Литература / UMTS-Report

Concept Group Delta

WB-TDMA/CDMA

System Description Performance Evaluation

Disclaimer:

“This document was prepared during the evaluation work of SMG2 as a possible basis for the UTRA standard. It is provided to SMG on the understanding that the full details of the contents have not necessarily been reviewed by, or agreed by, SMG2.”

Concept Group Delta

Wideband TDMA/CDMA

Evaluation Report - Part 3

V 2.0 b

Compared to the previous version of this report, the following improvements are included in this version:

∙the minimum mean square error equalizer instead of the zero forcing equalizer has been used for joint detection, which leads to a performance improvement at the same computational complexity,

∙in some cases, burst type 1 instead of 2 has been used, leading to a better performance,

∙the channel estimation has been optimized,

∙the receive filter has been optimized.

1 Evaluation results

1.1 Introduction

As a part of the work carried out by the ETSI/SMG2 concept group Delta, Wideband TDMA/CDMA, a performance evaluation of TD/CDMA is carried out by means of simulations.

The SMG2 document UMTS TR 30.03 [1] describes how this evaluation is to be made. It lists a large number of environments and services to be tested. In Tdoc SMG2 260/97 [2] and Tdoc SMG2 329/97 [5], a subset of all these test cases are listed as prioritised. Simulation results obtained so far are for these prioritised test cases. The prioritised simulation cases from Tdoc 260/97 and Tdoc 329/97 are shown in Table 1-1. In addition, a 2 Mbit/s circuit switched service is investigated in the Pedestrian environment.

Table 1-1. Required simulations according to Tdoc 260/97 and Tdoc 329/97

In this chapter, link level simulation results for TD/CDMA for the services in Table 1-1 and for the 2 Mbit/s circuit switched service in the Pedestrian environment are shown.

1.2 Abbreviations

1.3 Link level simulations

In the following, link level simulation results for TD/CDMA are presented. The results are valid for both FDD and TDD operation. However, in TDD operation the results can be further improved by making use of the reciprocal channel for e.g. open loop control and pre-equalization.

The circuit switched services, i.e., speech and LCD services, cf. Table 1-1, are implemented with forward error correction (FEC) and the packet services, i.e., UDD services, use automatic repeat request (ARQ) together with FEC. The basic assumptions and technical choices for the link level simulations are summarized in Table 1-2.

Table 1-2 Basic assumptions and technical choices for the link level simulations

In the simulations, all intracell interferers are modelled completely with their whole transmission and reception chains. Intercell interference is modelled as white Gaussian noise. In the following, bit error rates (BER) are shown as a function of the average Eb/N0 in dB (Eb is the energy per bit and N0 is the

one-sided spectral noise density) with the intracell interference, i.e., the number K of active users per time slot as a parameter. The relation between the Eb/N0 and the carrier to interference ratio C/I, with C

denoting the carrier power per CDMA code and with I denoting the intercell interference power, is given by

The expression log2M is the number of bits per data symbol and Q.Tc/log2M is the bit duration at the output of the encoder. One net information bit is transmitted in a duration of Q.Tc/(Rc.log2M). Therefore, (1-1) is equivalent to C/I = (Eb/Tb)/(N0.B), i.e., C = Eb/Tb and I = N0.B with Tb the duration of a net information bit. The carrier to interference ratio per user is Kc times the carrier to interference ratio per CDMA code, with Kc denoting the number of CDMA codes per time slot per user.

Compared to the previous version of this report, the following improvements are included in this version:

∙The minimum mean square error equalizer instead of the zero forcing equalizer has been used for joint detection, which leads to a performance improvement at the same computational complexity.

∙In some cases, burst type 1 instead of 2 has been used, leading to a better performance.

∙The channel estimation has been optimized by weighting the estimated taps by their reliability.

∙The receive filter has been optimized. A receive filter matched to the linearized GMSK pulse has been introduced.

1.3.1 Speech service

In this section, link level simulation results for the speech service are given. The system parameters for implementing the speech service are summarized in Table 1-3.

Table 1-3 System parameters for the speech service

The required values for Eb/N0 and C/I in order not to exceed a BER of 10-3 as defined for the speech service are summarized in Table 1-4. The values of C/I are obtained from the values of Eb/N0

according to (1-1) by subtracting 13.2 dB for the spread speech/data burst 2 and by subtracting 12.4 dB for the spread speech/data burst 1.

All the simulations have been performed with an equalizer which is not adaptive, i.e., which does not adapt to the time variations within one burst. By using an adaptive equalizer, the performance for the high speed cases can be improved. In all the simulation cases given in Table 1-4, no power control has been performed. For Pedestrian and Indoor environment, frequency hopping has been used if not

mentioned otherwise. For the Vehicular B channel, the midamble has been used which is designed for a maximum excess delay of the channel of 15 μs although the excess delay of the channel impulse response is 20 μs. The reason for this choice is that the power of the taps with long delay spread are rather weak in the Vehicular B channel.

In the following, the possible reduction of the required values for Eb/N0 and C/I in order not to exceed

a BER of 10-3 in the case of low mobile velocities by using an enhanced power control instead of frequency hopping is investigated theoretically. For the Indoor case with K = 4 active users, the effect of an enhanced power control has been investigated exemplarily. In a first, idealized investigation, the actual power is estimated for each burst and then the transmit power of the next burst is adjusted according to the power estimate obtained in the last burst. In the simulations, real noisy power estimation is performed. The transmit power in the next burst is adjusted based on the unquantized power estimate obtained from the previous burst, which will give an upper bound of the gain that can be achieved by enhanced power control. This upper bound is equal to 5 dB. In a second investigation, the actual power is also estimated by real noisy power estimation for each burst. Based on this estimate, the transmit power in the next burst is either increased or decreased by a fixed step size of 2 dB. The gain achievable is 2.5 dB as compared to the case of using frequency hopping. The purpose of these investigations is to show the basic potential of an enhanced power control.

Table 1-4 Required values for Eb/N0 and C/I for the speech service

Kc = number of codes per time slot per user, K = number of users per time slot, UL/DL = uplink/downlink, FH = frequency hopping

Corresponding bit error rate curves are shown in Figure 1-2 to Figure 1-6, where the coded BER (userBER) is depicted versus the Eb/N0.

In Figure 1-1, the dependence of the required values for Eb/N0 for speech 8 kbit/s in order not to

exceed a BER of 10-3 as a function of the number K of active users per time slot for Indoor A, Pedestrian A and Vehicular A in the downlink is depicted. Values of K between 1 and 12 are considered. There is a slight degradation with increasing number K of active users per time slot. This is due to the increase of intracell interference with increasing K. The degradation is less for the Indoor and Pedestrian channels which have less multipaths than for the Vehicular channel with more multipaths. When the number K of users approaches the spreading factor of 16, the degradation increases.

Figure 1-1. Dependence of the required values for Eb/N0 for speech 8 kbit/s in order not to

exceed a BER of 10-3 as a function of the number K of active users per time slot in the downlink for Indoor A, Pedestrian A and Vehicular A

1.3.2 LCD services

In this section, link level simulation results for the LCD services are given. The system parameters for implementing the LCD 144 kbit/s service are summarized in Table 1-5 and for implementing the LCD 384 kbit/s service in Table 1-6. Furthermore, an LCD 2048 kbit/s service is investigated, for which the system parameters are given in Table 1-7. For the LCD 144 kbit/s service, three alternatives are considered:

∙allocating 1 code in each of the 8 time slots to a user (LCD 144a),

∙allocating 9 codes in 1 of the 8 time slots to a user (LCD 144b),

∙allocating 3 codes in 4 of the 8 time slots to a user (LCD 144c).

For the LCD 384 kbit/s service, two alternatives are considered:

∙allocating 3 codes in each of the 8 time slots to a user (LCD 384a),

∙allocating 9 codes in 3 of the 8 time slots to a user (LCD 384b).

For the LCD 2084 kbit/s service, 9 codes are allocated in each of the 8 time slots to a user (LCD 2048).

Table 1-5 System parameters for the LCD 144 kbit/s service

To reach the BER requirement of 10-6, the LCD services (except for LCD 144a) use a concatenated coding scheme with an inner convolutional code and an outer Reed Solomon code. In the results given here, the BER at the output of the inner convolutional decoder is shown. It is expected that a BER of

about 10-4 at the output of the inner convolutional decoder will lead to a BER of 10-6 at the output of the outer Reed Solomon decoder. The results valid for the output of the Reed Solomon decoder are not

yet available due to the extremely long simulation times to measure a BER of 10-6 with sufficient accuracy. The required values for Eb/N0 and C/I in order not to exceed a BER of 10-4 at the output of

the inner convolutional decoder are summarized in Table 1-8 for LCD 144 kbit/s, in Table 1-9 for LCD 384 kbit/s and in Table 1-10 for LCD 2048 kbit/s. The values of C/I are obtained from the values of Eb/N0 according to (1-1) by subtracting 9.7 dB for LCD 144a, 10.4 dB for LCD 144b, 11.6 dB for

LCD 144 c, 10.4 dB for LCD 384a, 10.9 dB for LCD 384b and 7.9 dB for LCD 2048. The required Eb/N0 and C/I values for the downlink can be considerably reduced by using antenna diversity also in

the downlink. This would be a reasonable assumption for those applications which are executed e.g. on a laptop.

Table 1-8 Required values for Eb/N0 and C/I for the LCD 144 kbit/s service

Kc = number of codes per time slot per user, K = number of users per time slot, UL/DL = uplink/downlink, CC = at the output of the inner convolutional decoder, RS = at the output of the outer Reed Solomon decoder