GB2300093A

GB2300093A - Receiver for timing recovery and frequency estimation

Info

Publication number: GB2300093A
Application number: GB9607879A
Authority: GB
Inventors: Dariusz Andrzej Blasiak
Original assignee: Motorola Inc
Current assignee: Motorola Solutions Inc
Priority date: 1995-04-19
Filing date: 1996-04-16
Publication date: 1996-10-23
Anticipated expiration: 2016-04-16
Also published as: CN1139321A; BR9601239A; DE19609504A1; AU4574596A; JPH08307408A; GB2300093B; KR960039712A; GB9607879D0

Description

- l- 2300093 RECEIVER AND ASSOCIATED METHOD FOR TIMING RECOVERY AND

FREQUENCY ESTIMATION

Background of the Invention

1. Technical Field

The present invention relates to a signal receiver and, more specifically, relates to a signal receiver using a frequency and burst detector to detect the timing and frequency of a burst.

2. Descrintion of the Related A is A pulse communication receiver, such as a digital receiver or a radar receiver, must obtain both a time reference and a frequency reference to receive and decode a signal. Bursts can be detected in the received signal to provide the time and frequency reference. In a digital communication system, such as a TDMA (time division multiple access) communication system, frames of information are periodically received. A timing reference for a received frame can be obtained by detecting any expected burst at a deterministic position within the frame. For example, a burst occurring at the beginning or other location of a frame can be detected to obtain a time reference for decoding the received signal. Once a burst has been detected information can be extracted from the frame or other portions of the received signal. This information can also be used to obtain timing for subsequent frames. Such frame synchronization is required before detecting information to provide an output for the user of the receiver.

A frequency reference for the received signal can be obtained by comparing an oscillator frequency of the receiver with a carrier frequency of the received signal. If the frequency is too high or too low, the oscillator of the receiver can then be adjusted in a feedback arrangement. A frequency reference for the received signal can also be obtained by receiving a predetermined frequency correction portion of a burst. Based on received characteristics of the predetermined frequency correction portion of the burst, the oscillator frequency of the receiver can be reset.

In previous receivers, a received signal is often compared with an expected pattern to establish a timing and/or frequency references. Such a system requires transmission from a transmitter to a receiver of dedicated patterns consuming valuable frequency spectrum and restricting system capacity. Should a system be established without dedicated patterns for establishing a timing reference and/or correcting frequency, system capacity is increased and frequency spectrum conserved.

When the transmitter and receiver obtain large frequency differences, the above correlation technique becomes unreliable. This large frequency differences can be caused by differences in the transmitter and receivers reference frequency due to, for example, crystal errors.

Furthermore, this large frequency difference can be caused when the receiver moves relative to the transmitter at a large velocity. For example, an aircraft or a satellite is fast moving and typically would have Doppler frequency errors when communicating with a ground station or another aircraft or satellite. As the transmitter and receiver obtain a larger frequency difference, the received signal moves outside the range of correlation with the expected pattern. Thus, as the frequency difference increases, the received signal and expected pattern become increasingly decorrelated and hence more difficult to establish a timing reference and quickly estimate the frequency.

The performance of either of the above techniques also degrades as the signal to noise ratio decreases. As the signal to noise ratio decreases, noise is hard to distinguish from the expected patterns.

A radio receiver capable of quickly and reliably estimating frequency and detecting a time offset of a received burst is needed.

Brief DescriRtion of the DrawinU FIG. 1 illustrates a block diagram of a radio receiver according to the present invention; FIG. 2 illustrates a block diagram of an embodiment of a synchronization stage of a radio receiver for estimating the frequency and offset of a received burst according to the present invention; FIGS. 3 and 4 illustrate block diagrams of the autocorrelation circuit according to the present invention; FIGS. 5 and 6 illustrate block diagrams of the correlation filter according to the present invention; FIGS. 7 and 8 illustrate block diagrams of the peak detector according to the present invention; and FIG. 9 illustrates a block diagram of the frequency estimator according to the present invention.

Detailed Descril2tion of the Preferred Embodiments is FIG. 1 illustrates a block diagram of a radio receiver according to the present invention. Antenna 100 receives a radio frequency signal and a radio frequency (RF) stage 110 converts the radio frequency signal to an inphase signal (I) and a quadrature signal (Q). An analog to digital converter 120 samples the in-phase signal and the quadrature signal to produce a digital in-phase signal and a digital quadrature signal in response to a sample timing from a timing circuit 130. A frequency and burst detector 140 establishes a coarse timing reference T and a frequency estimate f in response to the digital in-phase signal and the digital quadrature signal from the analog to digital converter 120 and in response to the sample time from the timing circuit 130. The digital in-phase signal and the digital quadrature signal from the analog to digital converter 120 are stored in a buffer 150. Upon detection of a burst as indicated by the coarse timing reference T from the frequency and burst detector 140 the signals stored in the buffer 150 are transferred to a receiver 160. Thereafter, the receiver provides a fine timing reference to the timing circuit 130 and can deliver received data to a voice decoder, a data unit and a call processor 170, for example, of the radio receiver. A fine frequency adjustment is also made at the RF stage 110 by the receiver 160.

The present invention recovers the timing and estimates the frequency of the burst without requiring dedicated patterns to establish a reference. Because large Doppler shifts alter the deterministic portions of the dedicated patterns, this improved technique is needed to recover the carrier of a received signal. Reliable burst detection and frequency estimation by the present invention is possible even when a transmitter and receiver obtain large frequency differences due to crystal errors as well is as the Doppler shifts. In the present invention, the characteristics of the signal itself are recognized. For instance, a constant power transient characteristic can be recognized when the signal bursts. The present invention reliably recovers bursts, for example, because it is capable of using second order statistics of a receiver burst. The present invention also increases system capacity and conserves frequency spectrum by not requiring dedicated patterns to establish a timing reference or to establish a frequency estimate. The present invention will not degrade the signal to noise ratio due to false detection of correlation peaks. The present invention also avoids multiple receiver paths to establish a timing reference, such as in a Rake receiver, thus saving processing time.

The frequency and burst detector 140 continually provides frequency estimates f whenever new bursts are received. The receiver 160 uses these frequency estimates f to obtain a burst and derive a fine frequency adjustment to reset the frequency of an oscillator in the RF stage 110. The timing recovery and frequency estimates are provided with an accuracy within the tolerance of the receiver 160. Fine frequency adjustments can also be made by the receiver 160 on a periodic basis in the RF stage 110 based on subsequent reliable reception of, for example, a frequency correction burst or frame portion. Such frequency correction burst or frame portion could not have been received until the frequency estimate was used by the receiver 160 to receive the burst.

When the frequency and burst detector 140 detects a burst as indicated by the coarse timing reference T, the timing circuit 130 causes a mode change from a burst detection mode to a gated receive mode. While in the burst detection mode, a timing reference has not yet been obtained by the frequency and burst detector 140 and information can not yet be extracted to provide an output for the user of the receiver. After a timing is reference has been obtained by the frequency and burst detector 140, information can subsequently be obtained from the received signal by the receiver 160 under the assumption that the timing will be slowly varying. A mode switch 180 switches between the burst detection mode and the gated receive mode in response to the timing circuit 130. During the gated receive mode, slow variations in timing will be corrected by the receiver 160 via the fine timing reference. The receiver 160 generates the fine timing reference from its synchronization resulting from extracting information from the received signal to compensate for slow variations in timing.

The timing circuit 130 provides the sample time to clock the sampling by the analog to digital converter 120 and also provides the sample time for digital circuits of the frequency and burst detector 140. The timing circuit 130 could contain, for example, a latch and a counter. Upon detection of the burst as indicated by the coarse timing reference T, the latch will be triggered causing a mode change by the switch 180. The counter will reset and begin counting in response to the coarse timing reference T to generate the sample time for clocking of the analog to digital converter 120 and the frequency and burst detector 140.

FIG. 2 illustrates a block diagram of an embodiment of a synchronization stage of a radio receiver for estimating the frequency and offset of a received burst according to the present invention. An autocorrelation circuit computes an autocorrelation metric. The autocorrelation metric is combinations of autocorrelation sequence lags of the digital in-phase signal (1) and the digital quadrature (Q) signal received by the RF stage 110. A digital filter 220 filters the autocorrelation metric from the autocorrelation circuit 210 and produces a filtered signal. A digital filter 220 matched to the expected burst would provide the -7 maximum signal to noise ratio. The correlation filter 220 preferably is constructed as a finite impulse response filter (FIR) or an infinite impulse response filter (HR) having taps chosen such that an impulse response of the correlation filter is representative of a magnitude and duration of the expected burst. A peak detector 230 detects a peak of the filtered signal from the correlation filter 220. The location of a peak is indicated by the coarse timing signal T output of the peak detector 230. The peak detector 230 can use a maximum approach to determine a leading edge of the burst.

The peak detector 230 can also use pattern match approach. A frequency estimator 240 estimates the frequency of the burst based on the filtered signal from the correlation filter 220 and based on the location of the burst indicated by the coarse timing signal T from the peak detector 230. The peak indicates the time offset or location of the burst.

FIG. 3 illustrates a block diagram of the autocorrelation circuit according to the present invention. The illustrated autocorrelation circuit receives sampled inphase (1) and quadrature (Q) baseband signals from the radio frequency (RF) stage of a radio receiver. The autocorrelation circuit has two paths, a lead path and a lag path. A delay element 310 provides the delay of the lag path. The delay can be chosen to be greater than unit delay if the sampling rate R is sufficiently high such that the value of the delay divided by R is less than one over the maximum frequency offset of the burst. The delay is chosen so that gross errors in the frequency estimate are not suffered as a result of the discrete frequency offset of the burst measuring greater than 2n radians, which would result in a noninvertible phase wrapping. One of the two paths has a complex conjugate operation prior to multiplying the two paths in a complex multiplier 330. The complex conjugate block 320 preferably is provided in the lag path before the complex multiplier 330.

is FIG. 4 illustrates another block diagram of the autocorrelation circuit according to the present invention. In FIG. 4 the lead path has a lead filter 340 and the lag path has a lag filter 350. For purposes of computational simplicity, both filters 340 and 350 are linear and preferably finite impulse response filters (FIR). The lead filter 340 and the lag filter 350 have taps chosen such that the output of the autocorrelation circuit 210 provides an autocorrelation estimate having an improved signal-to-noise ratio over that achieved by the autocorrelation circuit in FIG. 3 and in addition exhibits linear phase, thus assuring an unbiased frequency estimate. An improved estimate of the position and frequency of the received burst is obtained with such lead and lag filters 340 and 350 added to the construction of FIG. 3.

FIGS. 5 and 6 illustrate block diagrams of two alternative embodiments for implementing the correlation filter 220 according to the present invention. The correlation filter 220 can be implemented in a complex finite impulse response filter (FIR) according to FIG. 5 or an infinite impulse response filter (HR) according to FIG. 6. The FIR filer of FIG. 5 has a series of delay stages 410, 420, 430 and 440. The number of delay stages is preferably one less than the number of samples L necessary to capture the entire length of an expected burst. Each delay stage and the input itself are multiplied in taps 450, 460, 470, 480 through 490 by values C1 through CL. The outputs of taps 450 through 490 feed a summer 495.

The taps 450 through 490 preferably have only real values C1 through CL.

Should the taps 450 through 490 have complex values, that is have both non-zero real and imaginary parts, then the weighting by the taps 450 through 490 must be accomplished using complex multiplication. In practice, only real values for the taps 450 through 490 will quite likely be needed. Nevertheless, taps requiring complex values could still be is avoided to maintain simple calculations and reduce current drain and processing time.

FIG. 6 illustrates the correlation filter 220 implemented in an infinite impulse response filter (HR). A summer 510 sums an input from the autocorrelation circuit 210 and outputs of taps 520, 530, 540 through 550. Delay stages 560, 570, 580 through 590 delay a result of the summer 510 and feedback delayed results to taps 520 through 550. The taps 520 through 550 multiply the delayed result by values Cl through Ck. The taps 520 through 550 preferably have only real values Cl through Ck. As above, should the taps 520 through 550 have complex values, that is have both non-zero real and imaginary parts, then the weighting by the taps must be accomplished using complex multiplication. In practice, only real values will quite likely be needed.

FIG. 7 illustrates a block diagram of the peak detector 230 for generating the coarse timing reference T according to one exemplary embodiment of the present invention. The exemplary embodiment of FIG. 7 illustrates a maximum technique. A magnitude of the complex number from the correlation filter 220 needs fo be determined during peak detection. A magnitude circuit 610 is preferably used by the peak detector 230. The magnitude circuit 610 derives a quantity from the output of the correlation filter 220 which is equivalent to the magnitude of that output. Only the magnitude of the output of the correlation filter 220 is needed to determine the location of a peak. The location of the peak establishes the location of the first sample of a received burst. The taps in the correlation filter 220 are related to the peak detector in the sense that they are chosen so that only one peak will occur at the output of the magnitude circuit 610 when a burst is present in noiseless conditions.

is The magnitude circuit 610 could be implemented as the sum of a square of the real part and of a square of the imaginary part of the input. Alternatively the magnitude circuit could be implemented as the sum of an absolute value of the real part and of an absolute value of the imaginary part of the input.

A maximum detector 620 detects the peak based on the output of the magnitude circuit 610. The maximum detector 620 preferably contains a threshold by which no peak will be identified unless its magnitude is greater than the threshold. The threshold is significantly above the noise power but significantly below the expected signal peak power to avoid falsing on noise. Any peaks which are detected below this threshold will not be mistaken for an actual burst location. Such a threshold can either be deterministic or dynamic based on current channel noise conditions.

FIG. 8 illustrates a block diagram of another peak detector 280 according to the present invention. A pattern match detector 720 pattern matches a shape of the signal output from a magnitude circuit 710 with an expected waveform such as, for example, the shape of the signal. The pattern match detector 720 takes into account more characteristics of the signal, such as, for example, the slope or shape of the signal.

FIG. 9 illustrates a block diagram of the frequency estimator 240 for providing a frequency estimate f according to the present invention. Although a frequency can be derived in many ways from a sequence of complex samples, a preferred embodiment of a frequency estimator is illustrated in FIG. 9. A complex conjugate of the output of the correlation filter 220 is delayed and multiplied by itself to produce a product. A delay stage 810 delays the output signal and a complex conjugate block 820 determines the complex conjugate before multiplying by multiplier 830. An argument operator 840 determines an arctangent of the imaginary part is of the product divided by the real part of the product. A sample and hold 850 gates the output of the argument operator 840 in response to the location of new bursts indicated by the coarse timing reference T from the peak detector 230. A scaling factor 860 converts a discrete frequency estimate in radians to a frequency estimate in Hertz by scaling the radians value by R/21c, where R is the sampling rate. The discrete frequency estimate must be further scaled by one over the number of delay units implemented by the delay element 310 in the autocorrelation circuit.

The signal processing techniques of the present invention disclosed herein with reference to the accompanying drawings are preferably implemented on a digital signal processor (DSP) or other microprocessor. Nevertheless, such techniques could instead be implemented wholly or partially as discrete components. Further, it is appreciated by those of skill in the art that certain well known digital processing techniques can be represented mathematically in different ways depending on the choice of implementation.

Although the invention has been described and illustrated in the above description and drawings, it is understood that this description is by example only, and that numerous changes and modifications can be made by those skilled in the art without departing from the true spirit and scope of the invention. Thus the outputs of the timing circuit 130 may be required by different circuits and not needed by all others. Although the present invention exhibits Doppler shift tolerance, the present invention provides additional advantages as mentioned herein and is thus applicable to all radio communications systems regardless of the need for Doppler shift tolerance such as paging, cellular and satellite communication system receivers.

Claims

What is claimed is:

Claims is 1. A synchronization stage in a radio receiver for both recovering timing and estimating frequency of a received burst, comprising: an autocorrelation circuit to provide an autocorrelation metric indicative of combinations of autocorrelation sequence lags of the received burst; a correlation filter, having an impulse response resembling a magnitude and duration of an expected burst, operatively coupled to said autocorrelation circuit, to filter the autocorrelation metric and provide a filtered signal; a peak detector operatively coupled to said correlation filter to detect a peak of the filtered signal and provide a coarse timing signal; and a frequency estimator operatively coupled to said peak detector and said correlation filter to estimate the frequency of the burst based on the filtered signal and the coarse timing signal.
2. A synchronization stage according to claim 1, wherein said correlation filter comprises a finite impulse response filter having taps chosen such that an impulse response of the finite impulse response filter is representative of a magnitude and duration of an expected burst.
3. A synchronization stage according to claim 1, wherein said correlation filter comprises an infinite impulse response filter having taps chosen such that an impulse response of the infinite impulse response filter is representative of a magnitude and duration of an expected burst.

is
4. A synchronization stage according to claim 1, wherein said autocorrelation circuit comprises a delay path for receiving the burst and providing an autocorrelation metric of real and imaginary parts of the burst based on a delayed burst.
5. A synchronization stage according to claim 1, wherein said peak detector comprises a magnitude circuit operatively coupled to said correlation filter to determine a magnitude of the filtered signal.
6. A synchronization stage according to claim 5, wherein said peak detector further comprises a maximum detector operatively coupled to said magnitude circuit to select a maximum of the magnitude of the filtered signal.
7. A synchronization stage according to claim 5, wherein said peak detector further comprises a pattern match detector operatively coupled to said magnitude circuit to pattern match a shape of the magnitude of the filtered signal.
8. A synchronization stage according to claim 1, wherein said synchronization stage further comprises an RF stage and an antenna of the radio receiver.
9. A method of both recovering timing and estimating frequency of a time offset of a received burst, comprising:

(a) generating an autocorrelation metric indicative of combinations of autocorrelation sequence lags of the received burst; (b) filtering the autocorrelation metric and providing a filtered signal using a filter having an impulse response resembling a magnitude and duration of an expected burst; (c) detecting a peak of the filtered signal and providing a coarse timing signal; and (d) estimating the frequency of the burst based on the filtered signal and the coarse g signal.