U.S. patent application number 14/589290 was filed with the patent office on 2015-07-02 for method for multi-carrier frequency division multiplexing transmission.
The applicant listed for this patent is Vodafone Holding GmbH. Invention is credited to Steffen Bittner, Gerhard Fettweis, Marco Krondorf.
Application Number | 20150188654 14/589290 |
Document ID | / |
Family ID | 42026378 |
Filed Date | 2015-07-02 |
United States Patent
Application |
20150188654 |
Kind Code |
A1 |
Fettweis; Gerhard ; et
al. |
July 2, 2015 |
METHOD FOR MULTI-CARRIER FREQUENCY DIVISION MULTIPLEXING
TRANSMISSION
Abstract
A general frequency division multiplex (GFDM) transmission
system is proposed. Vacant frequency ranges are detected and
subsequently used for transmission, wherein a single carrier
transmission system with cyclic prefixing is deployed. A
corresponding transmitter and receiver are disclosed.
Inventors: |
Fettweis; Gerhard; (Dresden,
DE) ; Krondorf; Marco; (Dresden, DE) ;
Bittner; Steffen; (Goethestr. 11, DE) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Vodafone Holding GmbH |
Dusseldorf |
|
DE |
|
|
Family ID: |
42026378 |
Appl. No.: |
14/589290 |
Filed: |
January 5, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12640428 |
Dec 17, 2009 |
8929352 |
|
|
14589290 |
|
|
|
|
Current U.S.
Class: |
370/344 |
Current CPC
Class: |
H04L 5/003 20130101;
H04J 11/0023 20130101; H04L 27/2653 20130101; H04L 27/2607
20130101; H04L 27/2637 20130101; H04L 5/06 20130101; H04L 5/0007
20130101; H04L 27/2614 20130101; H04J 2011/0009 20130101; H04L
27/0006 20130101 |
International
Class: |
H04J 11/00 20060101
H04J011/00; H04L 5/00 20060101 H04L005/00; H04L 27/26 20060101
H04L027/26; H04L 27/00 20060101 H04L027/00 |
Foreign Application Data
Date |
Code |
Application Number |
Dec 18, 2008 |
EP |
08022010.6 |
Claims
1. A method for data transmission by frequency division
multiplexing (FDM), comprising the steps of: detecting vacant
frequency ranges in a plurality of allocated frequency ranges of a
transmission frequency band; defining a center frequency for each
of said detected vacant frequency ranges as a frequency divisional
multiplex carrier frequency, and simultaneously generating
frequency divisional multiplex transmit signals for transmission
for each carrier frequency wherein said signal generation is
performed entirely within the digital domain, wherein each carrier
is modulated individually with an individual bandwidth and an
individual pulse shaping; adding the simultaneously generated
digital frequency divisional multiplex transmit signals to a single
digital transmit signal; and converting said single transmit signal
from digital to analog and transmitting said analog signal within
said detected vacant frequency ranges.
2. The method of claim 1, wherein said plurality of frequency
divisional multiplexed carriers are non-orthogonal to each
other.
3. The method of claim 1, wherein individual modulating comprises
selecting individual transmit filters for each carrier such as to
minimize mutual interference between said frequency divisional
multiplexed carriers.
4. The method of claim 1, wherein said step of generating transmit
signals comprises inserting a cyclic prefix.
5. The method of claim 4, wherein a tail-biting technique is
performed to shorten the length of the cyclic-prefix.
6. The method of claim 1, wherein said vacant frequency ranges
comprise unused frequency ranges in radio frequency bands such as
UHF television frequency band and sensor network bands.
7. The method of claim 1, wherein the step of detecting vacant
frequency ranges in a plurality of allocated frequency ranges of a
transmission frequency band is performed by a receiver
communicatively coupled to a transmitter.
8-13. (canceled)
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is a divisional of copending application
U.S. Ser. No. 12/640,428, filed on Dec. 17, 2009, now U.S. Pat. No.
8,929,352, which claims under 35 U.S.C. .sctn.119(a) the benefit of
European Application No. 08022010.6, filed Dec. 18, 2008, the
entire contents of which are incorporated herein by reference.
BACKGROUND
[0002] (a) Technical Field
[0003] The present invention relates to a method for data
transmission employing a frequency division multiplexing technique
and a corresponding apparatus for multi-carrier frequency division
multiplexing transmission.
[0004] (b) Description of the Related Art
[0005] The exponential growth of communication traffic of the
recent years that calls for an optimized utilization of the radio
frequency spectrum has drawn scientific and industrial attention on
a search for agile mechanisms that may use "holes" or "white
spaces", i.e. vacant frequency bands, in the radio spectrum for
radio transmission.
[0006] A major technical challenge in exploiting white spaces in
the radio spectrum, e.g. in the TV UHF bands, which are located in
close proximity to allocated spectrums is the fragmentation of the
spectrum. Typically, TV white spaces (TVWSs) are not consecutively
placed in the spectrum; rather, the UHF TV spectrum exhibits a
strong spectrum fragmentation. In order to efficiently exploit all
detected TV white spaces, a system is required which can cope with
strong spectrum fragmentation and which is able to perform
aggregation of several TVWS by one single wideband signal. Another
example for a radio spectrum that may exhibit unused spaces is in
low power sensor networks. The design of a transceiver architecture
and concept, allowing to opportunistically exploit white spaces of
allocated radio frequency bands for wireless data communications is
particularly difficult, mainly for two reasons. On the one hand,
signal generation should ensure ultra-low out-off-band radiation to
strictly avoid harmful interference to legacy TV signals. On the
other hand, the receivers and/or transmitters should exhibit high
sensitivity in order to explore white spaces, i.e., to sense even
very weak radio signals in allocated frequency bands.
[0007] Recent studies indicate the usage of multi-carrier systems
to flexibly exploit spectrum vacancies by switching sub-carriers on
and off. However, the proposed multi-carrier system OFDM
(Orthogonal Frequency Division Multiplexing) is known to cause
strong spectral leakage even when using pulse shaping techniques or
guard carriers.
[0008] What is needed in the art, therefore, is a transmission
scheme that combines the flexibility and simplicity of OFDM with
stronger interference reduction mechanisms.
SUMMARY
[0009] According to the invention there is provided a method for
data transmission by frequency division multiplexing (FDM). The
invention further provides a corresponding transmitter arrangement
and a receiver arrangement.
[0010] The inventive approach exhibits some attractive features
which are of particular importance for scenarios exhibiting high
degrees of spectrum fragmentation. Specifically, the generalized
digital multi-carrier transmitter concept according to the
invention provides a lower peak-to-average power ratio (PAPR)
compared to OFDM, a ultra-low out-of-band radiation due to
adjustable Tx-filtering, and last but not least a block-based
transmission using cyclic prefix insertion and efficient FFT-based
equalization. The novel transmission concept of the invention
enables frequency and time domain multi-user scheduling comparable
to OFDM and provides an efficient alternative for white space
aggregation even in heavily fragmented spectrum regions. Unlike in
OFDM, the FDM carriers are no longer necessarily orthogonal to each
other. Below, the novel transmission concept will also be referred
to as Generalized Frequency Division Multiplexing (GFDM).
BRIEF DESCRIPTION OF THE DRAWINGS
[0011] Additional features and advantages of the present invention
will be apparent from the following detailed description of
specific embodiments which are given by way of example and in which
reference will be made to the accompanying drawings, wherein:
[0012] FIG. 1 shows a fragmented TV UHV frequency spectrum with
white spaces as an example for frequency ranges that may be used
for transmission according to the concept of the invention;
[0013] FIG. 2 shows a schematic block diagram of a preferred
digital transmitter according to the invention;
[0014] FIGS. 3A-3D illustrate the principle of CP shortening by
tail biting;
[0015] FIG. 4 shows a schematic block diagram of a preferred
digital receiver for utilization in conjunction with the GFDM
concept;
[0016] FIG. 5 shows a graph of bit error rate performance for QPSK
modulation, assuming a AWGN channel; and
[0017] FIG. 6 shows a graph of the peak-to-average power ratio both
for OFDM and the GFDM concept according to the invention.
DETAILED DESCRIPTION
[0018] FIG. 1 illustrates a frequency range 100, which in one
embodiment may be the UHF frequency range traditionally used for
the transmission of television signals 110-112. These TV signals
may occupy an individual bandwidth and may be of individual signal
power as indicated by the height of the signal blocks. As
illustrated, the TV signals 110-112 are not bordering each other,
thus leaving vacant frequency ranges 113-115 between the TV
signals. These vacant frequency ranges 113-115 are also called TV
white spaces (TVWS) in the following.
[0019] TV white spaces 113-115 may be used for the transmission of
arbitrary signals, wherein the arbitrary signals must not interfere
with adjacent signals, i.e. the arbitrary signals transmitted in
the TV white spaces must not leak into adjacent frequency bands
used for other, e.g. TV, signals. Accordingly the signals
transmitted must exhibit a low out-off-band radiation to avoid
harmful interference to legacy signals of adjacent frequency
bands.
[0020] For finding vacant frequency bands, which could be used for
subsequent signal transmission, the transmitter or the receiver or
both may scan potential frequency ranges. Vacant frequency ranges
can be detected for example by investigating the energy transmitted
on a frequency. Note that the transmitter and/or receiver scans the
potential frequency ranges very carefully in order to detect even
weak signals and in order to prevent any interference with detected
signals.
[0021] Once vacant frequency ranges are detected, the transmitter
or the receiver determines the vacant frequency ranges to be used
for subsequent transmission. In one embodiment the transmitter and
the receiver both may scan potential frequency ranges and may then
coordinate the search for vacant frequencies to ensure that
frequencies are used, which have been detected as vacant by both
the transmitter and the sender.
[0022] Once a vacant frequency has been determined, the bandwidth
and the center frequency of frequency range are determined by
either the transmitter or the receiver or in cooperation. Since a
single carrier transmission method will be used, the center
frequency of the determined vacant frequency range determines the
carrier frequency and the bandwidth limits the baud rate of the
transmission.
[0023] In one embodiment the receiver informs the transmitter about
possible center frequencies and available bandwidths. On the other
hand the transmitter informs the receiver about details for the
following signal transmission accordingly. That is, the transmitter
informs the receiver about all the details necessary for
transmitting data, i.e. the modulation type, the baud rate, coding
mode etc. to be used for the following transmission. Accordingly an
established communication channel between the transmitter and the
receiver is assumed, which can be used for transmitting the
determined detail information of the subsequent transmission from
the receiver to the transmitter and back. Said established
communication channel in one embodiment may be a signaling channel
for distributing information to a plurality of receivers.
[0024] The transmission using the detected vacant frequency range
stops if at least one of the following criteria is matched. In one
embodiment, the transmitter may listen on the used frequency range
for unexpected signals, which indicate that there is another
transmitter using the frequency range. For this purpose the
transmitter may comprise corresponding means for listening. The
transmitter may inform the receiver correspondingly about the
detected interference using above mentioned established
communication channel. Alternatively, the transmitter may stop the
transmission on the detected frequency range, if the receiver
indicates that it cannot receive, i.e. decode, the transmitted
signals successfully, which may due to interfering signals at the
receiver. The receiver may indicate to stop the transmission of
signals on the vacant frequency range for example directly by a
signal sent on the above mentioned established communication
channel or by not sending an acknowledge message in the vacant
frequency range.
[0025] FIG. 2 shows a schematic block diagram of an embodiment of a
digital transmitter 200 capable of transmitting an incoming stream
210 of information bits using at least one or a plurality of
detected TV white spaces simultaneously.
[0026] Processing block 220 performs channel encoding on the
incoming stream of information thus producing a stream of channel
encoded bits. The channel encoding may comprise, for example, the
removal of redundant information or may comprise the addition of
error correcting information, for example a cyclic redundancy check
(CRC) code. The stream of encoded bits is then mapped to transmit
symbols, thus producing a stream of transmit symbols representing
the stream of channel encoded information bits, wherein said
symbols may be QAM or alternatively QPSK symbols.
[0027] The stream of transmit symbols is passed to processing block
230, which maps the symbols to at least one or a plurality of N
carrier frequencies, wherein the n carrier frequencies are center
frequencies of vacant frequency bands. In transmitter 200 a
plurality of n of the N parallel processing branches is used,
wherein n corresponds to the number of detected vacant frequencies.
Each of the symbols is then passed to one of a plurality of N
parallel processing branches 240. Since each of the branches
comprises the same processing blocks, the following description
relates to all branches.
[0028] Each branch comprises a processing block 241.sub.n for
adding a cyclic prefix (CP) to each symbol, an up-sampling block
242.sub.n and a digital filter 243.sub.n for pulse shaping each
symbol, processing block 244.sub.n for performing a tail-biting
technique and a complex multiplier 245.sub.n for digitally
frequency shifting a symbol relative to the at least one other
carrier frequency of another used parallel branch.
[0029] Every carrier is modulated individually, using some
arbitrary form of modulation. In the following the QAM symbol
stream on carrier n is denoted as s(n,k), where k represents the
symbol index. After upsampling in block 230, the symbol's index now
turns into the sample index k', representing the sample duration
T.sub.s.
[0030] Subsequently, cyclic prefix (CP) insertion is performed,
accounting for the filter length of the digital pulse shaping, the
filter length of the digital receive filter and the length of the
mobile channel impulse response. The insertion of cyclic prefixes
is used to allow for low complex equalization at the receiver side.
After cyclic prefix insertion, digital pulse shaping is performed
in blocks 242.sub.n, 243.sub.n carrier-wise, wherein each symbol is
up-sampled in block 242.sub.n and each up-sampled symbol s(n, k')
of the n-th carrier is convoluted with the transmit filter function
of the n-th branch g.sub.Tx(n,k'):
s(n,k')*g.sub.Tx(n,k') (1)
[0031] This pulse shaping is crucial for low out-off-band
radiation, the digital filters correspondingly exhibiting sharp
filter edges, which in turn necessitate high transmit (Tx) filter
orders. Large filter orders are generally problematic due to the
cyclic prefix, which has to be matched to the aggregate filter
lengths of all system filters involved. However, at least for the
digital Tx-filter, the principle of tail-biting as originally
introduced by H. Ma and J. Wolf in "On Tail Biting Convolutional
Codes"; IEEE Transactions on Communications, Vol. 34, Issue 2,
February 1986, for convolutional codes, can be optionally applied
in order to reduce the CP overhead as depicted in FIG. 3 and
explicated below.
[0032] After individual pulse shaping, each carrier n is digitally
frequency shifted to its relative carrier frequency f.sub.n by a
complex multiplier 245.sub.n, which is normalized to the signal
bandwidth B=1/T.sub.s, i.e. the symbols are frequency shifted
relative to each other. Note that this relative frequency shift is
maintained when subsequently mixing the signal to a carrier
frequency in the analog processing stage. A signal produced by one
branch is hence defined in the interval [-1/2:1/2] centered around
the carrier frequency. The resulting time domain signal x(k') hence
becomes
x ( k ' ) = n ( s ( n , k ' ) * g Tx ( n , k ' ) ) j 2 .pi. k ' f n
. ( 2 ) ##EQU00001##
[0033] The digital time domain signals of all used branches are
then combined by a complex adder 246 to form a single digital
signal representing the combination of all digital time domain
signals, which is passed to block 250.
[0034] Finally, in block 250 each time domain signal x(k') is
digital-to-analogue converted, mixed to the carrier frequency and
amplified to form one transmit signal. In case of using at least
two parallel branches 240, i.e. two carrier frequencies shifted
digitally relative to each other as mentioned above, the resulting
single transmit signal is a wideband transmit signal. Said wideband
transmit signal is transmitted via antenna 260.
[0035] Note that the used carriers may not be orthogonal to each
other. Rather, each carrier is modulated individually thus forming
a single carrier transmission using cyclic prefixing. Accordingly,
i.e. if at least two or more vacant frequency ranges are detected
and used, the wideband signal as transmitted by antenna 260
comprises a plurality of single carrier signals each using cyclic
prefixing.
[0036] It is well known that OFDM uses a cyclic prefix (CP) to
simplify signal equalization but on the other side exhibits a high
peak-to-average power ratio (PAPR) as one major drawback. A lower
PAPR and simple equalization is achieved when using a cyclic prefix
in conventional single carrier systems, leading to the SC-CP
(single carrier with cyclic prefix) system concept. Hence, from the
above given GFDM signal generation scheme the proposed frequency
division multiplexing concept may be interpreted as a system of
parallel single-carrier cyclic-prefixed signals(SC-CPs) realized in
the digital domain.
[0037] The resulting transmit signal exhibits the following
properties. Each carrier represents an independent SC-CP link,
which can be modulated individually having an own bandwidth and
pulse shaping.
[0038] In contrast to well known OFDM systems, the frequency
division multiplexed (FDM) carriers are not orthogonal but exhibit
mutual interference which can be adjusted by the individual
transmit (Tx) filters and signal bandwidths.
[0039] Due to the digital implementation of the signal processing
in the parallel branches, the GFDM concept combines the advantages
of the simple equalization with the ability of flexible allocation
of white spaces, i.e. vacant frequency bands, and controllable
out-off-band radiation. Furthermore the transmitter architecture
allows an easy equalization despite the wideband nature of the
transmit signal, a frequency agile white space allocation and
flexible signal bandwidth. Due to the digital implementation of
each branch 240, the requirements of the analogue front-end are
reduced.
[0040] FIGS. 3A to 3D schematically depict the tail-biting
technique for a digital transmit signal 300 as deployed in
processing block 244.sub.n of FIG. 2. Transmit signal 300 consists
of payload samples, i.e. portion 310, and samples of a cyclic
prefix 320, wherein conventional cyclic prefix samples are the last
samples of the payload copied to the beginning of the signal to
produce a symmetric signal. The length of cyclic prefix is
determined by the properties of the transmit filter, the transmit
channel and the receive filter as schematically indicated by
portions 330-350.
[0041] In contrast to a signal comprising a conventional cyclic
prefix 320 as depicted in FIG. 3A, the first portion 330 reflecting
the length of the transmit (Tx) filter is neglected, thus
shortening the length of the cyclic prefix as depicted in FIG.
3B.
[0042] FIG. 3C depicts the signal after having passed the transmit
filter processing block 243.sub.n, which appends some digital
samples 350 to the signal.
[0043] The samples 360 appended by the transmit filter processing
are cut off and superimposed to the first samples of the shortened
cyclic prefix 320, thus emulating a circular convolution to
assimilate the beginning of the signal block to its end.
[0044] FIG. 4 depicts a preferred embodiment of a receiver 400
according to the concept of the invention, which performs parallel
single-carrier cyclic-prefix (SC-CP) demodulation for each of the
carriers.
[0045] Receiver 400 comprises an antenna 410 for receiving an
incoming radio signal, an analog input stage 420 and a plurality of
N parallel digital processing paths 430 coupled to the output of
input stage 420, a carrier de-mapping block 440 coupled to the
output each of the N parallel processing paths and a decoder block
450 outputting a stream of information bits 460, which ideally are
those as input to a corresponding transmitter.
[0046] The signal as received by antenna 410, i.e. the receive
signal, is passed from antenna 410 to the analog input stage 420,
which amplifies the receive signal typically by using a low noise
amplifier (LNA), and down-converts the amplified receive signal to
a frequency suitable as input to a digital-to-analog converter
comprised in stage 420. Then the amplified and down-converted
receive signal is analogue-to-digital converted thus yielding a
digitized receive signal y(k').
[0047] Here, the digitized receive signal y(k') is given by the
discrete convolution of the transmitted signal (Tx-signal) x(k')
with the channel impulse response h(k'), corrupted by zero mean
additive white Gaussian noise (AWGN) n(k') with variance
.sigma..sub.n.sup.2:
y(k')=x(k')*h(k')+n(k'), (3)
[0048] The signal to noise ratio (SNR) y can be defined as:
.gamma. = E { x ( k ' ) * h ( k ' ) 2 } .sigma. n 2 .
##EQU00002##
[0049] The receive signal is then coupled to n of the N parallel
digital processing paths 430, wherein n is the number of used
carrier frequencies. Since all of the N parallel digital processing
paths 430 comprise the same functionality, the following
description relates to each. In each of the n paths the receive
signal is mixed to an individual carrier frequency f.sub.n,
individually in each of the n used parallel branches by deploying a
complex multiplication in multiplier 431.sub.n.
[0050] Each down-converted digital receive signal is filtered
subsequently by a filter 432.sub.n, thus yielding the signal
z(n,k'), which can be expressed mathematically as the convolution
of the received down-converted signal y(k')e.sup.j2.pi.kf.sup.n
with receive filter function g.sub.Rx(n,k'):
z(n,k')=(y(k)e.sup.j2.pi.kf.sup.n)*g.sup.Rx(n,k'). (4)
[0051] Here the receive filtering (Rx-filtering) is used to cancel
out the undesired adjacent channel interference. A high
selectivity, i.e. steep filter edges, will provide low inter
channel interference (ICI) and therefore a higher signal to noise
ratio (SNR). However, sharp filter edges typically require high
filter orders, which in turn have to be compensated for by an
appropriate cyclic prefix length which in turn decreases the
system's spectral efficiency. After digital Rx-filtering, the
signal is down-sampled using processing block 433.sub.n turning the
sample index k' back into the symbol index k.
[0052] In processing block 434.sub.n the cyclic prefix is removed
and the signal z(n,k') is transformed into the time domain using
FFT processing block 435.sub.n in order to yield the frequency bins
Z(n,l) of the n-th carrier signal. The signal model can now be
written as:
Z(n,l)=S(n,l)H(n,l)+(n,l), (5)
where S(n,l) is the FFT transformed version of the data signal s(n,
k) and W(n,l) is the l-th frequency bin of the receive (Rx-)
filtered noise and interchannel interference (ICI). The channel
transfer function H(n,l) denotes the FFT transformed effective
channel, consisting of the transmit (Tx-) filter, the receive (Rx-)
filter and the transmission channel:
H(n,l)=FFT{g.sub.Tx(k)*h(k)*g.sub.Rx(k)} (6)
[0053] Equalization can now be achieved in processing block
436.sub.n e.g. by zero-forcing the operation
{circumflex over (S)}(n,l)=Z(n,l)/H(n,l), (7)
where the equalized data signal s(n, k) is obtained from subsequent
IFFT block 437.sub.n. Subsequently all equalized data signals are
fed into processing block 440 for de-mapping the symbols from the n
carriers, thus producing a stream of digital symbols, which ideally
reflect the stream of digital symbols as output by processing block
230 in the transmitter. Lastly, the stream of digital symbols as
output by de-mapper block 440 is fed into the detector/decoder
stage 450, which outputs a stream of bits 460.
[0054] In the following the bit error rate (BER) performance of the
proposed system is compared with the corresponding bit error rate
of an OFDM system. The aim of this evaluation is to show the impact
of the non-orthogonal GFDM carriers which can be controlled by the
selectivity of the Tx and Rx digital system filters. The channel
properties in the frequency domain, i.e.
FFT{g.sub.Tx(k)*h(k)*g.sub.Rx(k)}, is assumed to be perfectly known
at the receiver where we consider an AWGN environment. This means
that the channel impulse response h(k') only consists of one tap
which is always set to h(0)=1.
[0055] For allowing a reasonable comparison between the systems,
both the OFDM and the GFDM system are designed to have the same
spectral efficiency, i.e., both system parameter sets are defined
according to Table I.
TABLE-US-00001 TABLE I SYSTEM PARAMETER SETS FOR GFDM AND OFDM
Parameter OFDM GFDM Signal Bandwidth 20 MHz 20 MHz Number of
carriers 64 4 Carrier bandwidth 312.5 kHz 5 MHz Symbol duration 3.2
.mu.s 0.2 .mu.s CP overhead 25% 25% CP length 16 Samples Variable
Pulse Shape rectangular FIR: Cos-roll r = 0.1
[0056] In our numerical example, we use a cyclic prefix (CP)
overhead of 25% for both the GFDM and the OFDM system. OFDM
inherently exhibits this overhead due to the cyclic prefix in each
OFDM symbol. In contrast to that, in the GFDM system the cyclic
prefix depends on the length of the receive (Rx-) filter, where the
length of the mobile channel can be neglected due to the AWGN
assumption. Assuming a cyclic prefix overhead of 25% and a receive
filter order of 25 symbols, the GFDM FFT block length is 100
symbols. Hence, the GFDM FFT is 6.25 times longer than in the OFDM
system.
[0057] FIG. 5 depicts the bit error rate of uncoded QPSK modulated
symbols for both OFDM and GFDM, where GFDM uses different filter
orders for transmit (Tx) and receive (Rx) filtering. The orders of
the digital Tx and Rx cosine roll-off filters are denoted by
O.sub.Tx and O.sub.Rx, respectively.
[0058] FIG. 6 shows a comparison of the peak-to-average power ratio
(PAPR) for OFDM and GFDM. Due to fewer carriers, GFDM exhibits a
superior peak-to-average power ratio (PAPR) performance compared to
conventional OFDM signaling.
[0059] It has thus been presented a multi-carrier system
architecture based on digitally implemented filter banks. In an
ex-post facto consideration the inventive transmitter
implementation reveals an allusion to a classical filter bank
approach of the late 1950ies having multiple branches for multiple
carriers (which has not been pursued due to the exhausting
requirements on the analog components), but which is now
implemented digitally. It has been shown that our novel GFDM
approach significantly reduces the requirements set on the analogue
front-end. In particular, GFDM combines both the advantages of a
specific carrier allocation and a low PAPR. Low PAPR allows to
reduce the hardware cost and power consumption, which is an
important point of sale for future wireless systems. Furthermore,
as each single carrier can be modulated individually, that provides
a high degree of flexibility in the system design, e.g. multi-user
scheduling. It has been shown that, in comparison to conventional
multi-carrier systems, GFDM is an efficient alternative for heavily
fragmented spectrums.
* * * * *