System and method for symbol clock recovery

Abstract

A system and method for symbol clock recovery independent of segment location recovery uses the frequency and phase information in the upper and lower band edges of a signal to generate a signal for correcting the symbol clock. A particular combination of raised-root cosine filters, low-pass filters, multipliers, and adders effectively uses the tails of a received signal in the frequency domain to correct phase errors.

Description

BACKGROUND

[0002] Traditionally, local communication was done over wires, as this presented a cost-effective way of ensuring a reliable transfer of information. For long-distance communications, transmission of information over radio waves was needed. Although this was convenient from a hardware standpoint, radio frequency (RF) transmission brought with it problems related to corruption of the information and was often dependent on high-power transmitters to overcome weather conditions, large buildings, and interference from other sources of electromagnetic radiation.

[0003] The various modulation techniques that were developed offered different solutions in terms of cost-effectiveness and quality of received signals, but until recently they were still largely analog. Frequency modulation and phase modulation provided a certain immunity to noise, whereas amplitude modulation was more simple to demodulate. More recently, however, with the advent of low-cost microcontrollers and the introduction of domestic mobile telephones and satellite communications, digital modulation has gained in popularity. With digital modulation techniques come all the advantages that traditional microprocessor circuits have over their analog counterparts. Problems in the communications link can be overcome using software. Information can be encrypted, error correction can ensure more confidence in received data, and the use of digital signal processing can reduce the limited bandwidth allocated to each service.

[0004] As with traditional analog systems, digital modulation can use amplitude, frequency, or phase modulation with different advantages. As frequency and phase modulation techniques offer more immunity to noise, they are the preferred techniques for the majority of services in use today.

[0005] A simple variation from traditional analog frequency modulation can be implemented by applying a digital signal to the modulation input. Thus, the output takes the form of a sine wave at two distinct frequencies. To demodulate this waveform, it is a simple matter of passing the signal through two filters and translating the resultant back into logic levels. Traditionally, this form of digital frequency modulation has been called frequency-shift keying.

[0006] Spectrally, digital phase modulation, or phase-shift keying, is very similar to frequency modulation. It involves changing the phase of the transmitted waveform instead of the frequency, these finite phase changes representing digital data. In its simplest form, a phase-modulated waveform can be generated by using the digital data to switch between two signals of equal frequency but opposing phase. If the resultant waveform is multiplied by a sine wave of equal frequency, two components are generated: one cosine waveform of double the received frequency and one frequency-independent term whose amplitude is proportional to the cosine of the phase shift. Thus, filtering out the higher-frequency term yields the original digital data.

[0007] Taking the above concept of phase-shift keying a stage further, the number of possible phases can be expanded beyond two. The transmitted “carrier” can undergo changes among any number of phases and, by multiplying the received signal by a sine wave of equal frequency, will demodulate the phase shifts into frequency-independent voltage levels.

[0008] An example of this technique is quadriphase-shift keying (QPSK). With quadriphase-shift keying, the carrier changes among four phases, and can thus represent any of four values per phase change. Although this may seem insignificant initially, it provides a modulation scheme that enables a carrier to transmit two bits of information per symbol instead of one, thus effectively doubling the data bandwidth of the carrier.

[0009] The mathematical proof of how phase-modulated signals, and hence QPSK, are demodulated is shown below.

[0010] Euler's relations characterize sine and cosine waves as follows: 1$\sin \omega t=\frac{e^{\mathrm{j\omega }t}-e^{-jt}}{2j}\cos \omega t=\frac{e^{\mathrm{j\omega }t}+e^{-\mathrm{j\omega }t}}{2}$

[0011] where j={square root}{square root over (−1)}. Thus, the multiplication of two sine waves of the same frequency and phase is given by: 2${\sin }^2\omega t=\frac{e^{\mathrm{j\omega }t}-e^{-\mathrm{j\omega }t}}{2j}\times \frac{e^{\mathrm{j\omega }t}-e^{-\mathrm{j\omega }t}}{2j}=\frac{e^{2\mathrm{j\omega }t}-2e^0+e^{-2\mathrm{j\omega }t}}{-4}=-\frac{1}{2}\left(\frac{e^{j\left(2\omega \right)t}+e^{-j\left(2\omega \right)t}}{2}\right)+\frac{1}{2}.$

[0012] Digital receivers implement this operation by mixing an incoming sinusoidal signal with an oscillator output. As the equations above show, the result is a sinusoidal output having a frequency double that of the input, and an amplitude half that of the input, superimposed on a DC offset of half the input amplitude.

[0013] Similarly, multiplying sin (ωt) by cos (ωt) gives: 3$\sin \omega t\times \cos \omega t=\frac{e^{2\mathrm{j\omega }t}-e^{-2\mathrm{j\omega }t}}{4j}=\sin 2\omega t.$

[0014] The result is an output sinusoid having a frequency double that of the input, with no DC offset.

[0015] It can be seen that multiplying the cosine wave by any phase-shifted sine wave yields a “demodulated” waveform with an output frequency double that of the input frequency, whose DC offset varies according to the phase shift, φ: 4$\begin{array}{c}\sin \omega t\times \sin \left(\omega t+\phi \right)=\frac{e^{\mathrm{j\omega }t}-e^{-\mathrm{j\omega }t}}{2j}\times \frac{e^{j\left(\omega t+\phi \right)}-e^{-j\left(\omega t+\phi \right)}}{2j}\\ =\frac{e^{j\left(2\omega t+\phi \right)}-e^{j\left(\omega t-\omega t-\phi \right)}-e^{j\left(\omega t+\phi -\omega t\right)}+e^{-j\left(2\omega t+\phi \right)}}{-4}\\ =\frac{\cos \left(2\omega t+\phi \right)}{-2}-\frac{e^{\mathrm{j\phi }}+e^{-\mathrm{j\phi }}}{-4}\\ =\frac{\cos \left(2\omega t+\phi \right)}{-2}+\frac{\cos \phi }{2}\\ =\frac{\cos \phi }{2}-\frac{\cos \left(2\omega t+\phi \right)}{2}\end{array}$

[0016] Thus, a carrier to which a varying phase shift is applied can be demodulated into a varying output voltage by multiplying the carrier with a sinusoidal output from a local oscillator and filtering out the high-frequency component. Unfortunately, the phase shift detection is limited to two quadrants; a phase shift of π/2 cannot be distinguished from a phase shift of −π/2. Therefore, to accurately decode phase shifts present in all four quadrants, the input signal needs to be multiplied by both sinusoidal and cosinusoidal waveforms, the high frequency filtered out, and the data reconstructed. Expanding on the equations above: 5$\cos (\omega t)\times \sin \left(\omega t+\phi \right)=\frac{e^{\mathrm{j\omega }t}+e^{-\mathrm{j\omega }t}}{2}\times \frac{e^{j\left(\omega t+\phi \right)}+e^{-j\left(\omega t+\phi \right)}}{2j}=\frac{e^{j\left(2\omega t+\phi \right)}-e^{j\left(-\phi \right)}+e^{j\left(\phi \right)}-e^{-j\left(2\omega t+\phi \right)}}{4j}=\frac{\sin \left(2\omega t+\phi \right)}{2}+\frac{\sin \phi }{2}$

[0017] However, removing the data from the carrier is not a simple process of low-pass filtering the output of the mixer and reconstructing four voltages back into logic levels. In practice, exactly synchronizing a local oscillator at the receiver with an incoming signal is not easy. If the local oscillator differs in phase from the incoming signal, the signals on the phasor diagram will undergo a phase rotation of a magnitude equal to the phase difference. Moreover, if the phase and frequency of the local oscillator are not fixed with respect to the incoming signal, there will be a continuing rotation on the phasor diagram. Therefore, the output of the front-end demodulator is normally fed into an analog-to-digital (A/D) converter, and any rotation resulting from errors in the phase or frequency of the local oscillator is removed in digital signal processing.

[0018] The Advanced Television Systems Committee (“ATSC”) has selected vestigial sideband (“VSB”) modulation as the transmission standard for digital television (“DTV”). In the ATSC standard, 8 VSB is the standard for terrestrial broadcast, while 16 VSB is used for cable transmission. (The International Telecommunications Union (“ITU”) standard defines five VSB modes: 2, 4, 8, 16, and 8T.)

[0019] Typically, 8 VSB uses three supplementary signals for synchronization. First, it uses a low-level RF pilot for carrier acquisition. Second, as shown in FIG. 1, a four-symbol data-segment sync is used once every 832 symbols—that is, once each segment—for synchronizing the data clock in both frequency and phase. (Typically, the four symbols are [1, −1, −1, 1], normalized.) Finally, an 832-symbol data-frame sync is used once every 313 segments for data framing and equalizer training. The data-frame sync also includes information identifying the signal as either 8 VSB, 16 VSB, or one of the other appropriate ITU modes.

[0020] The pilot signal has 0.3 dB power. Although the pilot recovery is typically reliable, it can fail under certain circumstances, such as strong, close-in, slow-moving multipathing situations.

[0021] Symbol clock recovery from the segment sync signal is relatively slow, and depends on successful carrier recovery and segment location recovery. Furthermore, although the segment sync signal is typically reliable once carrier recovery and segment location recovery are successfully performed, it can still fail under certain circumstances, including the kind of multipathing that might destroy the pilot signal (and even in specific instances where the pilot signal has not been affected by the multipathing). Because this kind of multipathing is relatively common in urban environments, where broadcast digital transmission is likely to be desirable, resolving this problem is important to the commercial development of digital television, and to the improvement of other digital transmission systems.

[0022] Therefore, a new system and method for symbol clock recovery is needed that can perform symbol clock recovery from an 8 VSB signal even when the segment sync signal is totally destroyed or severely altered, and that is independent of carrier recovery and segment location recovery. The present invention is directed towards meeting these needs, among others.

BRIEF DESCRIPTION OF THE DRAWINGS

[0023] FIG. 1 is a diagram of certain features of an 8 VSB data segment.

[0024] FIG. 2 is a frequency-domain diagram showing certain features of a typical VSB signal.

[0025] FIG. 3 is a block diagram of a circuit for carrier recovery according to the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

[0026] For the purpose of promoting an understanding of the principles of the invention, reference will now be made to the embodiment illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended. Alterations and modifications in the illustrated device, and further applications of the principles of the invention as illustrated herein are contemplated as would normally occur to one skilled in the art to which the invention relates.

[0027] A symbol clock recovery system according to the present invention provides robust recovery, even in an urban environment, where ghosts due to multipath interference are common. Prior art systems have generally used the segment sync signal for clock recovery. The symbol clock recovery of the present invention uses the band edges of the signal, so it is independent of segment sync, making it both faster and more robust than recovery from the clock segment sync of prior art systems. Furthermore, because the symbol clock recovery is independent of segment sync, it can be completed earlier in the demodulation process, which can, in turn, improve the performance of other parts of demodulation.

[0028] FIG. 2 shows certain features of the spectrum of a VSB signal, shown generally at 100. In this example, the primary portion 210 of the signal 200 is 5.38 MHz wide, including an unattenuated portion 205 within the 3 dB attenuated portion 210. However, the amplitude is not completely damped outside the main frequency domain. A substantial signal exists in this example for an additional 0.31 MHz above and below the primary portion 210 of the signal, this full band being indicated at 215. These “band edges” can be used for carrier recovery, as discussed hereinbelow.

[0029] FIG. 3 is a block diagram of a circuit according to the present invention, shown generally at 300. A signal is input to the circuit 300 at 301 from an A/D converter (not shown) preferably running at twice the symbol rate. It will be appreciated that sampling at twice the symbol rate is sufficient to satisfy the Nyquist condition. This upstream AID converter can sample its input signal at greater than twice the symbol rate, but increases in the hardware frequency beyond this point result in increases in the hardware cost without a corresponding increase in performance. The circuit 300 comprises a digitally controlled oscillator (“DCO”) 310, which produces two signals: sin(ωn), and cos(ωn), where “n” is the symbol count and ω=2π/ƒ. A first multiplier 302 multiplies the input signal by the cos(ωn) signal, and a second multiplier 304 multiplies the input signal by the sin(ωn) signal. The signals from the first and second multipliers 302 and 304 are then passed through first and second root-raised cosine (“RRC”) filters 320 and 330, respectively. The output of the first RRC filter 320 is multiplied by sin(πn/4) at a third multiplier 322, and by cos(πn/4) at a fourth multiplier 324. The output of the second RRC filter 330 is likewise multiplied by sin(πn/4) at a fifth multiplier 332, and by cos(πn/4) at a sixth multiplier 334.

[0030] The output of the sixth multiplier 334 is subtracted from the output of the third multiplier 322 by a first accumulator 340, and added to the output of the third multiplier 322 by a third accumulator 360. The output of the fifth multiplier 332 is subtracted from the output of the fourth multiplier 324 by a second accumulator 250, and added to the output of the fourth multiplier 324 by a fourth accumulator 370. The output of the second accumulator 350 is passed through a first low-pass infinite impulse response (“IIR”) filter 348, preferably having a −3 dB attenuation at about 70 KHz to filter out high-frequency components beyond the band edge.

[0031] The output of the IIR filter 348 passes through a first limiter 346. The first limiter 346 assigns a value of 1 to any positive input, and a value of −1 to any negative input. (Those skilled in the art will recognize this as a sign( ) function.) The output of the first limiter 346 is multiplied by the output of the first accumulator 340 using a seventh multiplier 380. It will be appreciated by those skilled in the art that the output of the seventh multiplier 380 has been multiplied by two RRC filters, so that the signal has been effectively multiplied by a plain raised cosine filter. Thus, the output of the seventh multiplier 380 represents the frequency and phase correction information obtained from the lower band edge.

[0032] The output of the fourth accumulator 370 is passed through a second low-pass IIR filter 368, preferably having a −3 dB attenuation at 70 KHz to filter out high-frequency components beyond the band edge. The output of the second low-pass IIR filter 368 passes through a second limiter 366. Like the first limiter 346, the second limiter 366 assigns a value of 1 to any positive input, and a value of −1 to any negative input. The output of the second limiter is multiplied by the output of the third accumulator 360 using an eighth multiplier 390. It will be appreciated that the output of the eighth multiplier 390 represents the frequency and phase correction information obtained from the upper band edge.

[0033] The output of the seventh multiplier 380 is then multiplied by a weight factor “k” using a ninth multiplier 385. The output of the eighth multiplier 390 is subtracted from the output of the ninth multiplier 385 by a fifth accumulator 395. The output of the fifth accumulator 395 is then passed through a third low-pass IIR filter 397 to generate the symbol clock adjustment signal 399, which is then returned to the symbol clock to complete the feedback loop.

[0034] Those skilled in the art will recognize that the lower band edge of a VSB signal contains the pilot signal. This is the reason for the weight factor applied by the ninth multiplier 385. Typically, when k is about 0.3 the upper and lower band edge contributions will be properly balanced.

[0035] It will further be appreciated that, because the frequency and phase information from the lower band edge is contained in the output of the ninth multiplier 385 and the frequency and phase information from the upper band edge is contained in the output of the eight multiplier 390, the output of the fifth accumulator is driven to zero when the upper and lower band edges are balanced, so that the output of the third low-pass IIR filter 397 can be used to complete a feedback loop that provides the symbol clock recovery.

[0036] Variations in the implementation of the invention will occur to those of skill in the art. For example, some or all of the generation and calculation of signals can be performed by application-specific or general-purpose integrated circuits, or by discrete components, or in software.

[0037] While the invention has been illustrated and described in detail in the drawings and foregoing description, the same is to be considered as illustrative and not restrictive in character, it being understood that only the preferred embodiment has been shown and described and that all changes and modifications that come within the spirit of the invention are desired to be protected.

Claims

1. A method of demodulating a received signal, comprising: receiving a stream of digital data comprising a sequence of data elements st representing the received signal sampled according to a symbol clock; selecting a local frequency ƒ; determining at=sin(πt/4)RRC(st cos(2πt/ƒ)); determining bt=cos(πt/4)RRC(st cos(2πt/ƒ)); determining ct=cos(πt/4)RRC(st sin(2πt/ƒ)); determining dt=sin(πt/4)RRC(st sin(2πt/ƒ)); providing a first output signal vt=bt+dt; and providing a second output signal ct=L3(k(at−ct)(sign(Lt(bt−dt)))−(at−ct)(sign(L2(bt+dt)))); wherein RRC is a root-raised cosine filter; and L1, L2, and L3 are infinite impulse response, low-pass filters having a predetermined pass band.

2. The method of claim 1, further comprising adjusting the symbol clock responsively to the second output signal.

3. A system for processing a received signal having an expected center frequency at 0, a 0 dB bandwidth b0, and a −3 dB bandwidth b3, comprising: an analog-to-digital converter configured to sample the received signal; and a digital signal processing means for generating a clock adjustment signal as a function of the frequency-domain components of the received signal having frequencies f1 and fh, such that ((b0/2)−b3)<ft<−(b0/2), and (b0/2)<fh<(b3−(b0/2)).

4. The system of claim 3, further comprising a clock responsive to the analog-to-digital converter to control the frequency and phase of sampling.

5. A method of demodulating a received signal, comprising: receiving a stream of digital data comprising a sequence of data elements representing the received signal sampled according to a clock, where the clock is subject to adjustment in frequency and/or phase by a clock adjustment signal; multiplying the sequence of data elements by a digital cosine wave of the target frequency, and passing the result through a first raised-root cosine filter to yield a first intermediate sequence; multiplying the sequence of data elements by a digital sine wave of the target frequency, and passing the result through a first raised-root cosine filter to yield a second intermediate sequence; multiplying the first intermediate sequence by a digital sine wave of one-quarter the target frequency to yield a third intermediate sequence; multiplying the first intermediate sequence by a digital cosine wave of one-quarter the target frequency to yield a fourth intermediate sequence; multiplying the second intermediate sequence by a digital cosine wave of one-quarter the target frequency to yield a fifth intermediate sequence; multiplying the second intermediate sequence by a digital sine wave of one-quarter the target frequency to yield a sixth intermediate sequence; subtracting the fifth intermediate sequence from the third intermediate sequence to yield a seventh intermediate sequence; subtracting the sixth intermediate sequence from the fourth intermediate sequence to yield an eighth intermediate sequence; obtaining a ninth intermediate sequence as the product of a predetermined constant k; the seventh intermediate sequence; and the sign of the result of passing the eighth intermediate sequence through an infinite-impulse-response, low-pass filter; adding the third intermediate sequence and the fifth intermediate sequence to yield a tenth intermediate sequence; adding the fourth intermediate sequence and the sixth intermediate sequence to yield an eleventh intermediate sequence; obtaining a twelfth intermediate sequence as the product of the tenth intermediate sequence; and the sign of the result of passing the eleventh intermediate sequence through an infinite-impulse-response, low-pass filter; adding the ninth intermediate sequence and the twelfth intermediate sequence to yield a thirteenth intermediate sequence; and adjusting the clock as a function of the result of passing the thirteenth intermediate sequence through an infinite-impulse-response, low-pass filter.