`
`Practical Considerations of Using Antenna Diversity in DECT
`Preben E. Mogensen and Steve Petersen
`Center for Personkommunikation
`
`Fredrik Bajersvej 7A, DK-9220 Aalborg ©, Denmark
`PAX: +45 98 15 15 83 E-mail: pn@kom.auc.dk
`
`Abstract: Performance of different antenna diver-
`
`sity techniques which can be applied in DECTis inves-
`tigated. Test-equipment that complies with the DECT
`physical-layer (radio interface) has been developed in
`order
`to do measurements under
`real conditions.
`Diversity measurements were performed in three office
`buildings and two industrial buildings. We have con-
`cluded that antenna diversity (of a kind) must be em-
`ployed in DECTto mitigate the effect of both fast fad-
`ing and time dispersion.
`I. INTRODUCTION
`
`Cordless equipment made according to the Digital
`European Cordless Telecommunication (DECT) standard
`is already commercially available from a number of manu-
`facturers. The cost of deploying a DECT based telephone
`system in a large business area with multiple Radio Fixed
`Parts (RFP) will be strongly affected by the coverage range
`of each RFP. In addition to the costs of RFP equipment,
`also expenses for wiring, installation, and cellular planning
`must be considered. The coverage range of DECT equip-
`ment is often mentioned to be in the range of 50 to 100
`meters. However, professiona] users are very demanding
`and will
`not
`only
`require
`100
`percent
`coverage
`(disregarding the quality), but 100 percent coverage with a
`speech quality equivalent to the wired telephones. This pa-
`per is concerned with the useful coverage range of DECT
`in a business application, and the potential improvement by
`employing antenna diversity.
`There has been some skepticism about the coverage
`range of DECT. The transmitted poweris fixed to 250 mW
`(+24 dBm)peak during a TDMA-packet, and the receiver
`sensitivity is specified to -83 dBm. Hence, the maximum
`path loss Lyi¢ for a non-fading radio channel is 107 dB
`(antenna gains are disregarded). Because of multipath
`propagation a large fading margin must be subtracted from
`the static link budget, and in practice the maximum
`tolerated mean path loss Leageg is much smaller than Leratc:
`Antenna diversity is a well-known method to reduce the
`short-term fading probability and consequently the required
`fading margin. The theoretical diversity gain in terms of
`signal distribution is thoroughly analyzedin e.g., Jakes [1]
`for narrow band Rayleigh-distributed signals.
`In DECTthe Bit Error Rate (BER) is not only affected
`by multipath propagationin termsof signal fading, but also
`in terms of time dispersion, which leads to Inter Symbol
`{nterference (ISI). The Data rate in DECT is 1.152 Mbit
`per second (corresponding to a bit duration of 868 ns).
`
`0—7803—1927-—3/94/$4.00 © 1994 IEEE
`
`1532
`
`Impulse response measurements in typical DECT business
`environments have indicated that rms delay spread values
`in the order of 100 ns may often be present, and values
`higher than 250 ns can be found in extreme situations.
`DECT link-simulations have shown that rms delay spread
`valucs of morc than 80-100 ns will severely limit the BER
`for a standard DECTreceiver [7][8]. From previous indoor
`radio channel investigations, we have found that antenna
`diversity not only reduces the fading probability, butit also
`mitigates the effect of time dispersion for a guasi-narrow-
`band radio channel (BT < 0.1). ‘This is due to a correlation
`(approx.
`-0.5)
`between
`instantaneous
`power
`and
`instantaneous rms delay spread [2]. The following section
`gives a brief description of the DECTradio interface as a
`background for understanding the dedicated antenna
`diversity solutions for DECT.
`
`TDMA/TDDframe {10 ms, 11520 bits)
`
`Radio Fixed Part Tranmit
`Portable Part Tranmit
`a
`
`
`
`
`
`
`
`
`
`
`O}1 1273 als 6|7
`
` 14|15|16|17 18|19)20)21]
`
`|
`
`>
`
`
`
`
`
`
`
`Contiol kee]
`Pe.
`[Sync]
`Dota fed
`frendGu1s
`|__4
`1
`
`HOI a8 6
`320 a a
`
`Full siot P32 (480 bits, 417us)
`i
`“
`—+
`
`Fig. 1: The DECT TDMA/TDDStructure and the DECTfuil-slotpacket.
`
`1]. THE DECT TDMA/TDD STRUCTURE
`
`Theallocated radio frequency band for DECTis 1880-
`1900 MHz. In this 20 MHz bandthere are 10 RF carriers
`with a channel separation of 1.728 MHz. The data trans-
`mission rate is 1152 kbit/s, and the modulation methodis
`Gaussian Frequency Shift Keying (GFSK) with a band-
`width bit period product (BT) of 0.5. A RF channelis time
`divided into TDMA frames of 10 ms duration. Each
`TDMAframeis sub-divided into 24 full-slots, where the
`Radio Fixed Part (RFP) normally transmits during thefirst
`12 time slots, and the Portable Part (PP) transmits during
`the last 12 time slots. A duplex transmission (e.g., for voice
`data) uses a pair of timeslots on the same RF-carrier
`separated by 12 full-slots (5 ms) for downlink and uplink
`transmission respectively. The access technique is a com-
`bination of Frequency Division and Time Division
`Multiple Access,
`and Time Division Duplex (FD-
`TDMA/TDD). The transmitted power during a TDMA-
`
`Authorized licensed uselimited to: Fish & Richardson PC. Downloaded on May 09,2022 at 13:18:26 UTC from IEEE Xplore. Restrictions apply.
`
`4
`
`SAMSUNG 1041
`
`1
`
`SAMSUNG 1041
`
`

`

`
`
`
`
`packet is +24 dBm for both RFPs and PPs. The receiver
`
`*CRC-controlledSwitchdiversity is the simplest type of
`
`diversity, which is used in most commercial DECT
`sensitivity is specified to -83 dBm for a static channel at a
`BER of 10°[4].
`equipment. The additional complexity of the RFP is
`limited to a second antenna and two RF-switches, see
`The DECTfull-slot (e.g., used for voice data) has a du-
`ration of 480 bit
`intervals, see Fig. 1. It
`includes: A
`Fig. 2(a). The RFP will choose an antenna branch for
`Synchronization (S)-field (32 bits), which aids clock and
`transmission and reception by default, and continue
`packet synchronization, an A-field (64 bits) for signaling
`using that branch until a CRC error occurs, whereupon
`both the transmitter and receiver will switch to the other
`and contro] information, which includes a R-CRCfield (16
`bits) from a Cyclic Redundancy Code, a B-field (324 bits),
`branch for use in the succeeding TDMA/TDD frame,
`and so on. At least one CRC error must occur before a
`where the first 320 bits are available for user data (e.g.,
`switching takesplace!
`voice data) and the last 4 bits (X-field) are used for a parity
`check on some defined bits in the data-field, a Z-field (4
`bits), which is a duplication of the X-field aimed forCRC-controllede Selectiondiversity requires two full
`
`
`
`detecting sliding collision, and lastly a guard space of 56
`receiver branches and is the most complex ofthe three
`bit intervals [3][4].
`suggested diversity techniques, see Fig. 2(b). The re-
`ceived signal from both antenna branches is demodu-
`I. ANTENNA DIVERSITY TECHNIQUES
`fated, and a branch selection takes place controlled by
`DECTis well-suited for antenna diversity implemen-
`the CRC-check. The selected antenna branchis used for
`tations. Because of the TDMA/TDD access scheme de-
`transmission in the succeeding TDMA/TDD frame.
`scribed in the previous section, antenna diversity shall only
`Apart from the costs of the second receiver there is an
`be implemented at the RFP to obtain the full advantage for
`additional drawback of this diversity technique: It re-
`both
`up
`and
`downlink
`transmission. A pair of
`quires a dedicated diversity Burst Mode Controller
`TDMA/TDD-packets in a duplex connection use the same
`(BMC), which is able to handle two Rx input-signals
`RF-carrier only separated in time by 5 ms. Asthe spatial
`and perform the branch selection. We are not aware of
`movement of a PP during this time interval is small com-
`such BMC's,though they should be simple to produce.
`pared to the wavelength, the radio channel is considered
`reciprocal (time-invariant). Applying e.g., selection di-
`
`
`*DelayedRSSI-controlledSelection diversity is a hybrid
`
`versity at the RFP, the antenna branch, which was selected
`between selection and switch diversity. The second re-
`(according to a given criteria) for reception, shall be used
`ceiver branch is only implemented to the point where
`for transmission 5 ms later. Switch diversity can be im-
`RSS] is measured, see Fig. 2(c). The RFP receives only
`plemented in a similar manner, but here the switching delay
`from one of antenna branch at a time, while the RSSI
`can be found to 10 ms for the PP to the RFP transmission.
`from both of the branches is measured simultaneously.
`The validity of the approach that the radio channelis time-
`If the RSSI from the antenna branch selected for the
`invariant during a TDMA-frame depends on the user
`moment goes below the monitored branch, an antenna
`speed. A time varying channel will degrade the antenna
`switching takes place, which is used in the following
`diversity performance significantly. Simulations have
`TDMA/TDDframe.
`shown that for user speeds faster than | m/s, the gain for
`switch diversity has nearly disappeared, whereas the gain
`for selection diversity remains nearly unaffected for user
`speeds up to a couple ofmeters per second [8].
`For true narrowband systems (BT<<0.1) the Received
`Signal Strength Indication (RSSD is often used as a selec-
`tion or switching criteria, because it is a good measure for
`the signal quality. Due to the relatively high bit rate in
`DECT,the link quality may however often be limited by
`time dispersion, and the RSSI criteria will only to some
`extend reflect the signal quality [2]. It seems therefore ob-
`vious to use the CRC-check fields provided in the DECT
`TDMA-packet for a selection or switching criteria. This
`criteria will include both types of channel
`impairments.
`However, there is a complexity versus performance trade-
`off, which has to be taken into account. We therefore con-
`sider three different diversity implementations of various
`complexity:
`
`
`
`
`
`
`
`(c)
`(b)
`(a)
`Fig. 2: Block diagram ofdifferent diversity configurationsjor DECT
`IV. HARDWARE TEST BED
`
`In order to investigate the performance of DECT under
`real conditions, we have developed a real-time DECT test
`
`1533
`
`Authorized licensed uselimited to: Fish & Richardson PC. Downloaded on May 09,2022 at 13:18:26 UTC from IEEE Xplore. Restrictions apply.
`
`2
`
`

`

`
`
`bed that complies with the DECT physical-layer (radio
`interface). In addition to propagation and diversity studies
`the test bed is also used for developing advanced DECT
`data-receivers. The test bed consists at present of one
`transmitter and two receivers. The RF-technology for
`DECT has been outside the field of our interest, and all
`RF-parts have been implemented by using standard
`modular components.
`including
`data-receiver
`and
`The
`data-transmitter
`modulation or demodulation, TDMA-timing, clock and
`packet synchronization, CRC-checks, scrambling, and etc.,
`were implemented in an ADSP-2111 Digital Signal
`Processor (DSP) by means of firmware. The digital to
`analog conversion of the GFSK modulated signal in the
`transmitter, and visa versa in the receiver, was performed at
`an IF of 4.608 MHz (4-1,), and the sampling frequency f,
`was 18.432 MHz (16-1). The implementation method has
`been described in Sollenberger [6]. As the used DSP could
`not make data transfer at the sampling rate f, , an external
`buffer was used for intermediate storage of the GFSK
`modulated TDMA-packet. The type of demodulator, which
`was implemented for the diversity measurements, was a
`non-coherent differential detector
`[5]. The
`receiver
`sensitivity was measured to approx. -89 dBm at a BER of
`10°,
`
`V. MEASUREMENTSETUP ANDSITES
`
`The DECTtest-transmitter was put on a trolley together
`with battery power-supply and a turntable. The Tx antenna
`was mounted on the turntable in a radius of 0.6 m, and ina
`height of 1.8 m. During measurements the turntable rotated
`with a steady speed of 10 seconds per round(i.e., the Tx
`antenna speed was 0.38 m/s). The two DECTtest-receivers
`were put on another trolley, where the antennas were
`mounted with a 0.2 meter horizontal separation and in a
`height range of 1.8 to 2.5 m. Both the Tx and Rx antennas
`wereverticallypolarized dipoles. For each Tx, Rx position,
`6000 TDMA-packets (spaced by 10 ms) were recorded
`simultaneously for both receivers, and they were stored on
`a PC for later post-processing. In order to minimize the
`amount of collected data, the important information was
`compressed to only four bytes per TDMA-packet. The
`stored information included: Packet synchronization error,
`R-CRC error, X-CRC error, RSSI, numberof bit errors in
`the data-field, etc. Evaluation of the different antenna
`diversity techniques was performedlateroff-line.
`The sensitivity requirements in DECTis specified to -
`83 dBm at a BER of 10%. From DECTlink simulations we
`found that the bit errors mostly appear in bursts for a
`Rayleigh faded signal. As there is no error protection of the
`data field when transmitting voice data, we believe that a
`packet
`failure rate is a more appropriate performance
`criteria. We have therefore utilized the powerful R-CRC
`check provided in the DECT TDMA-packet as a perform-
`ance criteria. Simulations have shown that a BER of 103
`
`approximately corresponds to a R-CRCfailure rate of 102
`[7][8]. In order not to measure with a receiver, which have
`a much better sensitivity level than what can be expected of
`commercial DECT equipment, a 5 dB attenuation in front
`of the receiver was in-calibrated.
`Three office buildings and two industrial buildings
`were used for diversity measurements. The office buildings
`were NOVI, AUC, and JTAS, and the two industrial
`buildings were Sanistal and Silvan. In this paper, due to the
`limited space, we have decided mainly to show detailed
`results from one of the buildings, Sanistal.
`VI. RESULTS
`
`Sanist4l is a steel company. The site is a large ware-
`house for steel and sanitary items. The building is 60 me-
`ters wide and 75 meters long with an internal wall at the x
`position, -55 meters, see Fig. 3. Rows of steel shelves
`raised to the roof carrying heavy pallets are situated in the
`hall. The walls are made of reinforced concrete, some with
`a steel facade. The numerous gateways are made of alumi-
`num with only a few small windows. As described the hall
`is very closed with respect to radio waves, which should
`give basis for severe time dispersion. The test-receivers
`were situated at the x,y position (-32.5 , 15.0), see Fig. 3.
`The Tx trolley was in a systematic manner placed at 52
`different positions both inside and outside the hall. The
`measured Tx positions are in Fig. 3 shown with a symbol,
`where the gray scale of the symbol indicates the measured
`signal quality (Frame Error Rate, FER = R-CRC-failure
`rate).
`At the office building NOVI we compared the meas-
`ured R-CRC failure rates achieved from the DECTtest-
`equipment with the voice quality of a commercial DECT
`product. From the comparison we made the coarse rela-
`tionship shown in Table I. A more thorough analysis of the
`speech quality is made in [10].
`
`TABLE1.
`RELATION BETWEEN R-CRC FAILURE RATE
`AND SPEECH QUALITY
`
`
`Measured R-CRC
`Voice Quality
`
`
`
`Failure Rate
`(FER
`
`
`
`
`
`No Connection
`>0.1
`Poor /Unacceptable Quality
`0.03 - 0.1
`
`0.01-0.03
`Acceptable/Poor Quality
`
`0.003-0.01
`Good/Acceptable Quality
`
` good Quali
`
` < 0.003 Very
`
`
`In Fig. 3(a) the measured failure rate is shown for No
`diversity. We estimate in this case that there will be full
`DECTcoverage (i.-e., connection can be established) inside
`the hall, but the speech quality will be unacceptable. If R-
`CRC Switch Diversity,
`is applied, see Fig. 3(b), we
`estimate that the hall is covered with a poor to an ac-
`ceptable speech quality. For RSSI Selection Diversity, see
`
`1534
`
`Authorizedlicensed uselimited to: Fish & Richardson PC. Downloaded on May 09,2022 at 13:18:26 UTC from IEEE Xplore. Restrictions apply.
`
`3
`
`3
`
`

`

`
`
`is covered with an
`Fig. 3(c), we estimate that the hall
`acceptable to good speech quality, and if using R-CRC
`Selection Diversity, see Fig 3(d), the entire hall and most of
`the outside positions are covered with a very good speech
`quality.
`In Fig. 4 the R-CRC failure rate is shown versus re-
`ceived signal
`strength. The two individual
`receiver
`branches are shown with triangles. It can be observed that
`the R-CRC failure rate crosses the 1 percent
`level at
`approx. -60 dBm, and that the failure rate remains just
`below 1 percent even for very high signal strengths. For R-
`CRC Switch Diversity the 1 percentlevel is reached at app.
`-65 dBm,and the errorfloor for high signal strengths is re-
`duced to approx. 0.3 percent. For RSS! Selection Diversity
`the performanceis slightly better than for R-CRC switch
`diversity, but there is no significant improvement. For R-
`CRC Selection Diversity the failure rate cross the 1 percent
`at approx. -75 dBm, and the failure rate declines fast down
`to 0.01 % or better for higher signal strengths. Note that the
`diversity gain is negligible for signal strengths below -80
`60
`
`60
`
`0
`
`x”
`
`Ez
`
`;
`
`2 20
`>
`10
`
`Q
`
`10
`2 0 “70
`
`®
`
`6
`
`e
`
`e
`
`e
`
`@
`
`&
`i)
`
`40 w 20
`60
`x distance [m]
`(b) R-CRC Switch Diversity
`
`10
`
`0
`
`10
`
`dBm. Fig. 5 shows a very severe position from the NOVI
`site, where the DECTlink-quality is limited by time dis-
`persion. The R-CRC failure rate is measured versus
`transmitted power for a fixed (Rx, Tx) position.
`The x-axis is the transmitted power in dB relative to
`24.5 dBm, which was full power, and the Y-axis is the
`measured R-CRC failure rate. Due to ISI the failure rate
`reaches an irreducible error floor of 8 percent
`if not
`applying diversity, but if using R-CRC Selection Diversity,
`the error floor is well below 1 percent. R-CRC Switch
`Diversity and RSSI Selection Diversity give similarfailure
`rates in an ISI limited situation, though, it should be men-
`tioned that the failure patterns are different, which may
`have impact on the subjective speech quality [10]. Similar
`types of measurements at other locations have shown that
`RSSI controlled selection diversity perform equivalent to
`R-CRC controlled selection diversity in situations, where
`the link quality is dominated by lack ofsignal strength.
`When antenna diversity is applied to DECT,the cover-
`range will mostly be
`limited by the received
`
`age
`
`-70
`
`40 3 20
`-60
`x distance {m]
`
`-10
`
`0
`
`10
`
`(c) RSSI Selection Diversity #0
`(d) R-CRC Selection Diversity
`
`(a) No-Diversity
`
`
`ad
`
`00
`
`10
`
`Q
`
`10
`
`40 0 -20
`-70 a 60
`x distance {m]
`Fig. 3: Measured R-CRC Failure Rates (FER) atthe test site Sanistal
`shownfor different antenna diversity techniques. The Building layoutis
`shownby thick solid lines.
`
`the affect of time
`signal strength, as diversity combat
`dispersion at nearly all the measured positions. A simple
`path loss model can therefore be a useful tool for planning
`
`1535
`
`Authorized licensed uselimited to: Fish & Richardson PC. Downloaded on May 09,2022 at 13:18:26 UTC from IEEE Xplore. Restrictions apply.
`
`4
`
`4
`
`

`

`
`
`a DECTcellular layout. Different indoor path loss models
`have been proposedin theliterature. In the model proposed
`by Devasirvatham [9] the path loss is modeled by free-
`space propagation plus a linear attenuation factor a, which
`can be adjusted depending on the building structure and
`materials:
`
`Pru = 24 dBm - 38.4- 20logd-a*d +Gt&+Grnx
`
`(1)
`I, = K, + 20logd+a-d
`K; is a frequency dependent constant anddis distance. In
`Fig. 6 this model has been applied to the measured data,
`shown for various a values. Except for someofthe 'outside'
`Tx positions, the model with an a value of 0.7 gives a good
`lower bound for the measured signalstrengths.
`
`[dBm] 40
`RecetvedSignalStrength
`
`
`
`Distance [m]
`Fig. 6: Measured RSSI comparedto the prediction model in Eq. (1)
`
`VII. CONCLUSIONS
`
`Based on the measurementresults shown in the paper
`together with similar results from the other measured sites
`we conclude the following:
`Antenna diversity (of some kind) must be applied in
`DECT, otherwise the range where a solid coverage can be
`expected, will be less than 15-25 meters in an indoor busi-
`ness environment. This range can be extended up to 30-45
`meters by applying CRC Selection Diversity.
`«
`Ina businessapplication, seen from a performance ver-
`sus price criteria, we believe that the optimum choice of di-
`versity technique is CRC Selection Diversity, as the
`performance gain over both fast fading and time dispersion
`is superior. We do not find it worthwhile to apply RSSI
`Selection Diversity, compared to using the simple CRC
`Switch Diversity technique as
`the gain difference is
`marginal.
`
`layer", DE/RES
`
`VIII.. REFERENCES
`[1] Jakes, William C, "Microwave Mobile Communications". John
`Wiley & Sons, Inc. 1974, ISBN 0-471-43720-4
`(2] Mogensen, P. E., “Wideband Polarization Diversity Measurements
`for Wireless Personal Radio Communications", Proc. of Nordic Radio
`——¥— Branch 1
`go
`:
`9
`—*— Branch 2
`Symposium, Aalborg 1992, pp. 49-53.
`
`t4 —®— R-CRC Selection|-
`[3] ETSI DECT recommendation Part 3: "Medium access control layer",
`—— R-CRC Switch
`:
`DE/RES 3001-3, Feb. 1992 .
`—?— RSSI Selection
`[4] ETSI DECT recommendation Part 2: "Physical
`3001-2, Feb, 1992
`[S] "DSP Implementation of a DECTreceiver", internal student report:
`S860Te, June 1993, Aalborg University ( in English).
`[6] Sollenberger, N. R. et al, " Architecture and Implementation of an
`Efficient and Robust TDMA Frame Structure for Digital Portable
`Communications", IEEE-VT, Vol. 40 No. 1, Feb. 1991.
`(7] Lopes, L. B. and M.R. Heath, "The Performance of DECT in the
`Outdoor 1.8 GHz radio channel", Proc. of IEE, Mobile Radio and
`Personal Communications Conference 1991, pp. 300-307.
`[8] Mogensen, P. E., "Practical considerations of using Antenna di-
`versity in DECT", COST231 TD(93)122.
`[9] Devasirvatham, D., et al., "Multi-Frequency Radiowave Propagation
`Measurements in the Portable Radio Environment”, Proc. of IEEE
`conference ICC '90, pp. 1334-1340
`[10]"Compensation for Click Noise in DECT caused by Multipath
`Fading", internal student report: $720, Jan 1994, Aalborg University (in
`English)
`
`0.0001
`1 2
`
`Branch 1
`Branch 2
`R-CRC Selection}
`R-CRC Switch
`RSSI Selection
`
`-
`|
`z
`
`0.1
`.
`
`(0.01
`
`0.001
`
`2
`«
`
`223
`
`uo
`oO
`
`«9@
`
`Received Signal Strength [dBm]
`Fig. 4: Measured R-CRCfailure rate versus received signal strengthfor
`various antenna diversity techniques.
`
`e o.
`
`£2‘
`
`auo
`
`o©
`
`0.001
`-30
`
`-16
`-20
`-26
`+10
`5
`Relative Transmitted Power [dB] (ref.: +24.5 dBm)
`Fig. 5: Measured R-CRCfailure raie versus transmittedpowerfor a
`Fixed Rx and Tx position
`
`If R-CRC Selection Diversity is used, a mean signal
`strength of approx. -75 dBm is required (9 dB fading
`margin), and the coverage range in this type of building can
`be predicted to be in the order of 40-45 meters. If no
`diversity is applied the coverage range where an acceptable
`speech quality can be achieved, will be less than 25-30
`meters. These coverage range predictions agree very well
`with results measured at the other industrial site, Silvan.
`For the three office building sites the coverage range is
`even shorter.
`
`1536
`
`Authorizedlicensed uselimited to: Fish & Richardson PC. Downloaded on May 09,2022 at 13:18:26 UTC from IEEE Xplore. Restrictions apply.
`
`5
`
`5
`
`