`
`Compressed Mode Techniques for Inter-Frequency Measurements in a
`Wide-band DS-CDMA System
`
`Maria Gustafsson*, Karim Jamalli’ and Erik Dahlrnan*
`
`aleEricsson Radio Systems AB, S-164 80 Stockholm, SWEDEN
`*Nippon Ericsson K.K., 5F Kioicho Kelton Bldg, 4-5 Kioicho, Chiyoda—ku, Tokyo 102 JAPAN
`
`e—mail: managustafsson @era—t.ericsson.se
`
`Abstract — In this paper we discuss Inter-Frequency Hand-
`Over (IFHO) in third generation cellular systems based on
`Wide-band DS-CDMA. In order to provide maximum flexibil-
`ity for future radio network design, it is important to develop a
`technique allowing reliable IFHO in the system. The Com-
`pressed Mode (CM) scheme first presented in [1] facilitates the
`use of the Mobile-Assisted HandOver (MAHO) technique com-
`monly used in all second generation systems, also for IFHO in a
`CDMA system. We present and analyze four different ways of
`achieving a CM mechanism in a CDMA system.
`It is found that methods based on variable spreading factor and
`coding-rate increment may be of most interest for practical use.
`
`to deal with the issue of inter-frequency HO (IFHO). There
`are several situations where IFHO may be desired:
`
`'
`
`Systems with multiple-carrier cells only in high-
`trafi‘ic areas: Handover at the borders of the high-
`traffic areas.
`
`
`
`I. INTRODUCTION
`
`- Hierarchical Cell Structures
`Handover between cell layers.
`
`(HCS)
`
`[2M3]:
`
`Tomorrow’s wireless mobile systems should be able to sup-
`port services that are significantly different from those sup-
`ported by systems in operation today, e.g. GSM and PDC. In
`particular, higher information bit-rates up to 2 Mb/s, packet
`transmission for e.g. Internet access, and multi—media ser-
`vices are of great importance. At the same time, low-rate ser-
`vices such as speech, must also be supported. Efficient
`handling of a mix of services with different characteristics is
`thus a key requirement.
`
`Wide-band DS-CDMA (WCDMA) is seen as a promising
`multiple access scheme candidate for third generation sys~
`terns, mainly due to its potential for efficient support of
`mixed services and multi-media applications. In order to
`fully exploit the capacity of a WCDMA system, the carrier
`frequencies should be re-used in every cell. This in turn leads
`to a need to use so called SOft HandOver (SOHO) near cell-
`borders to avoid excessive interference into neighbouring
`cells. A Mobile Station (MS) in SOHO is connected to a plu—
`rality of Base Stations (BS), giving a macro diversity effect.
`
`Assuming that only one carrier frequency is used in all cells
`throughout
`the system,
`the Mobile-Assisted HandOver
`(MAHO) technique, used for improved HO performance in
`all second generation systems, can relatively easily be used;
`The MS continuously scans the link-quality to the surround—
`ing cells and reports the measurements to the NetWork (NW)
`that uses the reports as basis for (SO)HO decisions. This
`makes the actual HO fast (low probability of data-loss), since
`at least some level of synchronization to the target BS can be
`determined prior to the HO execution, and reliable (low prob-
`ability of call loss) since a good target BS decision is facili»
`tated.
`
`With more than one carrier frequency in the system, we have
`
`is
`
`-
`
`Inter-system HO: Handover to other third genera-
`tion systems or second generation systems, e.g.
`GSM.
`
`
`
`
`
`WCDMA+GSM
`
`[:1 : GSM
`
`There are several possible ways to enable IFHO in a
`WCDMA system. The methods can be classified into two
`groups:
`
`1) Methods with forced intra-cell IFHO followed by
`inter-cell intra-frequency HO.
`
`2) Methods with true inter—cell IFHO
`
`The first group of methods is mainly applicable to the first
`case illustrated above, i.e. uniform cell types with different
`number of carriers in cells according to expected traffic. The
`example used here is to have f1 and f2 in high-traffic areas
`and only f1 in other areas. One solution is to insert f2 “bea—
`con” transmitters in all cells bordering an (f1,f2) cell. In this
`
`0-7803-3871-5/97/$10.00 © 1997 IEEE
`
`231
`
`LG E_0001 233
`
`LG Electronics Ex. 1013
`
`LG Electronics Ex. 1013
`
`LGE_0001233
`
`

`

`way, an MS connected to f2 at the border of the (f1,f2) area
`can hear a strong f2 beacon and report it to the network,
`which in turn forces an intra—cell IFHO to f1. After that, an
`inter—cell intra—frequency HO may be done to the low-traffic
`cell. Another related method is to introduce complete f2 car—
`riers (transmitters and receivers) along the borders of the
`
`(f1,f2) area. As soon as an inter-cell HO using fl is done into
`these cells, a forced intra—cell IFHO to fl is conducted.
`
`These methods have some disadvantages; First, a lot of extra
`fZ—hardware has to be installed in all the border cells. Second,
`the total HO reliability may be lowered due to the two-step
`procedure leaving less time for measurement averaging.
`Third, since forced H0 is done to f1 in both cases, the link
`quality of f1 may not be sufficiently good at all times. Fourth,
`these techniques do not really solve IFHO in HCS environ-
`ments.
`
`The second group of methods, where we have a true inter-cell
`IFI-IO, we rely on inter-frequency link-quality measure-
`ments.
`
`One way of achieving this is to introduce Network—Assisted
`HO (NAHO), where BSs are equipped with extra receivers
`that scan for MSs that are not connected to their own cells.
`
`However with such a solution, accurate MS synchronization
`(and thus MS identification) may be difficult to provide,
`especially with non—synchronized BSs. Thus this solution
`may add significantly to the complexity of the B85.
`
`Another way is to use the MAI-IO technique described ear—
`lier, which is readily used for intra-frequency H0 in a
`WCDMA system. However, there is a CDMA—specific prob—
`lem; Transmission and reception is normally continuous,
`leaving no idle time for the MS receiver to switch to other
`carrier frequencies.
`
`There are however different ways of getting around this and
`actually enabling “Mobile—Assisted IFHO” (MA—LFHO) in a
`WCDMA system. The most straightforward solution is to
`equip the MS with a second scanning receiver, designed to
`perform inter-frequency link measurements. However, this
`solution adds significantly to the complexity, and thus also to
`the size and power consumption, of the MS. If the implemen—
`tation is done using a wide-band receiver that receives the
`entire system’s
`frequency band,
`there would also be
`extremely high requirements on the dynamic range of the
`receiver due to near—far problems, especially in a HCS envi-
`ronment.
`
`The concept of Compressed Mode (CM) was introduced in
`[1] as a method to facilitate MA-IFHO without having to add
`any extra hardware. In CM, the BS “compresses” the infor—
`mation sent to an MS into a shorter time-frame than in nor-
`
`mal mode, thus creating idle time for the MS receiver to
`switch to other frequencies and perform link measurements.
`Since there will always be a certain performance loss in CM,
`a connection would only go into CM occasionally, e.g. by
`control of the network.
`
`In the remainder of this paper we describe four different ways
`of achieving a CM mechanism on a WCDMA physical layer.
`The merits and drawbacks of the different schemes are dis-
`
`cussed, and the performance of the schemes is determined
`
`through simulations. Finally, some conclusions are made as
`to the most promising CM schemes.
`
`II. COMPRESSED MODE
`
`We define CM as transmission with the same information bit-
`
`rate as in normal mode, but the actual transmission occurring
`only during part of a frame [1]. This is depicted in Figure 1.
`Clearly, the power has to be increased in order to compensate
`for the shorter transmission time, as indicated in the Figure.
`
`NORMAL MODE
`
`COMPRESSED MODE
`
`
`Figure 1. Compressed Mode
`
`Note that this assumes that the TX power amplifier is not
`using its maximum peak power already in normal mode, i.e.
`the amplifier has to be some what over-dimensioned.
`
`The compressed mode ratio, 7, is defined as the fraction of
`the frame left idle during CM. Note that a y of around 50% as
`shown in Figure 1 is only an example. Other values of 7, both
`larger and smaller than 50%, are equally possible.
`
`Depending on the RX/TX timing in the MS and the need to
`have a common RX/TX frequency synthesiser1 in the MS,
`different CM concepts are possible.
`
`COMMON RXfl'X SYNTHESISER
`SYNCHRONIZED
`
`SEPARATE RX/I'X SYNTHESISERS
`SYNCHRONIZED
`
`
`
`3;": RECEIVER IDLE TIME
`
`Figure 2. Various CM possibilities
`
`Figure 2 shows the receiver’s idle time periods achieved in
`the different cases. If the common RX/TX frequency synthe—
`siser requirement is relaxed, it is possible to have CM only in
`the DownLink (DL), regardless of the RX/TX synchronism
`in the MS. Such a solution would be very attractive as it con-
`siderably relaxes the peak-power requirements of the MS
`TX.
`
`CM can in principle be used for the following functions:
`
`1) Link-quality measurements on BSs on other fre—
`quencies (MAHO measurements).
`
`i.e. Random Access and
`2) Actual HO execution,
`Channel Allocation at a target BS during CM.
`
`1. Frequency Division Duplex (FDD) With a fixed duplex
`distance is assumed.
`
`LGE_0001234
`
`232
`
`LGE_0001234
`
`

`

`
`
`the frame, and can thus not generate correct or relevant PC
`commands to be transmitted in the DL. The MS on the other
`
`hand cannot receive these PC commands during the idle
`times anyway, since it is busy doing IF-MAHO measure—
`ments. This PC degradation will translate into a BER perfor-
`mance loss. Different additional performance losses are also
`expected with different CM methods. In the following, we
`describe the above mentioned four CM mechanisms and their
`
`characteristics in more detail. The performance of the meth-
`ods is then evaluated through simulations.
`
`A Variable Spreading Factor CM (VSF—CM)
`
`Changing the spreading factor, SF, i.e. the number of chips
`per channel symbol, is very easily done if long (non-orthog-
`onal) spreading codes are used, where an almost arbitrary SF,
`i.e. arbitrary y, can be achieved [1]. If short (orthogonal)
`spreading codes are used, it is also possible to change SF, but
`normally only by a factor of 2 at a time, due to orthogonality
`constraints. Thus in this latter case, only a Y of 1/2 would be
`of major interest.
`
`B Code-Rate Increased CM (CRI-CM)
`
`A method somewhat related to VSF—CM is to increase the
`
`channel coding rate, R. If convolutional coding is used, code-
`puncturing can advantageously be employed to achieve this.
`The coding rate, R, and thereby 7, may then be increased in
`relatively small incremental steps.
`
`C Multi-Code CM (MC-CM)
`
`The number of parallel physical CDMA code channels used,
`N, may be increased in order to implement CM. If N=1
`would be used in normal mode, 'y=1/2 would be the most
`interesting case. On the other hand, if multi—code transmis-
`sion is already in use in normal mode, e.g. N=2, it could be
`possible to use N=3 (7:1/3) as well as N=4 (7:1/2) for CM
`etc. Thus the “resolution” in ydepends on the value of N used
`in normal mode.
`
`D Higher Order Modulation CM (HOM-CM)
`
`If the modulation order is increased e.g. from M=2 (QPSK)
`to M=4 (16QAM) [4], a 'Y of 1/2 can be achieved, while a ‘y of
`1/3 is achieved by shifting to M=3 (8PSK). Clearly, also this
`method will have a relatively “coarse” y—increment since the
`normal mode would typically use M=2 or M:1 .
`
`Note that coded modulation could be used to achieve some-
`
`thing in-between HOM—CM and CRI—CM.
`
`IV. SIMULATIONS
`
`The performance losses of the four CM methods are evalu-
`ated by computer simulations with parameters according to
`Table 1.
`
`Table l
`
`3)
`
`SOHO in CM, i.e. the MS communicates simulta-
`neously (within one frame) with two base stations
`on two different carrier frequencies.
`
`Note that in the first case only DL CM is needed, while in the
`second and third case, CM is required in both links.
`
`Although IFHO would typically be performed rather seldom,
`it is important that it is reliable when it does occur. The first
`function above, i.e. MAHO measurements, clearly suffices
`for this and basically only requires a specification of the CM
`mechanism, the signalling to enable/disable CM and the link
`measurement report signalling. The second function would
`also require the specification of a special “shortened” layer-1
`signalling format for Random Access etc. in CM, and the
`third function would need a new macro-diversity combining
`scheme and some time-alignment management.
`
`Here we assume that CM will be used for inter-frequency
`measurements only, i.e. the first of the functions listed above.
`
`The value of yin CM depends on the physical structure of the
`DL common control channels of the system and is also a
`trade-off between measurement accuracy and performance
`degradation in CM.
`
`It should be noted that CM may be of primary interest for
`connections that carry low-delay services sensitive to data-
`loss, such as speech and video.
`
`III. BASIC COMPRESSED MODE MECHANISMS
`
`The information bit rate, Rb, can be written as:
`
`R.=(1—v)-N.
`
`RcvR-M
`SF
`
`(1)
`
`where RC is the chip-rate, R is the channel encoding rate, M is
`the modulation order (channel bits/channel symbol), SF is
`the spreading factor (chips/channel symbol) and N is the
`number of parallel physical CDMA code channels used (N>l
`corresponds to so-called multi-code transmission, see [4]).
`The compressed mode ratio is 7 which has values between 0
`and 1.
`
`In CM, the compressed mode ratio, 7, will become larger
`than 0, and since we want to keep Rb constant, we can either
`decrease SF or increase R, M or N in response to y. (The chip-
`rate, RC, is constant and the average TX power is kept con-
`stant by increasing the peak TX power in response to 7.)
`These methods correspond to Variable Spreading Factor CM
`(VSF-CM), Code-Rate Increased CM (CRI—CM), Higher-
`Order Modulation CM (HOM-CM) and Multi-Code CM
`(MC-CM) respectively. Note that when the lowest “allow-
`able” SF or highest allowable R, M or N is already used in
`normal mode (for very high bit-rate transmission), a different
`method would have to be used for CM.
`
`Assuming only intra-frame interleaving, all CM methods
`will necessarily result in a loss in Bit Error Rate (BER) per-
`formance due to a shorter interleaving depth. As we consider
`CM transmission according to the upper left part of Figure 2,
`there will also be some deterioration of the fast Power Con-
`
`trol (PC) performance. Assuming UL fast PC, the BS will not
`be able to estimate the received power during the idle part of
`
`LGE_0001235
`
`233
`
`
`
`I ' -. . Basierimming
`
`
` V
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`LGE_0001235
`
`

`

`Table 1
`
`V. IMPLEMENTATION ASPECTS
`
`FEC / Interleaving
`
`Convolutional, R=ll3 / 10 ms
`
`Spreading Codes
`
`Long (PN) + Short (Orthogonal)
`
`Basic Spreading Factor (SF)
`
`
`
`Frame Length/ Slot Length
`
`10 ms / 0.625 ms
`
`Fast Power Control
`
`1 PC symbol/slot (up/down), 1 dB step
`
`Channel Model
`
`4-ray Rayleigh. Equal average power
`
`Channel Estimation
`2 Pilot symbols in each end of each
`slot
`
`
`Detector
`
`4-finger coherent RAKE with known
`synchronization
`
`Note that we in all simulations assume CM in both UL and
`DL.
`
`Only y=ll2 is possible in VSF-CM since we use orthogonal
`codes and in MC—CM due to single code transmission in nor—
`mal mode. For the CRI-CM method, we achieve both 7:5/16
`and 7:1/2 by using different puncturing rates. With HOM—
`CM, 7:5/ 16 and 7:1/2 are achieved by using 8PSK and
`16QAM respectively. The performance measure is the Bit
`Error Rate (BER) after channel decoding as a function of Eb/
`No, where Eb is the information bit energy and NO is the
`background noise power spectrum density. We study the per-
`formance with and without fast PC and with Doppler spreads
`offD=7 Hz /fD=100 Hz, which we denote “slow” and “fast”
`fading respectively.
`
`Figures 3 and 4 show the performance of CRI-CM and
`HOM—CM with and without PC when y=5/16. The other
`methods are not simulated, since the bit-rate can only be
`altered in steps of two for those methods, as mentioned pre-
`viously. When no PC is used (Fig. 3), the losses that can be
`seen are mainly due to the shorter interleaving depth. This
`loss is mainly noticed in fast fading, since in slow fading the
`interleaving has a small effect even in normal mode. For
`CRI—CM there is also a reduced coding gain that results from
`the code-puncturing, and for HOM-CM, the uncoded BER
`and thus also the coded BER gets worsened due to the sym-
`bol points getting closer together. The losses with CM are
`about 0.5 dB in slow fading, but almost 1 dB in fast fading
`(@ BER=1E—3). Figure 4 also allows us to see the effect that
`CM has on the fast PC. The losses are slightly larger with PC,
`especially in fast fading, where we have 1.5 dB for HOM—
`CM and 2 dB for CRI—CM.
`
`Figures 5 shows all four methods with 7:1/2 without PC.
`Here we see a larger absolute loss (0.5—l dB in slow fading
`and 1.3—3 dB in fast fading) as well as a larger spread in per—
`formance among the methods (1 dB in slow fading and 1.5
`dB in fast fading). With PC (Figure 6) we see a similar trend
`with even larger spread among the methods. It is clear that
`HOM-CM has the worst performance of the four with up to 5
`dB loss in fast fading with PC. The VSF‘CM and MC—CM
`methods outperform the other methods for 7:1/2.
`
`LGE_0001236
`
`234
`
`Besides the different losses in BER performance for the var—
`ious CM mechanisms,
`there are also implications on the
`implementation of CM, both general and method—specific.
`
`To start with, CM in itself assumes that the TX peak power
`can be increased by at least lOlog(1/(l—'y)) dB. Assuming that
`CM is used in the UL, the MS power amplifier has to be over—
`dimensioned by a corresponding amount. This is a strong
`reason for having CM only in the DL.
`
`Besides the general implications of CM, there are also differ—
`ent aspects with different CM methods. For instance, the
`increased number of symbol-points of HOM-CM leads to
`increased phase linearity and phase resolution demands in
`both the transmitter and receiver. Moreover, HOM-CM as
`
`well as MC—CM will result in increased envelope variations,
`which sets higher linearity requirements on the power ampli-
`fier. If MC is already used or allowed in normal mode, this
`might be less of a problem though. On the other hand MC—
`CM may involve code—allocation functions in the DLI, so
`from an implementation point of view, the VSF—CM and
`CRI—CM methods seem preferable.
`
`VI. CONCLUSIONS
`
`Four methods for achieving slotted transmission in a Wide-
`band DS—CDMA system have been investigated in this paper.
`The methods, denoted Compressed Mode (CM) methods,
`facilitate inter-frequency measurements, and thus mobile—
`assisted inter-frequency handover.
`
`The methods are based on Variable Spreading Factor (VSF-
`CM), Coding Rate Increment (CRI-CM), Multi-Code (MC-
`CM) and Higher Order Modulation (HOM-CM). The Bit
`Error Rate (BER) performance of the methods has been sim-
`ulated to establish the losses of the various methods. There
`
`are method-independent losses in CM due to shorter inter-
`leaving and lost power control commands, but some methods
`also have method-specific losses which contribute to the total
`performance loss.
`
`It was found that for a compression rate of 1/2, i.e. half of the
`frame idle, the methods with the smallest performance loss
`compared to normal mode were VSF—CM and MC—CM (up
`to 1.5 dB without power control and 2.5 dB with power con—
`trol). The largest loss was seen with the HOM-CM method
`(up to 3 dB without power control and 5 dB with power con-
`trol).
`
`With a smaller compression rate of 5/16, all methods were
`not possible to use, due to limitations on bit-rate increment,
`so only CRI-CM and HOM—CM were evaluated. The losses
`were slightly smaller (up to 1 dB without PC and up to 2 dB
`with PC), and HOM-CM had slightly better performance
`than CRT-CM (up to 0.5 dB in fast fading with PC).
`
`Besides having differing performance, the methods also have
`different
`implementation implications. For example,
`the
`HOM—CM method leads to harsher phase linearity and reso—
`lution requirements as well as signal envelope variations.
`
`1. Assuming short (orthogonal) codes are used.
`
`LGE_0001236
`
`

`

`
`
`This affects both transmitter/receiver complexity and power
`consumption. Also the MC—CM method leads to increased
`envelope variations, but in that case one might assume that
`MC transmission is already included in the system specifica-
`tion for normal mode transmission. The MC—CM method
`
`may however also imply a need for code allocation in the DL
`when entering CM.
`
`It should be noted that the CM function in itself requires the
`TX power amplifier to be designed so that the peak power
`can always be increased when entering CM. For this reason,
`it is advantageous to have CM only in the downlink of the
`system.
`,
`
`In summary, when considering both performance and imple-
`mentation aspects, the following conclusions can be drawn:
`
`'
`
`'
`
`°
`
`-
`
`CM should be used only in the DL if possible
`
`For larger compression rates (around 1/2), the CR1-
`and HOM-CM methods are not suitable from a
`
`performance point of view
`
`For smaller compression rates (1/3 and below), the
`four methods have similar performance
`
`impact on
`The CM methods with the smallest
`implementation are VSF-CM and CRI-CM
`
`REFERENCES
`
`[1] A. Baier et. al, "Design Study for a CDMA-Based Third-
`Generation Mobile System," IEEE JSAC, vol. 12, no. 4, pp.
`733-743, May 1994
`
`[2] H. Eriksson et. 31., “Multiple Access Options for Cellular
`Based Personal Communications,” Proc. IEEE VI‘C’93, pp.
`957-962
`
`[3] CL. I et. al., “A Microcell/Macrocell Cellular Architecture
`for Low- and High-Mobility Wireless Users,” IEEE J-SAC,
`vol 11, no 6, pp. 885-891, Aug. 1993
`
`[4]
`
`F. Adachi et. al., “Coherent DS-CDMA: Promising Multiple
`Access For Wireless Multimedia Mobile Communications,”
`Proc. IEEE ISSSTA ’96, pp. 351-358
`
`7H1
`
`Normal
`CHLPFI=1l3
`
`' I
`'
`
`'
`
`' - - -
`'
`
`Normal
`CRI.PR=1/3
`HOM, BPSK
`
`.
`‘
`'
`
`
`
`
`BER
`
`5
`
`5.5
`Eb/No {dB}
`100 Hz
`
`s
`
`5.5
`
`Normal
`— - —
`CHLPR=1I3
`‘ 0—0 HOM, SPSK
`
`I
`'
`‘
`
`6
`
`6.5
`
`7
`
`7.5
`Eb/No 1113]
`
`8
`
`8.5
`
`9
`
`9.5
`
`10
`
`Figure 4. y: 30%, PC
`
`7H2
`
`Normal
`CR1. PFhl/Z
`MC
`HOM, ISOAM
`
`10.5
`
`11
`
`11.5
`Eli/No [as]
`100HZ
`
`12
`
`12.5
`
`13
`
`13.5
`
`14
`
`VSF, sf = 32 10
`VSF, $1 = 32 7
`
`Normal
`CR1, PR=1I2
`MC
`HOM, 160AM
`
`8
`
`9
`Eb/No [dB]
`
`10
`
`11
`
`12
`
`Figure 5. y: 50%, no PC
`
`7H2
`
`Normal
`CR1, PR=1I2
`MC
`HOM. IGQAM
`
`VSF, sf = 32 5
`
`
`5.5
`
`S
`EblNa [dB]
`
`6.5
`
`7
`
`7.5
`
`8
`
`Normal
`CRI. PR=II2
`MC
`HOM, 160AM
`VSF. $1 = 32
`
`7
`
`s
`
`9
`
`EblNo {as}
`
`10
`
`11
`
`12
`
`Figure 6. y: 50%, PC
`
`12
`
`13
`
`14
`
`..
`- - -
`"
`'l 6—-—0
`
`Normal
`ORLPR=1IS
`HOM, H’SK
`
`"
`"
`
`HOM. EPSK B
`
`
`11
`Eb/NoldB]
`100th
`
`
`10
`
`9
`
`
`S
`6.5
`7
`7.5
`8
`8.5
`9
`9.5
`10
`10.5
`11
`Eb/No [:18]
`
`Figure 3.
`
`'Y= 30%, no PC
`
`LGE_0001237
`
`235
`
`LGE_0001237
`
`