EP1203474A2

EP1203474A2 - Acs unit for a viterbi decoder

Info

Publication number: EP1203474A2
Application number: EP00958204A
Authority: EP
Inventors: Mario TRÄBER
Original assignee: Infineon Technologies AG
Current assignee: Infineon Technologies AG
Priority date: 1999-08-09
Filing date: 2000-08-02
Publication date: 2002-05-08
Also published as: CN1190902C; CN1369140A; US6813744B1; KR20020027535A; DE19937506A1; WO2001011842A3; JP3607894B2; WO2001011842A2; JP2003506969A

Abstract

The invention relates to an ACS unit for a Viterbi decoder. In order to determine the path metrics of two states of a time step in a trellis diagram, the inventive unit compares the difference (Η) between the path metrics of the two states (19, 20) of a previou s time step of the trellis diagram, which are linked to said states by means of branches in the form of a butterfly structure, with the differences (Μ) of the corresponding branch metrics. The structure of the ACS unit is simplified by evaluating the signs (sg(Η), sg(Μ)) of both differences.

Description

⁾ co N> P ¹

Ui o cπ O Cn o Cπ

ö 3 PHO <! S ö α o OS for Pi OS CJ H Hi? Σ φ> M £ S φ n ö t tu φ P ⁾ ti li fr O Φ Φ P P- P- ⁾ P- B P- o er Φ P ⁾ P ⁾ φ P- P- P oo P- o P- n Φ or> fr φ Φ ti rt HH Φ Φ tr φ Ό Φ er Φ for P ιQ PP er er P ä φ ω cn o tr ι- (ti n cn H l_l. P- HP pi rt Φ for P- P- φ Ό 1 n for lh P> (- ^{■ for} P- tr er φ Cn φ ⁾ rt PPH "cn <Φ P ⁾ <n tr

P- o P P-! Φ X P- Φ OPPP Hi M for OPO P- P Hi Hi HO P- H φ r-> cn cn r- ¹ tf Φ rt Φ P- tr P-> £) Hl PP>P> ou = »N Xι rt CP <! Φ HP Φ l- <^ XI 1 Φ 3 Φ P- Φ XX cn ιQ P iQ P fr cn P Φ Φ 3 PP P- 1 P ^J tr P- c φ Φ σ P- H 3 Hi P- Φ cn P > XP rt P- OHPH φ X Ϊ rt CΛ P- Φ er ti p ιQ P- P Φ er Φ XX P Φ er cn P Φ φ OP ⁾ er P- cn Φ Φ φ φ P- c

XX Φ P ⁾ ω z Φ P- Φ P> ω & Φ P- iQ H α ü P P- cn P- 3 H ι £) rt P er | rt φ P 1 P er P- φ P ιQ ω P P- Φ P 1 rt ^ Q n er φ φ iQ ti cn Φ H φ P- ö rt «Φ nt ⁾ Φ α P cn PP φ O rt ü Φ P Hi P- o PH) tu: o ti P> φ φ cn P> tr α P Φ J P- Hi XX HO Φ φ PP> pj: 1 rt α p: tr TJ P- ti n Ό P z X Φ N ω OO rt α H n P ¹ P o Φ H

Φ Z l-ti ιQ P- o J P- Φ cn cn tr Ü Z Φ P- o er Φ ιQ φ tf Φ Φ cn P ⁾ Φ XX fr Φ H P- Φ P> M φ 3 P- P φ £ fr Φ P Φ o P w Φ ti HP φ P- O n fr rt Λ P 3 P ⁾ h HP ⁾ p- XH o P> H P-

Φ for ^ & 3 φ tr o cn Ό P- tr rt for SCX Φ ⁾ Φ P for o Hi P

H X φ P> Φ H Φ <l p. Φ H Φ P- O P P- H P p- P- tr P- Φ c 3 3 rt P d Φ 3 P- o φ ti cn pi: α P- O P- er XX 3 Φ P P N φ P P

(D: H t <H cn Hi H fr Φ ti H N pι: P rt P 3 P- h rt cn C H fr fr

P N P- ιQ P> Φ l-i B iQ a <3 Φ Φ H Φ Φ P; <; rt α i T Φ fr Φ cn P fr rt tr rt Φ P- N tr CΛ Hi o 1 <! P rt H 3 P P-

Φ for P> cn HP φ P> Φ l- ¹ H Φ P- Φ C ti P Φ N o _^ ιQ rt

H P- P σ rt fr 3 H iQ B rt H P- cn ω PPP P- H fr er φ O Φ rt Φ 3 Φ ⁾ Φ P- H cn P> n φ Φ N fr? X X- s: P- P> 0 er er H ω cn rt Φ <pi: tr er cn H c φ cn P> Φ Φ ^ QHO Φ Φ Φ er z π P- rt rt Φ P- O cn Φ Φ tr iQ H P- PP ^J P P- α ^ f H rt P- φ li P- r + P H- Φ H rt tr NP cn I_l. O: Φ Xi P> P- rt P- P- er H 1 ti P φ NPH P- ≤ Φ C rt φ tr Hi P tr Φ P ⁾ cn φ P- Φ P- O fr p rt ιQ Φ rt n Φ h P- Hi 3 Φ fr Φ p: Φ PJ O o rt rt- lf P ιQ Hi φ

Φ α fr • 1 P er O P- O: tr O P ^ P o o • Φ • P cn H Hi n

P ω fr o for P- tr ti XX rt n • rt Φ fr for P P σ P- rt o

• P> os Φ P- P P- Φ 1 Φ fr * tr _^ NP P- 1 w XX φ Hi fr

XX o cn cn i-f P> rt ü P φ P- fr 03 P φ α P- S ω Hi φ P-

Φ α t 1 O ιQ Φ φ P O P »<Φ p <H P O <O P- Φ

0 P- P> O ü t H o ω tr tr o P- 3 p: Φ rt O er o P P H

Φ 3 Φ P- Pl n o Φ φ Φ ti er H Φ o <; P- ti α φ φ Φ α f-f n PJ Cd 3 o for P for P H for α Φ z P for P- Φ ω H

Φ rt o> Q Φ 3 for P- for P- ta Φ P- H (0 • _* P- rt Hi X tf ti Φ for HH P- Φ Φ φ cn P ⁾ H φ rt P Φ Φ P p ⁾ Φ

P- P- PJ Φ er φ H cn φ ⁾ c H P- H. ι-f PP gn cn

<CΛ Φ Φ 3 n Φ H Φ φ 3 P M cn P> Φ P Φ er? P> P- cn 1

P- Φ cn H 3 tr cn <! H Λ P- for P- n er XX rt cn H P- Φ P rt Ό M

Λ φ Φ cn ti o o P rt φ P tr Φ Φ er 1 3 n H p-

Φ C cn t - _" C tr H ζ φ cn Φ HP ö cn z Φ) o C- p ι- (Φ p H cn P ⁾ P XI Φ P rt D. P- cn φ Hi fr O tr tr ti d CL XX P- O tr N Λ PJ Φ HO Φ O: P- tr φ

P- N Hl α Φ φ tr O: £ P C: P> P α P H o P Φ Φ P-

1 ⁾ P- H cn er H ιQ Φ S> Φ tr er XX • fr fr υ-l er H cn rt α Φ O Φ P- φ ti rt PP ⁾ Φ c cn Φ P- P- φ rt

Φ cn PJ XX 3 Hi ti rt 3 NPH ti ^ ω HP φ H φ P " _^ cn XX er φ rt o cn O: Φ fr XX 3 Φ Φ H pt) rt ti t P z 3 Ό XI PP ⁾ 5" P- fr Φ lh

P Φ O: ^P>:) Φ ΪT BH P fr Φ H cn ^ α φ osp:

P D-

| P if P ⁾ Φ P- P- φ Φ P- - rt H 1 pt ⁾ O n α HP

P φ Φ P- lf rt Φ P- rt Φ t ⁾ tr tr rt φ PH cn Φ fr ti rt Φ cn rt HH

4 shows, by way of example, a trellis diagram, each with four different states at times t ... t + 3, which correspond, for example, to bit states '00', '10', '01' and '11'. A corresponding path is assigned to each symbol sequence in the trellis diagram. A path comprises a sequence of branches between two successive states. Each branch symbolizes a state transition between two successive states in time, the upper branch emanating from one state, for example, corresponding to a receive symbol with the binary '0' and the lower branch emanating from the same state corresponding to a receive symbol with the binary value '1'. Each of these state transitions to which a branch metric ('branch metric', BM) λ _{t is} assigned corresponds to a transmission symbol. The branch metric λ _t is defined as follows:

4 = | ^y ', -ι ² |

Here, r _t corresponds to the capture symbol at time t and y ' _{t to} the transmission symbol expected therefrom at time t.

In addition, a path metric γ _{t is} assigned to each path through the trellis diagram up to the time or time step t.

The trellis diagram shown in FIG. 4 is in particular a trellis diagram with a so-called 'butterfly' structure. This means that there are two

States of a time step t + 1 of the trellis diagram are assigned two states of the previous time step t, the branches of which lead to the first-mentioned states of the time step t + 1, two branch metrics of the branches originating from different states being identical. For example, the states shown in FIG. 4 form which the path metrics γ _t ^{ι: ι} , γ _t ¹³¹ , γ _{t +} ι ⁽²⁾ and γ _{t +} ι ^f3) are assigned such a 'butterfly' structure, the branch metric for the branch from the state with the path metric γ _t ⁽¹⁾ to the state with the path metric γ ₊ ι ⁾ the branch metric λ _t ^{l3) of} the branch from the state with the path metric γ _t ¹³¹ corresponds to the state with the path metric γ _{t +} ι ⁽³⁾ , while on the other hand the branch metric of the branch from the state with the path metric γ _t ⁽¹⁾ to the state with the path metric γ _{t +} ι ¹³¹ corresponds to the branch metric λ _t ^{(1) of} the branch from the state with the path metric γ _t ⁽³⁾ to the state with the path metric γ _{t +} ι ⁽²⁾ . In general, γ _t ^{lsl denotes} the ^{path metric} assigned to state s in time step t, while λ _t ^(s) denotes the branch metric of the state transition corresponding to signal s at time t.

The Viterbi decoder must now use the trellis diagram to determine the path that has the best path metric. As a rule, this is by definition the path with the smallest path metric.

Each path metric of a path leading to a specific state is composed of the path metric of a temporally preceding state and the branch metric of the branch leading from this previous state to the specific state. As a result, not all possible paths and path metrics of the trellis diagram have to be determined and evaluated. Instead, the path that has the best path metric up to this point in time is determined for each state and for each time step of the trellis diagram. Only this path, known as the 'Survivor Path', and its path metric must be saved. All other paths leading to this state can be ignored. Accordingly, during each time step there is a number of such survivor paths corresponding to the number of different states.

The above description makes it clear that the calculation of the path metric γ _{t +} ι ^ls) from the path metrics of the state s path metrics of the previous time step t connected via a branch. The calculation of the path metrics can therefore be implemented by a recursive algorithm, which is perceived in a Viterbi decoder by a so-called 'add-compare-select' unit (ACS unit).

5 shows the typical structure of a Viterbi decoder. In addition to the ACS unit, a 'Branch Metric Unit' (BMU) and a 'Survivor Memory Unit' are planned. The task of the 'Branch Metric Unit' is to calculate the

Branch metrics λ _t ^(s) , which are a measure of the difference between a receive symbol and that symbol which causes the corresponding state transition in the trellis diagram. The branch metrics calculated by the branch metric unit become the ACS unit

Determination of the optimal paths ('Survivor Paths'), whereby the 'Survivor Memory Unit' stores these 'Survivor Paths', so that finally a decoding can be carried out on the basis of the 'Survivor Path' which has the best path metric. The one associated with this path

Symbol sequence most likely corresponds to the sequence actually broadcast.

A processor element 1 of a conventional ACS unit can be constructed, for example, as shown in FIG. 6. It is assumed that each state of the trellis diagram is evaluated by a separate processor element 1. The task of processor element 1 is to use two competing paths that lead to a state of the trellis diagram to find the path with the best, i.e. select the smallest path metric. The saved values for the 'Survivor Path' leading to this state and its path metric are then updated.

As can be seen from the trellis diagram shown in FIG. 4, each state s at time t + 1 has an upper branch and a lower branch with a corresponding one previous state connected. To determine the survivor path corresponding to this state s, the path metric of the path leading to state s via the upper branch must therefore be compared with the path metric of the path leading to state s via the lower branch, ie the task of the path shown in FIG. Processor element 1 shown in FIG. 6 consists in determining the survivor path with the path metric γ _{t +} ι ^(s) either via the previous “upper” state with the path metric γ _t ^(u) and the upper “branch with the branch metric λ _t ^(u) leading path whose path metric corresponds to the sum γ0 ^ul + λ _t ^ty) , or the path leading via the 'lower' state with the path metric γ _t ^(1! and the 'lower' branch with the branch metric λ _t ^α) whose path metric corresponds to the sum γ _t ⁽¹⁾ + λ _t ⁽¹⁾ .

The previously described function of the processor element can consequently be performed, for example, by the circuit shown in FIG. 6, in which the possible path metrics are calculated with the aid of adders 14 and 15 and compared by means of a comparator 16, so that δ _s depends on the comparison result With the aid of a multiplexer 17, the smaller of the sums calculated by the two adders 14 or 15 is then output as the path metric γ _{t +} ι ^(s) .

FIG. 7 shows the overall structure and the connection of an ACS unit to the 'Branch Metric Unit' and the 'Survivor Memory Unit' for the trellis diagram shown by way of example in FIG. 4. Since each result γ _t ^(Ξ) calculated for a state s in the time step t is also the basis for the calculation of a path metric for a temporally subsequent state, the feedback of the processor elements 1 shown in FIG. 7 via an intermediate register 18 is necessary. The decision or signal values δ ₀ ... δ supplied by the individual processor elements 1 of the 'survivor path unit' enable this Selection and storage of the correct survivor paths and their path metrics.

The ACS unit described above is the most computationally intensive part of a Viterbi decoder. It has the largest space and energy requirements. In applications in mobile radio technology, the complexity of the ACS unit increases exponentially with the complexity of the code used for channel coding of mobile radio signals.

There is therefore a fundamental need to keep the circuitry for the ACS unit as low as possible. For example, in the publication "Low Power ACS Unit Design For The Viterbi Decoder", Chi-ying Tsui, Roger S.K. Cheng, Curtis Ling, Conference IEEE ISCAS, 1999, Orlando, in which an ACS unit is described according to the preamble of claim 1, proposed to determine a survivor path of a state, the difference in the path metrics of those previous states, each of which form a branch to the state under consideration, and compare it with the difference of the corresponding branch metrics. Depending on the comparison result, the 'Survivor Path' and thus the sum to be calculated for calculating the corresponding path metric can be inferred directly. This procedure has the advantage that the differences can be carried out jointly for the calculation of the path metrics of two states of a time step of the trellis diagram, assuming the already mentioned 'butterfly' structure, as a result of which the circuit complexity can be reduced.

However, even with this procedure, the separate formation of a comparison for determining the path metrics of the "upper" and "lower" states of a time step belonging to the respective "butterfly" structure is still required, for which purpose two comparators are required accordingly. M

H

3

P- rt rt

P "

P

P ιQ fr

Φ

H 3

Hi

P ⁾

Φ rt

H

P-

W

Φ

P α

Φ cn

- o he

Φ

H

Φ

P

-

P

P α

-

P

P rt

Φ

H

Φ

P

-

co Cd tV>> P> cπ o cπ o cπ O cπ

p> cn

Hi

P: t

H

P

X * cn er

Φ

P- cn

Ό

P- φ cn φ

H

PJ:

P rt

Φ l-f rt

•

1 shows a block diagram of a processor element of an ACS unit according to the present invention,

2 shows an illustration to explain the 'butterfly' structure of a trellis diagram,

3 shows an illustration to explain the principle on which the present invention is based when calculating path metrics,

4 shows an exemplary trellis diagram with four states,

5 generally shows the structure of a Viterbi decoder,

6 shows a block diagram of a processor element of a known ACS unit, and

FIG. 7 shows the construction of a known ACS unit with several of the processor elements shown in FIG. 6 and the connection of this ACS unit in a Viterbi decoder.

In order to explain the invention, the "butterfly" structure already explained, which is a prerequisite for the present invention, will be briefly discussed again below.

As can also be seen from the exemplary trellis diagram shown in FIG. 4, the trellis diagram has a certain regularity in its structure. This structure consists of so-called 'butterflies', such a 'butterfly' being shown in FIG. 2. According to the 'butterfly' structure, two states 19 and 20 of a time step t are connected to two states 21 and 22 of the subsequent time step t + 1 such that each of these states 19 and 20 has a branch to state 21 and a branch to the State 22 leads, the branch metric λ _t ^(u) of the branch leading from state 19 to state 21 of the branch metric of the branch leading from state 20 to state 22 and the branch metric λ _t ^{(1) of} the branch leading from state 19 to the branch 22 leading to state 22 corresponds to the branch metric of the branch leading from state 20 to state 21. 2 also shows the path metrics γt ^lu) , γ ^αι , γ _{t +} ι ^(u) and γ _{t + 1} ⁽¹⁾ corresponding to the individual states.

This regularity of the "butterfly" structure is used in the context of the present invention in order to reduce the number of arithmetic operations required to determine the "upper" path metric γ _{t +} ι ^<ul and the "lower" path metric γt + i ¹¹¹ .

To determine the path metrics γ _t + ι ^<u) and γ _{t +} ι ⁽¹⁾ , the cheapest path metrics of the theoretically possible path metrics must be determined. Each path metric of a state is made up of the path metric of a state of the previous time step and the branch metric of

Connecting branch of these two states together. Thus, to determine the path metric γ _{t + ι} ^(ul, the smaller of the theoretically possible path metrics γ _t ^(ul + λ _t ^(ul and γ _t ^αι + λ _t ⁽¹⁾ and to determine the path metric γt + ι ^αι the smaller of the theoretical possible path metrics γ _t ^(u) + λ _t ⁽¹⁾ and γ _t ⁽¹⁾ + λ _t ^{(u) are} determined, so the following comparison operations are required:

γ _t ^(ul + λ _t ^{, u)} ><γ _t ⁽¹⁾ + λ _t ^{! 1)} for the 'upper' state γ _t ^(u) + λ _t ⁽¹⁾ ><γ _t ⁽¹⁾ + λ _t ^(u) for the 'lower' state

Each comparison operation is represented by the operator '> <'.

The comparison operations to be carried out for the 'upper' state 21 and the 'lower' state 22 of the 'butterfly' structure are somewhat similar. The branch metrics and path metrics can be separated from one another by subtraction, so that a comparison between a path metric difference and a branch metric difference results for the 'upper' and 'lower' state:

λt ' ^u ' - λt ⁽¹⁾ X γ _t ^U »- - γ _t ^(u) for the 'upper' state λt ' ¹ ' - λt ' ^ul r><γ _t < ^l . - - γ _t ^(ul for the 'lower' state

If a path metric difference T _t = γ _t ^α) - γ _t ^(u) and a branch metric difference Λ _t = λ _t ⁽¹⁾ - λ _t ^{(u) are} defined for these differences, the following results for the comparison operations to be carried out:

Λ _t > r _t for the 'upper' state -Λ _t ><I \ for the 'lower' state

The two comparisons can be made by the absolute or absolute comparison IΛ _J IXIΓ ,! be solved. The result of this

Comparative establishes the relationship between | Λ, | and | r, | firmly. Since the metrics λ _t ' ^u) , λt' ¹ ', γ _t ^{(11 and} γ _t ' ^u) are positive by definition, none of the above inequalities have been changed by the previously made mathematical modifications. The inequalities originally established for the 'upper' state 21 and the 'lower' state 22 for determining the path metrics γ _t V ^u 'and γt ₊ ι ^(1!) Can thus be solved by evaluating the signs of Λ _t and I \, whereby it must be taken into account that | Λ, | for the inequality of the

'lower' state 22 must have a negative sign.

The result of these considerations is that to determine the path metrics γ _{t +} ι ^(u) and γ _{t +} ι ⁽¹⁾ only a single absolute comparison has to be carried out, the final decision then depending on this comparison result by evaluating the signs of Λ _t and T _{t is} hit. As will be explained in more detail below, this is possible with relatively simple circuitry means.

1 shows a block diagram of an exemplary embodiment for a processor element 1 of the ACS unit of a Viterbi decoder. The processor element 1 is used to calculate the path metrics γ _{t +} ι ^(u> and γt + ι ^U) of two states of a time step t + 1 of a trellis diagram and receives the path metrics γ _t ^{(1) and} γ _t ^{(ul of} those two states as input signals of the preceding time step t of the trellis diagram, which, together with the two states to be calculated, form the 'butterfly' structure shown in Fig. 2. Furthermore, the processor element 1 receives the branch metrics λ _t ^(ul and λ _t ⁽¹⁾⁾ as input signals The path metrics γ _t ⁽¹ '/ γ _t ^(ul and the branch metrics λ _t ' ^u) , λ _t ^(1! are each supplied to multiplexers 8 and 9 or 10 and 11, which are dependent supply one of its two input values from a control signal C ^u or C ^{1 to} an adder 12 or 13, which finally outputs as path metric γ _t + ι ^(u 'or γt + i' ¹ 'the smaller of the two possible sums Decision or tax Signals C ^u and C ¹ are also fed (inverted) to the 'Survivor Memory Unit'.

The control signals C ^u and C ¹ are from one

Evaluation unit 6 generated. A subtractor 2 and a comparator 4 are provided to control the evaluation unit 6, the subtractor 2 calculating the difference between the two path metrics γ _t ⁽¹⁾ and γ _t ' ^u) , while the comparator 4 calculates the resulting path metric difference Y with an external one supplied branch metric difference Λ compares and supplies the comparison result to the evaluation unit 6.

The branch metric difference Λ = λ _t ^(u '- λ _t ^ul is calculated in the' Branch Metric Unit '(BMU) and the ACS unit or processor element 1 in two different versions, namely once as a normal branch metric difference Λ + and once as a negated pointing metric difference Λ-. The selection of which of the two branch metric versions is to be used for the comparison with the path metric difference I ^" is made by a multiplexer 3, which is controlled depending on the sign bit sg (T) of the path metric difference. In a two's complement representation, the sign bit corresponds to the MSB ( Most Significant Bit) The distinction between the two branch metric versions, as will be explained in more detail below, is necessary in order to be able to compare values with the same sign and thus to be able to calculate a correct comparison result.

As has already been mentioned, the comparator 4 determines the result of the absolute comparison C = r _f | > | Λ, | , Since both 10 and Λ _{t can be} positive or negative, there are four possible combinations of values that could make an inversion of the determined comparison value C necessary in order to correctly model and overwrite the previously described inequalities for the 'upper' and 'lower' state the multiplexers 8, 9 and 10, 11 to be able to select the correct values for the path metric calculation.

The corresponding logic is shown in FIG. 3, the possible combinations of the signs of T and Λ as well as the respectively resulting values for C ^u and C ¹ depending on the one for the “upper” state and the “lower” state Absolute comparison result C are shown. A trailing minus sign corresponds to a negative sign, while a trailing plus sign corresponds to a positive sign, so that, for example, r- represents a negative path difference value and T + a positive path difference value.

From Fig. 3 it can be seen in particular that both for the 'upper' state and for the 'lower' state for two possible sign combinations by evaluating the Sign of T and Λ a determination of the values for C ^u and C ^{1 is} possible without actually carrying out the absolute comparison and thus having to calculate the value C. In these cases, the 'comparison result', ie the control signals C ^u and C ¹ , is available much earlier, since no comparison has to be carried out, which significantly reduces the glitch probability at the selection multiplexers 8, 9 and 10, 11 Consequence.

In all other cases of the signs of T and Λ, the values for C ^u and C ^{1 are} calculated analogously to the logic shown in FIG. 3 as a function of the value C of the absolute comparison result.

The determination of the values for C ^u and C ¹ depending on the signs sg (T) and sg (Λ) and the value of the absolute comparison C = | r, | > | Λ, | 1 takes place in the

Evaluation unit 6. As has been described above with reference to FIG. 3, the evaluation unit only has to evaluate the signs sg (D and sg (Λ) and, depending on the values C ^u and C ^1, either the fixed binary value '1', the fixed one Assign binary value '0', the unchanged value C of the comparison result or the negated value C of the comparison result. The multiplexer 3 ensures that in order to implement the

Absolute comparison | r, | > | Λ _r | always compare the correct values ± T _t > ± Λ _t .

The use of the logic shown in FIG. 3 enables only an absolute comparison C = | r, | > | Λ _r | must be evaluated in order to be able to calculate both the 'upper' path metric γ _{t +} ι ' ^u) and the' lower 'path metric γ _{t +} ι ^αι . If it is assumed that the complexity of a comparator essentially corresponds to that of an adder or subtractor, it can easily be seen by comparing the processor elements 1 shown in FIGS. 1 and 6 that with the aid of the present invention approx. 33% u> C tt P ¹ P ¹

Cπ o Cπ o cπ o cn

«N χι P> πd • n ιQ <<! P> X P. ö tsi P <Cd <πd P> P hd P Od ω P. _- d Φ ^ 1 fl α o P Φ for Hi P- Φ φ OPP ¹ O: P- PJ Z ⁾ o P- Φ Hi PP Hl φ O Φ φ i Od lh ι-f P- P- P ⁾ φ

3 ιQ 3 PP ⁾ χι N li PO cn P Φ H φ cn lf rt H P> cn H P> H lf P- H cn PP ⁾ Hi XX iQ PH

V φ pi: P- • Φ £ tr P cn P-? R er iQ XX o rt φ rt P-. • P> O tP φ ^ P- Φ Φ Φ ^ rt lQ tr OHP ^J & Φ tr ^ P Φ P. O P- rt & P P-

HHH Φ ^ j XX P Φ Φ P- P Pi: Φ p O: 3 Φ Φ φ cn Hi Φ φ P- Φ tr P Φ Φ P σ P ¹ - _^ P

P ⁾ P. t * l Φ rt rt PHHP • φ O Ό P- P- rt rt P: rt H P- 3 PH rt P - <rt P P- PH p: Φ φ Hi P ^J rt tr P ⁾ rt o H ⁾ tr H P> tr P ^ Hi HP • iQ P- ^ o φ iQ P- er P rt s P- Z σ Φ PH cn H φ tr P- H P- P ⁾ P- P P- P 'P - XX φ Φ P-

H rt cn X Φ Φ P Φ ⁾ P P- rt P> P rt X rt rt X rt lh P ιQ ιQ • <? cn ö XX ιQ

Φ P- ZH tυ PN P. P- cn "l ιQ XZ rt ^• » fr Φ • fr rt • - Φ XI P> φ • rυ PP * n φ H PJ P rt + fr Φ O Hi P- rt P- P> Φ ^ P φ lf P tr H ^• o H tr P Φ <<r> J P- HH φ p Pn Hi P P- Φ P »P. 3 p: p: P CTl cn Φ P- rt Φ NH IQ H φ> P Hi rt cn 3 Hi cn Hi cn S ^| Q lf Φ ^ pi: er er p.

O P- P fr Φ ^ Q φ Φ cn Φ H 1 Hi Hi p- Cπ rt Φ Φ H "Φ XX Φ • XX cn + tP φ φ P- iQ

Z P P- P- cn α ιQ cn iQ Φ XX N cn l-i P- ιQ H Φ P- TJ Φ Φ l-f H φ Φ

P- ιQ P. φ cn cn P ^J Φ Φ P- p p. H Φ P. PO Φ P Φ Φ cn rt Hi N rsi c PHN φ P ⁾ Φ p: rt o P- H 3 M Φ p P- Φ Φ cn 3 P 3 P rt Φ P ⁾ φ φ tr cn 1 Φ

PP dn Φ H cn pi: P- P- p. φ P ω H rt P- N Φ NP ⁾ H iQ P- P- d P- φ - * P- fr XX Hi XH Φ l P tP P n cn N P- P ⁾ Φ rt P P- P "er φ & XX rt PHPO

H P * P P> P rt P- N Φ rt cn P XX

Φ cn PP »XX σ P ιQ tr • -I Φ rt P- O ⁾ tr cn rt

P cn Φ p. φ P ¹ Φ P- rt ZP ⁾ cn 1 rt 3 P tP Φ Φ rt Φ ONP cn P Φ

Φ P 3 Λ- 00 3 P ⁾ Φ Φ P Φ i S Φ. -. Φ O:

H α f 3 rt for P- H tr - <φ cn σ P

£ P- Φ O Φ lQ H P- ιQ h SP ⁾ πd rt Ό NPH cn cn Cn XX> Φ P ιQ Φ ⁾ P- P Xi P- π cn • rt

P rt ₊ P- ιQ CΛ P • PHP rt P- P + rt ^ H P- cn P- PH ϋ

P ^J P- Hi H Ό p rt PJ O rt cn Φ cn ω cn P- Φ P cn 03 Φ 0 rt ~ O cn P PJ r + χ>P> P- Φ 3 Φ 3 tr P- er φ Φ - - - xι cn n H α P. rt P Φ N rt i ^ H rt P- tr O

P- Φ - 3 P- P. XX P P- φ O tr P- fr Φ - * φ P- Φ Φ Φ φ P- N

Ό P & ≤ rt INI P ¹ φ Φ P- ti P- er n P cn Φ rsi φ φ H P- πd P cn cn Hi P <N h φ Φ

Φ Φ s: cn P. φ P> cn Φ tr Φ rt ΪV H s P P h Φ cn p: Φ P 1 H cn

Φ cn rt H cn P- P ⁾ H rt H P>P> z φ tr INI ⁾ O cn O rt H 3 1 cn Cπ

XOH rt P- cn P 3 P- s: Φ - - P φ P- O: <! ≤ tP <N h + φ er OO φ tr P- Φ P n Hi φ nn N P- P- πd rt P P- tr O Φ o φ N Φ P "r ω α P Φ POH ι- (P ⁾ ^ P tr * «• tr t PH p- rt for Hi Φ cn Φ H P- Φ H cn P P- rt pi φ P. rt

P P- Φ P ⁾ • Φ H Φ 1 iQ α XX cn ⁾ P> rt σ φ Φ H iQ HX cn cn Φ er Φ Φ O lf φ O rt P t-> P P. Φ P w Φ 1 Φ rt s: o B o Ott 3 Φ cn tr H Ό o P ≤ cn φ H P- Hi P P- σ cn P- PH Φ 3 Φ Φ 3 h ⁾ φ H er rt O

P ^ i 3 Φ TJ H Φ rsi P p: Φ) 3 φ Φ Φ cn P P- H Ό rt P- Ό Φ P P Φ Φ P "Φ O«

P XX _ pi: vQ Φ Φ P tr tr H P- rt P- Ό rt X rt P ⁾ HPP ⁾ rt O P- P 3 O o

P cn c tr Φ P- P P. cn Φ H <H P P- Φ for P- H P- φ H Φ Φ tr P. XX Ό 3 IQ

P. Ό PO Φ rt P- rt Hl OP P- ^• • φ H P- ιQ P ⁾ f P Pi 3 cn P ¹ ti Φ P ¹⁾ Ό P-

Hl p P- tr * H pi: rt • P Φ H φ Λ "ω Hi Φ rt P. rt φ Φ Φ P Φ P. HP ⁾ φ

P> ⁾ p OP Φ PP cn cn ⁾ H fr α cn O Hi H o P- OP <(X rt P- PJ PJ li

P »PP tr P- H cn P- fr P- s: et P P- p ⁾ s: tr Φ Φ H Hi OH rt o P- INI n P rt P ⁾ φ φ Φ P XX P Φ P Φ XX cn tsi Hi 03 Φ P- HP Hi ι- (H XX ZP tr O rt P-

Hi P -i pi cn pi: H Φ s: Hi P- φ Φ Cπ Φ <cn P ¹ ιQ Φ cn N 3 HOP p: P Φ P. P ^ α s rt n er φ Φ φ cn fr P ω H Φ Φ P- H rt Φ P- Φ H XI tr Hi H Φ ιQ P- Φ P rt Φ P- h H φ N P- α φ h <P- cn O pi: P- rt P- Φ Φ

HZ "_" H Φ K3 cn tr P Φ P- φ Φ P- rt Φ P XX O cn Φ tr Φ P rt P P- cn

Φ Φ ϊö • pi: ιQ cn HP P- ι_ι. H ¹ P. O 3 cn HNH rt tr PP P- X tr P TJ

P cn p: Φ P ⁾ P cn Ό Λ- N φ φ tr P- P> φ iQ Φ N Φ XI P- φ tr

- * io OOHPP> φ ιQ Φ rt P- OPH Φ rt PH> P- Φ P> PP P- φ H tr 3 P> z P- P- Φ MP P- Φ Φ rt n cn H s: φ fr Hi rt P- rt

Φ P. X P- φ XX P Z Ό P XX P- H tr Φ cn P- pi: P- P- Φ Φ rt

PJ o rt O P- X tr Φ cn P ¹ P. P tr cn tr p φ p φ

N P Ό rt XX Φ φ z

HS φ P- P ⁾ p. φ rt H Φ

Ό Φ cn P- P P Φ

P P- (i φ 3 α O P Φ P • Φ cn Φ P- f * - N rt rt φ P- φ Φ tr fr H p cn p p- fr P rt P P p cn P l-f Φ P p φ

Φ P Φ CTi Φ rt H

3 P iQ P 1

A renormalization unit (not shown) can preferably be interposed, which avoids an overflow and renormalizes the feedback path metric values.

For each state of a time step of the trellis diagram, the cheapest, ie minimum, path metric determined in the manner described above is temporarily stored together with the corresponding path in the 'survivor memory unit'. The main difference between the use of the processor elements 1 shown in FIG. 1 in an ACS unit and the representation of FIG. 7 is only that the processor elements 1 according to the invention shown in FIG. 1 for calculating two path metrics γ _t + ι ^ιul and γt + i ¹¹ 'are provided.

Claims

claims

1. ACS unit for a Viterbi decoder, the ACS unit being provided for determining path metrics of a trellis diagram assigned to the Viterbi decoder, a first and a second state (21, 22) of a time step (t + 1) the trellis diagram is assigned a third and fourth state (20, 19) of the preceding time step (t) connected to it via a branch of the trellis diagram, with a first branch metric (λ _t ^(u) in the trellis diagram ) for the branch connecting the first state (21) with the fourth state (19) and the branch connecting the second state (22) with the third state (20) is the same size and a second branch metric (λ _t ⁽¹⁾ ) for the branch connecting the first state (21) with the third state (20) or the second state (22) with the fourth state (19) is the same size, and wherein the ACS unit has a processor element (1) for determining the first and second state (21, 22) zugeo rdneten

Path metrics (γt + ι ^(u) / Yt + ι ⁽¹⁾ ), which as input values the first and second branch metrics (λ _t ^(u) , λ _t ⁽¹⁾ ) and the third and fourth states (20, 19 ) assigned path metrics (γt ^(u '/ γt ⁽¹⁾ ) receives and comprises: - a differentiating device (2) for calculating a

Path metric difference (D between the path metrics assigned to the third and fourth states (19, 20) (γ _t ^(u) , γ _t ⁽¹⁾ ), and - an evaluation device (3-6) for comparing the path metric difference (D with a branch metric difference ( Λ) between the first and second branch metrics (λ _t ^<ul , λ _t ⁽¹⁾ ) and for determining the path metrics associated with the first and second states (21, 22) (γ _{t +} ι ^(ul , γ _t + ι ⁽¹⁾ ) as a function of the comparison result (C), characterized in that the evaluation device (3-6) is designed in such a way that it includes the path metrics (γ _t + ι ^(ul , γ _{t +} ι) assigned to the first and second states (21, 22) ⁽¹⁾ ) depending on the comparison result (C) with additional evaluation of the signs of the path metric difference (D and the branch metric difference (Λ).

2. ACS unit according to claim 1, characterized in that the evaluation device (3-6) is designed such that it a first control signal (C ^u ) depending on the comparison result and the sign of the path metric difference (D and the branch metric difference (Λ) for a first selection device (8, 9, 12) assigned to the first state (21) and a second control signal (C ^x ) for a second selection device (10, 11, 13) assigned to the second state (22), and that the first The selection device (8, 9, 12) is designed such that depending on the first control signal (C ^u ) as path metric (γ _{t +} ι ' ^u) ) of the first state (21) either the sum of the path metric (γ _t ^{( ul} ) of the fourth state (19) and the first branch metric (λ _t ^(u) ) or the sum of the path metric (γ _t ⁽¹⁾ ) of the third state (20) and the second branch metric (λ _t ⁽¹⁾ ) , and that the second selection device (10,11,13) is designed such that si e depending on the second control signal (C ¹ ) as path metric (γ _{t +} ι ^α) ) of the second state (22) either the sum of the path metric (γ _t ^{| u)} ) of the fourth state (19) and the second branch metric ( λ _t ⁽¹⁾ ) or the sum of the path metric (γ _t ⁽¹⁾ ) of the third state (20) and the first branch metric (λ _t ^{(u |} ) issues.

3. ACS unit according to claim 2, characterized in that the first and second selection means comprises a first and second control signal (C ^ C ¹ ) driven first and second multiplexer (8,9; 10,11), the first multiplexer as input signals the first and second branch metrics (λ _t ' ^u) , λ _t ⁽¹⁾ ) and the second multiplexer as input signals that correspond to the third and fourth states (20, 19) assigned path metrics (γ _t ' ^u) , γ _t ⁽¹⁾ ), and wherein the output signals of the first and second multiplexers are fed to an adder (12; 13).

4. ACS unit according to claim 2 or 3, characterized in that the evaluation device (3-6) is designed such that it assigns the first control signal (C ^u ) a fixed binary value without the comparison between the path metric difference (D and the To form a branch metric difference (Λ) if the path metric difference (D and the branch metric difference (Λ) have different signs, and that the evaluation device (3-6) is designed in such a way that it assigns a fixed binary value to the second control signal (C ¹ ), without making the comparison between the path metric difference (D and the branch metric difference (Λ) if the path metric difference (T) and the branch metric difference (Λ) have identical signs.

5. ACS unit according to claim 4, d a d u r c h g e k e n n z e i c h n e t that the evaluation device (3-6) the first control signal

(C ^u ) assigns a first binary value, by means of which the first selection device (8, 9, 12) uses the path metric (γ _{t +} ι ' ^{u |} ) of the first state (21) to sum the path metric (γ _t ' ^u) ) of the fourth state (19) and the first branch metric (λ ' ^ul ) is output if the path metric difference (D has a positive sign and the branch metric difference (Λ) has a negative sign, and that the evaluation device (3-6) the first control signal (C ^u ) assigns a second binary value corresponding to the inverted first binary value, by means of the first selection device (8, 9, 12) as the path metric (γ _{t +} ι ^(υ) ) of the first state (21) the sum of the path metric (γ _t ^αι ) of the third state (20) and the second branch metric (λ _t ^αι ) is output if the path metric difference (D a has a negative sign and the branch metric difference (aufweist) has a positive sign.

6. ACS unit according to claim 4 or 5, characterized in that the evaluation device (3-6) assigns the second control signal (C ¹ ) a first binary value by means of which the second selection device (10, 11, 13) as path metric ( γ _{t +} ι ^αι ) of the second state (22) the sum of the path metric (γ _t ^(u) ) of the fourth state (19) and the second branch metric (λt ' ¹ ') is output if the path metric difference (D and the branch metric difference (Λ) each have a positive sign, and that the evaluation device (3-6) assigns the second control signal (C ^u ) a second binary value corresponding to the inverted first binary value, by means of which the second selection device (10, 11, 13) the path metric (Yt + ι ⁽¹¹ ) of the second state (22) is the sum of the path metric (γ _t ⁽¹⁾ ) of the third state (20) and the first branch metric (λ _t ' ^u '), if the

Path metric difference (D and the branch metric difference (Λ) each have a negative sign.

7. ACS unit according to one of claims 2-6, characterized in that the evaluation device (3-6) is designed such that it assigns the binary value (C) of the comparison result to the first control signal (C ^u ) if the path metric difference ( D and the branch metric difference (Λ) each have a positive sign, while the

Evaluation device (3-6) assigns the inverted binary value of the comparison result to the first control signal (C ^u ) if the path metric difference (D and the branch metric difference (Λ) each have a negative sign.

8. ACS unit according to one of claims 2-7, characterized in that the evaluation device (3-6) is designed such that it assigns the binary value (C) of the comparison result to the second control signal (C ¹ ) if the path metric difference (D is a positive sign and the branch metric difference (Λ) is a negative sign has, while the evaluation device (3-6) assigns the inverted binary value of the comparison result to the second control signal (C ¹ ) if the path metric difference (T) has a negative sign and the branch metric difference (ein) has a positive sign.

9. ACS unit according to one of claims 2-8, d a d u r c h g e k e n n z e i c h n e t that the evaluation device (3-6) comprises:

- A multiplexer (3) which receives the branch metric difference (Λ +) and the negated value of the branch metric difference (Λ-) as input signals and from which the sign of the path metric difference (D formed by the differentiating device (2) (Dg (D) is controlled,

- A comparator (4) which receives the path metric difference (D and the output signal of the multiplexer (3), and

- An evaluation logic (6) which, as input signals, the output signal of the comparator (4), the sign of the

Receiving branch metric difference bit (sg (Λ)) and the bit indicating the sign of the path metric difference (sg (T)) and depending on it outputs the first and second control signals (C ^U , C ^X ).

10. ACS unit according to claim 9, characterized in that the evaluation device (3-6) comprises a pre-comparator (5) which performs a pre-comparison of certain bits of the path metric difference (D and the branch metric difference (Λ) and a corresponding comparison result of the Evaluation logic (6) supplies as a further input signal, and that the evaluation logic (6) takes into account the comparison result supplied by the pre-comparator (5) when generating the first and second control signals (C ^u , C ¹ ).

11. ACS unit according to claim 10, so that the pre-comparator (5) performs a pre-comparison of the more significant bits of the path metric difference (D and the branch metric difference (Λ).