CDMA), thus leading to smaller difference in the output bit rate of converters between the concepts. The modulators in Figure 1. are naturally similar in number, some additional burden, not shown in the figure, will be caused to the WB-TDMA/CDMA solution as several modulation formats, including 16QAM are used for making the UMTS services possible. The power amplifier part has always been an interesting part of the discussion, here the difference is in the dimensioning of the amplifier, as with WCDMA the maximum average TX power is very near the maximum peak TX power. With TDMA based solutions the difference between peak and average will be there more inherently unless all the slots are taken into use. The example case for the amplifier power needed during transmission is the 24 dbm (250 mw) of average power. If the terminal is expected to send that with 1/8 duty cycle then the peak power is obviously 2 Watts. If the amplifier is the same as used in the WB-TDMA/CDMA spectrum mask and guard band evaluation, having 12 percent efficiency (Tdoc SMG2 414/97), the amplifier needs to be dimensioned to such that when producing 2 Watts peak, 16 Watts is drained from the battery during operation. For multislot transmission as the same amplifier is used, the amplifier dimensioning for more continuos transmission is then not optimal although the basic functional difference as such to WCDMA gets smaller as more slots in a are used.

3 GSM900 DEMOD GSM1800 SYNTH 1 SYNTH 2 MOD CDMA / TD-CDMA DEMOD SYNTH 3 SYNTH 4 MOD Figure 1. GSM/DCS/UMTS multimode terminal RF parts. Dual mode terminal baseband parts The baseband processing for equaliser and channel decoding will take most of the effort and thus the call control processing will take in the order of few percents of the total processing power and is thus such a negligible component in the complexity analysis. Also when looking the slot//multi structures the most meaningful similarity is the use of GSM compatible multi structure and thus allowing similar timing for measurements, as in Figure 2. The use of GSM length compared to for example 10 ms length has not been shown to make any practical differences for dual mode terminal as such. When comparing generally different complexity estimates in terms of MIPS/MOPS (Millions of Instructions Per Second/Millions of Operations Per Second) one should also pay attention to the fact that in some cases if the claim is to reuse the processing hardware for GSM operation, the UMTS processing for the certain has to be done then in much sorter time domain than the expected averaging over the all N time slots in a would suggest. With WCDMA the low end terminal will experience only about 10% increase in the peak MOPS figure due to measurements with slotted mode, while with other solutions (such

4 as WB-TDMA/CDMA) if the same hardware is used, the increase might be in the order of 100 or 200 percent. Similarly this is the case if the claim is made at the same time that N MOPS is used and then said that power save can be used for the M-1 idle time slots. In practise this means that the time slot processing would need to be done during the single slot as well in real time thus resulting to M times higher peak MOPS capability requirement than counted by averaging over the length. When now observing the needed complexity for base band processing for WCDMA and WB- TDMA/CDMA, the following figures summarised in Table 1 have been presented in SMG2 from the concept groups Alpha (WCDMA) and Delta (WB-TDMA/CDMA). The figures for WB-TDMA/CDMA are based on having 3 slots for 384 kbits/s service and 8 slots for 2 Mbits/s service. Table 1. Multiple access dependant baseband complexity. Multiple Access / Service 384 kbits/s 2 Mbits/s WCDMA WB-TDMA/CDMA (Burst type 1) WB-TDMA/CDMA (Burst type 2) (In millions of real multiplications per second, source SMG2 WCDMA and WB-TDMA/CDMA concept group documents. WB-TDMA/CDMA figures assume averaging the processing over the whole with 384 kbits/s service. Figures for Wb-TDMA/CDMA per slot can be found from the SMG2 Copenhagen Q&A Workshop documentation.) These numbers are not in MIPS, the needed MIPS figure could be then 2 to 4 times higher for actual DSP operations. If then a 400 MIPS DSP is available at year 2002 with around 20 mw of power consumption, for the 2 Mbits/s case obviously a single devise will not be able to process 2 Mbits/s or barely even 384 kbits/s for TDMA/CDMA solution according to the figures given from the TDMA/CDMA documentation. In conclusion, it is important to minimise the DSP load from the multiple access dependent part, not only trust the DSP developments, as there will be large demands for prosessing power from the other signal processing functions for UMTS services. Slotted Mode with WCDMA As discussed also in the WCDMA evaluation document (Tdoc SMG 905/97), for WCDMA there exists basically two solutions for taking care of the measurements from GSM during WCDMA operation. The solution with slotted mode facilitates a partly shared receiver chain solution while dual receiver solution has a totally separate receiver chain for measurements on other frequencies than the operation frequency. The dual receiver solution is illustrated in Figure 1. What are then the advantages and disadvantages with slotted mode? Based on the ETSI WCDMA group simulations in both link and system level, the loss in the capacity is estimated to be max. 6 percent. This is the worst case scenario when all the mobiles are making measurements on GSM in every 10th 10 ms. In practise this is not likely to be the case as when and on which GSM carriers measurements need to be done can be determined and thus the measurement cycle is expected to be reduced. Further, it will not be necessary for all terminals to make GSM handover measurements all the time. Thus the expected capacity loss with slotted mode GSM handover measurements will be much less than 6%. The noticeable thing is that the coverage will hardly be affected by the slotted mode since the slotted mode is only used in the downlink.

5 GSM CCH F-burst S-burst multi- GSM TCH/F idle W-CDMA blank slot Figure 2. Measurements similarity between GSM and WCDMA multi solutions. For high rate WB-TDMA/CDMA terminals (multi-slot) an additional receiver is necessary to make handover measuremenst, both for GSM and UMTS handover measurements. Further, GSM handover measurements in WB-TDMA/CDMA will take UMTS handover measurement capabilities. Conclusions In conclusion, the WCDMA UMTS air interface has at least as good or better GSM/UMTS dual mode properties than any of the other candidates. This is true for RF complexity, baseband signal processing and GSM/UMTS handover performance as shown in this paper. Also, observing the reality of dual mode terminals between second generation systems and analog first generation cellular systems in several markets for the systems that had nothing in common in the design phase. Thus when standardising UMTS all possibilities are there to ensure that the dual mode equipment will be efficient to work both with UMTS and GSM and that seamless handover to the user between GSM and UMTS can be provided for ensuring sufficient coverage for UMTS especially in the roll-out phase.