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 4

V 2.0 b

System Level Simulations

General

This section describes the TDMA/CDMA system level simulations. The TDMA/CDMA approach has 3 dimensions: frequency, time and code dimension.

In general, only the downlink direction is considered. It was found that the downlink is limiting the system capacity, since for example antenna diversity can be used in the uplink to improve the soft blocking limit significantly.

The simulation results presented in this document rely on statistics gathered from about 5000 calls in the reference cells.

Used Models

All test environments including network structure, cell shape, antenna pattern, propagation models (path loss and shadowing, channel model), mobility models, traffic models and quality of service (QoS) criteria for Real Time (RT) users are modeled according to [1].

Traffic and quality of service (QoS) criteria for Non Real Time (RT) users is modeled according to [3].

In the macro environment a re-use pattern 1/3 is used. Statistics is collected within the central cluster surrounded by 3 interferer rings.

In the micro environment a re-use scheme of 1 and 3 is used within the Manhattan grid structure given in [1]. Statistics is gathered from the 6 cells according to [1].

For the results (given in table 1) in the pico environment a re-use scheme of 1 and 3 is used within the indoor office model given in [1]. Statistics is gathered from the 6 cells in the middle floor according to [1]. A clustered system for the mixed service in indoor environment is currently under investigation.

Resource Allocation

A resource unit (RU) in TDMA/CDMA is a triple consisting of frequency channel, time slot and code. For services that require more than one RU:

∙a number of codes (multi-code), or

∙a number of time-slots (multi-slot), or

∙a combination of both (mixed allocation) may be allocated.

For RT services the resources are allocated at session setup and are kept unchanged till session end. On intercell handover the same type of resource is allocated in the new cell.

For NRT services the allocation and de-allocation is done on block level driven by the number of data a user has in the buffer.

The resource allocation tries to distribute the allocated codes homogeneously over all frequencies and timeslots, i.e. it is searched for the time slot with minimum number of codes.

Frequency and Time Hopping

The simulator is capable of simulating both frequency and time hopping by using quasi-random hopping-sequences. The number of hopping frequencies corresponds to the number of carriers within a cell (e.g. 3 for the 1/3 cluster in macro environment). Frequency hopping is performed on a frame by frame basis. The presented simulation results have been produced without time hopping. Simulation results without any hopping are presented as well.

It should be noted, since resource allocation for non real time services is performed on a block basis the effect of frequency hopping is irrelevant for that type of service.

Power Control

Slow (0.5 sec control interval) level based power control is used with a dynamic range of 20 dB in the downlink and 70 dB in the uplink. No fast PC is used within the presented TDMA/CDMA simulation results.

Handover

The handover is based on GSM-like power budget (path loss difference between serving and neighbour cells) with a handover margin of 3 dB.

Link Adaptation

Link adaptation has not been used yet explicitly within the simulations. However, it may be applied in TDMA/CDMA to improve spectrum efficiency.

ARQ

For NRT services the first transmission of a block is done nearly uncoded. If the first transmission fails a second transmission contains the coding bits in a way that all blocks of the first and second transmission together result in a higher coding rate.

If the second transmission also fails the worst (raw-BER) burst is retransmitted and a maximum ratio combining is done between the original and the retransmitted burst. The code rate is not increased in this step. For all further retransmissions this maximum ratio combining is used.

If the number of retransmissions exceeds a given threshold (10 - 20) the session is dropped.

DTX

For speech services voice activity detection (VAD) and discontinous transmission (DTX) is used with an activity factor of 0.5 and mean speech (activity) periods of 3 seconds according to [1]. During the silent periods the transmit power is switched off, but the channels are not released, i.e. DTX is used to reduce interference and thereby to relieve the soft blocking limit.

Interface between System and Link Level Simulation

Link level simulations are done with fast (multipath) fading and the result is the BER-CIR relationship.

In a system with frequency / time hopping and resource allocation on block level, system simulations based on an average value interface result in a very low accuracy. This is why an actual value interface (AVI) has been used in the system level simulator to produce the results presented in this document.

Within the system level simulations fast fading has been taken into account to calculate the actual values of CIR experienced on each burst.

Due to the code dimension in TDMA/CDMA the CIR values are distinguished between intercell CIRinter and intracell CIRintra.

The CIRinter is the ratio between the wanted signal in the refernce cell and the sum of interfering signals from all other cells at the same frequency and time slot.

The CIRintra is the ratio between the wanted signal in the reference cell and the sum of the signals from all other users in the same cell at the same frequency and time-slot. Simulations have shown that the impact of the intracell interference is negligible (due to joint detection).

Within the AVI the burst CIR values are mapped on a raw BER on a burst. Beside on the CIRintra and CIRinter the raw BER also depends upon

∙the number of codes KC per user used with a certain service (code pooling)

∙the number of users K per time slot.

Depending on the interleaving depth assumed for each service, the average raw BER on a corresponding number of bursts constituting one block is calculated and subsequently mapped on a user BER value of the received block.

Actual Value Interface at system level

For ARQ there is an additional interface function that gives the relationship between raw BER and BLER (block erasure rate).

Results

The simulation results are summarized in Table 1. The variation in the values compared with the corresponding ones given in Table 1 of the Evaluation Report - Part 4 Version 1.0 is due to corresponding improvements on link level which are presented in Evaluation Report - Part 3 Version 2.0b. Furthermore the table has been updated with cluster 3 results for UDD and pedestrian speech services. More detailed description of the simulations carried out is presented in the following chapters.

Table 1: Summary of the system simulations results

∙hopping randomly over 3 frequencies (per cells) gives an average usage of 2 different frequencies within the interleaving depth of 4, hence the gain from frequency diversity is small;

∙on the other hand in a re-use scheme of 3 the interference is very low except for a small area on the crossings; hence the gain from interference diversity is small as well;

∙furthermore in the micro environment there is a high coherence bandwidth.

Environment: Indoor

The figures in table 1 are based on 14.4 MHz spectrum (9 carrier) without guard bands.

UDD2048

For the UDD services resources are allocated on demand, i.e. depending on the current packet size. In the TDMA/CDMA system the possible resource allocation granularity is 1 code. To speed up simulations a resource allocation granularity of 1 time slot has been used within the simulations (exception: for the cluster 3 with AD a granularity of ½ timeslot has bee used). From the point of view of the traffic model (usage of resource units) and interference averaging this is a pessimistic assumption. The maximum number of allocated resources is 8 slots x 9 codes (4 codes for the cluster 3 with AD).

Results are presented for both cluster 1 and 3.

Mixed Speech and UDD384

For this case a resource allocation granularity of 1 code has been used within the simulations. At the moment results are available for cluster 1 only. Results for cluster 3 will be available in 2 weeks. However first estimations show that the figures for cluster 3 are nearly the same as for cluster 1. Simulations have been performed with and without frequency hopping. However, the results do not differ significantly (less than 5%). This is due to the following facts:

∙the UDD services provide inherently interference averaging since resource allocation is done at block level.

∙furthermore in the pico environment there is even a higher coherence bandwidth than in micro.

Summary

Simulation results concerning spectrum efficiency of TDMA/CDMA have been presented for real time and non real time services. The presented values do not include the impact of signaling which however is estimated to be about 10%.

On the other hand there are some options which may significantly improve the spectrum efficiency values:

∙link adaptation

∙fast power control

∙power control based on quality

∙time slot hopping

∙smart antennas

∙floating carrier.

These options and the corresponding tradeoff between capacity increase and complexity issues are under investigation.

References

[1]UMTS TR 30.03 “Selection procedure for the choice of radio transmission technologies of the

[2]ETSI SMG2 UMTS ad hoc #3 TDoc 73, Rennes, 1997.

[3]CR on UMTS: 30.03 A002, Oct. 1997