This application claims priority from German application 10 2004 025 826.0 filed May 24, 2004, which is hereby incorporated by reference.
The invention relates to a method and apparatus for decoding a bit sequence from QPSK or QAM symbols
There are generally two known fundamental techniques for decoding QPSK or QAM symbols-hard demapping and soft demapping. In the hard demapping approach, the individual received QPSK or QAM signals (received vectors) are assigned based on an unambiguous decision to a constellation point (symbol vector) (see
This soft decision decoding, which weights the demodulated data by the error probability of the data, results in an improved forward error correction. For communication systems that utilize M-level QAM, the receiver thus requires a decoding algorithm that uses a two-dimensional (complex) receive signal to calculate the corresponding soft decision values as the input signals for the channel decoder. The prerequisite to ensure the reliability of this type of system is that the correct occurrence probability parameters for a given symbol are used as the basis for calculating the corresponding soft decision values.
As a rule, the receiver operates according to the maximum likelihood principle in which the individual probabilities are each multiplied and the receive sequence with the highest overall probability is selected. The main approach to determining the required individual probabilities is to use the Euclidean distance between receive vector and the nearest ideal symbol vector. In addition, it is generally assumed that the transfer channel shows a Gaussian amplitude distribution for the noise. Given a high signal-to-noise ratio, the logarithmic maximum likelihood function for this transfer channel is assumed to be approximately represented by:
LLR˜(CTF(i)2/σ2*(min[r(i)−α0]−min[r(i)−α1]2)
where:
In coded orthogonal frequency division multiplexing (COFDM) systems, the soft information for the forward error correction (FEC) should for this reason be computed from the energy of the given carrier, the detected noise energy, and the probability of the corresponding constellation point.
The prior-art approach to accomplishing this starts with a fixed noise energy.
Calculation of the soft information is frequently implemented using mapping or lookup tables, see U.S. Pat. No. 6,115,435. Using this approach, the handling of the various constellations or hierarchy modes, such as those supporting, for example, DVB-T (Digital Video Broadcasting—terrestrial), specifically, 16-QAM, 64-QAM, a non-hierarchical constellation, a hierarchical constellation, et cetera, is difficult.
If, on the other hand, the decoding characteristic is calculated explicitly, the implementation is often either too complex, or significant approximation errors occur. For example, although U.S. Pat. No. 6,424,685 provides a comparatively simple calculation of the decoding characteristic from polar coordinates, considerable effort is required to adapt to the different constellations or hierarchy modes.
To simplify the decoding process, recent publications propose a transformation of the received constellation vectors into a simpler constellation arrangement. The term used here is “remapping”. For example, U.S. Pat. No. 6,661,282 describes a remapping by subtraction of an offset. However, this procedure is suitable only for the 16-QAM method. However, U.S. Pat. No. 6,226,333 describes the decoding of QAM symbols from a single quadrant by employing a rotator.
Therefore, there is a need for a technique of decoding QPSK or QAM symbols in which different constellations and hierarchy modes are easily implementable. In addition, the technique should have a high degree of reliability in predicting the decoded QPSK or QAM symbols.
According to an aspect of the invention, a bit sequence (b, b′) from QPSK or QAM symbols received following transmission over a channel is decoded, and an associated receive probability (w, w′) is assigned to each receive bit (b, b′). The receive probability (w, w′) is adaptively determined as a function of the transfer properties of the channel.
These and other objects, features and advantages of the present invention will become more apparent in light of the following detailed description of preferred embodiments thereof, as illustrated in the accompanying drawings.
In the decoding system 1 includes a circuit to decode QPSK or QAM symbols by a “hard” decision (hard-decision-output demapper) 2, a circuit 3 to determine the receive probability w for a bit, and a circuit 4 for additional weighting of the probability w by a factor G.
The hard-decision-output demapper 2 receives signal vectors r with coordinates I/Q on a line 100. The hard-decision-output demapper 2 provides an output HD (hard decision) on a line 102 to tap a so-called “hard” decision b. The hard-decision-output demapper 2 provides a second output on a line 104 to the circuit 3 to determine the receive probability w of a bit. An output of this circuit 3 on a line 106 is in turn connected to the circuit 4 to weight the receive probability by a factor G. The weighting circuit 4 provides an output SD (soft decision) on a line 108 from which “soft” decision information g can be tapped. The circuits 3, 4 have control inputs on a line 110 through which time-variant and, in the case of multicarrier systems, carrier-dependent information about the carrier energy S, noise energy N, and/or interference can be supplied.
The demapping procedure is described below for a QPSK constellation:
An input vector r, which has in-phase coordinate I and quadrature coordinate Q and is input on the line 100 to the hard-decision demapper 2, is assigned internally to an ideal symbol vector α and an associated bit sequence b. This bit sequence b is output from the hard-decision-output demapper 2 on the line 102. In addition, the Euclidean distance a of the received signal vector r relative to the decision threshold 7, 8, (
g=G*W(a)
The weighting factor G which may preferably be employed here is the ratio SINR of the instantaneous signal energy S relative to the sum of the instantaneous noise and interference energies N, IF for the associated channel.
The decoding of a signal vector is described below using the example of a 16-QAM constellation:
An input vector r on the line 46 with in-phase coordinate I and quadrature coordinate Q is supplied to the remapper 45 and resolved step-by-step into sub-constellations. In a first step, input vector r is passed directly on to the hard-decision-output demapper 42. The hard-decision-output demapper 42 makes a hard decision by assigning the receive vector r to the two most-significant bits bh of the closest ideal symbol vector α. The soft information is determined in a procedure analogous to that described for QPSK.
After the initial hard decision, only a subset of possible ideal symbol vectors a remain. This remaining sub-constellation is selected in the remapper 45. Through appropriate transformation, this sub-constellation is transformed to a constellation symmetrical with the origin. This transformation involves a shift and, as necessary, a subsequent reflection.
If one starts with a non-hierarchical 16-QAM constellation as found, for example, in
If one assumes that, as in
If one assumes that, as in
If one starts with a hierarchical 16-QAM constellation, the result may be, for example, the transformation shown in
To implement the decision of the least significant bit, the receive vector r′ transformed into this constellation with in-phase coordinate I′ and quadrature coordinate Q′ is supplied to the input of the hard-decision-output demapper 42.
The transformed receive vector r′ is assigned internally to a transformed ideal symbol vector α′ and to an associated bit sequence b′. This bit sequence b′ can be tapped as a hard decision at the output HD of hard-decision-output demapper 42.
In addition, within the hard-decision-output demapper 42, the Euclidean distance a′ for the now transformed received signal vector r′ is determined relative to the decision threshold 7, 8 used for hard decision b′—as shown in
The value a′ is once again subjected to a soft-decision procedure. Referring still to
The log-likelihood ratio (LLR), generally employed to determine the receive probability for a bit, is ideally a function of the in-phase coordinate I and the quadrature coordinate Q, and thus a two-dimensional function.
The LLR characteristic for bits b of varying significance does vary considerably. However, it turns out that by appropriately shifting the individual characteristics for bits of different significance, a uniform overall characteristic is obtained which represents a sufficient approximation within the relevant control range (
This appropriate shift is implemented by the remapping procedure described above. As a result, a function that is uniform for all bits can be employed as demapping characteristic W, that is, the log-likelihood ratio (LLR) of the most significant bit MSB.
The characteristic W may be implemented, for example, by combining linear segments.
In this embodiment, output signal w′ generated by the circuit 3 is also weighted by a quantity G dependent on the carrier energy S and/or the noise energy N and/or the interference energy IF. The output quantity obtained is:
g′=G*W(a′)
The ratio SINR of the instantaneous signal energy S relative to the sum of the instantaneous noise and interference energies N, IF, of the associated channel may again be employed as the weighting factor G.
The combined de-/remapper 1202, 1205 may, first of all, be of similar design to that of the embodiment of
It is also possible to have the remapper 1205 connected on the output side of the demapper 1202, or to combine the remapper 1205 and demapper 1202 in a single circuit.
In addition, it is possible to implement the circuit 3 for determining the receive probability w of a bit and the circuit 4 for weighting receive probability w within a single circuit.
The following possibilities may be considered in regard to the control quantities S, N, IF for the adaptation of the circuits 3 and/or 4:
Case 1: Noise energy N is assumed to be constant; only signal energy S is determined from channel transfer function (CTF): S˜abs(CTF)2; interference IF is neglected.
Case 2: Noise energy N is a function of carrier i; signal energy S is determined from channel transfer function CTF: S˜abs(CTF)2, interference IF is neglected.
Case 3: Noise energy N is constant; signal energy S is determined from channel transfer function CTF: S˜abs(CTF)2; interference IF is determined for each carrier, or possibly estimated.
Case 4: noise energy N is a function of carrier i; signal energy S is determined from channel transfer function CTF: S˜abs(CTF)2; interference IF is determined for each carrier, or possibly estimated.
Although the present invention has been shown and described with respect to several preferred embodiments thereof, various changes, omissions and additions to the form and detail thereof, may be made therein, without departing from the spirit and scope of the invention.
Number | Name | Date | Kind |
---|---|---|---|
5377133 | Riggle et al. | Dec 1994 | A |
5968198 | Hassan et al. | Oct 1999 | A |
6115435 | Harada et al. | Sep 2000 | A |
6226333 | Spalink | May 2001 | B1 |
6347125 | Dent | Feb 2002 | B1 |
6424685 | Messel et al. | Jul 2002 | B1 |
6442212 | Kratochwil | Aug 2002 | B1 |
6529559 | Reshef | Mar 2003 | B2 |
6578173 | Alamouti | Jun 2003 | B2 |
6611551 | Jones et al. | Aug 2003 | B1 |
6661282 | Ha et al. | Dec 2003 | B2 |
6757337 | Zhuang et al. | Jun 2004 | B2 |
6990627 | Uesugi et al. | Jan 2006 | B2 |
7093188 | Maiuzzo et al. | Aug 2006 | B2 |
7099403 | Dagdeviren | Aug 2006 | B1 |
7120213 | Gatherer et al. | Oct 2006 | B2 |
7280840 | Murakami et al. | Oct 2007 | B2 |
Number | Date | Country |
---|---|---|
199 62 162 | Jul 2000 | DE |
1 024 634 | Aug 2000 | EP |
2 355164 | Apr 2001 | GB |
WO 0016528 | Mar 2000 | WO |
Number | Date | Country | |
---|---|---|---|
20050259765 A1 | Nov 2005 | US |