The present invention relates to a method and an apparatus to cancel additive sinusoidal disturbances of a known frequency in OFDM receivers.
Orthogonal frequency division multiplexing (OFDM) has become a popular transmission method for high-speed wireless radio transmission, due to its potential for low complexity of transmitters and receivers, paired with robustness under severe multi-path conditions. A more detailed discussion on OFDM in found in S. B. Weinstein and P. M. Ebert: Data transmission by frequency-division multiplexing using the discrete Fourier transform. IEEE Trans. Communication Technology, COM-19(5):628-634, October 1971. The wired counterpart, known as discrete multi-tone (DMT) employs similar techniques. The transmitter uses an inverse discrete Fourier transform (IDFT) to generate a multi-carrier signal, and the receiver applies the Discrete Fourier Transform (DFT) to demodulate the data.
Integrated radio receivers need a large gain and a low noise figure to achieve a high sensitivity. Clock signals which are present for frequency reference, mixer control, and A/D converter control, as well as harmonics and mixing products of these periodic signals, may couple into some point in the receiver chain and appear as rotating complex exponentials superimposed to the complex baseband receive signal. If the level of such tones becomes too high, they may degrade the receiver sensitivity. The frequencies of such disturbing tones originating from the RF receiver itself are directly related to the clock frequencies occurring in the receiver.
As stated above, unwanted tones superimposed to the received signal may reduce the receiver sensitivity. The safest approach to prevent this problem is to directly avoid the occurrence of such tones. Even the coupling mechanism may be known and a re-spin of the receiver design may be able to reduce the coupling. However, in highly integrated receiver systems the effort to achieve this can be quite high, possibly requiring detailed modelling, design modifications and additional verification.
A general object of the present invention, therefore, is to mitigate such additive disturbing tones in an OFDM baseband receiver, while achieving low implementation complexity.
According to an aspect of the present invention there is provided a method for cancelling additive sinusoidal disturbances in OFDM receivers. According to a further aspect of the present invention there is provided an apparatus. The inventive method and apparatus obtain an estimation of an amplitude and phase of a disturbing superimposed tone, whose frequency is known, and use such amplitude and phase estimation values to cancel the tone such that receiver sensitivity degradation is avoided.
In accordance with the invention the implementation is made in a way to achieve a low complexity, which translates into low overhead power consumption in applying the method.
Additional features and advantages of the present invention will be apparent from the following detailed description of specific embodiments which is given by way of example and in which reference will be made to the accompanying drawings, wherein:
Before referring to
fT is the frequency of the disturbing tone, normalized to the sampling frequency;
NDFT is the length of the discrete Fourier transform in samples;
NGuard is the length of the guard interval in samples;
NSym=NDFT+NGuard is the number of time-domain samples per OFDM symbol;
k is the sampling time index;
y(k)=r(k)+z(k) is the complex baseband receive signal input into OFDM demodulator 40; with
r(k) being the actual receive signal including other disturbances like noise; and
z(k)=AT·exp(j2π·fT·k+φT) being the disturbing superimposed complex exponential; with
fT being the known frequency, and AT and φT being amplitude and phase, respectively, of the disturbing complex exponential, which are to be estimated. Assuming that 0≦fT<1, the periodic spectrum of a digital signal allows to map any possible tone onto this range.
Transformation of a complex exponential z(k)|0≦k<NDFT via DFT yields the values
where
n denotes the element index in the resulting vector.
Rewriting this equation as
Z(n)=[AT·exp(j·φT)/NDFT]·W(n)
splits it into the amplitude/phase factor (AT·exp(j·φT)/NDFT) and the weighting pattern
which is only determined by the frequency of the disturbing tone (when treating NDFT as given). We assume that the frequency of the disturbing tone is known, and we need to estimate the amplitude and the phase of the tone.
Now consider the receive signal after DFT, which ideally consists only of a superposition of data symbols disturbed by the channel fading and additive noise. Let y(k)|K·NSym≦k<K·NSym+NDFT denote the DFT input samples of OFDM symbol number K and
YK(n) denote the associated DFT output vector, with 0≦n<NDFT.
If no additive complex exponential is present, we assume that the output of all OFDM sub-carriers during reception is a zero-mean random process, i.e.,
E{YK(n)}=0∀K,n.
Furthermore we assume that distinct DFT output symbols are statistically independent, i.e.,
E{YKn1≠n2
Three key ideas are applied for estimation of a superimposed disturbing complex exponential:
An offset phasor FK=exp(j2π·fT·NSym), here with the constant FK=F∀K is input to an offset phasor accumulation unit 510 and is cumulatively multiplied to obtain a sequence of start phasors
RK is fed into a back rotation unit 520 where the complex conjugates of these values are multiplied with the amplitude/phase estimator output values
from an amplitude and phase estimation unit 570.
Here {tilde over (W)}K(n) is the estimation pattern, which equals W(n) in a first embodiment of the invention, which may however be simplified in another embodiment, as explained below. Furthermore, the estimation pattern may vary from OFDM symbol to OFDM symbol, which is denoted by the index K.
The obtained back-rotated amplitude/phase estimates QK=RK·P*K are fed into an Infinite Impulse Response (IIR) linear low-pass filter 530 with a DC gain of one (“History averaging”), controlled by the factors cK with 0<cK<1. In a first embodiment of the invention, the factors cK are constant over time, irrespective of K.
The output values
The output value
ZK(n)=YK(n)−VK(n).
In another embodiment of the invention the condition E{YK(n)}=0 may not be satisfied for some pairs (K,n), which is the case if pilot tones are included in the OFDM signal. To prevent the amplitude/phase estimate from becoming biased, the concerned pairs (K,n) shall not be considered in the estimator.
This is achieved by a modified arrangement shown in
In the arrangement of
Thus, the functionality of pilot symbol replacement unit 580 may be described by the equation
Another embodiment of the invention exploits a degree of freedom in the choice of the pattern W(n), which is to rotate the phase of the entire vector in the complex plane, in order to obtain real-valued coefficients W(n), which reduces the computational complexity. This can also be achieved using the equation
In still another embodiment of the invention the complexity of the amplitude/phase estimator 570 is reduced by exploiting the fact that most of the energy of the disturbing rotating exponential of known frequency is concentrated on a few bins at the DFT output. In this embodiment only a subset of DFT output bins, indexed by the set
NEst={n1, n2, . . . , nN
is used, and the estimation pattern is determined by
Here, the subscript K indicates that W(n) may vary from OFDM symbol to OFDM symbol. The set NEst is typically defined such as to collect most of the energy with a limited number of bins, which is achieved by using only the coefficients with the largest absolute values in W(n). In an extreme case, only a single value out of W(n) is used.
In still another embodiment of the invention the complexity of the pattern weighting/spur subtraction units, 550 and 560, respectively, is reduced by exploiting the fact that most of the energy of the disturbing rotating exponential of known frequency is concentrated on a few bins at the DFT output, eliminating the need to subtract negligibly small disturbances. In this embodiment, only a subset of DFT output bins indexed by a set
NCancel={n1, n2, . . . , nN
is used, and the cancellation pattern is defined as
Again the subscript K indicates that W(n) may vary from OFDM symbol to OFDM symbol. The set NCancel is typically defined to apply to all elements in W(n) where an unacceptable excessive disturbance is expected to occur. In an extreme case, only a single value out of W(n) is addressed.
In another embodiment of the invention, a fast ring-in of the history averaging low-pass is realized by time-variation of the filter coefficients cK. For example, when the first amplitude/phase estimate is performed at OFDM symbol with K=1, a good choice of a sequence is
This results in an equal weighting of all incoming samples until the history averaging low-pass has rung in. After ring-in, weighting of filtered samples decays exponentially over time.
In another embodiment of the invention, each vector of samples subjected to DFT is first cyclically shifted before the DFT is processed, due to the OFDM receiver design. For a cyclic shift by NShift samples, the weighting pattern becomes
All other principles of the invention are applied as described before.
In another embodiment of the invention the frequency of the disturbing tone changes over time, possibly due to some adaptation of the mixer frequency in the radio front-end. To cope with this, the offset phasor FK as well as the estimation pattern {tilde over (W)}K(n) and the cancellation pattern ŴK(n) are adapted accordingly.
In another embodiment of the invention, where multiple disturbing sinusoids shall be cancelled, a plurality of spur cancellers, as described above, may be implemented. In this case all amplitude/phase estimations are performed in parallel on the DFT output data, whereas the subtractions of the estimated tones occur sequentially, tone by tone.
As an example, consider a DVB-H receiver implementation with NDFT=4096, NGuard=1024, NSym=5120, NShift=512, with a sampling frequency fsample=48/7 MHz, which is disturbed by a spurious tone at a frequency fSpur=1 MHz. The normalized frequency of the tone is fT=fSpur/fSample=7/48. The tone frequency corresponds with the OFDM sub-carrier index nT=fT·NDFT=597⅓. The offset phasor is determined as
The estimation pattern is defined as
the cancellation pattern is defined as
and the minimum filter constant after ring-in is set to
Applications of the Invention
The various embodiments of the invention as detailed above may be applied separately or in combination in an OFDM receiver for wireless or wired transmission including, but not limited to, receivers in wireless local area network (WLAN) applications, e.g., according to the IEEE 802.11 standard, in wireless personal area network (WPAN) applications, e.g., according to the IEEE 802.16 standard, in digital TV receivers for, e.g., DVB-T, DVB-H, T-DMB, DMB-T, DAB, in ultra-wideband (UWB) receivers according to the multi-band OFDM alliance (MBOA) standard proposal, etc.
Number | Date | Country | Kind |
---|---|---|---|
08100957 | Jan 2008 | EP | regional |
08103777 | Apr 2008 | EP | regional |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/IB2009/050150 | 1/15/2009 | WO | 00 | 7/19/2010 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2009/093156 | 7/30/2009 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
5410750 | Cantwell et al. | Apr 1995 | A |
6052421 | Richardson et al. | Apr 2000 | A |
20040218520 | Aizawa | Nov 2004 | A1 |
20080273615 | Song et al. | Nov 2008 | A1 |
20100322364 | Bogdan | Dec 2010 | A1 |
Number | Date | Country |
---|---|---|
1310894 | Aug 2001 | CN |
1937429 | Mar 2007 | CN |
1 349 337 | Oct 2003 | EP |
2000-165343 | Jun 2000 | JP |
2004-140739 | May 2004 | JP |
2004-328440 | Nov 2004 | JP |
0011823 | Mar 2000 | WO |
0019598 | Apr 2000 | WO |
2004114617 | Dec 2004 | WO |
Entry |
---|
Weinstein, S. B., et al; “Data Transmission by Frequency-Division Multiplexing Using the Discrete Fourier Transform,” IEEE Trans. Communication Technology, COM-19(5), pp. 628-634 (Oct. 1971). |
Number | Date | Country | |
---|---|---|---|
20100296568 A1 | Nov 2010 | US |