IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 48, NO.

1, FEBRUARY 1999 39

Wideband Radio Channel

Measurement System at 2 GHz
Jarmo Kivinen, Timo O. Korhonen, Pauli Aikio, Ralf Gruber,
Pertti Vainikainen, Member, IEEE, and Sven-Gustav Häggman

Abstract— This paper describes the wideband radio channel laboratory measurement devices is not preferred. For these
sounding techniques for mobile radio channel measurements. reasons the DS method was chosen to be used in the measure-
Implementation of the cross-correlation method using both a ment system described below. Both sliding correlator [4] and
sliding correlator and a matched filter detector is presented.
Limitations and accuracy of radio channel measurements are matched filter (MF) detectors [5], [6] can be used as correlation
discussed. Typically, delay resolution of about 20 ns is achieved detectors.
with 100 MHz bandwidth. With a sliding correlator, a dynamic The sliding correlator (SC) is the simplest way of imple-
range of 25 dB was obtained with maximum Doppler bandwidth menting a wideband correlator, because it eliminates the need
of 25 Hz and maximum excess delay of 19 s. Digital matched for rapid sampling and digital processing. Stationarity require-
filtering with a maximum sampling rate of 125 MHz can be used
in real-time measurements with Doppler shifts of several kilohertz ments of the channel between consecutive IR measurements
and 30 dB dynamic range. Using matched filter deconvolution as limit the usefulness of the SC, because of the inherent slowness
a resolution enhancement technique is discussed. Examples of the of it. Matched filtering was realized by both surface acoustic
time-variant complex impulse response measurements are given. wave (SAW) delay lines and by digital signal processing (DSP)
Index Terms— Antenna measurements, correlators, data pro- methods. The use of SAW delay lines degraded the dynamic
cessing, deconvolution, microwave measurements, radio commu- range of the measurement system, and they have not been used
nication, radio propagation, spread spectrum communication. in our field measurements. Improvement to the dynamic range
and flexibility of SAW-devices has been achieved lately by
I. INTRODUCTION using the SAW convolver [7], but actually no advantage has
been achieved over the DSP-receiver, which does not suffer

N EW RADIO systems need new frequency ranges to in-

crease the capacity, especially in big cities. In Europe, the
new mobile systems are developed in the UMTS framework.
from the nonidealities of analog components.
This paper is organized as follows. In Section II we consider
the requirements that are set by channel modeling to the
For UMTS frequencies around 2 GHz will be utilized. The sounding system. In Section III we introduce the measurement
HIPERLAN (wireless LAN) system will be using frequencies system at both the conceptual and practical level. The perfor-
around 5.2 and 17 GHz and MBS (mobile broadband system) mance of the concept is discussed in Section IV, followed
40 and 60 GHz. Another way of increasing the system capacity by measurement examples in Section V and conclusions in
is the use of adaptive antennas. Complex impulse response Section VI.
(IR) measurements in different propagation environments and
at different frequencies are needed for radio channel models to
characterize the channel for the design of radio systems and II. SYSTEM DEFINITION
adaptive antennas. Wideband radio channel models are either empirical,
Various wideband radio channel measurement technologies semideterministic or deterministic. Empirical modeling
have been discussed during recent years [1], [2]. Direct se- is based on statistical analysis of a large number of
quence (DS) methods are popular because of the inherent measurements. Deterministic models have two approaches: IR
processing gain achieved by averaging the continuous wave derived from the simplified environment with electromagnetic
(CW) signal in the cross-correlation process. Thus the re- simulation by, e.g., finite difference time-domain (FDTD)
quirement of transmitter peak power is lower compared to methods, and the ray-tracing approach, where dominant
conventional pulse methods (e.g., [3]). Another advantage is propagation paths are first predicted. Semideterministic
the simplicity of implementation compared to other methods. modeling uses empirical modification of deterministic models.
The transportability of the measurement system is important The radio channel can be defined by its time-variant com-
for mobile channel measurements, and thus the use of bulky plex impulse response , where and denote the
measurement time and delay, respectively, by
Manuscript received April 15, 1995; revised December 10, 1998.
J. Kivinen and P. Vainikainen are with the Helsinki University of Technol-
ogy, IRC/Radio Laboratory, FIN 02015 HUT, Espoo, Finland.
T. O. Korhonen and S.-G. Häggman are with the IRC/Helsinki University of
(1)
Technology, Communications Laboratory, FIN 02015 HUT, Espoo, Finland.
P. Aikio and R. Gruber are with Nokia Telecommuncations, FIN 00045
Nokia Group, Espoo, Finland. where is the complex amplitude of a signal arriving via a
Publisher Item Identifier S 0018-9456(99)02840-5. discrete propagation path which has an excess delay . Thus
0018–9456/99$10.00  1999 IEEE

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 1, 2009 at 06:22 from IEEE Xplore. Restrictions apply.
 40 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 48, NO. 1, FEBRUARY 1999

(1) implies the finite impulse response (FIR) model character-

ized by taps. In a statistical channel model, the distributions
of the delays, phases and amplitudes in a certain environment
are found out by measurements. Another possibility is to use
a recorded channel in radio system simulations. In this case,
real measurement results of the propagation environment are
needed.
Simple concepts that describe the physical properties of
the radio channel are often used. For example, rms delay
spread, coherence bandwidth and polarization correlation can Fig. 1. Transmitter of the measurement system.
be reported as results of a measurement. These are seldom
adequate for estimating thoroughly the performance of the
radio systems, but can be useful indicators of characteristics
of the radio channel.
Complex IR measurement results can be Fourier trans-
formed to frequency domain, delay-Doppler domain and
Doppler domain as shown in [8]. The measured maximum
Doppler frequency is upper bounded by the Nyquist
sampling theorem by the time interval between the
consecutive sampled IR’s

(2) Fig. 2. Receiver block diagram.

Generally, it is assumed, that gives the B. Receiver

maximum mobile speed , which gives the In the receiver (Fig. 2) RF front-end, the antenna signal is
ms when speed is 100 km/h and wavelength filtered ( MHz), amplified in a low-noise preamplifier
m. For mobile measurements, real vehicle speed has and downconverted to 300 MHz intermediate frequency (IF)
to be used in the measurements, because the effects of moving with the total single-sideband noise figure of 2.5 dB. The IF-
objects nearby cannot be neglected, and the Doppler spectrum stage includes automatic gain control (AGC) with computer-
extends beyond . In addition, rapidity is essential, be- controlled digital step attenuators having a dynamic range of
cause measurements with multiple antennas for, e.g., direction 72 dB, and a major part of the signal amplification.
of arrival or polarization correlation estimations, presume that The performance of the DS channel sounder is based on
the subsequent IR’s can be considered simultaneous. baseband waveform autocorrelation properties. The autocor-
The longest excess delays of significant multipaths that relation of the periodic signal with period is
can be expected in terrestrial mobile propagation seldom defined as
exceed 30 s. The delay range requirement fulfills also the
resolution requirement in the transfer function of the channel. (3)
The measurement bandwidth of the system is limited by the
physical bandpass filters to 100 MHz, which is assumed to be 1) Sliding Correlator Configuration: In the SC, a replica
adequate for most of the channel measurement applications. of the transmitted m-sequence is generated in the receiver
with a chip rate . The scaling factor is defined by
the difference of the TX and RX chip rates
III. MEASUREMENT SETUP
The measurement system consists of a pseudo-noise (PN) (4)
coded transmitter (TX) and a correlation receiver (RX), which
is controlled by a PC. Both RX and TX can be used as a The RX sequence is upconverted to the IF and correlated
mobile due to moderate size and weight. with the received signal. The complex output of the SC
is the convolution of the impulse response and the
cross-correlation of the TX and RX baseband signals
A. Transmitter
(5)
In the transmitter (Fig. 1) the 2.154 GHz carrier is modu-
lated by the m-sequence generated in the PN generator. The PN where is a function of time: and “ ” denotes
generator consisting of feedback shift registers can generate convolution. Bandwidth compression by the amount of is
31-2047 chip m-sequences. The chip rates between achieved at the cost of the measurement time. Thus
2.5 MHz and 60 MHz can be generated by a digital PLL. is limited to a few tens of hertz at maximum. Because small
Hence the delay range and resolution of the measurement can values of result in distortion [9], the SC has -dependent
be varied according to the environment. A double-balanced trade-off between and dynamic range of one IR. The
microwave mixer is used as a 2-PSK modulator. Linear output measurement thermal noise floor is also reduced by . In the
power of 40 dBm can be transmitted to the channel. measurements presented in Section V, resulted in

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 1, 2009 at 06:22 from IEEE Xplore. Restrictions apply.
 KIVINEN et al.: WIDEBAND RADIO CHANNEL MEASUREMENT SYSTEM 41

25 dB IR dynamic range and 120 dBm thermal noise floor.

The inphase (I) and quadrature (Q) time/bandwidth scaled IR
components are A/D-converted to 12 bit with maximum 2
500 kHz sampling rate. The data can be stored in real
time to the mass memory of the PC. Thus the length of the
measurement route is not limited in practice.
2) DSP Configuration: In the DSP receiver configuration,
the IF signal is converted to the baseband in the IQ-
demodulator, and the baseband signal is sampled with 2
125 Ms/s maximum sampling rate 8-bit dual-channel sampling
board. The information is then stored in the 2 4 MB buffer
memory of the board. The data sampling is performed in Fig. 3. Principle of the data acquisition in the DSP-configuration of the
receiver.
intervals lasting longer than the sequence length, allowing
optimization for a certain environment and vehicle speed. The
buffer memory data must be transferred through the PC bus to IV. LIMITATIONS AND ACCURACY OF MEASUREMENTS
the mass memory when the buffer is full. Thus the buffer mem-
ory size restricts the route length , which can be written as A. Delay Resolution and Dynamic Range
(6) The theoretical minimum delay resolution of a radar is
considered the reciprocal of the bandwidth of the measure-
where is the number of samples per wavelength , ment system
and is the number of antenna elements used in a multi-
plexed measurement. In (6), the sampling rate is considered (8)
equal to . Typically data for a few IR’s is sampled at a
time and signal-to-noise ratio is improved by averaging, which which can be achieved by the rectangular frequency domain
results in a larger value of . The complex IR is then obtained signal (as in FMCW-radar), which results in a sinc-shaped
from the raw data by off-line processing by MF. The received time-domain waveform having 13 dB sidelobes [10]. In the
complex baseband signal is convolved with the impulse DS method, the transmitted spectrum is sinc-shaped, giving
response , which is matched to the used m-sequence the best dynamic range in time domain for CW-signals with
(7) delay resolution equal to the chip pulse width of the PN code.
The dynamic range of one IR is obtained from the auto-
The thermal noise floor is reduced by the sequence length in
correlation properties of the transmitted waveform (3) giving
the convolution process. This is referred to as code gain. The
for m-sequence of length . In practice, this dynamic
noise floor is also reduced by the number of averaged IR’s.
range is not achieved. In the following, we consider some of
C. Data Acquisition and Control System the nonidealities of the system.
1) Band Limiting of the Signal: The measured IR can be
The data acquisition and control system consists of a
written as cross-correlation
portable computer (100 MHz Pentium, 32 MB RAM, 2.1 GB
hard disk) that holds three ISA bus expansion boards: one low- (9)
speed multifunction board (A/D 1 Msps and I/O functions),
a high-speed sampling board and a digital programmable Hence the measured IR is not only convolved with the auto-
timer board used to generate trigger and synchronization correlation of the used waveform, but also with the IR
signals. Additional synchronization logic circuits complete the of the RF parts (mainly band-limiting filters) of the
hardware involved in the acquisition and controlling tasks. system. In Fig. 4, we show an example of this. The simulated
The acquisition and control system performs the following autocorrelation of a 127-chip m-sequence is convolved with a
tasks. It initializes and configures the PC boards and receiver five-stage Butterworth band pass filter with 3 dB bandwidth of
hardware. It generates the synchronization signals, that are . The filter causes spurious peaks near the mainlobe. In
needed in the high-speed sampling, and samples and stores practice, the nearest peak is seen as widening of the mainlobe,
the data using the high-speed sampling board as described when the sampling rate is limited.
in Fig. 3. The low-speed sampling and data storing in SC Phase noise effects to the dynamic range are considered neg-
configuration is performed with the multifunction board. It also ligible at 2 GHz, because the phase noise standard deviation
takes care of the AGC being able to give the new AGC value in IR measurement is clearly less than 1 in the measurement
to the IF-stage every 3 ms. The AGC values are stored with system [11].
the sampled data, which allows path loss measurements with 2) Quantization Noise: Dynamic resolution of the 8-bit
estimated ±1 dB uncertainty. Also the positioning information A/D converter determines the size of the dynamic window that
can be stored. The received signal can be monitored in real is addressed by the AGC to 48 dB. Assuming a 6 dB maximum
time with the graphical user interface. In off-line processing, difference between the maximum components of consecutive
the original LabVIEW measurement files are converted into IR sets, the minimum dynamic range in the DSP-configuration
Matlab MAT-files. is predicted to be 36 dB.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 1, 2009 at 06:22 from IEEE Xplore. Restrictions apply.
 42 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 48, NO. 1, FEBRUARY 1999

Fig. 5. Comparison between (a) MF and (b) MFD in radio channel IR

estimation. Vehicle speed is approximately 50 km/h and environment type
Fig. 4. Simulated cable IR measurement, which is band limited with is typical urban.
five–stage Butterworth band pass filter, compared to a measurement with
infinite bandwidth (Tchip = 1=fchip ).

3) Matched Filter Mismatch in the SC: In (9) cross-

correlation instead of the autocorrelation has to be used
in dynamic range prediction because of the overlap caused by
slightly different chip frequencies. Generally and
. We have achieved a dynamic range over 30 dB
when and in back-to-back calibration.
Spurious correlation peaks are caused by the nonidealities
of the m-sequence generators. We have achieved a spurious
level of 40 dB with our in-house generators.
4) Distortion of the Channel: It is notable, that in practice
the dynamic range gets worse as the number of resolvable
paths increases. The measured channel affects the dynamic
range of the measurement system as is seen from (9). It can
be seen from system simulations, that equally strong and
resolvable paths reduce the achievable dynamic range of the Fig. 6. Sample of space variant IR measured at tunnel entrance with a
measurement to . Therefore, the dynamic ranges directive antenna in TX and an omnidirectional antenna in RX using the
DSP-configuration of the receiver. The excess delay range is 4.2 s. For
in cable (or back-to-back) calibration and in real measurement visualization, only 280 ns is shown here.
situations are different.
external lowpass filtering is used [14]. The difference between
B. Resolution Enhancement in the DSP-Configuration using MF (7) and MFD (10) in IR estimation is presented
in Fig. 5, where . The waveform is a 31-chip m-
Post-processing methods can be used to improve the usual
sequence with MHz and the sampling rate 100
limit of the DS-channel sounder delay resolution without
MHz.
increasing the clock frequency of the PN-code (e.g., [12]),
when is not limited by (8).
C. Doppler Resolution and Arrival Time Uncertainty
We have introduced a matched filter deconvolution (MFD)
method to be used in resolution enhancement in the DSP- The Doppler resolution, meaning the minimum difference of
configuration of the sounder [13]. In the MFD scheme the Doppler frequencies of paths with the same excess delay
is the inverse Fourier transform (IFFT) of the spectral density that can be resolved, of the measurement is theoretically the
reciprocal of the measurement time of the IR set. The un-
(10) certainty of the Doppler frequency measurement is determined
by the Allan variance of the primary standards used [15].
where capital letters refer to the Fourier transforms of the For the standards used in the system, is announced to be
respective signals, is the power spectral density of the between time intervals 0.1 s to 100 s. The
transmitted waveform and is the deconvolution constant. The standard deviation of the measured Doppler frequency is
improvement in the delay resolution is paid by the smaller given by
dynamic range and by reduction of the code gain. This trade-
(11)
off can be regulated by using external filtering or by adjusting
. Delay resolution is maximized for a very small value of . where is the carrier frequency and the factor is caused
The MFD method has code gain only for larger values of if by the quadratic summing of the errors of the two similar and

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 1, 2009 at 06:22 from IEEE Xplore. Restrictions apply.
 KIVINEN et al.: WIDEBAND RADIO CHANNEL MEASUREMENT SYSTEM 43

Fig. 7. RMS delay spread from measurement in a HST tunnel with omnidirectional antennas using the SC-configuration of the receiver.

independent primary standards. At 2.15 GHz, s, thus fulfilling the Nyquist criterion (2) for a static channel.
Hz for two consecutive IR sets ( s) with no deadtime An example of measurements performed with the SC is given
between them. Therefore, coherent measurements without a in Fig. 7, where the rms delay spread along a 400 m long
cable connection between the RX and TX can be made during route in an HST tunnel is shown. Measurement parameters
a few seconds. were , and MHz.
Similarly, the uncertainty of absolute arrival time
after time period after the calibration of the RX and TX VI. CONCLUSIONS
primary standards can be derived from the Allan variance. In this paper, the wideband radio channel sounder concepts
The is estimated constant in the flicker frequency range based on cross-correlation are investigated. Two detection
( s s) [16]. Hence methods, analog correlation with the sliding correlator and
s , which gives about 2.4 ns when off-line processing using the DSP are implemented, and the
is 10 min. performance of the sounder is analyzed considering both detec-
tion methods. The advantage of the described implementation
V. EXAMPLES OF MEASUREMENTS is the versatility to various environments. Using the DSP
The measurement examples are from a campaign performed configuration, the Doppler bandwidth is practically not limited.
in France along the TGV high-speed train (HST) line in The sliding correlator-based sounder is a very widely used
different environments like tunnels, cuttings and in open instrument due to its simplicity, but in practice, it is preferable
area. About 66 000 IR’s were measured. Omnidirectional to use the described DSP receiver in mobile radio channel
antennas were vertically polarized discone antennas having measurements. In addition, when, e.g., polarization diversity
3 dB beamwidth of 108 in the vertical plane [17]. The is under study, only rapid measurements can be used in mobile
directive antenna was a pyramidal horn with 10 dB gain. A measurements. Especially for high sampling rates, the draw-
sample of space-variant IR in an entrance of a tunnel is back of the DSP configuration is the limited route length due to
presented in Fig. 6. It is measured with the DSP-configuration the limited memory of the sampling boards. A dynamic range
having , MHz and the A/D sampling rate of 30 dB was achieved in real measurements with the DSP
of 100 MHz per channel. IR components that are more than configuration. However, when high speed is not required, the
30 dB below the highest peak are cut, which is seen as the SC is well adaptable, and due to the bandwidth compression
noise floor of the IR. The vehicle speed (RX was mobile) gives good resolution. Hence, the use of the SC can be argued
was 24 km/h, and the channel sampling interval was 0.01 for if the bandwidth of the measurement is not limited.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 1, 2009 at 06:22 from IEEE Xplore. Restrictions apply.
 44 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 48, NO. 1, FEBRUARY 1999

The future development of the measurement system includes Timo O. Korhonen was born in Kajaani, Finland,
two- and three-dimensional direction of arrival (DOA) estima- in August 1960. He received the B.Sc. degree in
telecommunications engineering from the Technical
tions with multi-element antennas [18]. Measurements have Institute of Turku, Finland, in 1986, and the M.Sc.
also been performed at higher frequencies [19]. The described degree in telecommunications engineering from the
concept has already been used at 5.3 GHz, and frequency Tampere University of Technology, Tampere, Fin-
land, in 1990.
extension to 60 GHz is also under study. From 1990 to 1992, he was with the Asian
Institute of Technology, Division of Telecommuni-
cations, Bangkok, Thailand, as a Research Scientist,
REFERENCES and beginning in 1992, he was with the Helsinki
University of Technology, Communications Laboratory, as a Researcher and
[1] A. Hewitt and E. Vilar, “Selective fading on LOS microwave links: Lecturer. His main areas of interest are the OFDM and CDMA wireless
Classical and spread-spectrum measurement techniques,” IEEE Trans. cellular communications and fiber optic communications.
Commun., vol. 36, pp. 789–796, July 1988.
[2] J. D. Parsons, D. A. Demery, and A. A. D. Turkmani, “Sounding
techniques for wideband mobile radio channels: A review,” Proc. Inst.
Elect. Eng., vol. 138, pp. 625–635, Oct. 1991.
[3] T. S. Rappaport, S. Y. Seidel, and R. Singh, “900-MHz multipath
propagation measurements for U.S. digital cellular radiotele-phone,” Pauli Aikio was born in Utsjoki, Finland, in 1959.
IEEE Trans. Veh. Technol., vol. 39, pp. 132–139, May 1990. He received the M.Sc. degree in technology from
[4] D. C. Cox, “Delay Doppler characteristics of multipath propagation Helsinki University of Technology (HUT), Espoo,
at 910 MHz in a suburban mobile radio environment,” IEEE Trans. Finland, in 1996 in electrical engineering.
Antennas Propagat., vol. AP-20, pp. 625–635, Sept. 1972. From 1996 to 1998, he was a Research Engineer,
[5] J. D. Parsons and A. S. Bajwa, “Wideband characterization of fading Radio Laboratory, HUT. Since 1998, he has been a
mobile radio channels,” Proc. Inst. Elect. Eng. F, vol. 129, pp. 95–101, Network Planning Engineer at Nokia Telecommu-
Apr. 1982. nications.
[6] P. C. Fannin, A. Molina, S. S. Swords, and P. J. Cullen, “Digital signal
processing techniques applied to mobile radio channel sounding,” IEE
Proc. F, vol. 138, pp. 502–508, Oct. 1991.
[7] W. Pietsch, “Measurement of radio channels using an elastic convolver
and spread spectrum modulation—Part I: Implementation,” IEEE Trans.
Instrum. Meas., vol. 43, pp. 689–694, Oct. 1994.
[8] P. A. Bello, “Characterization of randomly time-variant linear channels,”
IEEE Trans. Commun. Syst., vol. C–11, pp. 360–393, Dec. 1963.
[9] J. Talvitie and T. Poutanen, “Self-noise as a factor limiting the dynamic Ralf Gruber was born in Mindelheim, Germany, in 1969. He received the de-
range in impulse response measurements using sliding correlation,” in gree of Diplom-Ingenieur (FH) in electrical engineering from Fachhochschule
Proc. ISSSTA’94, Oulu, Finland, July 4–6, 1994, pp. 619–623. Kempten in 1995. From 1995 to 1997, he worked as a Research Engineer at
[10] D. R. Wehner, High Resolution Radar. Norwood, MA: Artech House, the Communications Laboratory at HUT. Since 1997, he has worked as an
1987, p. 472. SW Design Engineer at Nokia Telecommunications, Espoo, Finland.
[11] J. Kivinen and P. Vainikainen, “Phase noise in a direct sequence based
channel sounder,” in Proc. PIMRC’97, Helsinki, Finland, Sept. 1–4,
1997, pp. 1115–1119.
[12] T. Manabe and H. Takai, “Superresolution of multipath delay profiles
measured by PN correlation method,” IEEE Trans. Antennas Propagat.,
vol. 40, pp. 500–509, May 1992. Pertti Vainikainen (M’91) was born in Helsinki,
[13] T. Korhonen and S.-G. Häggman, “Location of subchip radio channel Finland, in 1957. He received the M.Sc. degree in
multipath components from direct sequence sounding data,” in Proc. technology, the Licentiate of Science in technology,
IPCWC’97, Bombay, India, 1997, pp. 43–47. and the Doctor of Science degree in technology from
[14] , “Time-resolution-signal-to noise ratio trade-off in deconvolved Helsinki University of Technology (HUT), Espoo,
and Wiener filtered radio channel sounding,” in Proc. ICUPC’97, San Finland, in 1982, 1989, and 1991, respectively.
Diego, CA, 1997, pp. 833–837. He was with the Radio Laboratory, HUT, from
[15] D. W. Allan, “Statistics of atomic frequency standards,” Proc. IEEE, 1981 to 1992, mainly as a Teaching Assistant and
vol. 54, pp. 221–230, Feb. 1966. researcher. From 1992 to 1993, he was Acting Pro-
[16] , “The measurement of frequency and frequency stability of
fessor of Radio Engineering; since 1993, Associate
precision oscillators,” NBS Tech. Note 669, p. 28, May 1975.
Professor of Radio Engineering; and since 1998,
[17] T. S. Rappaport, “Wide-band test antennas,” RF Design, vol. 11, pp.
37–41, Apr. 1988. Professor in Radio Engineering. From 1993 to 1997, he was the director of the
[18] K. Kalliola and P. Vainikainen, “Characterization system for radio Institute of Radio Communications (IRC), HUT. His main fields of interest
channel of adaptive array antennas,” in Proc. PIMRC’97, Helsinki, are RF techniques in radio communications and industrial measurement
Finland, Sept. 1–4, 1997, pp. 95–99. applications of radio waves. He is the author or coauthor of two books and
[19] J. Kivinen and P. Vainikainen, “Wideband indoor radio channel measure- about 50 refereed international journal or conference publications and the
ments at 5.3 GHz,” in Proc. EUMC’97, Jerusalem, Israel, Sept. 8–12, holder of four patents.
1997, pp. 464–469.

Jarmo Kivinen was born in Helsinki, Finland, in

1965. He received the M.Sc. degree in technology Sven-Gustav Häggman received the diploma, licentiate, and doctor degrees in
and the Licentiate of Science degree in technology communications engineering from Helsinki University of Technology (HUT),
from the Helsinki University of Technology (HUT), Espoo, Finland, in 1971, 1979, and 1991, respectively.
Espoo, Finland, in 1994 and 1997, respectively. He has been with the Communications Laboratory, HUT, since 1971
Since 1994, he has been a Research Engineer in and was appointed Associate Professor in radio communications in 1991
the Radio Laboratory, HUT. For one and a half and Professor in 1998. His activities have included research in microwave
years he was an RF Design Engineer at Nokia terrestrial radio link system planning and l.o.s. channel measurement and
Telecommunications. His main fields of interest are modeling. Presently, he is conducting research on mobile radio channel
wideband radio channel measurement and modeling measurement and modeling, radio interface techniques, and cellular radio
techniques and RF techniques in radio communica- network planning methods within the Institute of Radio Communications,
tions. HUT.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR. Downloaded on September 1, 2009 at 06:22 from IEEE Xplore. Restrictions apply.