Title of Invention

METHOD FOR DEMODULATING A CARRIER WAVE MODULATED USING A DIGITAL SYMBOL SEQUENCE

Abstract A method is described for demodulating a carrier wave that is modulated using a digital symbol sequence and that is transmitted over a noise-impacted channel, the ideal edge shapes of possible transitions between two symbols being known and stored in memory (reference edges), and a received edge being scanned and digitalized using a scanning frequency which is a multiple of the frequency of the symbol sequence. According to the present invention, it is provided that for detecting a received and scanned edge, all scanning values are used for the calculation of Euclidean distances from at least two reference edges, and the reference edge having the lowest Euclidean distance is selected. In one preferred embodiment, it is provided that a Viterbi algorithm is applied not, as is the case in known CPM demodulation methods, to an estimated symbol sequence, but rather to a sequence of estimated edges, the specific Euclidean distances between the edge received in one symbol period and the reference edges being considered as the costs of one trellis branch of a Viterbi algorithm.
Full Text

METHOD FOR DEMODULATING A CARRIER WAVE MODULATED USING A
DIGITAL SYMBOL SEQUENCE
The invention relates to a method for demodulating a carrier wave modulated using a digital symbol sequence and transmitted over a noise-impacted channel, the ideal edge shapes of possible transitions between two symbols being known and stored in memory (reference edges), and a received edge being scanned and digitalized using a scanning frequency which is a multiple of the frequency of the symbol sequence.
In particular, continuous phase modulation (CPM) is used to transmit (data) packets. In this context, by varying the phase angle (relation, position), a plurality of different symbols; e.g., four;, can be transmitted. Since modulation using square-wave (rectangular) pulaes all having the same symbol length would lead to a very broad spectrum, modulation is often carried out using pulses that extend over the length of two symbol periods and that are represented as cosine-shaped --to avoid steep edges, A modulation of this type is also known as 2RC-CPM,
From the conference proceedings,
"K.'H. Tietgen, 'Numerical Modulation Methods Applied in the FD/TDMA System S 900 D,' Second Nordic Seminar on Digital Land Mobile Radio Communications, Stockholm,
it is known not to use symbols themselves for modulation, but rather to numerically modulate the known edges between two

Figure 1 shows the block diagram of a transmitter that is well-suited to use the known method. In this context, the system parts shaded in gray operate in the high system clock.
Figure 2, for a four-value symbol alphabet, shows the conceivable edges in the transition from one specific symbol to the next in a symbol sequence to be transmitted. The idea in the known modulation methods is to describe the instantaneous frequency (i.e., the phase derived from time) of a (for example) four-step signal given a pulse shape of a length 2T as a side-by-side arrangement (consecutive alignment) of 16 possible edges fn(t) having length T. Instead of transmitting overlapping, shifted elementary pulses Σ1d(l)g(t - iT), is also possible to directly transmit a non-overlapping sequence of edges ΣifN[t - iT), neighboring symbols d{I) and d[I + 1)T in each case deciding which of the 16 edges in interval iT In the cited literature reference, a method of this type is termed CP-4FSK. To simplify the comparison with other methods, in what follows, it will be termed Num2RC 4st, derived from a four-step method, in which modulation takes place numerically and the edge shapes are based on the application of a "raised cosine" lasting for two symbols.
To receive a data sequence modulated in this way, a receiver has been proposed in the seminar report,
"D,E. Pfitzmann and H.-p, Ketterling, A New CP-4FSK Sampling Demodulator for the FD/TDMA System S 900 D,'

Second Nordic Seminar on Digital Land Mobile Radio Communications, Stockholm, 1986."
The same receiver is also known from German Patent 36 2 8 993
The mode of operation of the receiver described above is a method of the species. Using the receiver described above, a received edge is scanned and digicalized using a scanning frequency which is a multiple of the frequency of the symbol sequence. In the example cited, scanning is carried out between two symbols, for example, 16 times. The digitalised values are compared with each other and, in the event that the difference is small becween two consecutive values, are combined into a symbol average (center). Ey averaging consecutive values, it is attempted to elimunace the influence of noise.
Classical CPM receivers and demodulation methods are also termed "discriminator Integrate & Dump," Therefore, in what follows, when a signal edge shape of the "raised cosine" lasting for two symbol lengths is used, the term 2RCdsI&D is used, being supplemented, as appropriate, by the addition 2st (i.e., two-step, two symbols) or 4st {i,e., four-step, four symbols).
In general, it is desirable for a receiver if each individual symbol is received in a highly reliable manner. In the classical CPM receiver method, it is necessary to filter the signal that has become noisy in passing through the real transition channel. Conventional filters operate as an integrator, for example, over two symbol periods, to average the noise. In the above-mentioned 2RC signal shapes,

integration leads to intersymbol interference,abbreviated as ISI. To avoid intersymbol interference, it is known to apply a Viterbi algorithm to the received symbol sequence-Disadvantageous in this context is the great computing effort that places correspondingly great demands on the hardware,
Therefore; the present invention is based on the objective of indicating a demodulation method in which, based on a method according to the species, it is possible to have small computing effort while maintaining the error rate at as close to the same level as possible. In this manner, the design effort for a receiver is simplified without relinquishing the advantages of "Continuous Phase Modulation" (CPNI) such as low complexity and a compact spectrum, which, as was already mentioned, is of decisive importance with respect co good bandwidth utilization.
The objective is achieved through the present invention in that, to detect a received and scanned edge, all of the scanning values are used, in each case, to calculate Euclidean distances from at least two reference edges, and the reference edge having the lowest Euclidean distance is selected.
In contrast to the cited classical receivers, in a demodulation method of this type, instead of the symbols to be estimated, the transition edges between the symbols are made central to the detection process. These edges are scanned multiple times and are compared with the original edges. Subsequently; the decision is made in favor of the edge whose distance, in the Euclidean sense, from the received edge is the smallest. This decision is carried out edge-by-edge independently of the neighboring edges, the decision hypothesis therefore not necessarily guaranteeing the

continuity of the derivation of the phase of the CPM signal.
Therefore, an inventive refinement is preferred in which there is a continuous phase,
Therefore, according to the present invention, it is preferably provided that a Viterbi algorithm be applied to a number of consecutive edges, the respective Euclidean distances between the edges received in one symbol period and the reference edges being considered as the cost of a trellis branch of the Viterbi algorithm.
The properties of the CPM modulation thus make it possible to carry out an equalization of the data signals very efficiently,
The proposed maximum-likelihood estimation of the received data sequences using a Viterbi algorithm not on the symbol sequences but on the received edges has the consequence that a model can be used whose number of states is smaller by a factor of M in comparison to the classical equalizer. In this context, M designates the step-quantity of the signal. Therefore, the computing effort required can be kept small, as a result of which the hardware implementation becomes more attractive, as is described below.
The present invention is described in greater detail below on the basis of the drawing. The contents of the drawing are as follows;
Figure 1 depicts the block diagram of a transmitter using
numerical modulation of a digital symbol sequence on a carrier wave,

Figure 2 depicts a representation of the conceivable edges between two consecutive symbols, using a 2RC pulse and a four-step symbol alphabet (Num2RC 4st) ,
Figure 3 depicts a block diagram of a receiver according to
the present invention,
Figure 4 depicts a table by analogy no Figure 2 for a two-step modulation,
Figure 5 depicts a representation of a reference edge; an
actually received sdge, and a trellis segment havin assigned costs,
Figure 6 depicts an SER curve (Symbol Error Rate) for the
two-step Num2RC, having and not having Viterbi
detection,
Figure 7 depicts a comparison according to Figure 6 for the
four-step Num2RC; having and not having Viterbi detection,
Figure 8 depicts a comparison for the two-step case of the method according to the present invention having Viterbi detection of the edges and a classical C?M demodulation method (Integrate & Dump), and
Figure 9 depicts a comparison according to Figure 8 for the
four-step case.
Figure 1 shows a generally known transmitter- for the numerical modulation of a digital symbol sequence. The baseband model of the transmitter is composed of a serial-parallel transducer

(converter) (S/?), a look-up table (LUX), in which essentially the contents of Figure 2 are stored, and a baseband FM modulator.
The four-step signal is extracted from the input bic scream by combining, in each case, two bits. As a result of a delay of one symbol period, a total of four bits then address the LUT depicted in Figure 1, containing the 16 symbol transitions fN(t) depicted in Figure 2, The number of interpolation points (supporting values) can be freely determined, the number 16 being proposed in the related art cited in the specification. This number is a compromise, because although a high number of incerpolation points is desirable to achieve good noise averaging, nevertheless a low number keeps the computing effort down. The parts of the transmitter that operate in the high system clock are shaded in gray.
Figure 3 shows the receiver according to the present invention.
The baseband model is composed of a band-limiting low pass TP, a baseband FM demodulator having parasite AM suppression, an edge detector, and a parallel-serial transducer (P/S). The parts of the receiver that operate in the high system clock are again shaded in gray.
According to the present invention, in the edge detector, the determination of the edge that most likely corresponds to the transmitted edge is carried out in an edge-by-edge manner.
Figure 4 depicts, in the form of a table, the two-step case in conceivable edges, the two-step case being fundamental for the

observation in Figure 5. Expansion to the four-step case or that of an even higher-step proceeds by analogy with no limitation on generalisability.
Figure 5, in the lafthand diagram, the solid line depicts a received, noisy edge, and the dot-dash line depicts one of the reference edges according to Figure 4 [Num2RC 2st). The (for example) IS values determined for each edge are used according to the present invention to calculate the sum of the quadratic differences (Euclidean difference). On the basis of the smallest Euclidean difference (right-hand side in Figure 5), the reference edge is identified which most closely corresponds to the transmitted edge. Each individual detected edge determines two data symbols: the output symbol and the final symbol,
According to one preferred embodiment of the present invention, provision is made to assure the continuity of the detected symbol transitions using a Maximum-Likelihood-Sequence estimator (i.e., preferably using a Vicerbi algorithm). Therefore, as is depicted in Figure 5 on the right-hand side on the basis of a trellis segment, according to the present invention, the distances of the received edge from all valid reference edge© are used as costs for the Viterbi algorithm. The Viterbi detector makes use of the following parameters; the trellis depth corresponds to step-quantity M of the signal (2RC pulse formation: memory of the length of one symbol period), the number of transitions is M-. According to this, the path unification length should be chosen at > 5 symbols.
For the purpose of illustrating the processes, the following description is limited to the two-step case without

restricting the generalizability-
The state transitions and the edges that are generated in the process are represented in the table according to Figure 4. The calculation of the cumulative path costs is made in the familiar manner: in every current state, the path costs are calculated arising from the cost contributions of the branches of the trellis diagram leading to this state and their points of origin. The minimum is determined and is stored on the underlying branch. The cost contributions of the individual branches of a trellis segment are calculated -- as was already explained -- as distances of the current received and noisy edge from the corresponding edges determined by the model.
The distances of the received edge from the possible edges are preferably calculated as the sum of the quadratic distances (Euclidean distance). However, it is also conceivable to use the sum of the absolute distances or a different suitable calibration.
The calculation of the cost contributions of one edge is explained on the basis of a numerical example. Figure S
depicts a -1 -> 1 edge according to the table in Figure 4, the edge being distorted as a consequence of interference. The distances from, all possible reference edges are entered on the corresponding edges, and therefore the right half of the image represents the trellis segment that belongs to this received edge. The "distance" of the received edge is smallest from the
-1 -> 1 edge, and therefore, in this trellis segment, this represents the most favorable branch.
After running through all of the trellis segments, the

decision is made in the usual way in favor of one path in the trellis structure; the path having the minimal costs is selected, i.e., the path that is closest to the received edge sequence. The symbols are then derived from the selecced path using the model. The difference from the known Viterbi algorithms can be seen in the fact that the Viterbi algorithm is applied not to a symbol sequence but to an edge sequence-As a result of the system-imminent redundancy; which is based an the fact that two neighboring edges have one symbol in common, the computing effort is significantly reduced. Although, when a classical Viterbi equalizer is used on symbol sequences; the ISI (intersymbol interference) can also be eliminated, nevertheless the memory length to be taken into account is increased to two symbol periods, which resulcs in an M-fold increase in the number of states and; as a result; a drastic increase in the computing effort. In contrast thereto, the computing effort in the method according to the present invention, i.e., the Num2RC rsceiver, is smaller by a factor of M.
In the four-step case, for example in the NumSRC dem.odulation according to the present invention using Viterbi detection; only one quarter of the computing effort is necessary/ in comparison to the classical CPM receiver method.
Figures 6 through S depict the advantages that are achieved by che method according to the present invention.
Figure 6 depicts a comparison of the demodulation method according to the present invention (Num2RC 2st} with the same method having the additional application of a Viterbi algorithm on edge sequences according to Figure 5 (Num2RC 2st vit) .

Figure 7 depicts an analogous comparison for the four-step case. On the vertical axis, the error race is indicated (SER, Symbol Error Rate) .
The comparison of Figure 6 and Figure 7 indicates that already in the four-step case, a significant reduction of the error rate can be observed.
Figures 8 and 9 depict a comparison of the method according to the present invention, having additional Viterbi detection of the received edges, with a classical CPM demodulation method, in which an integration is carried out over a symbol interval (period), i.e., the so-called "Integrate & Dump," to achieve an average of the channel noise, a Viterbi algorithm being additionally used on the detected symbols to identify the most probable symbol sequence and thus to reduce intersymbol interference.
The Figures show that the error rate of the method according to the present invention is slightly worse both in the two-step as well as in the four-step case. However, in this context, it should be noted that in the two-step case according to Figure 8, the computing effort comes to only one half, whereas in the four-step case, according to Figure 9, only one quarter of the computing effort is required.
Therefore, especially in the four-step case, it is possible to design receivers using a method according to the present invention, which are produced on the basis of significantly lower hardware costs and which are therefore more cost-effective to manufacture.









Patent Claims
1. A method for demodulating a carrier wave that is modulated using a digital symbol sequence and that is transmitted over a noise-impacted channel, the ideal edge shapes of possible transitions between two symbols being known and stored in memory (reference edges) , and a received edge being scanned (sampled) and digitalized using a scanning [sampling) frequency which is a multiple of the frequency of the symbol sequence, characterized in that, for detecting a received and scanned edge, all scanning values are used for the calculation; in each case, of Euclidean distances with respect to at least two reference edges, and the reference edge having the lowest Euclidean distance is selected.
2. The method as recited in Claim 1, characterized in that a Viterbi algorithm is used on a number of sequential edges, the Euclidean distances between the edge, received in one symbol period, and the reference edges being considered as the costs of one trellis branch of the Viterbi algorithm,
3. The method as recited in Claim 1 or 2, characterized in that, in a scanning frequency of n/T and a signal step-quantity of M, the reference edges are described by a total of nM3 stored values.
4, The method as recited in one of the preceding claims,
characterized in that the modulation is carried out as a CPM
(continuous phase modulation).
5. The method as recited in Claim 4, characterized in
that the modulation includes a numerical formation of the
derivation and of the phase of one cosine signal (2RC-CPM)

extending over two symbol periods.
6_ The method as recited in one of the preceding claims.
characterised in that the trellis depth of the Viterbi
algorithm corresponds to the step-quantity M of the signal
(Ncrelling := M) .

6. A method for demodulating a carrier wave substantially as herein described with reference to the accompanying drawings.
Dated this 6 day of November 2000


Documents:

abs-in-pct-2000-609-che.jpg

in-pct-2000-609-che-abstract.pdf

in-pct-2000-609-che-claims filed.pdf

in-pct-2000-609-che-claims grand.pdf

in-pct-2000-609-che-correspondence others.pdf

in-pct-2000-609-che-correspondence po.pdf

in-pct-2000-609-che-description complete filed.pdf

in-pct-2000-609-che-description complete grand.pdf

in-pct-2000-609-che-drawings.pdf

in-pct-2000-609-che-form 1.pdf

in-pct-2000-609-che-form 19.pdf

in-pct-2000-609-che-form 26.pdf

in-pct-2000-609-che-form 3.pdf

in-pct-2000-609-che-form 5.pdf

in-pct-2000-609-che-pct.pdf


Patent Number 210541
Indian Patent Application Number IN/PCT/2000/609/CHE
PG Journal Number 50/2007
Publication Date 14-Dec-2007
Grant Date 08-Oct-2007
Date of Filing 06-Nov-2000
Name of Patentee M/S. ROBERT BOSCH GMBH
Applicant Address Postfach 300220 D-70442 Stuttgart
Inventors:
# Inventor's Name Inventor's Address
1 CINKLER, Kalman Heinrich-Böll-Strasse 105 D-28215 Bremen,
2 KAMMEYER, Karl-Dirk Finkennest 6 D-21224 Buchholz,
PCT International Classification Number H04L25/06
PCT International Application Number PCT/DE1999/001041
PCT International Filing date 1999-04-07
PCT Conventions:
# PCT Application Number Date of Convention Priority Country
1 198 15 701.0 1998-04-08 Germany