Tsoulos George (ред.) MIMO System Technology for Wireless Communications

Подождите немного. Документ загружается.

344 MIMO System Technology for Wireless Communications

Equation 11.9, where the capacity is averaged over the number of subcarriers.

Effectively, the capacity is averaged over the number of independent paths

in the channel, which is intuitive. Large numbers of paths with signiﬁcant

separation in the delay domain are typical for urban outdoor scenarios where

the line of sight between the base station on a rooftop of a higher building

and the mobile terminal in a street canyon may be frequently blocked or

shaded so that the received signal is mostly due to multipath propagation.

The more independently faded paths are resolved, the more such a broad-

band MIMO system performs as a wired link, despite the fading in the radio

channel.

In Figure 11.18 (top), the measured bit error rates (BERs), spatially resolved

along the measurement track, are individually plotted for the three spatially

multiplexed streams using 64-QAM modulation, which is the 1 Gbit/s mode.

At the bottom of Figure 11.18, the coded error rate for the ﬁrst data stream

is plotted for the same measurement. Due to the fading, the un-coded error

rates vary along the track between 10

–3

and 3 · 10

–2

. The regions with lower

and higher error probability are reproducible also for the 16-QAM and QPSK

modes, of course at substantially reduced error numbers. Approximately at

–2

, the coded 64-QAM data are error free. Only at three positions, a minor

outage occurs in the ﬁrst stream, but the coded error rate always remains

below 10

–6

. Note that there are substantial regions along the track where all

three data streams have an un-coded error rate below 10

–2

. In these regions,

the good-put (i.e., the number of correctly detected bits per second) is larger

than 1 Gbit/s, and all three encoded streams may simultaneously show no

outage, if the channel coding is fully implemented.

In Figure 11.19 (top), we have plotted the average un-coded bit error rates

for BPSK, 16-QAM, and 64-QAM modulation vs. attenuation at the trans-

mitter. The Tx power at all antennas is set equal (±1 dB) to 0 dBm at ﬁrst.

The automatic gain control (AGC) is then regulated at a central position on

the track and then held ﬁxed for all measurements.* The Tx is moved along

the whole track for each attenuation at the transmitter.

At no attenuation, the average bit error rate in the 64-QAM mode is about

–2

, and the coded data of the ﬁrst stream have an average error rate of

1.2  10

–8

, indicating the low outage probability despite the statistical fading

in the channel. For the 3 Tx and 5 Rx antenna conﬁguration, with the MMSE

detector, one would expect a spatial diversity order of 3 in the Rayleigh

fading channel, which is comparable with the slope of the un-coded bit error

rate curve for QPSK in the high SNR region, as expected. The coded bit error

rate curves are, of course, much steeper, which is partly attributed to the

additional multipath diversity.

* The AGC is set jointly for all antennas such that the antenna signal with the largest received

power is below a threshold empirical determined by minimizing the bit error rate in all streams

due to clipping of peak amplitudes at the AD converters.

4190_book.fm Page 344 Tuesday, February 21, 2006 9:14 AM

Gigabit Mobile Communications Using Real-Time Signal Processing 345

11.10 Conclusions

We have presented a real-time concept for mobile communications that is

suitable for data rates of 1 Gbit/s, based on MIMO-OFDM. Perhaps the most

striking feature of MIMO-OFDM is that it is easily scaled to higher bandwidth,

as demonstrated in this chapter, whereby the required signal processing

complexity increases just linearly with the bandwidth (in the same channel).

FIGURE 11.18

(a) Un-coded bit error rates of the three data streams along the measurement track in 64-QAM

mode. (b) Coded error rate for stream 1 in the same measurement.

−1

−2

−3

(a)

(b)

−4

un-co

it error rate

Stream 1

Stream 2

Stream 3

Coded bit error rate

−6

−7

−8

Distance(m)

4190_book.fm Page 345 Tuesday, February 21, 2006 9:14 AM

346 MIMO System Technology for Wireless Communications

For this reason, the complexity of the spatio-temporal coding can be handled

with currently available signal processing components. Additionally, the

same fundamental approach as used for narrow-band can be applied for

broad-band applications as well. The concept may, therefore, be a good

starting point for the next generation of mobile systems, when it is further

equipped with the required multi-user functionality.

We have shown how the elementary base-band signal processing for

MIMO-OFDM can be implemented efﬁciently on a reconﬁgurable real-time

signal processing platform and integrated with the MIMO-OFDM core in a

fully functional radio system.

Mobile transmission experiments have been conducted in an indoor envi-

ronment, and the system allowed reliable transmission of data at a gross

FIGURE 11.19

Measured un-coded and coded bit error rates vs. the attenuation at all Tx antennas.

Measured uncoded BERMeasured BER after error correction

–2

–3

–4

–1

–5

–2

–3

–4

–1

–5

–2

–3

–4

–1

–5

–2

–3

–4

–1

–5

QPSK

16-QAM

64-QAM

Attenuation at all Tx antennas(dB)

4190_book.fm Page 346 Tuesday, February 21, 2006 9:14 AM

Gigabit Mobile Communications Using Real-Time Signal Processing 347

data rate of up to 1 Gbit/s. After error correction coding, the link is free of

error at almost all locations along the measurement track. Mobile data and

video transmission was also demonstrated with the experimental system.

The excellent performance of the experimental system may be attributed

to three major reasons. The indoor channel offers sufﬁciently rich scattering.

The inﬂuence of channel estimation errors is negligible due to the high

estimator gain, and the two excess antennas at the receiver enable a perfor-

mance similar to a 3 Tx, 3 Rx MIMO system using the optimal maximum-

likelihood detection instead of the linear MMSE.

But there are some points noteworthy for future research. The overall

MIMO-OFDM physical layer is currently feasible in real time for low mobility

with a moderate number of subcarriers. The limiting factor here is the large

number of matrix inversions or singular value decompositions to be per-

formed in the DSP in a fraction of the channel coherence time. With the same

number of antennas but 20 times the number of subcarriers (as recently

discussed in the 3GPP Long Term Evolution study item), about 20 times the

DSP power would be needed for the adaptation to the time-variant channel

compared to our experimental system. In the FPGA, just the memory effort

would be increased by factor 20. Fortunately, a single user is assigned to

only a fraction of the bandwidth and so the computational effort at the

terminal may still be in a realistic region for the down-link. For the up-link

at the base station, however, the full processing power must be available. A

practical way out is to perform those matrix computations in a chunkwise

manner, i.e., for subgroups of subcarriers, with multiple DSPs operating in

parallel in order to meet these challenging requirements. In this way, the

transfer times of channel estimates and weights can be reduced as well.

Also the MAC functions must be further simpliﬁed and partly imple-

mented in hardware, similar to the sample rate processing for MIMO-OFDM,

to allow such high data rates at the interface of the application in the future.

11.11 Summary

We develop a concept and describe the realization of an experimental mobile

communication system that can transmit data continuously up to 1 Gbit/s

in mobile scenarios by means of real-time base-band signal processing.

To allow for mobility, omni-directional antennas are used. This is in con-

trast to ﬁxed wireless links, which normally use directional antennas and

line-of-sight connections. A coded OFDM scheme is used to overcome the

resulting multipath fading effects. The spectral efﬁciency is boosted with

multiple transmit and multiple receive antennas. In order to identify at the

receiver the signals from multiple transmit antennas, we apply a novel

preamble enabling precise channel estimates on each OFDM subcarrier by

4190_book.fm Page 347 Tuesday, February 21, 2006 9:14 AM

348 MIMO System Technology for Wireless Communications

means of a highly parallel correlation circuit. Spatial multiplexing is indi-

vidually applied on each OFDM subcarrier. Techniques to speed up the

adaptation of the base-band processing to the time-variant channel are

described, and it is shown how the spatially multiplexed streams can be

instantaneously reconstructed.

Key functions of the concept have been realized and tested in an experi-

mental system. The real-time base-band processing is implemented on a

hybrid FPGA/DSP platform. Using a few elementary processing compo-

nents, the implementation is easily scalable with respect to the numbers of

subcarriers and antennas. For the gross data rate of 1 Gbit/s, the experimen-

tal system is conﬁgured with three transmit and ﬁve receive antennas, and

it uses 48 out of 64 OFDM subcarriers. All carriers cover a bandwidth of

100 MHz at an RF carrier frequency of 5.26 GHz.

Finally, the system is used for mobile transmission experiments. Using the

built-in channel estimator, we investigate the channel properties of an exem-

plary mobile indoor scenario. Then we report on over-the-air bit error rate

measurements with the experimental system for various modulation formats.

11.12 Acknowledgments

The authors wish to thank Stefan Schiffermueller and Clemens von Helmolt

(both from Fraunhofer HHI); Frank Luhn, Marian Pollock (from IAF GmbH);

and Matthias Lampe, Joseph Eichinger, and Egon Schulz (from SIEMENS) for

their valuable contributions in completing this project. We are grateful to the

Bundesministerium für Bildung und Forschung (BMBF) and SIEMENS for

ﬁnancial support in research projects Coverage, HyEff, 3GeT, and WIGWAM.

Appendix 11A

Here we describe an elementary interpolation scheme to improve the fre-

quency-domain channel estimates. It is based on the elementary relation

(Equation 11.2) between frequency- and time-domain channel coefﬁcients

and h

, respectively) in OFDM systems. Equation 11.2 is a set of N

equations with L variables, where L is the number of resolved multipaths.

To solve for h

, the values H

are stacked in a (1 × N) vector H where N is

the total number of subcarriers. Similar to the discrete Fourier transform,

Equation 11.2 is rewritten as a multiplication of a (N × L) matrix W with a

vector containing the time-domain channel coefﬁcients stacked in the (1 × L)

vector h

(11A.1)HWh= 

4190_book.fm Page 348 Tuesday, February 21, 2006 9:14 AM

Gigabit Mobile Communications Using Real-Time Signal Processing 349

where the elements of the matrix W are given as

(11A.2)

In practice, some values in H are not available, due to a spectrum mask

or the presence of pilot carriers. The available estimates are, hence, described

by a reduced form of Equation 11.2 as

(11A.3)

where the index red means that the rows corresponding to the missing

estimates in H and W are ﬁlled with zeros. Also, there is a vector N

red

Equation 11.A.3 describing the estimation error. In a ﬁrst step, the time-

domain estimates are obtained by using the pseudo-inverse W

red

(11A.4)

when W

red

is well conditioned. This is true if N

red

v L (interpolation rule),

where N

red

is the number of subcarriers on which channel estimates are

available. In a second step, Equation 11.2 is used to interpolate the channel

coefﬁcients.

This way we also get useful results for channel coefﬁcients on those sub-

carriers where estimates are not available. Note that the product W  W

red

has

dimension (N

× N

) but rank L, which explains the ﬁltering effect observed

in Figure 11.5 and Figure 11.6.

Appendix 11B

Here we like to mention a particular phenomenon when the signal processing

and the radio frequency (RF) front ends are combined to operate the OFDM

link over the air. Beside the IQ imbalance problems already mentioned in

the text, in addition the subcarrier assignment is mixed up at ﬁrst sight.

The proper assignment for a given RF transmitter can be tested in the lab.

Use an arbitrary waveform generator (like the SMIQ- or SMU-ARB from

R&S or a ﬂexible transmitter signal processing platform) to create time-

domain OFDM test signals from MATLAB, for instance. Load the test signal

into the generator, feed the signal into the complex-valued base-band IQ

inputs of the transmitter front end, and observe the spectrum of the trans-

mitted signal with a spectrum analyzer (as the FSQ form R&S). With our RF

front ends, we have observed the following:

= 

exp 2U

HWhN

red red red

=  +

hW H= 

red red

4190_book.fm Page 349 Tuesday, February 21, 2006 9:14 AM

350 MIMO System Technology for Wireless Communications

• Starting from MATLAB, the subcarrier index n varies from 1 to N.

• The subcarrier with index 1 is centered at the RF carrier frequency.

Due to the instable DC offset at the receiver* it is normally not used

for communication.

• The subcarriers with indices 2 f n f N/2 are mapped “en block” to

frequencies larger than the RF carrier.

• The subcarriers with indices N/2+1 f n f N are mapped “en block”

to frequencies smaller than the RF carrier such that the ﬁrst subcar-

rier in this group with index N/2+1 has the smallest RF frequency

and the last subcarrier with index N is close to the RF carrier. Intu-

itively this assignment is clear from the periodicity of the discrete

Fourier transform: if y(n) is the FFT of Y(k), then y(n) = y(n + N).

Starting from the natural placement of the subcarrier with n = 1 at

the RF carrier frequency, the observed carrier placement is explained

by continuing the FFT spectrum periodically to low RF frequencies.

• If base-band I and Q inputs are exchanged, the carrier assignment

is reversed.

• Once subcarriers are properly addressed at the transmitter such that

the rectangular spectrum of OFDM signals is realized, the order of

signals at the receiver can be checked accordingly. If the order is not

correct, changing the sign of the Q branch in time domain may be

helpful.

We have matched the OFDM signal processing according to our RF front

ends such that subcarriers are placed properly both in the RF domain at the

transmitter and in the base-band domain at the receiver. In general, take care

at the base-band-to-RF interface; design the OFDM signals accordingly. Do

not confuse base-band I and Q inputs or outputs.

* It is related to the “carrier-feedthrough” signal from the IQ modulator at the transmitter, which

is more or less faded in the radio channel and then down-converted to a very unstable DC signal

at the receiver.

4190_book.fm Page 350 Tuesday, February 21, 2006 9:14 AM

Gigabit Mobile Communications Using Real-Time Signal Processing 351

References

1. 3GPP TSG RAN Future Evolution Workshop, November 2–3, Toronto, CA,

ftp://ftp.3gpp.org/workshop/2004_11_RAN_Future_Evo/.

2. P.F. Driessen and L.J. Greenstein. 1995. “Modulation techniques for high-speed

wireless indoor systems using narrowbeam antennas,” IEEE Transactions on

Communications, Vol. 43, No. 10, Oct. 1995.

3. C. Evci, A. de Hoz, R. Rheinschmitt, M. Araki, M. Umehira, M.A. Beach,

P. Hafezi, A. Nix, Y. Sun, S. Barberis, E. Gaiani, B. Melis, G. Romano, V. Palestini,

and M. Tolonen. 1999. “AWACS: system description and main project achieve-

ments,” Proceedings ACTS Mobile Communications Summit, Sorrento, Italy,

June 8–11, 1999.

4. G. Grosskopf, A. Norrdine, D. Rohde, and M. Schlosser. 2004. “Gigabit Ethernet

Transmission Experiments at 60 GHz,“ Proceedings 13th International Plastic

Optical Fibres Conference, Nürnberg, Germany, Sept. 2004 (on CD-ROM).

5. D.R. Wisely. 1996. “A 1 Gbit/s optical wireless tracked architecture for ATM

delivery,” Proceedings IEE Colloquium on Optical Free Space Communication Links,

London, pp. 14/1–7.

6. C.B. Papadias and H. Huang. 2001. “Linear space-time multiuser detection for

multipath CDMA channels,” IEEE Journal of Selected Areas in Communications,

Vol. 19, No. 2, pp. 254–265.

7. V. Jungnickel, H. Chen, and V. Pohl. 2005. “A MIMO RAKE receiver with

enhanced interference cancellation,” Proceedings Vehicular Technology Conference

(VTC Spring), Stockholm, Sweden, May 30–June 2, 2005 (on CD-ROM).

8. G.G. Raleigh and J.M. Ciofﬁ. 1998. “Spatio-temporal coding for wireless com-

munications,” IEEE Transactions on Communications, Vol. 46, No. 3, March 1998.

9. A. van Zelst and T. C.W. Schenk. 2004. “Implementation of a MIMO OFDM-

based wireless LAN system,” IEEE Transactions on Signal Processing, Vol. 52,

No. 2, Feb. 2004, pp. 483–494.

10. W. Xiang, D. Waters, T.G. Pratt, J. Barry, and B. Walkenhorst. 2004. “Implemen-

tation and experimental results of a three-transmitter three-receiver

OFDM/BLAST testbed,” IEEE Communications Magazine, Vol. 42, No. 12, Dec.

2004, pp. 88–95.

11. C. Dubuc, D. Starks, T. Creasy, and Y. Hou. 2004. “A MIMO-OFDM prototype

for next generation wireless WANs,” IEEE Communications Magazine, Vol. 42,

No. 12, Dec. 2004, pp. 82–87.

12. K. Mizutani, K. Sakaguchi, J. Takada, and K. Araki. 2004. “Development of

4 × 4 MIMO-OFDM system and test measurements,” Proceedings 12th European

Signal Processing Conference (EUSIPCO), Vienna, Austria, Sept. 6–10, 2004 (on

CD-ROM).

13. S. Häne, D. Perels, D.S. Baum, M. Borgmann, A. Burg, N. Felber, W. Fichtner,

and H. Bölcskei. 2004. “Implementation aspects of a real-time multi-terminal

MIMO-OFDM testbed,” Proceedings IEEE Radio and Wireless Conference

(RAWCON), Atlanta, GA, Sept. 2004 (on CD-ROM).

14. V. Jungnickel, T. Haustein, A. Forck, U. Krueger, V. Pohl, and C. von Helmolt.

2004. “Over-the-air demonstration of spatial multiplexing at high data rates using

real-time base-band processing,” Advances in Radio Science, Vol. 2, pp. 135–140,

available: http://www.copernicus.org/URSI/ars/ARS_2_1/135.pdf.

4190_book.fm Page 351 Tuesday, February 21, 2006 9:14 AM

352 MIMO System Technology for Wireless Communications

15. T. Ylämurto. 2003. “Frequency domain IQ imbalance correction scheme for

OFDM Systems,” Proceedings IEEE Wireless Communications and Networking Con-

ference (WCNC), New Orleans, LA, March 16–20, 2003 (on CD-ROM).

16. G.L. Stuber, J.R. Barry, S.W. Mclaughlin, Y.G. Li, M.A. Ingram, and T.G. Pratt.

2004. “Broadband MIMO-OFDM wireless communications,” Proceedings of the

IEEE, Vol. 92, Feb. 2004, pp. 271–294.

17. T. Haustein, A. Forck, H. Gäbler, V. Jungnickel, and S. Schiffermüller. “Real

time signal processing for multi-antenna systems: algorithms, optimization and

implementation on an experimental test-bed,” accepted for EURASIP Special

Issue on MIMO Implementation Aspects, 2005.

18. F. Gantmacher. 1986. Matrizentheorie, Berlin: Springer-Verlag.

19. J. Wang and B. Daneshrad. “Performance of linear interpolation-based MIMO

detection for MIMO-OFDM systems,” Proceedings IEEE Wireless Communications

and Networking Conference, Atlanta, GA, March 21–25, 2004 (on CD-ROM).

20. V. Jungnickel, T. Haustein, A. Forck, S. Schiffermueller, H. Gaebler, C. von

Helmolt, W. Zirwas, J. Eichinger, and E. Schulz. 2004. “Real-time concepts for

MIMO-OFDM,” Proceedings CIC/IEEE Global Mobile Congress, Shanghai, China,

October 11–13, 2004.

21. M. Borgmann and H. Bölcskei. 2004. “Interpolation-based efﬁcient matrix in-

version for MIMO-OFDM receivers,” Proceedings 38th Asilomar Conf. Signals,

Systems, Computers, Paciﬁc Grove, CA, Nov. 2004 (invited).

22. D. Cescato, M. Borgmann, H. Bölcskei, J. Hansen, and A. Burg. 2005. “Interpo-

lation-based QR decomposition in MIMO-OFDM systems,” Proceedings 6th

IEEE Workshop on Signal Processing Advances in Wireless Communications

(SPAWC), New York, June 2005 (invited).

23. T. Haustein, S. Schiffermueller, V. Jungnickel, M. Schellmann, T. Michel, and

G. Wunder. 2005. “Interpolation and noise reduction in MIMO-OFDM — a

complexity driven perspective,” Proceedings ISSPA, Sydney, Australia, Aug.

28–Sept. 1, 2005.

24. http://www.prodesign-europe.com/ce/CHIPitGoldEdition.htm

25. http://www.iaf-bs.de/products/add-on-boards/

26. A.J. Paulraj, D.A. Gore, R.U. Nabar, and H. Bölcskei. 2004. “An overview of

MIMO communications — a key to Gigabit wireless,” Proceedings IEEE, Vol. 92,

No. 2, Feb. 2004, pp. 198–218.

4190_book.fm Page 352 Tuesday, February 21, 2006 9:14 AM

353

Network Planning and Deployment Issues

for MIMO Systems

Thomas Neubauer, Ernst Bonek, and Christoph Mecklenbräuker

CONTENTS

12.1 Network Planning .....................................................................................353

12.1.1 Introduction to Network Planning.............................................357

12.1.1.1 Planning and Optimizing for Advanced 3G

Technologies ...................................................................358

12.1.1.2 Advanced Planning and Optimization Methods.....359

12.1.2 Coverage and Capacity Enhancement Methods......................360

12.1.2.1 Fewer Base Stations.......................................................360

12.1.2.2 Savings in CapEx and OpEx .......................................360

12.1.2.3 Best Deployment of Budget.........................................361

12.1.2.4 Most Cost-Effective Technology Deployment ..........362

12.1.3 Base Stations with Downlink Transmit Diversity and

Beamforming..................................................................................362

12.2 Deployment ................................................................................................364

12.3 Smart Antenna Planning Example .........................................................364

References.............................................................................................................368

12.1 Network Planning

Network planning for MIMO systems is still in its infancy while the planning

of radio networks with beamforming enhancements is more mature. The

possible gain obtained from MIMO transmission and reception is currently

being intensely investigated by information theorists and baseband signal

processing experts. From a purely physical layer perspective, MIMO systems

provide three different types of gain that can be traded off against each other:

4190_book.fm Page 353 Tuesday, February 21, 2006 9:14 AM