Signal receiver and mobile communication device
The invention relates to a method of transferring information, the method comprising generating a signal, transferring the signal and receiving the signal, the signal comprising at least one preamble followed by a synchronization word and a data frame, the preamble having a predefined pattern, the preamble being provided for bit synchronization and the synchronization word being provided for frame synchronization. The invention further relates to a device for receiving the signal. The invention further relates to a device for generating the signal. The invention relates to the field of providing frame synchronization for recovering the information from the data frame. Detection words are used for correlating with a received signal for detecting the synchronization word.
The document US 6,480,559 describes a communication system in which a mobile receiver performs synchronization word (or unique word) detection by correlating a detection word with a received signal. Frame synchronization is effected by using real correlation coefficients which are not equal to the binary synchronization word. The correlation coefficients in the detection word are also not equal to any shift or scaling of the synchronization word, but are dependent on the synchronization word. In some embodiments, the correlation coefficients are dependent both on the synchronization word and also on the bit_sync pattern used in a pre-amble before the synchronization word. In particular in US Patent 6,480,559, real- valued detection words (called X) are considered and shown in Fig.l (and Fig.12) thereof, which have the same length as a synchronization word (called word U). It is further considered that, given a synchronization word U, X may be a detection word of real- valued filter coefficients, which is used in a filter structure to recognize synchronization word U. The optimal values for X are considered for optimization of the receiver. However, due to noise, the synchronization word may not be reliably detected.
It is an object of the invention to provide a system for transferring information via a signal having a synchronization word preceded by a preamble having an improved detection of the synchronization word. For this purpose, according to a first aspect of the invention the method as described in the opening paragraph comprises the steps of detecting the synchronization word by correlating the signal with a detection word, the detection word having a length different from the length of the synchronization word and being dependent on the synchronization word and a final part of the preamble that precedes the synchronization word. For this purpose, according to a second aspect of the invention the device for receiving a signal as described in the opening paragraph comprises a synchronization detector for detecting the synchronization word, the synchronization detector comprising a correlator for correlating a detection word with the signal, the detection word having a length different from the length of the synchronization word and being dependent on the synchronization word and a final part of the preamble that precedes the synchronization word. For this purpose, according to a third aspect of the invention the device for generating a signal as described in the opening paragraph comprises a synchronization generator for generating the synchronization word, the synchronization word being dependent on a detection word to be used for correlating with the signal in a receiver, and the detection word having a length different from the length of the synchronization word and being dependent on the synchronization word and a final part of the preamble that precedes the synchronization word. For this purpose, according to a fourth aspect of the invention the signal for transferring information as described in the opening paragraph comprises at least one preamble followed by a synchronization word and a data frame, the preamble having a predefined pattern, the preamble being provided for bit synchronization and the synchronization word being provided for frame synchronization, the synchronization word being dependent on a detection word to be used for correlating with the signal in a receiver, and the detection word having a length different from the length of the synchronization
word and being dependent on the synchronization word and a final part of the preamble that precedes the synchronization word. The measures have the effect that a longer (or shorter) detection word is allowed under the assumption that the synchronization word is preceded by a pre-amble (also called bit_sync pattern). Herewith, with a given synchronization word (U), a detection word (X) may be considered such that Lx ≠ Lu, Lx being the length of X and Lu being the length of U. More limited, with a given synchronization word U, a binary detection word X may be considered such that Lx > Lu. Advantageously the reliability of detecting the synchronization word is increased by selecting a length of an optimized detection word taking into account the preamble that precedes the synchronization word. In an embodiment of the device the detection word comprises the synchronization word preceded by the last bits of the pre-amble. Hence, with a given synchronization word U a binary X may be considered such that Lx > Lu, with the further limitation that X is comprised of U preceded by the last bits of the pre-amble described earlier. This has the advantage that a practical value for the detection word can be easily determined. In accordance with a further aspect of the present invention, a simple sheer was proposed that precedes the actual receiver. In an embodiment of the device the correlator is a binary correlator, and the device comprises a sheer for converting the signal to a binary signal coupled to the binary correlator. This has the advantage that, surprisingly, in the presence of a limited amount of noise, the correlator proves to be more reliable than a real valued correlator. This aspect invention is also based on the following recognition. In the known solutions the signals including noise are processed in the correlator. Hence noise values are accumulated in the correlation result. By adding the sheer before the correlator noise values are eliminated, which for a high signal to noise ratio, surprisingly proves to be a more reliable evaluation. It is noted that more complicated slicers may be used, e.g., a three level sheer that is defined as follows: slicer(x) = 1 if x>eps, slicer(x) = -1 if x<-eps, and slicer(x)=0 if -eps<x<eps. In an embodiment of the device the synchronization word is selected for providing a substantial detection distance between a correlation peak and correlation side
lobes during said correlating with the detection word. Effectively the synchronization word itself is adapted to the presence of the pre-amble, to provide an improved correlation result, i.e. an optimal detection distance between a high peak and low correlation. Note that at the peak the high correlation is with the complete synchronization word and possibly the last bits of the preamble. Before that, providing the correlation side lobes, the correlation is either only with the preamble or with the end of the preamble and beginning of the synchronization word, but not with the entire synchronization word. Hence, when the corresponding detection word having a different length is correlated with the synchronization word itself (providing the correlation peak) or part of the pre-amble (providing the correlation side lobes). This has the advantage that a gain is achieved over the standard synchronization words selected to be optimal only when correlation is based on the synchronization word itself. Further preferred embodiments of devices according to the invention are given in the appended claims, disclosure of which is incorporated herein by reference.
These and other aspects of the invention will be apparent from and elucidated further with reference to the embodiments described by way of example in the following description and with reference to the accompanying drawings, in which Fig.l shows block diagram of a frame synchronization system, Fig.2 shows optimal pairs of synchronization word and detection word patterns with lowest correlation side lobes, Fig.3 shows a binary correlation receiver, Fig.4A shows a comparison of error probabilities for a system with and without sheer, Fig.4B shows a comparison of error probabilities in detail, Fig.5 shows a schematic diagram of an optical storage system, Fig.6 shows a frame format, Fig.7 shows a block diagram of a frame synchronization receiver, Fig.8 shows an example of the output of the correlator, and Fig.9 shows a state transition diagram of a sync detect block. Corresponding elements in different Figures have identical reference numerals.
Fig.l shows block diagram of a frame synchronization system. A correlation receiver 11 provides an output signal 15 indicating if a synchronization word is detected. A source signal comprising at least one data frame f is combined with noise n in an adder 12 to constitute a received signal r. The received signal is input signal for a correlator 13, which correlates the input signal with detection word w. A correlation output q is coupled to a threshold detector 14 to generate the output signal 15. Background knowledge about correlation receivers can be found in "R.H. Barker, "Group Synchronizing of Binary Digital Systems," in W. Jackson (Ed.): Communication Theory. (Butterworth, 1953), pp. 279-287". Note that frame synchronisation is analyzed for a packet transmission system where the sync pattern is preceded by a known preamble. We allow the correlation receiver to use an extended or shortened ersion of the sync pattern, which is called detection word or detection pattern. As shown later this leads to a new criterion for designing optimum pairs of sync and detection patterns. The obtained patterns result in a normalized improvement of up to 3.91 dB over the known optimum patterns. The system applies a common packet transmission scheme, where every frame contains a sync pattern prior to the user data. We assume an additive white Gaussian noise (AWGN) channel. The frame synchronisation system at the receiver consists of a correlator followed by a threshold detector as shown in Fig.l. The correlator outputs a correlation sequence and whenever an element of this sequence surpasses a given threshold, the detection of the sync pattern is flagged. In practice, a common scenario is the one where the sync pattern is preceded by a known preamble and the search for the sync pattern is enabled once the preamble has been detected. A sync pattern design criterion for this case is proposed in "Driessen, P. F.: "Binary frame synchronisation sequences for packet radio", Electron. Lett., 1987, 23, p.l 190". It consisted in selecting the sync patterns that minimised the maximum (absolute) value of the correlation sequence before the correlation peak. The correlation sequence is the sliding correlation of the sync pattern with the concatenated preamble/sync pattern. The obtained patterns performed significantly better than the known optimum autocorrelation sequences (sync patterns minimising the maximum (absolute) value of the autocorrelation sequence before the autocorrelation peak) due to the fact that the optimum autocorrelation sequences are only optimal if the sync pattern is not preceded by a preamble.
In the current invention it is shown that the performance can still be improved significantly by considering the joint optimisation of sync pattern and correlation receiver. We allow the correlator to use extended detection words, i.e. to extend the sync pattern to a detection word of a different length, e.g. by preceding the sync pattern with the last bits of the known preamble. We also allow shortened detection words i.e. to consider only the last part of the sync pattern. We denote the pattern used by the correlator as the detection word or detection pattern. The detection pattern is unambiguously defined by its length, the sync pattern and the preamble (in the case it is an extended sync pattern). Our sync pattern design criterion is essentially the same as the one proposed above but for each possible sync pattern we consider a range of possible lengths for the detection pattern. We select the pairs of sync patterns and detection pattern lengths that minimise the maximum value of the correlation sequence before the correlation peak. Other criteria of optimality are also possible like the above mentioned where the minimization of the maximum absolute value of the correlation sequence before the correlation peak is sought. The possibility of considering detection patterns of different length than the sync pattern has not been exploited before. For explaining the frame format bold letters denote a sequence x of length M and regular letters denote its elements x[n] for n = 0, ..., M— 1 (indexing starts at zero). Each frame f consists of three parts: preamble p, sync pattern s and data d (in this order). The frame can alternatively be specified as f = (p, s, d). Let N, Np, Ns and Nd, denote the frame, preamble, sync pattern and data lengths respectively, so that N = NP + NS + Nd- We consider f to be binary i.e. fe {-1, 1}N. In the correlation receiver the correlator generates qs,w, the correlation sequence of the received sequence r = f + n with a stored sequence w (the detection pattern) of length L. We stress the dependence of the correlation sequence on the choice of s and w by using them as sub indices. We define the w for I, > Ns as w = sL ≡ (p[Np - (L- Ns)],...,p[Np -1), s) for L > Ns. For L ≤NS we define w as w = sL ≡ ( [Ns -L],..., s[Ns -1]) for L ≤NS. This means that for L > Ns the sync pattern is extended by prefixing it with the last bits of the preamble to form w, whereas for L ≤NS the sync pattern is shortened by considering the beginning (s[0],..., s[Ns -L-l]) as the last part of the preamble. We write the correlation sequence qs,w as:
qs,w M = ∑ k] ■ r[i + k -(L-ϊ)] for i=0, ...,N-1.
Since r = f + n we can write q
s,
w = c
S)W + n
w where c
S)W is the correlation sequence of the frame f with w i.e.
∑w[k] . f[i + k - (L -ϊ)] for i=0,...,N-l,
and n
w is the correlation sequence of the noise sequence n with w. We consider the elements of n to be statistically independent Gaussian random variables with zero mean and standard deviation σ. Therefore n
w is additive Gaussian noise of standard deviation
at the output of the correlator. We define the detection distance as the normalised difference of the value of c
s,
w at the correlation peak and the maximum value at any prior instant (also called side lobes) i.e.: D - N /vL . (cs, w [io] max { cs, w [i] : i < io }) where i
0 = N
P + N
s — 1 is the position of the last bit of the sync pattern in the frame. The sync and detection patterns are designed so that c
SjW [i
0] > max { c
S;W [i] : / < i
0 }.
The normalisation factor
*
s introduced in order to be able to make a fair comparison between detection distances D corresponding to different lengths L. This factor compensates for the different noise variances at the correlator output corresponding to different detection pattern lengths L (see eq. (1)). The correlation peak c
s,
w [i
0] =L since it is the autocorrelation of the detection pattern. Therefore the detection distance is written as
D = * NS ~ / L (L - max { cs, w [i] : i < io } ). Note that D is a function of the preamble, the sync pattern and the detection pattern. Subsequently the synchronization word and the detection word are selected for providing a substantial detection distance D between the correlation peak and the correlation side lobes. Hence we proceed to search for sync patterns s and detection pattern lengths L that maximise D. The maximum detection distance that can be achieved for a given sync pattern length Ny and range of detection pattern lengths a < L < b is given by:
D : i < i
0 })
or equivalently by:
The above maximisation can be performed by exhaustive search for practical sync pattern lengths Ns. Fig.2 shows optimal pairs of synchronization word and detection word patterns with lowest correlation side lobes. The optimal pairs of sync and detection patterns according to the optimisation criterion are for example found by exhaustive search for Ns= 8, 12, 16, 20 and 1 < L < 40. The first column 21 indicates the basic period pattern of the preamble, where we have used "+' (resp. '-') as a shorthand notation for signal values of '+1' (resp. λ-l'). We consider the preamble to be built up by repeating an integer number of times the basic period. The second column 22 gives the length Ns of the synchronization word Sync p. s in the third column 23. The fourth column 24 gives the length L of the corresponding Detection p. w in the fifth column 25. The next column 26 lists the achieved detection distances and between parentheses the detection distances obtained when performing the optimisation for L — Ns i.e. when conventionally the sync pattern is used as detection pattern. In the last column 27 we observe gains of up to 201ogιo(9.414/6)=3.91 dB. In an embodiment an improved correlation receiver for frame synchronization is considered. We study a modification of the classical correlation receiver for frame synchronization. The modification consists in introducing a sheer before the correlator. It turns out that apart from having a lower complexity, the modified receiver outperforms the classical receiver for practical values of the probability of synchronization error. Fig.3 shows a binary correlation receiver. In addition to elements of Fig.l the binary correlation receiver 31 has a sheer 33 coupled before a binary correlator 32. A communication system similar to the system described with Fig.l may have a transmitted sequence built up as a succession of frames. The preamble may be a periodic sequence used for bit synchronization. The sync-word is a short sequence (typically not longer than 20 bits) used to mark the beginning of the data part. The task of the frame synchronization is to identify the beginning of the data in a frame. This is done by detecting the end of the sync- word. We introduce the End of Sync-word Instant (ESI id), which is the instant at which
detection should ideally occur. The end of sync- word instant id is the position of the last bit of the sync- word in the frame, i.e. i = Np + Ns - 1. When designing the frame synchronization, an important measure of performance is the achieved Probability of Synchronization Error PSE- Before defining PSE we identify the causes for synchronization errors. Two frame synchronization error events can be identified: the false alarms and the missed detection. Probability of False Alarm at instant i < idPFAβ]-' The probability of false alarm at instant i < id, PFA[ΪJ, is the probability that the sync-word is declared as found at an instant i prior to the ESI id. Probability of Missed Detection PMD-' The probability of missed detection
PMD is the probability that the sync-word is not detected at the ESI id. Probability of Synchronization Error PSR-' The probability of synchronization error PSE is the probability that frame synchronization fails because either a false alarm occurred at any instant before id or the detection of the sync-word was missed. It is assumed that the frame synchronization receiver starts searching for the sync- word at the beginning of the preamble, although in practice a receiver starts searching for the sync- word at an unknown position of the preamble. However, for the sake of simplicity, we assume that it starts at the beginning, and write:
PSE =
- P
rA [/]).
(1) ' In order to find a simple expression for P
SE we assume that the P
FA is only non-negligible at one unique instant (this assumption will be true for a well designed sync- word). Therefore we define P^ = max
;<w P
FA [i] and approximate expression (1) by:
^ ≥ I - (I - ^ ) - (I - ^7 ) - ^D + ^7 - (2) A correlator, for example the correlator 13 as shown in Fig.l, functions as described above. Now we consider the design of a binary correlator 32. The threshold detector 14 flags detection whenever the output sequence of the correlator qw = cw + nw surpasses a given threshold which we now proceed to design. We require detection to occur at the ΕSI id, therefore we need that c [Id] > max{cw[i] : i<id} and choose a threshold T satisfying cw[Id] > T > max{cw[i] : i<id}.
The presence of the additive Gaussian noise n
w results in a P
MD of
and a PFA[Ϊ] of
1 °° where Q(x) = —== \e
~r dr . From expressions (3) and (4) we see that choosing T to 2π decrease either of the probabilities increases the other. Since both error events result in a frame synchronization error, we choose T in order to minimize the maximum of P
MD and P
FA[Ϊ]- The optimal threshold T
opt in the above sense is the one yielding P^ = P^j where
This results in
Topt = (cw[id] + max{cw [i] : i< id})/2 (6). and
where D is the detection distance. The detection distance is the difference of the value of c
w at the ESI and the maximum value at any prior instant i.e.
D — cw[id] -max {cw[i] : < id}.
Note that D is a function of the preamble p and the sync-word s. We now use (7) in (2) and obtain an approximate expression for the probability of synchronization error
The receiver depicted in Fig.3 converts the received sequence r into binary by means of the sheer 33. This implies that the correlation can be computed by adding binary values solely and therefore the correlation receiver has a very low complexity. We denote this correlation receiver as the binary correlation receiver. The sheer converts the real valued sequence r = f+n into the binary sequence b as shown in Fig.3. Therefore we write for b:
i.e. positive values are detected as l's an negative values as -l's. Note that alternatively {0,1 } or any other binary alphabet could have been chosen. We can now write b = f + e, where e e {2, 0, 2}^ is the error sequence. A value ef j - 2 (resp. 2) at an instant signals that bit f[i
0] = 1 (resp. 1) was detected erroneously as b[io] = 1 (resp. 1). Note that efiq] = 2 (resp. 2) is not allowed when ffioj = 1 (resp. 1), since b[zø] e {1, 1}. The bit error probability (bEP) β i.e. the probability that a bit is detected erroneously due to the AWGN (of mean 0 and variance er
2) is β = P(e[i] ≠ O) = Q (l/σ) foτI = 0, . . . , N- l. The binary sequence b enters the correlator. At the output of the correlator we will have the correlation sequence q
w = c
w + eκ
s where e
Λfs e {-2N.
S, . . . , -2, 0, 2, . . . , 2N
S}
N is the correlation sequence of e with w € {- 1 , 1 }
Ns of length N
s. Note that the only assumption we make for w is to be binary. Note that now the noise sequence takes only discrete values. We derive general expressions for error probabilities for binary correlation receiver P
MD and PF
A in the next section. Given a binary sequence of length L and a bit error probability Q, we compute Pρ(L, l,j) the probability of having errors more in any subsequence of / bits than in the remaining L - l bits. We write .
+y(ι-δ)
M'
'+y) (10)
where the first binomial term corresponds to having i errors in the subsequence ofL - l bits and the second binomial term corresponds to having +j errors in the subsequence of 1 bits. The probability distribution for each element of e^s e {2N
J3 . . . , 2, 0, 2, . . . , 2N
S}
N can be written as:
where l[i] is the number of bits that coincide between the detection- word w and
( [/- (N
s - 1)], f[i]) (the subsequence of the frame with which w is correlated at instant i)
and β is the bEP. The above expression states that a bit of the frame that coincides with the corresponding bit of the detection- word will contribute with -2 to e whenever it flips due to an error and a bit that does not coincide will contribute with +2 whenever it flips. Now we can express P
MD and
in function of P(e
Ns[i]). We write:
for i < id where [ϊ| is the ceil function and T e N is the threshold level. For = s we have that I [i
d] = N
s in expression (11). Furthermore we assume that the detector threshold is given by (6) which implies that c
w[id] - T
opt = D/2; where D is the detection distance. This enables us to compute P
MD and approximate it by N
s p+2 MD p + 2
■ Q ( l (with sheer) (14)
σ Now we will compare the performance of the binary correlation receiver (with sheer) with the classical receiver (without sheer). Since Prøis approximately proportional to P
MD (see (8)), we base our comparison on P
MD in order to facilitate the analysis. Comparing (14) with (7) we see that the introduction of the sheer results in an asymptotic gain (in dB) given by: 4N D + 2
G = 101og (15) D1
We have plotted the exact expressions of PMD for both receivers in Fig.4A and 4B for a sync-word with Ns = 13 and D = 12. Fig.4A shows a comparison of error probabilities for a system with and without sheer. The plot also includes a curve 41 corresponding to the error probability PMD of a maximum likelihood (ML) receiver versus signal to noise ratio (SΝR). The x-axis represents the SΝR in dB which is defined as SΝR = -20 log 10 σ. A middle curve 42 shows a curve of the current receiver with sheer of error probability PMD versus SΝR. An upper curve 43 shows a curve without sheer of error probability PMD versus SΝR. We see that, for
this particular case, at high SNR values (corresponding to practical values of PMD) the receiver with sheer outperforms the classical receiver by about G = 1.60 dB. The reason for the good performance of the binary correlation receiver is that at high SNRs the sheer works as a simple bit detector eliminating much of the noise before the correlator, whereas in the system without sheer the noise values are added by the correlator. However, we expect the classical receiver to be better at low SNRs where the sheer is a poor bit detector and will effectively increase the noise (introducing bit errors) at the input of the correlator, as shown in Fig.4B. Fig.4B shows a comparison of error probabilities in detail. The plot includes simulation results of both receivers like in Fig.4A, which match the predicted behavior. We see that, for this example, the sheer starts to pay-off at an SNR of about 4.3 dB. Hence the binary correlation receiver has a lower complexity and also has a gain of G dB (see (15)) with respect to the classical correlation receiver for values of PMD of practical interest (e.g. PMD < 10-1 ). Fig.5 shows a schematic diagram of an optical storage system. The system includes on a record carrier 50 containing marks representing information. A read device includes a reading unit 53 for converting a signal from the marks on the record carrier 50 via a head 52. The optical storage device may include a recording unit 51 for generating a signal to be recorded via a head 52. The recording unit includes a synchronization generator for generating the synchronization word. As explained above the synchronization word is dependent on a detection word to be used for correlating with the signal in a receiver. Note that the detection word has a length different from the length of the synchronization word and is dependent on the synchronization word and a final part of the preamble that precedes the synchronization word. In the following part a practical embodiment of the optical storage system is described. A two dimensional optical system (TwoDOS) is used as an example of a system that requires frame synchronization. The TwoDOS system provides a number of parallel tracks that are simultaneously read by a number of read detectors, resulting in a number of read signals (called channels). For synchronization, a number of channels have the same content and may be added. In TwoDOS the syncword s marks the beginning of the user data. All M = 7 rows use the same syncword. The TwoDOS format proposes the binary string: s = 001 (probably later s = 001001). Therefore Ns = 3. In TwoDOS the preamble p has a length of Np = 256 and is built by concatenating 64 times the sequence 1100, i.e. /? = 1100...11001100.
Fig.6 shows a frame format. A TwoDOS frame consists of three parts: a preamble 61, syncword 62 and data 63. Let N, Np, Ns and Nd denote the frame, preamble, syncword and data lengths respectively, so that N = NP +NS +Nd. Fig.7 shows a block diagram of a frame synchronization receiver. A frame synchronization receiver 71 is provided with an adder 72 to add multiple input signals that have a combined frame sync pattern like in TwoDOS. A sheer 73 converts the added input signal samples a[n] to a binary sequence b[n] based on an adaptive threshold calculated by threshold unit 75. The sequence b[n] is coupled to sync detector 74, which provides the output signal d[n]. A further control signal provided by a preamble detector (Preamble) has been omitted in the Figure. The frame synchronization receiver is activated by the output signal of the preamble detector i.e. SyncSearch is set to TRUE whenever Preamble = TRUE. SyncSearch is set back to FALSE once the syncword has been found (SyncFound == TRUE), deactivating the receiver. This may be implemented in a software implementation of the receiver. In Fig.7 we identify three sequential operations that are applied on the replay signals. In the adder 72 the channel output signals ChOut[n] sampled at 1/T of the M= 7 rows are added up to form signal a[n]. The sheer 73 is used to convert the real valued signal a[n] into the binary signal b[n]. The Sync Detector 74 flags the ideal detection instant, i.e. the position of the last bit of s, by putting SyncFound = TRUE. Since the optical channels may have a DC-offset, it is necessary to first set the threshold of the sheer to the DC-offset of signal a[n]. A variable TD represents the value of the threshold and the flag SlicerThdSet indicates if the value of the threshold has been set. Because of the availability of the preamble pattern and the characteristics of the optical channel it is easy to determine the DC-offset during the preamble. The threshold TD can be determined during the preamble by computing the mean value during an interval I. Note that the first part of the preamble may intentionally be skipped since the input signal levels are not reliable prior to convergence of the control loops. The sync detector 74 performs the actual search for the syncword. In an embodiment the digital correlator consists of two basic blocks: a correlator and a threshold detector (similar to Fig.1). The correlator compares a so-called detection-word w of length Nw with the Nw most recent bits. The correlator is implemented as a filter with impulse response h[n] = w[Nw - l - n] for n = 0, . . . , Nw - l and h[n] = 0 for n ≠ 0, . . . , Nw - 1. If the correlation surpasses a given threshold T the syncword is declared found by setting
SyncFound = TRUE. If not, the next bit of the input is read and a new correlation with w is computed. This procedure is repeated until the syncword is found. The detection-word w is usually chosen to be equal to the syncword and is lengthened by prefixing it with the last bits of the preamble. In the example system Nw = 5 i.e. the detection-word is w = 00001 since s = 001 and the last two bits of the preamble are 00. Fig.8 shows an example of the output of the correlator. A curve 81 shows the correlator output of the example according to the specific values above, where the sequence b[n] contains no errors. A threshold 82 is shown at TD = 2. Note that at the ideal detection instant id the detection- word is compared to itself leading to a peak output equal to Nw = 5. The maximum value of the correlation for < z^is 1. The difference between the output at id and the maximum for i < i is denoted detection distance D and is a measure of robustness against errors and therefore related to the probability of synchronization error PSE. In our case D = 4. The threshold detector will set SyncFound = TRUE whenever the output of the correlator surpasses a given threshold TD. Both error events, false alarms and missed detection, result in a frame synchronization error, therefore we choose TD in order to minimize, at an arbitrary instant /, the maximum of PMD and PFA[Ϊ]- Since choosing TD to decrease either of the probabilities increases the other, the optimal threshold in the above sense is the one yielding PMD = P£T where ^ = maxi<w PFA[i]. A threshold 1 < TD < 3 achieves PMD ~ 7 an(l TD = 2 is optimal i.e. the probability of more than one error at id is approximately equal to the probability of error for i < id. For a practical selection of TD the worst case influence of the user data following the synchronization word may be taken into account. Hence the synchronization word and/or the detection word are selected for providing a substantial detection distance between the correlation peak and correlation side lobes during the correlating with the detection word. Subsequently detecting the synchronization word comprises applying a detection threshold at substantially half the detection distance. It is noted that the sync detection may also be implemented by a state machine or in software as elucidated below. Fig.9 shows a state transition diagram of a sync detect block. A similar diagram may be used for calculating the sheer threshold. Input to the synchronization word detection block (DetectSync) is the output of the optical source channels. We assume for the source channel signals fixed-point parameter values, e.g. sampled real values. In a sequence
of steps the following functions are performed and flags are used:
- SyncFound: Output signal of DetectSync. If the syncword was found it will be TRUE (T), otherwise it will be FALSE (F).
- SyncSearch: This flag indicates if DetectSync is in search mode or not. SyncSearch is set to FALSE at instantiation. Whenever signal Preamble is TRUE, SyncSearch will be set to
TRUE and DetectSync enters into search mode. DetectSync stays in search mode until the syncword is found, and then sets SyncSearch back to FALSE. The reason of introducing this flag is that the Preamble signal can (will) be set FALSE before the end of the preamble. We introduce therefore SyncSearch in order to keep DetectSync in search mode until the syncword is found. Note that there is no counter to check for a missed detection, but this may be added.
- SlicerThdSet: Indicates that SlicerThd is set and therefore the sheer can be used to determine DetBit. At first step 90 (In) starting values are set for SyncSearch = F, SlicerThdSet=F, DetectionThd = DET_THRESHOLD (default value), and
Correlator.Coefficient= detection word. We use a binary correlator i.e. a filter with binary coefficients, taking values -1 and 1. The coefficients may be programmable. In step 91 SyncFound = F. In step 92 the presence of the preamble is detected. If so, in step 93 the sync detect mode flag is set to logical value True (SyncSearch = T). In step 94 the process is finished if the SyncSearch = F. In step 95 the input signal is computed, for example the signal Summedlnput is the sum of all 7 signals (rows) of the input source channels, and therefore is represented with 3 bits more than the sample input values. In a next step 96 it is tested if the sheer threshold is set (SlicerThdSet=T); if not the sheer threshold is computed in step 97. In step 98 the detection bit signal value is computed. The signal DetBit corresponds to a sliced version of Summedlnput. It is a binary signal taking values -1 and 1. In step 99 the correlation value of the detection word and the sliced input value signal DetBit is computed as explained above. In next step 100 it is detected if the correlation value exceeds the threshold used to detect the presence of the syncword. The threshold depends on the choice of the syncword and detection-word. Its range coincides with the range of correlation and therefore has the same parameters. It may be a programmable variable for different input signal types. If the synchronization word is detected in step 101 flags are set: SyncSearch = F and SyncFound = T. In a step 102 the flag
SlicerThdSet = F, which implies that the sheer threshold is determined for each frame separately. The process is finished at step 103 (Out), which provides the output SyncFound. Although the invention has been explained mainly by embodiments based on an optical storage frame format, similar synclironization detection may be applied in any communication system, for example mobile phone signals. Further it is noted, that in this document the word 'comprising' does not exclude the presence of other elements or steps than those listed and the word 'a' or 'an' preceding an element does not exclude the presence of a plurality of such elements, that any reference signs do not limit the scope of the claims, that the invention may be implemented by means of both hardware and software, and that several 'means' may be represented by the same item of hardware. Further, the scope of the invention is not limited to the embodiments, and the invention lies in each and every novel feature or combination of features described above.