Description
Title of Invention: TRANSMISSION APPARATUS, RECEPTION
APPARATUS, TRANSMISSION METHOD, RECEPTION
METHOD, AND METHOD FOR GENERATING MULTI-
DIMENSIONAL CONSTELLATIONS
Technical Field
[0001] The present invention relates to digital data communication, in particular to methods
for generating multi-dimensional constellations for digital data modulation, methods
for modulating and transmitting data on the basis of multi-dimensional constellations,
and a corresponding apparatus.
Background Art
[0002] Fading is one of the major problems in communication systems. It represents random
fluctuations in the amplitude of the received signal due to multi-path propagation. If
the delay spread of the channel is larger than the symbol period of the signal, the
fading is also frequency selective. The amplitude of fading is usually approximated by
a Rayleigh distribution. Such fading is referred to as Rayleigh fading.
[0003] In digital communication systems, information is encoded as a sequence of symbols
belonging to a discrete alphabet, referred to as a constellation. Such a constellation has
N dimensions and encodes B information bits per dimension. The number of possible
values, also referred to as constellation points, is therefore 2N*B. The number of bits per
dimension B directly determines the spectral efficiency of the transmission, given in
bits/Hz. The number of dimensions N has no effect on the spectral efficiency. An
example constellation with N = 2 and B = 1 is illustrated in FIG. 1A.
[0004] Traditionally, for example in a quadrature amplitude modulation (QAM) con-
stellation shown in FIG. 1A, each transmitted bit affects only one dimension. Referring
to FIG. 1A, "b," of each constellation point "b,b2" (= "00", "01", "10" and "11") affects
only the dimension represented by the horizontal axis, whereas "b2" of each con-
stellation point "bib2" affects only the dimension represented by the vertical axis. If the
dimension affected by the transmitted bits undergoes a deep fading, all bits that
modulate this dimension will be extremely unreliable, which increases the error
probability. This effect is illustrated by the errors in FIG. 1 A. For example, if the
channel represented by the vertical axis fades away, the constellation points "00", "01",
" 10" and "11" will approach the horizontal axis (along the solid arrows of FIG. 1 A). As
a result, the constellation points "00" and "01", as well as the constellation points "10"
and "11", will be indiscernible.
[0005] If the constellation is modified such that each bit affects all dimensions, the resilience
to fading is increased. A deep fading on one of the dimensions will affect all the bits of
the constellation; however, this effect would not be as detrimental as in the con-
ventional case, so that on average, the error probability decreases. This is referred to in
the literature as modulation diversity.
[0006] (Rotated Constellations)
One way to achieve modulation diversity is to rotate a (hyper-cubic) constellation to
spread the effect of a channel fading over all its dimensions. This is illustrated in FIG.
IB for the case where N = 2 and B = 1. For example, as shown in FIG. IB, if the
channel represented by the vertical axis fades away, the constellation points "00", "01",
" 10" and "11" will approach the horizontal axis (along the solid arrows of FIG. IB).
However, these constellation points will still be discernible in the dimension rep-
resented by the horizontal axis. As such, the constellation points "00", "01", "10" and
"11" remain discernible even after a deep fading of the channel represented by the
vertical axis.
[0007] A multi-dimensional rotation can be achieved by multiplying the N-element signal
vector by an N*N square matrix. The necessary and sufficient condition for a square
matrix to be a rotation matrix (or a reflection matrix) is for it to be orthogonal, i.e., to
satisfy the equation of the following Math. 1.
[Math.l]

Note that in the above Math. 1, the matrix

is a square matrix, the matrix

is a transpose matrix of the matrix
R
, and the matrix
I
is a unit matrix.
[0008] This means that with regard to the above Math. 1, the row/column vectors must be
orthogonal unit vectors, i.e., satisfy the equation of the following Math. 2.
[0010] This preserves the Euclidean distance between any two points of the constellation,
and ensures that the performance in channels with additive white Gaussian noise
(AWGN channels) is not affected.
[0011] Obviously, not all rotations yield the effect of improved modulation diversity. From
NPL 1, it is known that the optimum rotation angle

for 16-QAM satisfies the equation shown in the following Math. 3. The corre-
sponding 2-D (two-dimensional) rotation matrix

satisfies the equation shown in the following Math. 4.
[Math.3]

[0012] Finding the optimum rotation for constellations of more than two dimensions is more
complicated, because there is no single optimization parameter such as the one
pertaining to the rotation angle in a 2-D constellation. In the case of a 4-D
(four-dimensional) constellation, for example, there are six independent rotation
angles, each with its own partial rotation matrix. The partial rotation angles are also
called Givens angles in NPL 2. The final 4-D rotation matrix is obtained by mul-
tiplying the six Givens rotation matrices, namely the six matrices shown in the
following Math. 5.
[Math.5]
From NPL 2, it is known that the optimization may be carried out over the vector
having the six elements shown in the following Math. 6.
[Math.6]
[0013] According to NPL 2, the resulting optimum rotation angles for a 4-D constellation
with two bits per dimension have the values shown in the following Math. 7.
. [0014] The disadvantage of this method is the number of parameters, specifically for a large
number of dimensions. For N dimensions, the number of partial rotation angles is
equal to the number of possible combinations of two from a set of N, i.e., the value
given by the following Math. 8.
[Math. 8]

[0015] Hence, the number of rotation angles increases with the square of the number of di-
mensions, so the optimization problem becomes very difficult when the number of di-
mensions is large.
[0016] NPL 3 discloses two different approaches, relying on the use of the algebraic number
theory, which have the advantage of a reduced number of parameters.
[0017] The first approach allows the construction of rotation matrices by applying the
"canonical embedding" to an algebraic number field. Two methods are proposed. The
first method produces lattices with diversity L = N/2 for the number of dimensions N =
2e23e3, with e2, e3 = 0, 1, 2,.... Diversity means the minimum number of different
values in the components of any two distinct points of the constellation. The second
method produces lattices with diversity L = N. The possible values of N are very
limited, such as 3, 5, 9, 11, and 15.
[0018] A variant of this method for generating N-dimensional rotated constellations is also
known from NPL 3. The rotation matrix

is expressed by the following Math. 9.
[Math.9]

[0019] Note that the superscripted letter "T" denotes the transpose of a matrix.
[0020] For N = 4, the value of the rotation matrix

is given by the following Math. 10.
[Math. 10]

[0021] Although the resulting rotation matrix is a rotation matrix that is orthogonal for any
N, the full modulation diversity is only achieved when N is a power of two.
[0022] Each of these methods can guarantee a certain degree of diversity. However, the
resulting rotation matrix is fixed, having no parameter that allows the optimization for
different constellation sizes. Therefore, a severe disadvantage of these methods is that
the effect of modulation diversity cannot be maximized in accordance with different
constellation sizes.
[0023] The second approach first constructs rotation matrices with two and three di-
mensions, which can be used as base matrices for constructing matrices with more di-
mensions using a Hadamard-like stacked expansion shown in the following Math. 11.
[Math. 11]

[0024] The base 2-D and 3-D (three-dimensional) rotation matrices have a single in-
dependent parameter which is chosen so that the product distance of the constellation
is maximized. A 4-D rotation matrix is constructed from two 2-D rotation matrices
according to the above Math. 11. Because of the relative small dimension, it is possible
to find an algebraic relationship between parameters of the two 2-D rotation matrices,
so that the product distance is maximized. For larger dimensions, such an optimization
becomes intractable, which is the primary disadvantage of the second approach.
[0025] (Mapping constellation components to ensure independent fading)
Another aspect concerns the separation and mapping of the N dimensions of the
rotated constellation so that they experience independent fading. This is a key aspect
necessary for achieving the expected diversity performance.
[0026] The N constellation components, which are obtained by separating the N-
dimensional rotated constellation on a per-dimension basis, can be transmitted over
different time slots, frequencies, transmitter antennas, or combinations thereof. Further
signal processing is possible before transmission. The critical aspect is that fading ex-
perienced by each of the N dimensions must be different from, or ideally uncorrelated
with, fading experienced by any other one of the N dimensions.
[0027] The spreading of the N dimensions across different time slots, frequencies and
antennas can be achieved for example through appropriate interleaving and mapping.
[0028] (Mapping constellation components to transmitted complex cells)
Another aspect concerns the mapping of the N real dimensions of the rotated con-
stellation to complex symbols for transmission. In order to ensure the desired diversity,
the N dimensions must be mapped to different complex symbols. The complex
symbols are then spread as described earlier, e.g. through interleaving and mapping, so
that at the reception, fading experienced by each of the N dimensions is uncorrelated
with fading of any other one of the N dimensions.
[0029] FIG. 2 is a block diagram of a transmission apparatus.
[0030] The transmission apparatus is composed of an FEC encoder 210, a bit interleaver
220, a rotated constellation mapper 230, a complex symbol mapper 240, a symbol in-
terleaver/mapper 250, modulation chains 260-1 to 260-M, and transmitter antennas
270-1 to270-M.
[0031] The FEC encoder 210 performs forward error correction (FEC) encoding on the input
thereto. Note that the best FEC codes known so far, which are also the most used in
new standards, are the turbo codes and the low-density parity check (LDPC) codes.
[0032] The bit interleaver 220 performs bit interleaving on the input from the FEC encoder
210. Here, the bit interleaving can be block interleaving or convolution interleaving.
[0033] The rotated constellation mapper 230 maps the input from the bit interleaver 220 to
the rotated constellation.
[0034| Generally, the input to the rotated constellation mapper 230 is the output of the FEC
encoder 210 via the bit interleaver 220 that performs optional bit interleaving. The bit
interleaving is usually required when there are more than one bit per dimension (B >
1). The FEC encoding performed by the FEC encoder 210 introduces redundant bits in
a controlled fashion, so that propagation errors can be corrected in the reception
apparatus. Although the overall spectral efficiency decreases, the transmission
becomes overall more robust, i.e., the bit error rate (BER) decays much faster with the
signal to noise ratio (SNR).
[0035] Note that regarding the original mapping of the information bits on the non-rotated
hyper-cubic constellations, each dimension is modulated separately by B bits, using
either binary or Gray mapping, so the number of discrete values is 2B and the number
of constellation points is 2B*N.
[0036] The complex symbol mapper 240 maps each of N constellation components, which
represent N-dimensional rotated constellation symbols input from the rotated con-
stellation mapper 230, to a different one of complex symbols.
[0037] There are multiple possibilities for the mapping performed by the complex symbol
mapper 240, i.e., the mapping of each of N constellation components, which represent
N-dimensional rotated constellation symbols, to a different one of complex symbols.
Some of such possibilities are illustrated in FIG. 3. The essential function of the
complex symbol mapper 240 is to map each of N constellation components of one
rotated constellation symbol to a different one of complex symbols.
[0038] By way of example, FIG. 3 shows the case of four dimensions. Referring to FIG. 3,
the boxes showing the same number (e.g., "1") represent a group of 4-D rotated con-
stellation symbols. The number shown by each box indicates the group number of the
corresponding group. Also, each box indicates a constellation component of one
dimension.
[0039] Shown below "Constellation symbols" in FTG. 3 is a state where six groups of 4-D
rotated constellation symbols are aligned. Shown below "Complex symbols" in FIG. 3
are twelve complex symbols, which are obtained by rearranging the six groups of 4-D
rotated constellation symbols shown below "Constellation symbols" in FIG. 3. Note
that FIG. 3 shows three forms of "Complex symbols" as examples. At the time of
actual transmission, a pair of two constellation components that are vertically aligned
below "Complex symbols" (the result of rearrangement) is modulated and transmitted
as one complex symbol.
[0040] The symbol interleaver/mapper 250 performs symbol interleaving on the complex
symbols input from the complex symbol mapper 240, and thereafter maps the complex
symbols to different time slots, frequencies, transmitter antennas, or combinations
thereof. Here, the symbol interleaving can be block interleaving or convolution in-
terleaving.
[0041] The modulation chains 260-1 to 260-M are provided in one-to-one correspondence
with the transmitter antennas 270-1 to 270-M. Each of the modulation chains 260-1 to
260-M inserts pilots for estimating the fading coefficients into the corresponding input
from the symbol interleaver/mapper 250, and also performs various processing, such as
conversion into the time domain, digital-to-analog (D/A) conversion, transmission
filtering and orthogonal modulation, on the corresponding input. Then, each of the
modulation chains 260-1 to 260-M transmits the transmission signal via a corre-
sponding one of the transmitter antennas 270-1 to 270-M.
[0042] (Receiver Side)
On the receiver side, the exact inverse steps of the steps performed by the
transmission apparatus must be performed. FIG. 4 shows a block diagram of a
reception apparatus corresponding to the transmission apparatus whose block diagram
is shown in FIG. 2.
[0043] The reception apparatus is composed of receiver antennas 410-1 to 410-M, de-
modulation chains 420-1 to 420-M, a symbol demapper/deinterleaver 430, a complex
symbol demapper 440, a rotated constellation demapper 450, a bit deinterleaver 460,
and an FEC decoder 470.
[0044] The demodulation chains 420-1 to 420-M are provided in one-to-one correspondence
with the receiver antennas 410-1 to 410-M. Each of the demodulation chains 420-1 to
420-M performs processing such as A/D conversion, reception filtering, and or-
thogonal demodulation on the signal transmitted by the transmission apparatus of FIG.
2 and received by a corresponding one of the receiver antennas 410-1 to 410-M. Then,
the demodulation chains 420-1 to 420-M estimate (i) the amplitude values (fading co-
efficients) of the channel characteristics by using the pilots and (ii) noise variance, and
output the estimated amplitude values and noise variance together with the phase-
corrected received signal.
[0045] The symbol demapper/deinterleaver 430 performs the inverse processing of the
processing performed by the symbol interleaver/mapper 230 in the transmission
apparatus on the inputs from the demodulation chains 420-1 to 420-M.
[0046] The complex symbol demapper 440 performs the inverse processing of the
processing performed by the complex symbol mapper 240 in the transmission
apparatus on the input from the symbol demapper/deinterleaver 430. Through this
processing, N-dimensional rotated constellation symbols can be obtained.
[0047] The rotated constellation demapper 450 performs demapping processing on the N-
dimensional rotated constellation symbols, and outputs a decision result of each bit
included in the N-dimensional rotated constellation.
[0048] The bit deinterleaver 460 performs the inverse processing of the processing
performed by the bit interleaver 220 in the transmission apparatus on the input from
the rotated constellation demapper 450.
[0049] The FEC decoder 470 performs FEC decoding on the input from the bit deinterleaver
470.
[0050] Below, further explanations of the rotated constellation demapper 450 are given.
[0051] The rotated constellation demapper 450 can perform the processing of demapping N-
dimensional rotated constellation symbols in the following two ways (i) and (ii).
(i) First de-rotate the constellation, then extract the bits for each dimension
separately,
(ii) Decode the bits of all dimensions in one step.
[0052] Although the first solution (the above (i)) is the most simple, its performance is
suboptimal and even worse for rotated constellations than for non-rotated con-
stellations. Due to its simplicity, this solution may be used in some low-cost reception
apparatuses.
[0053] Although the second solution (the above (ii)) is more complex, it offers much better
performance in terms of BER at a given SNR. In the following, the second solution
will be described in greater detail.
[0054] As with the transmission apparatus, a preferred embodiment of the reception
apparatus includes the FEC decoder 470 after the rotated constellation demapper 450,
with the optional bit deinterleaver 460 in between, as shown in FIG. 4. More exactly,
the rotated constellation demapper 450, which performs the rotated constellation
demapping, receives N-dimensional symbol vectors (y(,..., yN) and the estimated
fading coefficient vectors (h,,..., hN), and extracts data of N*B bits (b,,..., bN»B) from
each symbol, as shown in FIG. 5.
[0055] When FEC decoding is used, the processing of demapping the N-dimensional rotated
constellation symbols can no longer be performed by way of a hard decision, because
the performance of the error correction would be suboptimal. Instead, "soft bits" must
be used, either in the form of probabilities or in the form of log-likelihood ratios
(LLRs). The LLR representation is preferred because probability multiplications can
be conveniently expressed as sums. By definition, the LLR of a bit bk is shown in the
following Math. 12.
[Math. 12]

[0056] Note that in Math. 12,

are the a-priori probabilities that bk = 0 and bk = 1 were transmitted when the symbol
vector

is received. According to the known theory, the LLR of a bit bk of a constellation has
the exact expression shown in the following Math. 13.
[Math. 13]

[0057] Note that in Math. 13, k is the bit index,

is the received symbol vector,

is the diagonal matrix having the associated (estimated) fading coefficients as
elements on the main diagonal,

is a constellation point vector,

is the squared norm, and

is the noise variance.
[0058] For an N-dimensional constellation, the squared norm represents the squared
Euclidean distance from the received symbol vector

to the faded constellation symbol vector

in the N-dimensional space. The squared norm can be expressed by the following
Math. 14.
[Math. 14]

[0059] Each bit bk divides the constellation into two partitions of equal size, Sk° and Sk', cor-
responding to those points for which bk is 0 and 1, respectively. Examples are shown in
FIGs. 6A and 6B for a classical 16-QAM constellation with Gray encoding. FIG. 6A
shows the constellation encoding and FIG. 6B shows the two partitions for each bit bk.
[0060] The exact expression for the LLR (the above Math. 13) is difficult to calculate due to
the exponentials, divisions and the logarithm. In practice, the approximation shown in
the following Math. 15 is made, called max-log, which introduces negligible errors.
[Math. 15]

[0061] By using the above Math. 15, the above Math. 13 leads to a much more simple ex-
pression for the LLR, which is shown in the following Math. 16.
[Math. 16]

[0062] For each received symbol vector

, the distances to all 2B*N constellation points must be calculated, and the corre-
sponding minimum for each partition is determined.
[0063] FIG. 7 shows a preferred hardware implementation of an LLR demapper (one
example of the rotated constellation demapper 450 shown in FIG. 4) for a 16-QAM
rotated constellation (N = 2, B = 2).
[0064] The LLR demapper is composed of a counter 710, a rotated constellation mapper
720, a squared Euclidean distance calculator 730, minimizers 740-1 to 740-4, and
adders 750-1 to 750-4.
[0065] For each received symbol vector

, the counter 710 repeatedly generates all 24 = 16 constellation points, and outputs
four bits bl5 b2, b3 and b4 indicating the constellation points to the rotated constellation
mapper 720.
[0066] The rotated constellation mapper 720 selects the 2-D rotated constellation point from
a look-up table by using the counter values provided by the counter 710 as an indexes,
and outputs two constellation components s, and s2 obtained through this selection to
the squared Euclidean distance calculator 730.
[0067] The squared Euclidean distance calculator 730 calculates the squared Euclidean
distances (see FIG. 8).
[0068] For each bit, the minimizers 740-1 to 410-4 maintain the corresponding minimum
squared Euclidean distances for the two partitions (see FIG. 9). The two constellation
partitions for each bit are simply indicated by the corresponding bit of the counter 710.
[0069] Each of the adders 750-1 to 750-4 subtracts the output of mini (corresponding to bit
1) from the output of minO (corresponding to bit 0), the mini and minO being provided
in each of the minimizers 740-1 to 740-4. Thereafter, the adders 750-1 to 750-4 output
the results of the subtraction as L(bO to L(b4), respectively.
[0070] FIG. 8 is a circuit diagram of a squared Euclidean distance calculator that calculates
an N-dimensional squared Euclidean distance. Note that the circuit structure of the
squared Euclidean distance calculator 730 has been modified from the one shown in
FIG. 8 so as to satisfy N = 2.
[0071] The squared Euclidean distance calculator is composed of multipliers 810-1 to
810-N, adders 820-1 to 820-N, multipliers 830-1 to 830-N, an adder 840, and a
multiplier 850.
[0072] The multipliers 810-1 to 810-N multiply h, to hN by s, to sN, respectively. The adders
820-1 to 820-N subtract h1S1 to hNsN from y1 to yN, respectively. The multipliers 830-1
to 830-N multiply (y1 - h1S1) to (yN - hNsN) by (yt - h1S1) to (yN - hNsN), respectively.
[0073] The adder 840 adds together the outputs of the multipliers 830-1 to 830-N. The
multiplier 850 multiplies the output of the adder 840 by

. The output of the multiplier 850 is the N-dimensional squared Euclidean distance.
[0074] FIG. 9 is a circuit diagram of the minimizers 740-1 to 740-4 that each calculate the
minimum squared Euclidean distances for each bit. The 1-bit subset (or partition) input
indicates the current position.
[0075] Each of the minimizers 740-1 to 740-4 is composed of a comparator 910, a selector
920, an inverter 930, D flip-flops 940-0 and 940-1, and a selector 950.
[0076] The following describes the operations to be performed in the situation of FIG. 9
when the subset value (the value input from the counter 710) is "0".
• [0077] From among the output of the D flip-flop 940-0 and the output of the D flip-flop
940-1, the selector 950 selects and outputs the former.
[0078] The comparator 910 compares din (A), which indicates the squared Euclidean
distance calculated by the squared Euclidian distance calculator 730, with the output
(B) of the selector 950. In a case where B is smaller than A, the comparator 910
outputs "0". In this case, from among din and the output of the selector 950, the
selector 920 selects and outputs the latter based on "0" received from the comparator
910. On the other hand, in a case where A is smaller than B, the comparator 910
outputs "1". In this case, from among din and the output of the selector 950, the
selector 920 selects and outputs the former based on " 1" received from the comparator
910. Note that in a case where A is equal to B, the same result will be obtained whether
the selector 920 selects din or the output of the selector 950. Accordingly, in this case,
the comparator 910 may output either one of "0" and "1".
• [0079] The inverter 930 inverts the subset value "0". Thus, "1" is input to the enable
terminal of the D flip-flop 940-0. As the D flip-flop 940-0 is enabled, it latches the
output of the selector 920. Meanwhile, "0" is input to the enable terminal of the D flip-
flop 940-1. As the D flip-flop 940-1 is disabled, it does not latch the output of the
selector 920.
[0080] The following describes the operations to be performed in the situation of FIG. 9
when the subset value is "1".
[0081] From among the output of the D flip-flop 940-0 and the output of the D flip-flop
940-1, the selector 950 selects and outputs the latter.
[0082] The comparator 910 compares din (A) with the output (B) from the selector 950. In a
case where B is smaller than A, the comparator 910 outputs "0". In this case, from
among din and the output of the selector 950, the selector 920 selects and outputs the
latter based on "0" received from the comparator 910. On the other hand, in a case
where A is smaller than B, the comparator 910 outputs "1". In this case, from among
din and the output of the selector 950, the selector 920 selects and outputs the former
based on " 1" received from the comparator 910. Note that in a case where A is equal to
B, the same result will be obtained whether the selector 920 selects din or the output of
the selector 950. Accordingly, in this case, the comparator 910 may output either one
of"0"and"l".
[0083] " 1" is input to the enable terminal of the D flip-flop 940-1. As the D flip-flop 940-1
is enabled, it latches the output of the selector 920. Meanwhile, the inverter 930 inverts
the subset value "1". Thus, "0" is input to the enable terminal of the D flip-flop 940-0.
As the D flip-flop 940-0 is disabled, it does not latch the output of the selector 920.
[0084] A significant improvement in performance of the reception apparatus can be
achieved by using iterative decoding. As shown in FIG. 10, the reception apparatus
configured to utilize such iterative decoding is composed of a rotated constellation
demapper 1010, a bit deinterleaver 1020, an FEC decoder 1030, an adder 1040, and a
bit interleaver 1050. Here, the rotated constellation demapper 1010 and the FEC
decoder 1030 are connected in a loop.
[0085] The rotated constellation demapper 1010 performs demapping processing on N-
dimensional rotated constellation symbols, and outputs L (see FIG. 11). The bit dein-
terleaver 1020 performs the inverse processing of the processing performed by the bit
interleaver 220 in the transmission apparatus on the input from the rotated constellation
demapper 1010. The FEC decoder 1030 performs FEC decoding on the input from the
bit deinterleaver 1020.
[0086] The adder 1040 subtracts the input from the FEC decoder 1030 from the output of
the FEC decoder 1030. The bit interleaver 1050 performs the same processing as the
processing performed by the bit interleaver 220 in the transmission apparatus on the
output of the adder 1040, and then outputs LE. LE, also referred to as extrinsic in-
formation, is fed back to the rotated constellation demapper 1010 in order to aid the
demapping processing performed by the rotated constellation demapper 1010, i.e., the
processing of demapping the N-dimensional rotated constellation symbols. In this case
it is essential that the FEC decoding produces soft bits, e.g. in the form of LLRs.
[0087] As known in the literature, the formula for calculating the LLR for bit bk is given by
the following Math. 17.
represents the K = N*B bits associated with each constellation point, and Xk° and XK1
represent the two constellation partitions associated with bit k, each constellation point
being represented by the N*B bits instead of the N bits of integer coordinates. Fur-
thermore,

is expressed as

and represents the constellation mapping function.
[0089] For example, X3° and X31 are shown in the following Math. 18.
[0090] FIG. 11 shows an example of the structure of the rotated constellation demapper
1010 for iterative decoding. Note that the rotated constellation demapper 1010 for
iterative decoding is similar to a rotated constellation demapper for non-iterative
decoding. Below, the elements that are the same as those described above are assigned
the same reference numerals thereas, and a detailed description thereof is omitted.
[0091] The rotated constellation demapper 1010 is composed of a counter 710, a rotated
constellation mapper 720, a squared Euclidean distance calculator 730, minimizers
740-1 to 740-4, adders 750-1 to 750-4, logical AND operators 1110-1 to 1110-4, an
adder 1120, adders 1130-1 to 1130-4, and adders 1140-1 to 1140-4.
The logical AND operators 1110-1 to 1110-4 perform logical AND operations on the
outputs of the bit interleaver 1050, namely LE(b 1) to LE(b4), and the outputs of the
counter 710, namely bl to b4 . The adder 1120 adds together the outputs of the logical
AND operators 1110-1 to 1110-4. Each of the adders 1130-1 to 1130-4 subtracts, from
the output of the adder 1120, the output of a corresponding one of the logical AND
operators 1110-1 to 1110-4. Each of the adders 1140-1 to 1140-4 subtracts, from the
output of the squared Euclidean distance calculator 730, the output of a corresponding
one of the adders 1130-1 to 1130-4. Then, each of the adders 1140-1 to 1140-4 outputs
the value obtained through the subtraction to din of a corresponding one of the
minimizers 740-1 to 740-4.
Citation List
Non Patent Literature
[0092] NPL 1: K. Boulle and J. C. Belfiore. "Modulation Scheme Designed for the Rayleigh
Fading Channel." Presented at CISS 1992.
NPL 2: B. D. Jelicic and S. Roy. "Design of Trellis Coded QAM for Flat Fading and
AWGN Channels." IEEE Transactions on Vehicular Technology, Vol. 44. Feb. 1995.
NPL 3: J. Boutros and E. Viterbo. "Signal Space Diversity: A Power- and Bandwidth-
Efficient Diversity Technique for the Rayleigh Fading Channel." IEEE Transactions on
Information Theory, Vol. 44. Jul. 1998.
NPL 4: M. O. Damen, K. Abed-Meraim, and J.C. Belfiore. "Diagonal Algebraic
Space-Time Block Codes." IEEE Transactions on Information Theory, Vol. 48. Mar.
2002.
Summary of Invention
Technical Problem
. [0093J As described above, although a wide range of proposals have been made regarding
rotation matrices for rotating a constellation, the proposals that have been made so far
do not provide any efficient method of generating a multi-dimensional rotated con-
stellation (a multi-dimensional rotation matrix) for digital modulation with a high
degree of modulation diversity with respect to various constellation sizes.
[0094] NPL 2 introduces an approach that makes use of a Givens rotation. The problem with
this approach is that the number of parameters for generating an optimal multi-
dimensional rotated constellation increases by the order of the square of the number of
dimensions in the constellation.
[0095] NPL 3 introduces two approaches. The first approach makes use of canonical
embedding. According to this approach, the method of generating a multi-dimensional
rotation matrix is uniquely determined based on the number of dimensions, and does
not have a parameter enabling the optimization for different constellation sizes.
Therefore, the problem with this approach is that it does not allow maximizing the
effect of modulation diversity for various constellation sizes.
[0096] The second approach introduced by NPL 3 generates a multi-dimensional rotation
matrix having a larger number of dimensions by using stacked expansion where 2-D
and 3-D rotation matrices are stacked. The problem with this approach is that the
algebraic relationships between the stacked rotation matrices become more com-
plicated as the number of dimensions increases, rendering the optimization difficult.
[0097] It is the object of the present invention to provide an efficient method of generating a
multi-dimensional rotated constellation (a multi-dimensional rotation matrix) for
digital transmission with a high degree of modulation diversity with respect to various
constellation sizes. It is also the object of the present invention to provide a
transmission apparatus and a transmission method for transmitting data based on the
multi-dimensional rotated constellation obtained by using the above method, and a
reception apparatus and a reception method for receiving data based on the multi-
dimensional rotated constellation obtained by using the above method.
Solution to Problem
[0098] A transmission apparatus of the present invention transmits a block of data over a
plurality of transmission channels. The transmission apparatus comprises: a modulator
operable to select one of a plurality of constellation points in accordance with the block
of data to be transmitted, each of the plurality of constellation points having a plurality
of components; and a transmitter operable to transmit each component of the selected
constellation point over a different one of the plurality of transmission channels,
wherein (i) the plurality of constellation points are defined by positions thereof within
an N-dimensional space, the positions being obtained by applying an orthogonal trans-
formation to a subset of

, which is an N-dimensional integer lattice, (ii) N is a multiple of four, and (iii) the
orthogonal transformation has an N-by-N matrix representation with absolute values of
all elements on a main diagonal equal to a first value, and with absolute values of all
elements not on the main diagonal equal to a non-zero second value.
Advantageous Effects of Invention
[0099] The above transmission apparatus allows efficiently generating a multi-dimensional
rotated constellation (a multi-dimensional rotation matrix) for digital transmission with
a high degree of modulation diversity with respect to various constellation sizes. Due
to the multi-dimensional rotated constellation obtained by using the generated multi-
dimensional rotation matrix, the above transmission apparatus also enables data
transmission that yields the effect of a high-degree of modulation diversity.
Brief Description of Drawings
[0100] [fig.lA]FIG. 1A illustrates an example constellation in 2D and the effect of fading.
[fig.lB|FIG.lB illustrates an example constellation in 2D that is obtained by rotating
the constellation of FIG. 1A and the effect of fading.
f fig.2]FIG. 2 shows a block diagram of a conventional transmission apparatus.
Tfig.3]FIG. 3 is a schematic drawing illustrating the mapping of constellation symbols
to complex symbols.
[fig.4]FIG. 4 is a block diagram of a conventional reception apparatus.
[fig.5]FIG. 5 is an illustration of the inputs to and the outputs of a rotated constellation
demapper.
[fig.6A]FIG. 6A shows an example of a conventional 16-QAM constellation with Gray
encoding.
[fig.6B]FIG. 6B illustrates the two partitions for each bit of the constellation of FIG.
6A.
[fig.7]FIG. 7 shows an example hardware implementation of an LLR demapper for a
16-QAM rotated constellation.
[fig.8]FIG. 8 shows an example hardware implementation for a squared Euclidean
distance calculator that calculates the N-dimensional squared Euclidean distance.
[fig.9]FIG. 9 shows an example hardware implementation for a minimizer that
calculates the minimum squared Euclidean distances.
[fig. 10]FIG. 10 shows a block diagram of a circuit that performs iterative decoding,
[fig. 11]FIG. 11 shows an example hardware implementation of the rotated con-
stellation demapper for iterative decoding.
[fig. 12]FIG. 12 shows a block diagram of a transmission apparatus according to an em-
bodiment of the present invention.
[fig.l3]FIG. 13 shows a block diagram of a reception apparatus according to an em-
bodiment of the present invention.
[fig. 14]FTG. 14 is a block diagram of the rotated constellation demapper shown in FIG.
13.
Description of Embodiments
[0101] The present invention provides a first transmission apparatus for transmitting a block
of data over a plurality of transmission channels, the first transmission apparatus
comprising: a modulator operable to select one of a plurality of constellation points in
accordance with the block of data to be transmitted, each of the plurality of con-
stellation points having a plurality of components; and a transmitter operable to
transmit each component of the selected constellation point over a different one of the
plurality of transmission channels, wherein (i) the plurality of constellation points are
defined by positions thereof within an N-dimensional space, the positions being
obtained by applying an orthogonal transformation to a subset of
Z*
, which is an N-dimensional integer lattice, (ii) N is a multiple of four, and (iii) the
orthogonal transformation has an N-by-N matrix representation with absolute values of
all elements on a main diagonal equal to a first value, and with absolute values of all
elements not on the main diagonal equal to a non-zero second value.
[0102] The present invention also provides a first transmission method for transmitting a
block of data over a plurality of transmission channels, the first transmission method
comprising the steps of: selecting one of a plurality of constellation points in ac-
cordance with the block of data to be transmitted, each of the plurality of constellation
points having a plurality of components; and transmitting each component of the
selected constellation point over a different one of the plurality of transmission
channels, wherein (i) the plurality of constellation points are defined by positions
thereof within an N-dimensional space, the positions being obtained by applying an or-
thogonal transformation to a subset of

, which is an N-dimensional integer lattice, (ii) N is a multiple of four, and (iii) the
orthogonal transformation has an N-by-N matrix representation with absolute values of
all elements on a main diagonal equal to a first value, and with absolute values of all
elements not on the main diagonal equal to a non-zero second value.
[0103] The above transmission apparatus and transmission method allow efficiently
generating a multi-dimensional rotated constellation (a multi-dimensional rotation
matrix) for digital transmission with a high degree of modulation diversity with respect
to various constellation sizes. Due to the multi-dimensional rotated constellation
obtained by using the generated multi-dimensional rotation matrix, the above
transmission apparatus and transmission method also enable data transmission that
yields the effect of a high-degree of modulation diversity.
[0104] The present invention also provides a second transmission apparatus and a second
transmission method, which are the first transmission apparatus and the second
transmission apparatus, respectively, wherein instead of the N-by-N matrix repre-
sentation, the orthogonal transformation has a matrix representation obtained by
permuting rows and/or columns in the N-by-N matrix representation.
[0105] The above structure produces the same effect as the effect produced by the N-by-N
matrix representation with absolute values of all elements on the main diagonal equal
to a first value, and with absolute values of all elements not on the main diagonal equal
to a non-zero second value.
[0106] The present invention also provides a third transmission apparatus, which is the first
transmission apparatus further comprising a mapper operable to map each component
of the selected constellation point to the corresponding one of the plurality of
transmission channels over which the component is to be transmitted, such that fading
of each of the plurality of transmission channels is uncorrected with fading of any
other one of the plurality of transmission channels.
[0107] The present invention also provides a third transmission method, which is the first
transmission method further comprising the step of mapping each component of the
selected constellation point to the corresponding one of the plurality of transmission
channels over which the component is to be transmitted, such that fading of each of the
plurality of transmission channels is uncorrected with fading of any other one of the
plurality of transmission channels.
[0108] The above structure can optimize the transmission performance, even in the presence
of fading.
[0109] The present invention also provides a fourth transmission apparatus, which is the first
transmission apparatus wherein the transmitter is adapted for transmitting each
component of the selected constellation point over a different one of a plurality of time
slots, frequencies, transmitter antennas, or combinations thereof.
[0110] The present invention also provides a fifth transmission apparatus and a fourth
transmission method, which are the first transmission apparatus and the first
transmission method, respectively, wherein the plurality of transmission channels
comprise a plurality of different carriers in an orthogonal frequency-division mul-
tiplexing scheme.
[0111] The present invention also provides a sixth transmission apparatus and a fifth
transmission method, which are the first transmission apparatus and the first
transmission method, respectively, wherein the plurality of transmission channels
comprise a plurality of different symbols in an orthogonal frequency-division mul-
tiplexing scheme.
[0112] The present invention also provides a first reception apparatus for receiving a block
of data over a plurality of transmission channels, the first reception apparatus
comprising: a receiver operable to receive a plurality of component signals over the
plurality of transmission channels; and a demodulator operable to select one of a
plurality of constellation points in accordance with the plurality of received component
signals, wherein (i) the plurality of constellation points are defined by positions thereof
within an N-dimensional space, the positions being obtained by applying an orthogonal
transformation to a subset of

, which is an N-dimensional integer lattice, (ii) N is a multiple of four, and (iii) the
orthogonal transformation has an N-by-N matrix representation with absolute values of
all elements on a main diagonal equal to a first value, and with absolute values of all
elements not on the main diagonal equal to a non-zero second value.
[0113] The present invention also provides a first reception method for receiving a block of
data over a plurality of transmission channels, the first reception method comprising
the steps of: receiving a plurality of component signals over the plurality of
transmission channels; and selecting one of a plurality of constellation points in ac-
cordance with the plurality of received component signals, wherein (i) the plurality of
constellation points are defined by positions thereof within an N-dimensional space,
the positions being obtained by applying an orthogonal transformation to a subset of

, which is an N-dimensional integer lattice, (ii) N is a multiple of four, and (iii) the
orthogonal transformation has an N-by-N matrix representation with absolute values of
all elements on a main diagonal equal to a first value, and with absolute values of all
elements not on the main diagonal equal to a non-zero second value.
[0114] The above reception apparatus and reception method allow efficiently generating a
multi-dimensional rotated constellation (a multi-dimensional rotation matrix) for
digital transmission with a high degree of modulation diversity with respect to various
constellation sizes. Due to the multi-dimensional rotated constellation obtained by
using the generated multi-dimensional rotation matrix, the above reception apparatus
and reception method also enable data reception that yields the effect of a high-degree
of modulation diversity.
[0115] The present invention also provides a second reception apparatus and a second
reception method, which are the first reception apparatus and the first reception
method, respectively, wherein instead of the N-by-N matrix representation, the or-
thogonal transformation has a matrix representation obtained by permuting rows and/or
columns in the N-by-N matrix representation.
[0116] The above structure produces the same effect as the effect produced by the N-by-N
matrix representation with absolute values of all elements on the main diagonal equal
to a first value, and with absolute values of all elements not on the main diagonal equal
to a non-zero second value.
[0117] The present invention also provides a third reception apparatus and a third reception
method, which are the first reception apparatus and the first reception method, re-
spectively, wherein the plurality of transmission channels comprise a plurality of
different carriers in an orthogonal frequency-division multiplexing scheme.
[0118] The present invention also provides a fourth reception apparatus and a fourth
reception method, which are the first reception apparatus and the first reception
method, respectively, wherein the plurality of transmission channels comprise a
plurality of different symbols in an orthogonal frequency-division multiplexing
scheme.
[0119] The present invention also provides a first generation method for generating a multi-
dimensional constellation for a digital modulation scheme in a data communication
system, the first generation method comprising the steps of: receiving a plurality of
vectors of a multi-dimensional vector space; and obtaining constellation points of the
multi-dimensional constellation by applying an orthogonal transformation to the
plurality of vectors received, wherein (i) the orthogonal transformation is adapted for
increasing a minimum number of different values in components of any two distinct
multi-dimensional constellation points relative to a minimum number of different
values in components of any two distinct vectors received, and (ii) the orthogonal
transformation has an N-by-N matrix representation, N being a multiple of four, with
absolute values of all elements on a main diagonal equal to a first value, and with
absolute values of all elements not on the main diagonal equal to a non-zero second
value.
[0120] The above generation method allows efficiently generating a multi-dimensional
rotated constellation (a multi-dimensional rotation matrix) for digital transmission with
a high degree of modulation diversity with respect to various constellation sizes.
[0121] The present invention also provides a second generation method for generating a
multi-dimensional constellation, the second generation method being the first
generation method wherein instead of the N-by-N matrix representation, the or-
thogonal transformation has a matrix representation obtained by permuting rows and/or
columns in the N-by-N matrix representation.
[0122] The above structure produces the same effect as the effect produced by the N-by-N
matrix representation with absolute values of all elements on the main diagonal equal
to a first value, and with absolute values of all elements not on the main diagonal equal
to a non-zero second value.
[0123] The present invention also provides a third generation method for generating a multi-
dimensional constellation, the third generation method being the first generation
method further comprising the steps of: selecting a rotation factor r as a real number
between 0 and 1; calculating the first value, a, by evaluating an expression

; calculating the second value, b, by evaluating an expression
; and determining the orthogonal transformation by selecting a sign value sio for each
element (i, j) of a matrix representation
, such that the matrix representation is orthogonal.
[0124] With the above structure, the orthogonal transformation can be easily determined.
[0125] The present invention also provides a fourth generation method for generating a
multi-dimensional constellation, the fourth generation method being the third
generation method wherein the selected rotation factor r maximizes the minimum
number of different values in the components of any two distinct multi-dimensional
constellation points.
[0126] The above structure makes it possible to achieve a high-degree of modulation
diversity and therewith increased robustness in the presence of fading, while
preserving spectral efficiency.
[0127] The present invention also provides a fifth generation method for generating a multi-
dimensional constellation, the fifth generation method being the first generation
method wherein the plurality of vectors received represent a subset of

, which is an N-dimensional integer lattice.
[0128] The above structure is useful in a straightforward numerical implementation.
[0129] The following describes an embodiment of the present invention with reference to
the drawings.
[0130] First, a description is now given of proposed multi-dimensional rotation matrices.
' [0131] Multi-dimensional rotation matrices have a single independent parameter and a
structure that is as regular as possible. The parameter can be configured in order to
minimize the error probability for various constellation sizes. Specifically, the
following two conditions (i) and (ii) are imposed on the multi-dimensional rotation
matrix employed for obtaining a multi-dimensional rotated constellation.
(i) Each output must have a dominant input,
(ii) The remaining inputs must have equal weights.
[0132] The above conditions (i) and (ii) are fulfilled if the multi-dimensional rotation matrix
is of the form shown in the following Math. 19 (for N = 4), or more generally, of the
form shown in the following Math. 20. Note that the multi-dimensional rotation matrix
shown in Math. 20 is an N-by-N matrix.
[Math. 19]
[0133] Here, a and b denote real parameters, with each sign value su satisfying

. Note that values of the parameters a and b that fulfill the above conditions (i) and
(ii) satisfy a relational expression a > b > 0.
[0134] Obviously, the same advantages can be achieved by permuting rows and/or columns
of the multi-dimensional rotation matrix shown in the above Math. 20. Therefore, the
matrix shown in Math. 20 can be used as the multi-dimensional rotation matrix. Alter-
natively, it is also possible to use a matrix obtained by permuting rows and/or columns
of the matrix shown in Math. 20 as the multi-dimensional rotation matrix. The matrix
shown in Math. 20 and the matrix obtained by permuting rows and/or columns of the
matrix shown in Math. 20 have the following features: (i) each row contains an
element having a real parameter a; (ii) each column contains an element having a real
parameter a; and (iii) the rest of the elements in each row/column have a real parameter
b.
[0135] The following describes normalization of the multi-dimensional rotation matrix
shown in the above Math. 20. Note that similar normalization can be performed on a
matrix (a multi-dimensional rotation matrix) obtained by permuting rows and/or
columns of the matrix shown in Math. 20.
[0136] The normalization condition establishes the relationship shown in the following
Math. 21 between parameters a and b.
[Math.21]

[0137] Therefore, the multi-dimensional rotation matrix has only one independent
parameter. In the following Math. 22, we define a "rotation factor" r between 0 and 1.
[Math.22]

[0138] Therefore, the parameters a and b can be expressed in terms of the "rotation factor" r
as shown in the following Math. 23.
[Math.23]

[0139] The advantage of using the "rotation factor" r is that the range is always 0 to 1 re-
gardless of the number of dimensions. The optimal value for the "rotation factor" r
depends on the constellation size, that is, the number of dimensions N and the number
of bits B per dimension for square/cubic constellations. Note that the value of r
satisfying the above conditions (i) and (ii) is greater than 0 and smaller than 1.
[0140] The multi-dimensional rotation matrix for rotating a multi-dimensional constellation
may be normalized or unnormalized.
[0141] The only open issue is what values the sign matrix

should take. The sign matrix

is defined by the following Math. 24.
[Math.24]

[0142] A necessary condition, which is not sufficient however, is that the sign matrix

must be orthogonal, up to a scaling factor. Such matrices are known in the literature
as the Hadamard matrices. Because a and b in the multi-dimensional rotation matrix

are different, the additional condition shown in the following Math. 25 must be
imposed.
[Math.25]

[0143] This condition ensures that any a*b product cancels out with the corresponding b*a
product.
[0144] If all elements on the main diagonal have the same sign, and each pair of elements
that are symmetrical with respect to the main diagonal have opposite signs, this
condition is fulfilled. Examples of such particularly preferred sign matrices for the 4-D
and 8-D (eight-dimensional) cases are shown in the following Math. 26 and Math. 27,
respectively.
[Math.26]
[0145] It is to be noted that Hadamard matrices are only possible for sizes that are multiples
of four. Therefore, multi-dimensional rotation matrices exist only for numbers of di-
mensions that are multiples of four. Thus, the number of dimensions of a constellation
according to the present invention is preferably a multiple of four (e.g., 4, 8, 12 and
16).
[0146] Once the sign matrix

has been fixed, the resulting multi-dimensional rotation matrix

may be optimized for a certain constellation size, i.e., the number of bits or con-
stellation points per dimension, by performing the following steps: selecting the
"rotation factor" r accordingly; and calculating parameters a and b by substituting the
selected "rotation factor" r into the above Math. 23. To this end, any suitable opti-
mization algorithm may be employed. As an optimization target, the minimum number
of different values in the components of any two distinct multi-dimensional rotated
constellation points may be employed. Other optimization targets may be used as well.
According to a preferred embodiment of the present invention, a cost function is
defined that takes the minimum absolute differences between corresponding
components of any two distinct multi-dimensional rotated constellation points into
account. An example of such a cost function calculates the minimum over all N
absolute differences between corresponding components of two multi-dimensional
rotated constellation points and sums these minimum values, or their squares over all
pairs of multi-dimensional rotated constellation points.
[0147] The multi-dimensional rotated constellation may already be useful if the minimum
number of different values in the components of any two distinct multi-dimensional
rotated constellation points is larger than that pertaining to the multi-dimensional
unrotated constellation. Also, the multi-dimensional rotated constellation may already
be useful if the minimum absolute difference of two corresponding components of any
two distinct multi-dimensional rotated constellation points is larger than that pertaining
to the multi-dimensional unrotated constellation.
[0148] In a preferred embodiment of the present invention, the entire transmission process
including the transmission channel and the decoder is simulated in order to determine
the bit error rate. The "rotation factor" r may then be adapted so as to minimize the de-
termined bit error rate.
[0149] Hence, the present invention allows generating a multi-dimensional rotated con-
stellation that can be used for modulating and transmitting data over a plurality of
fading (sub-) channels or slots at optimum spectral efficiency. To this end, a con-
ventional hyper-cubic constellation with the desired number of dimensions N and the
desired number of bits per dimension (i.e., the number of constellation points per
direction) is set up, for instance, by selecting an appropriate subset of

, which is the N-dimensional integer lattice. Here,

is the set of all points of the N-dimensional space having integer coordinates. This
hyper-cubic constellation may, for instance, be a generalization of a conventional
regular QAM constellation to N dimensions. However, other initial constellations may
be used, such as generalizations of circular constellation to N dimensions, and so on.
[0150] Once the initial constellation is fixed, it may be subjected to a rotation by applying
the above defined multi-dimensional rotation matrix

to each of the initial constellation points so as to obtain a rotated set of constellation
points, i.e., a multi-dimensional rotated constellation. The multi-dimensional rotated
constellation may be more favorable than the initial constellation in terms of the degree
of modulation diversity provided, depending on the particular choice of the "rotation
factor" r. The "rotation factor" r, and therewith the rotated constellation, may be varied,
as described above, so as to obtain a constellation that provides maximum modulation
diversity, or at least a certain minimum degree of modulation diversity, as required by
the specific application.
[0151] The present invention also provides a method and an apparatus for efficiently
transmitting and receiving data over a plurality of fading (sub-) channels or slots on the
basis of a modulation scheme that employs a multi-dimensional rotated constellation as
obtained by the above described method. The inventive method or apparatus may
either perform the above described method in order to obtain the desired multi-
dimensional rotated constellation, or use a set of predefined and prestored constellation
points of the multi-dimensional rotated constellation that have been calculated using
the above described method. In the latter case, the inventive method or apparatus may
access a storage means, wherein information indicating the positions of at least some
of the constellation points is stored.
[0152] Another aspect of the present invention concerns the separation and mapping of the
N dimensions of the N-dimensional rotated constellation so that they experience in-
dependent fading during transmission. This is a key aspect necessary for achieving the
expected diversity performance.
[01531 Generally, this can be achieved by transmitting each of the N components of a con-
stellation point of an N-dimensional rotated constellation over a different one of a
plurality of transmission channels, provided that fading of each of these transmission
channels is uncorrected with fading of any other one of the transmission channels.
Here, the phrase "a different one of a plurality of transmission channels" may refer to a
different one of a plurality of time slots, frequencies, transmitter antennas, or com-
binations thereof. In the context of orthogonal frequency-division multiplexing
(OFDM), the phrase "a different one of a plurality of transmission channels" may in
particular refer to a different one of a plurality of active carriers, OFDM symbols, or
combinations thereof. In the context of a single carrier system, the phrase "a different
one of a plurality of transmission channels" may in particular refer to a different one of
a plurality of symbols or time slots.
[0154] Further signal processing is possible before transmission. The critical aspect is that
fading experienced by each of the N dimensions must be different from, or ideally un-
correlated with, fading experienced by any other one of the N dimensions.
[0155] The spreading of the N dimensions across different time slots, frequencies, and
transmitter antennas can be achieved for example through appropriate interleaving and
mapping.
[0156] Another aspect of the present invention concerns the mapping of the N real di-
mensions of the N-dimensional rotated constellation to complex symbols for
transmission. Since fading of the in-phase component and the quadrature component of
a given channel is typically identical, a complex symbol may not be made up of two
different components of the same constellation point. Instead, the N components of a
constellation point must be mapped to different complex symbols in order to ensure the
desired diversity.
[0157] The complex symbols generated in this manner are then spread in a conventional
manner over the available time slots, frequencies, and/or antennas, e.g. through in-
terleaving and mapping, so that fading experienced by each of the N dimensions is un-
corrected with fading experienced by any other one of the N dimensions.
[0158] The following describes an example flow of a method for generating a multi-
dimensional constellation for a digital modulation scheme in data transmission. This
flow is achieved by, for example, a computer system. Each of the following steps is
executed by a central processing unit (CPU).
[0159] (Step 1) A plurality of vectors of an N-dimensional vector space are received. Note,
for example, the plurality of received vectors represent a subset of

, which is an N-dimensional integer lattice.
[0160] (Step 2) Sign values s^ of the sign matrix shown in the above Math. 24 are de-
termined, such that the N-dimensional rotation matrix

shown in the above Math. 20 is orthogonal.
[0161] (Step 3) A "rotation factor" r is selected as a real number between 0 and 1. It should
be noted that the "rotation factor" r, for example, is selected so that it maximizes the
minimum number of different values in the components of any two distinct multi-
dimensional constellation points. However, the present invention is not limited to this.
Alternatively, the "rotation factor" r may be selected so that it increases a minimum
number of different values in components of any two distinct N-dimensional rotated
constellation points relative to a minimum number of different values in components of
any two distinct vectors received in Step 1.
[0162] (Step 4) Values of the parameters a and b are calculated by substituting the value of
the "rotation factor" r, which has been selected in Step 3, into the above Math. 23.
[0163] (Step 5) The N-dimensional rotation matrix

is determined from the above Math. 20 by using (i) the sign matrix

having the sign values Sy determined in Step 2, and (ii) the values of the parameters a
and b calculated in Step 4.
[0164] (Step 6) A constellation point of the N-dimensional rotated constellation is obtained
by applying the N-dimensional rotation matrix

determined in Step 5 to the plurality of vectors of the multi-dimensional vector space,
which have been received in Step 1.
[0165] FIG. 12 is a block diagram of a transmission apparatus according to an embodiment
of the present invention, which is similar to the one shown in FIG. 2. The elements that
are the same as those described above are assigned the same reference numerals
thereas, and a detailed explanation thereof is omitted.
[0166] The transmission apparatus of FIG. 12 differs from that of FIG. 2 in that the rotated
constellation demapper 230 is replaced with a rotated constellation demapper 1230.
The rotated constellation demapper 1230 performs processing on the basis of an N-
dimensional rotated constellation that has a plurality of constellation points defined by
positions thereof within an N-dimensional space, the positions being obtained by
applying either the N-dimensional rotation matrix shown in the above Math. 20, or an
N-dimensional rotation matrix obtained by permuting rows and/or columns of the N-
dimensional rotation matrix shown in the above Math. 20, to a subset of

, which is the N-dimensional integer lattice. To be more specific, this processing is to
map the output of the bit interleaver 220 to the rotated constellation.
[0167] FIG. 13 is a block diagram of a reception apparatus according to an embodiment of
the present invention, which is similar to the one shown in FIG. 4. The elements that
are the same as those described above are assigned the same reference numerals
thereas, and a detailed explanation thereof is omitted.
[0168] The reception apparatus of FIG. 13 differs from that of FIG. 4 in that the rotated con-
stellation demapper 450 is replaced with a rotated constellation demapper 1350. The
rotated constellation demapper 1350 performs processing on the basis of an N-
dimensional rotated constellation that has a plurality of constellation points defined by
positions thereof within an N-dimensional space, the positions being obtained by
applying either the N-dimensional rotation matrix shown in the above Math. 20, or an
N-dimensional rotation matrix obtained by permuting rows and/or columns of the N-
dimensional rotation matrix shown in the above Math. 20, to a subset of

, which is the N-dimensional integer lattice.
[0169] FIG. 14 shows an example hardware implementation for the rotated constellation
demapper 1350 of FIG. 13 for a 16-QAM rotated constellation (N = 2, B = 2). The
rotated constellation demapper 1350 of FIG. 13 includes a rotated constellation mapper
1420, instead of the rotated constellation mapper 720 shown in FIG. 7. The rotated
constellation mapper 1420 maps the outputs bj to b4 from the counter 710 to an N-
dimensional rotated constellation that has a plurality of constellation points defined by
positions thereof within an N-dimensional space, the positions being obtained by
applying either the N-dimensional rotation matrix shown in the above Math. 20, or an
N-dimensional rotation matrix obtained by permuting rows and/or columns of the N-
dimensional rotation matrix shown in the above Math. 20, to a subset of

, which is the N-dimensional integer lattice. Then, the rotated constellation mapper
1420 outputs the resulting constellation components s1 to s4 to the squared Euclidean
distance calculator 730.
[0170] It should be noted that the structures of the transmission apparatus and the reception
apparatus are not limited to those described above. For example, the reception
apparatus may have either one of the structures shown in FIGs. 10 and 11. In this case,
the rotated constellation demapper 1010 or 720 performs processing on the basis of an
N-dimensional rotated constellation that has a plurality of constellation points defined
by positions thereof within an N-dimensional space, the positions being obtained by
applying either the N-dimensional rotation matrix shown in the above Math. 20, or an
N-dimensional rotation matrix obtained by permuting rows and/or columns of the N-
dimensional rotation matrix shown in the above Math. 20, to a subset of

, which is the N-dimensional integer lattice.
[0171] The present invention relates to digital data communication and provides an efficient
method for generating multi-dimensional constellations for digital data modulation
with a high degree of modulation diversity, a method for transmitting and receiving
data on the basis of such constellations, and a corresponding apparatus. This is
achieved by considering only multi-dimensional rotation matrices with all elements on
the diagonal having the same first absolute value and all other elements having the
same second absolute value. In this manner, multi-dimensional rotation matrices can
be generated having a single independent parameter and a structure that is as regular as
possible. The independent parameter can be configured in order to minimize the error
probability for various constellation sizes.
Industrial Applicability
[0172] The present invention is applicable to a communication apparatus that performs
modulation/demodulation by using a constellation.
Reference Signs List
[0173] 210 FEC encoder
220 bit interleaver
1230 rotated constellation mapper
240 complex symbol mapper
250 symbol interleaver/mapper
260-1 to 260-M modulation chain
270-1 to 270-M transmitter antenna
410-1 to 410-M receiver antenna
420-1 to 420-M demodulation chain
430 symbol demapper/deinterleaver
440 complex symbol demapper
1350 rotated constellation demapper
460 bit deinterleaver
470 FEC decoder
We Claim:
[Claim 1] A transmission apparatus for transmitting a block of data over a
plurality of transmission channels, the transmission apparatus
comprising:
a modulator operable to select one of a plurality of constellation points
in accordance with the block of data to be transmitted, each of the
plurality of constellation points having a plurality of components; and
a transmitter operable to transmit each component of the selected
constellation point over a different one of the plurality of transmission
channels, wherein
the plurality of constellation points are defined by positions thereof
within an N-dimensional space, the positions being obtained by
applying an orthogonal transformation to a subset of

, which is an N-dimensional integer lattice,
N is a multiple of four, and
the orthogonal transformation has one of (i) an N-by-N matrix
representation with absolute values of all elements on a main diagonal
equal to a first value, and with absolute values of all elements not on
the main diagonal equal to a non-zero second value, and (ii) a matrix
representation obtained by permuting rows and/or columns in the
N-by-N matrix representation.
[Claim 23 The transmission apparatus according to claim 1, further comprising
a mapper operable to map each component of the selected constellation
point to the corresponding one of the plurality of transmission channels
over which the component is to be transmitted, such that fading of each
of the plurality of transmission channels is uncorrected with fading of
any other one of the plurality of transmission channels.
[Claim 5] The transmission apparatus according to claim 1, wherein
the transmitter is adapted for transmitting each component of the
selected constellation point over a different one of a plurality of time
slots, frequencies, transmitter antennas, or combinations thereof.
[Claim4] The transmission apparatus according to claim 1, wherein
the plurality of transmission channels comprise a plurality of different
carriers in an orthogonal frequency-division multiplexing scheme.
[Claim 5] The transmission apparatus according to claim 1, wherein
the plurality of transmission channels comprise a plurality of different
symbols in an orthogonal frequency-division multiplexing scheme.
[Claim 6] A reception apparatus for receiving a block of data over a plurality of
transmission channels, the reception apparatus comprising:
a receiver operable to receive a plurality of component signals over the
plurality of transmission channels; and
a demodulator operable to select one of a plurality of constellation
points in accordance with the plurality of received component signals,
wherein
the plurality of constellation points are defined by positions thereof
within an N-dimensional space, the positions being obtained by
applying an orthogonal transformation to a subset of

, which is an N-dimensional integer lattice,
N is a multiple of four, and
the orthogonal transformation has one of (i) an N-by-N matrix
representation with absolute values of all elements on a main diagonal
equal to a first value, and with absolute values of all elements not on
the main diagonal equal to a non-zero second value, and (ii) a matrix
representation obtained by permuting rows and/or columns in the
N-by-N matrix representation.
[Claim7] The reception apparatus according to claim 6, wherein
the plurality of transmission channels comprise a plurality of different
carriers in an orthogonal frequency-division multiplexing scheme.
[Claim 8] The reception apparatus according to claim 6, wherein
the plurality of transmission channels comprise a plurality of different
symbols in an orthogonal frequency-division multiplexing scheme.
[Claim 9 ] A transmission method for transmitting a block of data over a plurality
of transmission channels, the transmission method comprising the steps
of:
selecting one of a plurality of constellation points in accordance with
the block of data to be transmitted, each of the plurality of constellation
points having a plurality of components; and
transmitting each component of the selected constellation point over a
different one of the plurality of transmission channels, wherein
the plurality of constellation points are defined by positions thereof
within an N-dimensional space, the positions being obtained by
applying an orthogonal transformation to a subset of

, which is an N-dimensional integer lattice,
N is a multiple of four, and
the orthogonal transformation has one of (i) an N-by-N matrix
representation with absolute values of all elements on a main diagonal
equal to a first value, and with absolute values of all elements not on
the main diagonal equal to a non-zero second value, and (ii) a matrix
representation obtained by permuting rows and/or columns in the
N-by-N matrix representation.
[Claim 10] The transmission method according to claim 9, further comprising the
step of
mapping each component of the selected constellation point to the
corresponding one of the plurality of transmission channels over which
the component is to be transmitted, such that fading of each of the
plurality of transmission channels is uncorrected with fading of any
other one of the plurality of transmission channels.
[Claim 11] The transmission method according to claim "9, wherein
the plurality of transmission channels comprise a plurality of different
carriers in an orthogonal frequency-division multiplexing scheme.
[Claim 12.] The transmission method according to claim 9, wherein
the plurality of transmission channels comprise a plurality of different
symbols in an orthogonal frequency-division multiplexing scheme.
[Claim 13] A reception method for receiving a block of data over a plurality of
transmission channels, the reception method comprising the steps of:
receiving a plurality of component signals over the plurality of
transmission channels; and
selecting one of a plurality of constellation points in accordance with
the plurality of received component signals, wherein
the plurality of constellation points are defined by positions thereof
within an N-dimensional space, the positions being obtained by
applying an orthogonal transformation to a subset of

, which is an N-dimensional integer lattice,
N is a multiple of four, and
the orthogonal transformation has one of (i) an N-by-N matrix
representation with absolute values of all elements on a main diagonal
equal to a first value, and with absolute values of all elements not on
the main diagonal equal to a non-zero second value, and (ii) a matrix
representation obtained by permuting rows and/or columns in the
N-by-N matrix representation.
[Claim 14] The reception method according to claim 13, wherein
the plurality of transmission channels comprise a plurality of different
carriers in an orthogonal frequency-division multiplexing scheme.
[Claim 151 The reception method according to claim 13, wherein
the plurality of transmission channels comprise a plurality of different
symbols in an orthogonal frequency-division multiplexing scheme.
[Claim 16] A generation method for generating a multi-dimensional constellation
for a digital modulation scheme in a data communication system, the
generation method comprising the steps of:
receiving a plurality of vectors of a multi-dimensional vector space;
and
obtaining constellation points of the multi-dimensional constellation by
applying an orthogonal transformation to the plurality of vectors
received, wherein
the orthogonal transformation is adapted for increasing a minimum
number of different values in components of any two distinct
multi-dimensional constellation points relative to a minimum number of
different values in components of any two distinct vectors received,
and
the orthogonal transformation has one of (i) an N-by-N matrix
representation, N being a multiple of four, with absolute values of all
elements on a main diagonal equal to a first value, and with absolute
values of all elements not on the main diagonal equal to a non-zero
second value, and (ii) a matrix representation obtained by permuting
rows and/or columns in the N-by-N matrix representation.
[Claim 17] The generation method according to claim 16, further comprising the
steps of:
selecting a rotation factor r as a real number between 0 and 1;
calculating the first value, a, by evaluating an expression

calculating the second value, b, by evaluating an expression

determining the orthogonal transformation by selecting a sign value Sij
for each element (i, j) of a matrix representation

, such that the matrix representation is orthogonal.
[Claim 18] The generation method according to claim 17, wherein
the selected rotation factor r maximizes the minimum number of
different values in the components of any two distinct
multi-dimensional constellation points.
[Claim 19] The generation method according to claim .16, wherein
the plurality of vectors received represent a subset of

, which is an N-dimensional integer lattice.

ABSTRACT
The present invention relates to digital data
communication and provides an efficient method for
generating multi - dimensional constellations for
digital data modulation with a high degree of
modulation diversity, a method for transmitting and
receiving data on the basis of such constellations, and
a corresponding apparatus. This is achieved by
considering only multi - dimensional rotation matrices
with all elements on the diagonal having the same first
absolute value and all other elements having the same
non-zero second absolute value. In this manner, multi - dimensional rotation matrices can be generated having
only a single independent parameter and a structure
that is as regular as possible. The independent
parameter can be configured in order to minimize the
error probability for various constellation sizes.