Beruflich Dokumente
Kultur Dokumente
9
B. Developed Network Models
where R is the resistance, L the inductance, C the capacitance To demonstrate the simulation scheme, a network model
and G the conductance of the mode of propagation. It is noted based on representative 11 kV overhead networks is
that both and are frequency dependent. R and L are constructed in ATP-EMTP. The actual network is based on a
themselves frequency dependent and complicate the series of typical networks known as the UK Generic
calculation of and over wide frequency ranges. Distribution Network (UKGDS) [13]. The UKGDS is a suite
The most popular frequency dependent model within the of reference networks designed to embody the main features
ATP-EMTP is known as the J.MARTI frequency dependent of distribution networks in the UK. The emphasis in this study
model [10]. The model requires the calculation of from DC is on rural overhead networks because this represents a part of
up to a frequency where it is constant and up to a frequency the power system network operators have least real time
where it becomes negligibly small. An obvious source of error knowledge of and is particularly vulnerable to outages. For the
is in the ATP-EMTP treatment of the frequency dependence of large rural network of the UKGDS is chosen. This is a large
overhead network with branches. It is noted that some branch
R and L. For R, the skin depth will have an ever-increasing
lengths in the reference network are identical, leading to the
influence as frequency increases. To calculate the skin depth,
possibility of unrealistic multipath conditions. To make the
the ATP-EMTP assumes a tubular conductor and the forced
network more realistic, all branch lengths are randomised to
approximation that a stranded conductor is equivalent to a 10% of their original length. The test network is shown in
tubular conductor of the same cross sectional area. For Fig. 2.
stranded aluminum or copper conductors, for instance those
commonly used in MV overhead lines, the actual skin depth 0.815 km
calculation is less trivial and non negligible errors are to be C
expected if the ATP-EMTP tubular approximation is used at 0.390 km
0.745 km 0.515 km
frequencies exceeding 5-10 kHz [11]. For frequencies
exceeding 5 kHz, the ATP-EMTP uses the Galloway formula 0.510 km 0.445 km
0.790 km 1.515 km
[12] which provides a more accurate model of the skin effect
at higher frequencies. With the current state of knowledge, it 0.410 km 0.434 km
be used in interpreting the results of the presented simulation 0.521 km 0.552 km 0.535 km
0.634 km 0.593 km 0.760 km
scheme in simulations where attenuation is an important 0.815 km 0.815 km
factor.
0.634 km 0.760 km
The actual geometry of the overhead line is based on 0.773 km 0.795 km
standard wood pole designs commonly deployed in the UK.
0.815 km 0.807 km
The dimensions of the overhead line model and the ATP-
D
EMTP computations of the transformation matrix and modal 0.876 km
0.770 km
propagation constants are shown in Table 3.
0.800 km
C. Magnitude and Phase Response of Test Network The impulse responses used to derive Fig. 3 can also be
A method is developed to determine the magnitude and used to calculate the root mean squared (RMS) delay spread.
phase responses between point A and points B, C and D The first moment of the power delay profile (PDP) is defined
respectively. An impulse is injected into the network at point as:
A and the measured signal at point B, C and D is recorded.
Phase to phase coupling is used resulting in a signal that
contains only mode 2. An FFT is applied to the received 10
signals at B, C and D. The magnitude and phase of the FFT
operation can be regarded as the magnitude and phase where is alternatively called the mean excess delay. The
response of the channel. square root of the second central moment of the PDP is given
by:
Magnitude Response (Position A to Position B)
50
Attenuation (dB)
0
11
-50
-100
Another measure of multipath delay spread is the coherence
0 1 2 3 4 5 6 7 8 9 10
x 10
5
bandwidth:
Phase Response (Position A to Position B)
4
2 1
Phase (Radians)
12
0 5
-2
0
listed in Table 4.
-50
-100
0 1 2 3 4 5 6 7 8 9 10
TABLE 4
5
x 10 RMS DELAY SPREAD AND COHERENCE BANDWIDTH
Phase Response (Position A to Position C)
4
2 B C D
Phase (Radians)
-50
under various conditions. For AWGN channels (no multipath),
-100
the length of the characteristic plateau is equal to the cyclic
-150
0 1 2 3 4 5 6 7 8 9 10
5
prefix length. This is clearly shown in Fig. 4. A weakness of
the Schmidl and Cox method is the inherent uncertainty in the
x 10
Phase Response (Position A to Position D)
4
2
starting time caused by the plateau. It was observed that the
Phase (Radians)
0
start time can be chosen anywhere within the cyclic prefix
without a significant loss in performance. For multipath
-2
channels, the plateau is shortened due to the delay spread of
-4
0 1 2 3 4 5
Frequency(Hz)
6 7 8 9 10
5
the channel. This is demonstrated in Fig. 5. Problems were
x 10
encountered when using the Schmidl and Cox algorithm in
(c) noisy channels and channels with large delay spreads, even
Fig. 3. (a) Magnitude and phase response between positions A and B. (b) when the condition that the cyclic prefix is greater in length
Between position A and C. (c) Between positions A and D. than the RMS delay spread of the channel. False peaks falling
outside of the characteristic plateau led to inaccuracies. A
possible solution to this follows from a suggestion published
in [14]. A moving average filter can be used to smooth the
timing metric, improving the accuracy of the algorithm. In
6
0.9
1 Moving Average Filter
C. Effect of position on network
0.8 The effect of positioning of the remote node is examined.
0.7
The BER is obtained for communication between the base
node, A and the remote positions B, C and D. A carrier
Timing Metric
0.6
0.5
0.4
frequency of 275 kHz is chosen, with a sampling frequency of
0.3 333 kHz, FFT size of 512 and 320 used subcarriers with
0.2
DBPSK modulation. These parameters lead to a useful symbol
0.1
0
duration of 1.5 ms, a sub-carrier spacing of 650 Hz and a data
400 500 600 700 800 900 1000
Sample Number
1100 1200 1300 1400 bandwidth of approximately 200 kHz. Fig. 7, Fig. 8 and Fig. 9
show the results of the simulation. It is observed that the BER
Fig. 4. Schmidl and Cox timing metric in AWGN noise channel and the climbs to 0.5 at particular frequencies, even in channels with
response of the modified metric using a moving average filter. an extremely high SNR. This is true at all three positions,
though position A shows a better performance.
Schmidl and Cox: Multipath channel and noiseless
0.7
Moving Average Filter
1
Unchanged Schmidl and Cox metric
0.9
0.6
0.8
0.7
0.5
Timing Metric
0.6
0.5
0.3
0.3
0.2
0.1
0.2
0
500 600 700 800 900 1000 1100 1200 1300
Sample Number
0.1
Fig. 5. Schmidl and Cox timing metric in multipath channel and the modified
0
moving average metric. 0 50 100 150 200 250 300 350
Subcarrier Number
practice, a back-off time was required due to the delay of the Fig. 7. BER for the channel between positions A and B. Cyclic prefix >>
filter and the tendency of the peak of the filtered metric to fall RMS delay spread
at the very latest part of the plateau. 0.8
the centre of the line and extraction at some arbitrary point 0.3
away from the injection point. The OFDM scheme uses
0.2
DBPSK, DQPSK or D8PSK with all subcarriers used to aid
0.1
comparison with theoretical BER curves. The results are
shown in Fig. 6. 0
0 50 100 150 200
Subcarrier Number
250 300 350
0
10 Fig. 8. BER for the channel between positions A and C. Cyclic prefix >>
RMS delay spread
-1
10
0.7
0.6
-2
10
Bit Error Rate
0.5
Bit Error Rate
0.4
-3
10
0.3
0.2
-4
10 DQPSK
D8PSK 0.1
DBPSK
-5 0
0 50 100 150 200 250 300 35
10
0 2 4 6 8 10 12 Subcarrier Number
Eb/N0 (dB)
Fig. 9. BER for the channel between positions A and D. Cyclic prefix >>
Fig. 6. BER curves obtained from the simulation setup using DBPSK, RMS delay spread
DQPSK and D8PSK on an AWGN channel (no multipath). Averaged over
2000 symbols.
7
more frequency selective and gives larger and more sudden 0.5
0.4
reflected by higher BERs.
0.3
setup, the BER for various cyclic prefix lengths are obtained Fig. 11. BER comparison of DBPSK, DQPSK and D8PSK for a carrier
and shown for three carrier frequencies in Fig. 10. It is frequency of 275 kHz, occupied bandwidth of ~200 kHz (175 kHz-375 kHz),
observed that the BER minimises at a certain cyclic prefix sampling frequency of 333 kHz, FFT size of 512 and between position A and
B.
length. For the simulated system, this occurs at around 200
samples (1.2 ms). This is around twice the RMS delay spread Bit Error Rate Comparison: SNR=50 dB, Carrier Frequency=440 kHz
for the channel (see Table 4). 0.7
D8PSK
DQPSK
0.6 DBPSK
0.5
0.16
Bit Error Rate
0.4
0.14
400 kHz
0.3
300 kHz
0.12 500 kHz
0.2
0.1
0.1
Bit Error Rate
0.08 0
50 100 150 200 250 300
Subcarrier Number
0.06
0.04 Fig. 12. BER comparison of DBPSK, DQPSK and D8PSK for a carrier
frequency of 440 kHz, occupied bandwidth of ~200 kHz (340 kHz 540
0.02 kHz), sampling frequency of 333 kHz, FFT size of 512 and between position
0
A and B.
0 50 100 150 200 250 300 350 400 450
Cyclic Prefix Length (samples)
Fig. 10. Cyclic prefix length versus BER for the channel between A and B
and at carrier frequencies of 300 kHz, 400 kHz and500 kHz. The sampling
frequency is 333 kHz
8
F. Number of Subcarriers power lines, IEEE Trans. Pow. Del., vol. 22, no.1, pp. 129-134, Jan.
2007.
It is generally recognised that the sub-carriier spacing should [5] M. Gotz, M. Rapp and K. Dostert, "Po ower line channel characteristics
be less than the coherence bandwidth of the channel. For and their effect on communication systeem design," IEEE Commun. Mag.
narrowband PLC, this condition may be challenged. For vol. 50, no. 4 pp. 78-86, Apr. 2004.
[6] Y.H. Ma, P.L. So and E. Gunawen, "Performance analysis of OFDM
instance, for a system using a 512 FFT and a sampling systems for broadband power line co ommunications under impulsive
frequency of 333 kHz, the sub-carrrier spacing is noise and multipath effects," IEEE Tra ans. Pow. Delivery., vol. 20, pp.
approximately 650 Hz. This is close too the calculated 674-682, Apr. 2005.
coherence bandwidth of the channel. One waay to decrease the [7] P. Amirshahi, S.M. Navidpour and M Kavehrad,
K "Performance analysis
of uncoded and coded OFDM broadban nd transmission over low voltage
sub-carrier spacing is to use a longer FFT pperiod. For every power-line channels with impulsive noise," IEEE Trans. Pow. Delivery.,
doubling of the FFT period, the sub-carrier spacing halves. vol. 21, pp. 1927-1934, Apr. 2006.
Simulations are run to examine the efffect on BER of [8] J. H. Manton, "Dissecting OFDM: thee independent roles of the cyclic
increasing the FFT period from 512 to 10244, 1536 and 2048. prefix and the IDFT operations," IEEE E Commun.. Letters, vol. 5, pp.
474-476, Dec. 2001.
Fig. 13 shows the simulation result. [9] T.M. Schmidl and D.C. Cox, R Robust frequency and timing
synchronization for OFDM, IEEE Tra ans. Commun., vol. 45, no.12, pp.
BER Versus Number of Subcarriers (500 kHz Carrier)
0.25 1613-1621, 1997.
DQPSK
DBPSK [10] J.R. Marti, Accurate modeling of freequency dependent transmission
D8PSK
0.2
lines in electromagnetic transient sim mulations, IEEE Trans. Pow,
App.and Sys., vol. PAS-101, Issue 1, Jann. 1982.
[11] P. de Arizon and H.W Dommel, Co omputation of cable impedances
0.15
based on subdivision of conductors, IEEEE Trans. Power Delivery., vol.
PWRD-2, no. 1, Jan. 1987.
BER
[1] J. T. Tengdin, "Distribution line carrier: a historiccal perspective," IEEE H. Griffiths joined the UK electricity indu ustry in 1978 as an engineering
Trans. Power Delivery, vol. 2, pp. 321-329, Apr. 1987. apprentice and obtained a B.Sc. degree fro om the Polytechnic of Wales in
[2] A. Treytl, T. Sauter, and G. Bumiller, "Real timee energy management 1982. Between 1983 and 1990, he worked d at the South Wales Electricity
over power-lines and internet," in Proc. 8th Int. Symp. on Powerline Board and the Central Electricity Generating Board (CEGB) as an engineer in
Communications and its Applications, Zaragosa, Spain, 2004, pp. 306- distribution and transmission system design. In 1990, he was appointed to the
311. lecturing staff at Cardiff University, where hee obtained the Ph.D. degree. He is
[3] M. Zimmerman and K. Dostert, "A multi-path signal model for the currently a Senior Lecturer in the HIVES S Group. His research interests
power line," IEEE Trans. Commun., vol. 50, pp. 5553-559, Apr. 2002. include earthing systems and transients. Hee is currently a member of BSI
[4] I.S. Xiaoxian, Z. Tao, Z. Baohui, H. Zonghong, C. Jian and G. Zhiqiang, PEL/99 and chair of BSI GEL/600. He is a chartered
c engineer and a member
Channel model and measurement methods for 100-kV medium-voltage of IET.