GB2386307A

GB2386307A - A multi-carrier signal with reduced or varying pilot sub-carriers

Info

Publication number: GB2386307A
Application number: GB0313896A
Authority: GB
Inventors: Wieslaw Jerzy Szajnowski
Original assignee: Mitsubishi Electric Information Technology Corp
Current assignee: Mitsubishi Electric Information Technology Corp
Priority date: 1998-10-30
Filing date: 1999-10-26
Publication date: 2003-09-10
Anticipated expiration: 2019-10-26
Also published as: GB2386307B; GB0313896D0

Abstract

The application relates to a multi-carrier signal with pilot sub-carriers lower than other sub-carriers, a multi-carrier signal with pilot sub-carriers varying in time or frequency and to a method of determining the frequency offset of a multi-carrier signal by using a function of the powers or amplitudes of subcarrier (eg pilot) signals.

Description

MULTICARRIER COMMUNICATIONS SYSTEMS The present invention relates to the field of frequency synchronisation in multicarrier communication systems, for example those which, for synchronisation purposes, utilise pilot subcarriers with increased power. The invention is particularly, but not exclusively, concerned with the tuning of receivers for OFDM signals, for example digital video broadcast receivers.

A multicarrier communication system utilises a large number of equally spaced subcarriers for data transmission and for other auxiliary functions. In order to perform the demodulation process correctly, the receiver developed for such a system must be frequency-locked with a very small residual frequency offset. To facilitate frequency locking, a class of multicarrier communication systems employs a set of pilots transmitted at selected subcarriers with an increased power. These selected subcarriers should form a pilot insertion pattern with optimal autocorrelation characterised by low sidelobe values.

Fig. 1 shows an example of the power associated with the DFT coefficients reconstructed by a receiver in an ideal case when there is no noise or interference and the communication channel is distortionless. However, in practical applications the received signals are corrupted by wideband noise as well as by strong narrowband signals generated at some frequencies by various interfering sources. Furthermore, when the transfer function of the

communication channel has not been corrected, the channel itself will introduce both magnitude and phase distortions. Fig. 2 illustrates the distortions resulting from a combined effect of frequency-selective fading, noise and strong interference occurring at some subcarrier frequencies. As seen, in this case discrimination between pilot subcarriers and data subcarriers is much more difficult which will result in a degraded performance of any system used for frequency locking.

Fig. 3 shows the two components of the total frequency offset Af comprising the fractional part Wc and the coarse frequency offset J*Afc where J* is 2 in the illustrated example, and where Ate is the subcarrier spacing.

Several time-domain methods are available for estimating and correcting the fractional frequency offset in a multicarrier communication system. However, many methods proposed for estimating the coarse frequency offset are based on coherent processing which utilises known phase relationships between signals to be discriminated. Such methods are not well suited for initiating the frequency acquisition process when the channel transfer function has not yet been corrected.

Aspects of the invention are set out in the accompanying claims.

The preferred embodiment of the invention enables the estimation of a coarse frequency offset by suitably processing received signals which may have been distorted severely by the uncorrected communication channel and

may have been corrupted by wideband noise and strong narrowband tn interference.

In accordance with the preferred embodiment, a signal processing apparatus determines a frequency offset for tuning purposes by obtaining, for each of a plurality of candidate frequency offsets, the sum of the powers of a set of subcarriers. If certain conditions are met, the frequency offset associated with the largest of these sums is used for adjusting tuning.

Preferably, for each candidate frequency offset, the power sum excludes the U largest of the subcarrier powers, where U is an integer of one or more. It has been found that this technique reduces the probability of selection of an incorrect frequency offset as a result of strong interfering signals.

Preferably, a frequency offset is selected for tuning purposes only if its associated power sum bears a predetermined relationship with at least some of the power sums associated with the other frequency offsets. Preferably, the largest sum must bear a predetermined relationship with the average of the next L largest power sums, where L is an integer of one or more.

Preferably, the power sum associated with each frequency offset is combined (e. g. integrated) over a plurality of symbol periods, for more reliable determination of the desired frequency offset. Preferably, this operation ceases when a prescribed criterion is met, this event indicating with a high probability that the correct frequency offset has been determined. This

means that the number of symbol periods needed for reliable frequency offset estimation is not fixed in advance, thus resulting in as short a time-to-lock value as possible.

To allow for variable observation periods, the criterion which establishes that the desired frequency offset has been determined is a function of the number of symbol periods used for observation, and how the power sums change during observation.

The present invention can thus provide a signal processing apparatus which utilises a sequential decision procedure to minimise the time required for determining a reliable estimate of a coarse frequency offset.

An arrangement embodying the invention will now be described with reference to the accompanying drawings, wherein: Figure 1 shows an example of the power associated with the DFT coefficients reconstructed by a receiver in an ideal case when there is no noise or interference and the communication channel is distortionless; Figure 2 shows a typical example of the power associated with the DFT coefficients reconstructed by a receiver when the received signal is corrupted by noise and strong interference and also distorted by the channel due to frequency-selective fading; Figure 3 illustrates the two components of the total frequency offset comprising the fractional part and the coarse frequency offset;

Figure 4 is a block diagram of a receiver in accordance with the invention; Figure 5 is a block diagram of an apparatus used in the receiver for non-coherent estimation of a coarse frequency offset ; Figure 6 is a block diagram of a trimming and summing unit (TSU) of the apparatus; Figure 7 is a block diagram of a sorting and storing register (SSR) of the apparatus; and Figure 8 is a block diagram of a decision block (DB) of the apparatus.

An embodiment of the invention, in the form of a digital broadcast receiver, will be described with reference to FIG. 4, which is a block diagram of the digital broadcast receiver.

The conventional part of this digital communication receiver, which may be a broadcast receiver, comprises an antenna 1 that receives an OFDM multicarrier broadcast signal, a radio-frequency amplifier 2 that amplifies the received broadcast signal, a mixer 3 that down-converts the amplified signal to an intermediate frequency signal, an intermediate-frequency amplifier 4 that amplifies the intermediate-frequency signal, an analog-to-digital converter (ADC) 5 that converts this signal to a digital signal, a real-tocomplex converter (also known as an IQ generator) 6 which takes the input signal and outputs a complex signal centred on zero frequency by digital signal processing, thus producing in-phase (1) and quadrature (Q) baseband

signals, a fast-Fourier-transform (FFT) processor 7 that executes a discrete Fourier transform on these signals to obtain the sub-symbol data for each subcarrier, an error-correcting processor 8 that detects and corrects errors in the sub-symbol data, an output terminal 9 to which the sub-symbol data are supplied, and a voltage-controlled oscillator (VCO) 10. The voltagecontrolled oscillator 10 is used as a local oscillator that supplies the mixer 3 with a signal tuned to a frequency differing from the OFDM broadcast frequency by a fixed amount. The oscillator 10 receives a control signal from a tuning controller 101 coupled to receive the baseband output from real-tocomplex converter 6 and the output of the FFT processor 7.

It is assumed that the fractional frequency offset has been corrected by employing one of the available time-domain techniques well known to those skilled in the art, using a fine frequency control circuit 102 coupled to receive the baseband output. It is similarly assumed that the symbol timing has been estimated so that the received signals can be sampled efficiently with minimal timing error. The coarse frequency control circuit 103 is then used to ensure thai the respective subcarriers are located at the correct frequency positions by adjusting the frequency of the VCO 10. In the figure, it is assumed that all timing is performed using the VCO 10; however, at least the fine tuning may alternatively be performed instead by applying digital rotation to the signals outputted by the converter 6 and at least the coarse tuning could be achieved by use of the FFT processor 7. It is also assumed in the figure that the output

of the FFT processor is used for coarse frequency control; instead, however, the output of error-correcting processor 8 could be used.

Assume that there are P pilots transmitted at subcarriers with indices k, k2,..., kp. The ordered set of these indices will be referred to as the pilot

insertion pattern. It is also assumed that the range of possible coarse frequency C : l offsets, measured in multiples of the subcarrier spacing, is (-Jmin,'Jmin'Mj...) -1, 0, 1,..., Jmax'l, Jmax)Therefore, the total number of possible offsets is equal to (nun+ Jmax+l).

Fig. 5 is a block diagram of the coarse frequency control circuit 103 for non-coherent estimation of a coarse frequency offset. The outputs of the FFT processor 7, performing the discrete Fourier transform (DFT) for demodulation purposes, are connected to the inputs of an output selector and multiplexer (OSM) 20. The OSM has (Jmin+Jmax+l) output channels and each channel provides P complex DFT coefficients corresponding to a respective frequency-shifted version of the pilot insertion pattern. Thus, each channel corresponds to a possible frequency offset, and provides the DFT coefficients representing the pilot signals, assuming that that frequency offset is correct.

The coefficients are presented to (Jmn+Jmax+I) trimming and summing units (TSU's) 22, each associated with a respective output channel, for processing the DFT coefficients which represent the original pilot insertion

pattern and its (Jnn+Jmax) versions shifted in frequency by all possible coarse frequency offsets- A functional block diagram of each TSU 22 is presented in Fig. 6. The TSU implements the following operations.

First. at block 222 each complex DFT coefficient is used to calculate q=x2+y2 where x and y are the real and imaginary part of the DFT coefficient, respectively. All the power values q are summed by an accumulator at block 2 :'+. These power values are also sorted at block 226. The largest U values (U is an integer of 1 or preferably more) are summed by an accumulator at 228.

The result is then deducted, at 230, from the sum produced at 224. Thus, a "trimmed" sum S is formed by disregarding the U largest values of q and adding together the remaining (P-U) values. The purpose of trimming is to increase the statistical distance between distributions representing different classes of signals thereby facilitating their discrimination. The effect is to avoid erroneous assumptions that the correct frequency offset has been found due to detection of large powers. where these large powers are instead a result of. for example. strong interfering signals. The value U can be chosen empirically.

Each trimmed sum is effectively a weighted sum, with some of the weighting coefficients set to 1, and others to 0. Other possibilities include

having a more gradual progression of coefficients, such that the coefficient increases for smaller power values, at least over part of the range of powers.

Because the TSUs operate in parallel, (Jnun+Jmax+l) trimmed sums are fed simultaneously to the inputs of a sorting and storing register (SSR), shown at 24 in Fig. 5.

Fig. 7 is a functional block diagram of an SSR 24. The operations performed by the SSR can be summarised as follows: at block 300 sort the trimmed sums S from the TSUs ;

at block 302 select the greatest of the trimmed sums Smax ; 4 : 1 at block 304 store the value of the shift, say J*, corresponding to this greatest trimmed sum; at block 306 calculate the average AVL of the L next-largest trimmed sums. supply the three above values to the decision block (DB) shown at 26 in Fig. 5.

A block diagram of the decision block (DB) 26 is shown in Fig. 8. The decision block utilises the information provided by the SSR 24 and by a symbol period counter (SPC) 28 (Fig. 5) at block 400 to determine the value of the decision threshold according to the formula : TH=AVJI+h/ (LM)]

where M is the number of symbol periods used for observation, h is a constant, preferably greater than 0.6, L is an integer of 1 or preferably more and AVL is the average of the L next-highest power sums after the largest sum Smax. The value of h may be chosen empirically, to achieve a good compromise between lowering the probability of obtaining an incorrect estimate (leading to false locks) and increasing the total observation time required to make a decision (time to lock).

Next, at block 402, the DB compares the value of Smax to the calculated threshold value TH. If Smax exceeds the threshold TH, then at block 404 the value of the shift J* is used as an estimate of the coarse frequency offset, measured in multiples of the subcarrier spacing. However, if the threshold TH has not been exceeded, then the information available is not sufficient to determine a reliable estimate of the offset and additional observations obtained from the next symbol period will have to be processed.

This sequential decision procedure is terminated when the threshold TH has been exceeded by Smax or the procedure is aborted when the total number of observed symbol periods M has reached a predetermined maximum value Mmax. Therefore, in extreme cases, especially for small values of Mmax and severe spectrum distortions, it is possible to terminate the sequential

procedure without obtaining an estimate of the frequency shift. On the other I hand. the frequency shift estimate is obtained as soon as enough information c

has been received, possibly after processing of only one symbol. Incidentally, although it is preferable that the apparatus consider every successive symbol period, this is not essential; selected periods could, for example, be disregarded.

The operation of the blocks and units described above is initiated by a control and timing unit (CTU) 30 in Fig. 5 which resets all accumulators, registers and counters. The CTU also determines when the sequential decision procedure is to be terminated or aborted.

The symbol period counter (SPC) 28 determines the number M of symbol periods which have been processed since the initialisation up to the current stage of the sequential estimation procedure. This information is used by the decision block (DB) to determine the value TH of the adaptive decision threshold as described above. The current value of M is also used by the CTU 30 to decide whether to abort the sequential procedure without producing a frequency shift estimate.

In a particular embodiment for use with a 2K OFDM signal (i. e. having 1705 active subcarriers, including P = 45 pilots), it has been found particularly desirable to have Jnun = Jmax = 20, and thus there are 41 possible frequency offsets, U = 8 or preferably 4, and L = 12, although of course each of these could independently be altered if desired.

The embodiment described above may be implemented entirely in

hardware, using for example an ASIC with appropriately designed logic gates. t

Alternatively. some or all of the functions may be performed by one or more appropriately-programmed general-purpose processor units. If several functions are to be performed by individual processor units, it may be desirable or necessary for these functions to be performed in serial, rather than in parallel.

Although the invention has been described in the context of multicarrier signals wherein the pilot subcarriers have increased power, by suitable modification it could be used with signals in which the pilots have decreased, or preferably zero, power (by looking for minima of the calculated powers). In such an arrangement, it may be preferred for the weighting to be such that both the largest power values and the smallest power values are disregarded, the smallest being disregarded to avoid erroneously judging an offset to be correct as a result of signal fading.

In a further alternative embodiment, the signals have pilots whose powers vary according to time and/or subcarrier index, and preferably form a specific two-dimensional pattern in the time/frequency domain. The frequency offset could be found by looking for predetermined distributions of the calculated powers. For example, it would be possible to check whether, for each possible frequency offset. the sum of the powers of a predetermined set of subcarriers varies from symbol period to symbol period in a predetermined manner.

Instead of using fixed weighting coefficients, as in the above-described embodiment, it would be possible to use coefficients which vary in the time and/or frequency domain.

However, some aspects of the invention can be performed even when all the weighting coefficients are equal to unity, i. e. when there is no effective weighting.

It is believed that the concept of using multicarrier signals where the pilot subcarriers have decreased or no power, or where the powers vary according to a pattern in the time and/or frequency domain, is independently inventive, and the invention extends to such signals, and to methods and apparatus for receiving and for transmitting such signals. Such arrangements may require less transmission power.

Claims

Claims:

1. A multicarrier signal having a plurality of frequency-spaced subcarriers of which a predetermined subset comprises pilot subcarriers which, for synchronisation purposes, have zero power, or diminished power with respect to other subcarriers.

2. A multicarrier signal having a plurality of frequency-spaced subcarriers of which a predetermined subset comprises pilot subcarriers which, for synchronisation purposes, have powers which vary in the time and/or frequency domain.

3. A signal as claimed in claim I or claim 2, the signal being an OFDM signal.

4. A method of transmitting a signal as claimed in any one of claims 1 to 3.

5. Apparatus arranged for transmitting a signal as claimed in any one of claims 1 to 3.

6. A method of receiving a signal as claimed in any one of claims 1 to 3.

7. Apparatus arranged to receive and synchronise with a signal as claimed in any one of claims 1 to 3.

8. A method of determining a frequency offset between a set frequency and a desired frequency for synchronisation with a multicarrier signal, the method comprising: (a) selecting a frequency offset, and, for that offset, determining a weighted sum of powers of a predetermined set of subcarriers ; (b) performing step (a) for other selected offsets; and (c) providing a signal representing the frequency offset associated with a selected one of said sums in response to that sum meeting a predetermined criterion.

9. A method as claimed in claim 8, wherein said selected sum is the largest of the sums.

10. A method as claimed in claim 9, wherein the weighting coefficients used to determine the weighted sum of powers for each frequency offset are smaller for the U largest of the powers, where U is an integer of 1 or more, than for other powers.

11. A method as claimed in claim 10, wherein the coefficients are such that the U largest of the powers are disregarded.

12. A method as claimed in claim 10 or 11, wherein U is 4 or more.

13. A method as claimed in any one of claims 9 to 12, wherein the predetermined criterion is a function of the L next largest of said sums,

wherein L is an integer of 1 or more. tp

14. A method as claimed in claim 13, wherein the predetermined criterion is a function of the average AVL of the L next largest of said sums.

15. A method as claimed in any one of claims 8 to 14, wherein the power sums are combined over a plurality M of symbol periods.

16. A method as claimed in claim 15, including the step of repeatedly checking whether said predetermined criterion exists for increasing numbers of symbol periods, and providing said signal representing the frequency offset when the predetermined relationship is found to exist, whereby the number M is variable.

17. A method as claimed in claim 16, wherein said predetermined criterion is a function of M.

18. A method as claimed in claim 17 when dependent on claim 14, wherein the predetermined criterion is determined by comparing the largest sum with: AVL [l+h/ (LM)] where h is a predetermined constant.

19. A method of determining a frequency offset between a set frequency and a desired frequency for synchronisation with a multicarrier signal, the method comprising, for each of a plurality of frequency offsets, determining the sum of the powers of a predetermined set of subcarriers, each sum being repeatedly recalculated to take into account additional symbol periods until one of said sums meets a predetermined criterion, whereupon a signal is provided representing the frequency offset associated with said one sum.

20. A method according to claim 8 or claim 9 of determining a frequency offset, the method being substantially as herein described with reference to the accompanying drawings.

21. A method of tuning a receiver to a multicarrier signal, the method comprising performing coarse tuning by using a frequency offset determined by a method as claimed in any one of claims 8 to 20.

22. A method as claimed in claim 21, wherein the coarse tuning is performed after performing a fine tuning operation in order substantially to bring the tuning frequency into a predetermined relationship with a multicarrier signal frequency, the coarse tuning then being performed to alter the subcarrier frequency to which the tuning frequency is matched.

23. A method as claimed in claim 22, wherein the fine tuning operation is performed in the time domain.

24. A method according to claim 21 of tuning a broadcast receiver, the method being substantially as herein described with reference to the accompanying drawings.

25. A multicarrier signal receiver, the receiver having a tuning control operable to perform a tuning operation according to any one of claims 21 to 24.

26. A receiver as claimed in claim 25, suitable for receiving digital 0 video broadcast signals.

27. A digital receiver according to claim 25 and substantially as

herein described with reference to the accompanying drawings. c