METHOD AND APPARATUS FOR CHANNEL ESTIMATION IN OFDM

RECEIVERS

CROSS-REFERENCES TO RELATED APPLICATIONS [0001] The present application claims benefit under 35 USC 1 19(e) of U.S. provisional application number 60/829131 , filed October 1 1 , 2006, entitled "Method And Apparatus For Channel Estimation In OFDM Receivers ^" , the content of which is incorporated herein by reference in its entirety.

BACKGROUND OF THE INVENTION

|0002] Signal processing systems such as communication receivers often must recover a desired signal which has been transmitted through a channel degraded by multipath. In order to compensate for the signal impairment introduced thereby, receivers can use signal processing techniques which estimate the channel conditions. However, this poses challenges for channels which are changing quickly, as is the case, for example, when the receiver is mobile and moving at high speed, i.e., due to high Doppler conditions.

[0003] Orthogonal frequency division multiplexing (OFDMA) systems, such as DVB-T often provide pilot tones for the purpose of making channel estimation easier. However, the sparseness of these tones renders it difficult to estimate the channel quickly and with efficient memory usage and calculations.

[0004] Figure 1 is a high-level block diagram of an OFDM system which employs channel estimation, as known in the prior art. Data modulated by modulator 120 and pilot tones 122 are inserted in inverse Fourier transform block 102. Block 104 adds cyclic prefix to the output of inverse Fourier transform block 102 and supplies its output to Rayleigh channel 106. The pilot tones are interspersed periodically in the subchannels to enable channel estimation block for the Rayleigh channel 106. Signal s _m received by Rayleigh channel 106 for transmission may be defined as shown below:

s _m (t) = mT _s ≤ t < (m ₊ l)T _s (1) [0005] After the Rayleigh channel, the received signal is defined as below:

where the channel response /?(/) is defined as: h(t, τ) = ∑χ _k (t)δ(τ - τ _k ) (3)

[0006) In the frequency domain, the received signal can be expressed as

^X l «, = ^H l n, ^a i n, + »/ „, ( ⁴ ) where

H. ^ ∑λe-' ² *" (5)

is the frequency response and w, _m represents the Additive White Gaussian Noise (AWGN).

For the pilot subcarriers, the temporal channel can be obtained in accordance with the following expression:

H _{1 n} , = x, _m a _{l m} = H, _m + * _{l m} a _{l m} (6)

[0007] Block 108 removes the cyclic prefix from the received signal. Fast Fourier transform block 1 10 converts the time domain signal supplied by block 108 to a frequency domain signal. Block 1 12 extracts the pilot symbols from the signal supplied by block 1 10 to estimate the channel. In the above representation, the channel is assumed to be an ideal channel plus noise. To estimate the channel, the noise is suppressed and pilot interpolation is performed. [0008] Figure 2 shows how pilot channels are distributed among subchannels in the DVB- T/H system. DVB-T supports 2K and 8K OFDM subchannels, whereas DVB-H also supports a mode with 4K subchannels. With each successive symbol, the location of most of the pilot tones (alternatively referred to herein as pilots) changes, with the exception of the pilot tones at locations called continuous pilot locations. In this system, the pilots are inserted every 12 subcarriers in the frequency domain; this density of pilot tones is insufficient for noise suppression. One method for increasing the density of channel estimates is to use pilot tones located nearby in time and frequency to interpolate the values in between. This can be done using previous and future OFDM symbols to fill the pilots from 1/12 total subcarrier density to 1/3 total subcarrier density. However, conventional interpolation techniques are inefficient

and require excessive amount of memory space to store data required to pertorm such interpolations.

BRIEF SUMMARY OF THE INVENTION

5 |0009] In accordance with one embodiment of the present invention, to estimate a channel associated with 2k and 4k modes, diagonal pilot positions are first used to obtain the pilots at the center between each two original pilots in accordance with the expression shown below:

P = C P +CP

10 [0010) Subsequently, using two pilots, linear interpolation is performed in accordance with the expressions shown below:

P = C P +CP P - C P -4- C P

15 P -CP +CP

/ = 3 * (/ mod 4) + 12n , for n integer, « ≥ 0.

where / represents an index of a symbol and / represents the position of the pilot tones within the symbol. 20 [0011] In accordance with another embodiment of the present invention, to estimate a channel associated with an 8k mode, the following expressions are used to interpolate pilot positions:

P = C P +CP

25 P =C P +CP

P = C P +CP

Where

3 ^J 0 ^{U J} P/+l ;+9 = C ^-1 ^J P /+1 ! + ^■ >, + ^τ CP

z = 3 * (/ mod 4) + 12« , for n integer, n ≥ 0

BRIEF DESCRIPTION Oh I Hh URAWlNUS

[0012] Figure 1 is a high-level block diagram of an OFDM system which uses channel estimation, as known in the prior art.

[0013] Figure 2 shows the distribution of pilot channels among subchannels in accordance with the DVB-T/H system, known in the prior art.

[0014] Figure 3 shows the pilot symbols used to perform interpolation, in accordance with one embodiment of the present invention.

[0015] Figure 4 shows the pilot symbols used to perform interpolation, in accordance with another embodiment of the present invention. [0016] Figure 5 shows the pilot symbols used to perform interpolation, in accordance with another embodiment of the present invention.

[0017] Figure 6 shows the pilot symbols used to perform interpolation, in accordance with another embodiment of the present invention.

[0018] Figure 7 shows various blocks of a pilot interpolation system, in accordance with one embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

[0019] A channel estimation system and method, in accordance with the present invention, is highly efficient and provides good performance under high Doppler conditions, particularly suitable for OFDM systems such as DVB-T. A number of different pilot symbols interpolation techniques are disclosed that enable the estimation of channel OFDM systems (such as DVB-T), well suited for mobile conditions via which signals are transmitted. The present invention may be used in digital receivers in standards which use OFDM systems. Such systems include, but are not limited to, DVB-T, ISDB-T, T-DMB, or DMB-T/H. The present invention enables reception of, for example, digital television signals based on these standards while the viewer is moving at high speeds relative to the transmitter.

[0020] In the following, both causal and non-causal interpolation techniques are described to increase pilot density in the frequency domain. Each techniques uses only a few pilot tones, resulting in simplicity of calculation, efficient implementation, and consumption of less memory space. In the following: (i) symbol λ is used to indicate an estimated pilot

symbol, (ii) black circles represent pilot tones, (in) white circles represent data tones (iv) cross-hatched circles represent estimated symbols, (v) / is the index of the symbol shown along the y-axis, (vi) and / is the pilot position in the current symbol shown along the x-axis.

Interpolation Technique No. 1

[0021] In accordance with one embodiment of the present invention and as shown in Figure 3, previous and future pilots in time domain are used to do linear interpolation, in accordance with the following expressions:

^ ₁₊₁ = C ₁ ^ _{1 1+1} + C ₂ ZJ _{+1 1+1} (7) P, , _+b = C ₁ ^ _{2 1+6} + C ₄ P ₁₊₂ , _+b (8)

P — C P + C P (9) i = 3 * (/ mod 4) + 12/2 , for n integer, n ≥ 0.

[0022] In other words, pilot position / + 3 of a current symbol, i.e. P _{1 1+1} , is estimated as being equal to a weighted average of the similar pilot positions of the succeeding symbol (the symbol whose index is one higher than the index of the current symbol), i.e. P _{l+] ;+1} , and the third preceding symbol (the symbol whose index is one less than the index of the current symbol by 3), i.e. P ₁ ^ _i+ , . For example, pilot 204 ( /J ₁₊₁ ) is estimated by interpolating pilot symbols 206 ( P _{l+] i+} , ) and pilot symbol 202 ( P _t _ _{λ i+} , ) in accordance with expression (7) shown above.

[0023] Pilot position / + 6 of a current symbol, i.e. P _{1 1+6} , is defined as being equal to a weighted average of the similar pilot positions of the second succeeding symbol (the symbol whose index is two higher than the index of the current symbol), i.e. P _{1+2 i+b} , and the second preceding symbol (the symbol whose index is two less than the index of the current symbol), i.e. P _{1-2 ι+b} . For example, pilot 214 ( P _{1 ι+b} ) is estimated by interpolating pilot symbols 216 ( P _{1+2 ι+ύ} ) and pilot symbol 202 ( P _t _ _{2 ι+b} ) in accordance with expression (8) shown above.

[0024] Pilot position i + 9 of a current symbol, i.e. P _{1 /+9} , is defined as being equal to a weighted average of the similar pilot positions of the third succeeding symbol, i.e. P ₁₊ ^ _;+9 ,

and the preceding symbol, i.e. P ₁ , _(+l) . For example, pilot 224 (P _n+9 ) is estimated by interpolating pilot symbols 226 ( P ₁₊ ^ _i+9 ) and pilot symbol 222 ( P ₁ ^ _ι+9 ) in accordance with expression (9) shown above. If the interpolated pilot position is also a continuous pilot position, the result of the interpolation is ignored and the transmitted continuous pilot is used as received.

[0025] In one exemplary embodiment, coefficients C ₁ , C ₂ , C, , C ₄ , C ₅ and C _b have the values shown below:

^ ₁₊ , =0.25/». _3l+3 +0.75/ ^> _+1I+3 ^ _/(+6 =0.5^ ₂₍₊₆ +0.5^ ₊₂₍₊₆ / ^> _I+9 =0.75/ ^> _ ₁₍₊₉ +0.25i ^> _+3l+9

/ = 3 * (/ mod 4) + 12«, for n integer, n ≥ 0

Interpolation Technique No.2

[0026] In accordance with another embodiment of the present invention and as shown in Figure 4, first diagonal pilot positions are used to obtain the pilots at the center between each two original pilots in accordance with the expressions (10) shown below:

L, ₊ ,= C ₂ P ₁₊₂ , _+b +C,P _n+]2 (10)

[0027] The diagonal interpolation is a linear interpolation. The pilots so obtained are shown as dotted circles. Subsequently in time domain, using three pilots, polynomials are used to do perform interpolation, in accordance with the expressions shown below.

¹ P I , + 1 - ~C *-| Q ¹ P l-λ , + λ + ^T C H r P/-l j + 3 + ^C ^l 2- P"/ + I i + l π V ¹ i ¹ "J>

¹ P I i+9-C *~l 2 ^J P/-1 i+9 + ^τ C ^l rP /+l ;+9 + ^τ C ^l 0 ^J P / + T , + 9 ( \M ^{l z} -)J z = 3 * (/ mod 4) + \2n, for n integer, n ≥ 0 (13)

[0028] In other words, pilot position i + 3 of a current symbol, i.e. P ₁₁₊ , is estimated as being equal to a weighted average of the similar pilot positions of the succeeding symbol, i.e. P _l+ i _I+ , , an estimate of the preceding symbol, i.e. P ₁ ^ _/+ , , and the third preceding symbol, i.e. P,_ _{λ /+1} . For example, pilot 308(/^ ₁₊₁ ) is estimated by interpolating pilot symbol 302 ( P ₁ ^ _i+1 ),

estimated pilot symbol 304 ( / ^> _, _i+ , ), and pilot symbol 306 ( P ₁ ^ _/+1 ) in accordance with expression (1 1 ) above. It is understood that during the interpolation process associated with index / , the pilot positions of all indices / of previous symbol / - 1 are known.

[0029] Pilot position i + 9 of a current symbol, i.e. P _{l l+t)} , is estimated as being equal to a weighted average of the similar pilot positions of the preceding symbol, i.e. P ₁ _ _{λ (+9} , an estimate of the succeeding symbol, i.e. P _{1+1 1 +9} , and the third succeeding symbol P ₁₊ ^ ₁₊₀ ■ For example, pilot 318 ( P _{1 J+Q} ) is estimated by interpolating pilot symbols 312 ( P ₁ ^ ₍₊₉ ), estimated pilot symbol 314 ( ^ ₊ , , ₊ ,, ), and pilot symbol 316 ( P ₁₊ ^ ₊₉ ) in accordance with expression (12) shown above. [0030] In one embodiment, coefficients C _{1 0} , C _n and C _{1 2} are respectively equal to -0.125,

0.75 and 0.375. If the interpolated pilot position is also a continuous pilot position, the result of the interpolation is ignored and the transmitted continuous pilot is used as received.

Interpolation Technique No. 3 [0031] In accordance with another embodiment of the present invention and as shown in Figure 5, first diagonal pilot positions are used to obtain the pilots at the center between each two original pilots in accordance with the expression (14) shown below:

P - C P + C P (14]

[0032] The pilots so obtained are shown as dotted circles. Subsequently in time domain, using two pilots, linear interpolation is performed in accordance with the expressions shown below:

i = 3 * (/ mod 4) + 12« , for n integer, n ≥ O .

[0033] Pilot position / + 3 of a current symbol, i.e. P _{1 ι+λ} , is estimated as being equal to a weighted average of the similar pilot positions of an estimate of the preceding symbol, i.e.,

P _f _ _{t i+1} , and a succeeding symbol, i.e. P _{1+1 /+1} . For example, pilot 404 ( /^ ₇₊₁ ) is estimated by interpolating pilot symbol 406 ( P _{1+1 /+1} ) and estimated pilot symbol 402 ( P ₁ ^ _/+1 ) in accordance with expression ( 15) shown above.

[0034] Pilot position / + 6 of a current symbol, i.e. P _{1 ι+b} , is estimated as being equal to a weighted average of pilot position i + 9 of an estimate of the preceding symbol P ₁ ^ _ι+t) and pilot position / + 3 of a succeeding symbol, i.e. P _{1+1 ;+} , . For example, pilot 414 ( P _{l l+b} ) is estimated by inteφolating estimated pilot symbol 422 ( / ^> _, ₍₊₁ ) and pilot symbol 406 ( P _{M (+1} ) in accordance with expression ( 16) above.

[0035] Pilot position / + 9 of a current symbol, i.e. P _{l l+<)} , is estimated as being equal to a weighted average of the similar pilot positions of an estimate of the preceding symbol i.e.

P ₁ ^ ₇₊₉ , and the succeeding symbol, i.e. P _{1+ 1 7+g} . For example, pilot 424 ( P ₁ , ₊₉ ) is estimated by inteφolating estimated pilot symbols 422 ( P ₁ ^ _ι+!) ) and pilot symbol 426 ( P _{l+] ι+t)} ) in accordance with expression (17) above.

[0036] Initially, the estimated channel for a preceding symbol is not available but may be obtained using a number of techniques. For example, in one technique, the receiver waits until it receives and saves symbols for each of the pilots (l-l,i), (1-1J+3), (l-l,i+6) and (1-1, i+9) locations. In other words, by knowing the estimated channel for the preceding (/-7) symbol, the present symbol (/), and two succeeding (future) symbols (1+1) and (1+2), all inteφolations can be performed. Since, in accordance with the present invention, only the indices associated with a present symbol and two succeeding symbols are stored, substantially smaller amount of memory location is required to perform the inteφolations. Following the initialization, inteφolation is performed to obtain the pilots for all bins/subcarriers, in accordance with equations (14)-(17).

[0037] In some embodiment, coefficients C ₁ - C ₈ have values varying between 0.25 and 0.75. In one specific embodiment, each of coefficients C ₁ - C ₈ has a value of 0.5.

Interpolation Technique No. 4

|0038] In the interpolation techniques 1 -3 that are applicable to 2k and 4k modes, a minimum of two future symbols is used. For 8k mode, in accordance with another embodiment of the present invention, only one future symbol is used, thereby saving additional memory space.

[0039] The following define the interpolation expressions in accordance with technique no. 4:

^ ₁₊ , = cλ, ₁₊ , + C ₂ P _{M l+λ} (18) P _{1 1+6} = Cj _u * + C _A P _{M l+λ} (19)

P = C P + C P ( 20")

Where

P _{M l+9} = C _η P _{M l+} , + C _g / ^> _+l , ₊₁ , (21)

/ = 3 * (/ mod 4) + 1 In , for n integer, n ≥ 0 (22)

[0040] Pilot position / + 3 of a current symbol, i.e. P _{1 ι+lι} , is estimated as being equal to a weighted average of the similar pilot positions of an estimate of a preceding symbol, i.e.,

P _f _ _{t ι+} , , and a succeeding symbol, i.e., P ₁ ^ _i+ , . For example, pilot 504 ( P _{1 ι+λ} ) is estimated by interpolating an estimate of preceding pilot symbol 502 (P _{1-1 /+1} ) and succeeding pilot symbol 506 ( P _{M /+} , ) in accordance with expression (18) above.

[0041] Pilot position i + 6 of a current symbol, i.e. P _{1 ι+b} , is estimated as being equal to a weighted average of pilot position i + 9 of an estimate of the preceding symbol, i.e., i}_, , ₊₉ , and pilot position / + 3 of the succeeding symbol, i.e. P _{1 +] (+} , . For example, pilot 514 ( P _{1 ι+b} ) is estimated by interpolating pilot symbol 522 ( / ^> _, _;+g ) and pilot symbol 506 ( P _{M ι+1} ) in accordance with expression (19) shown above.

[0042] Pilot position i + 9 of a current symbol, i.e. P _n+9 , is estimated as being equal to a weighted average of the similar pilot positions of an estimate of a preceding symbol, i.e.

P,_ _{t ι+9} , and an estimate of the succeeding symbol, i.e. P _{1+1 ι+9} . For example, pilot 524 ( P _n+9 ) is estimated by interpolating pilot symbols 522 ( /}_, _{M 9} ) and pilot symbol 526 ( P _{1+1 ι+9} ) in accordance with expression (20) above. As seen from expression (21), / ^> _, _ι+g is a weighted average of P _{l+] i+1} and P _{1+ 1 1+ I} , . In some embodiments, coefficients C ₁ - C ₈ are equal to 0.5. |0043] In some embodiment, coefficients C ₁ - C ₈ have values varying between 0.25 and 0.75. In one specific embodiment, each of coefficients C ₁ - C ₈ has a value of 0.5.

[0044] If (/ mod 4) = 1 , a pilot position at the end of the symbol is interpolated as:

/ ^> ₆₈₁₃ = 0.667/} _{6816 +} 0.333> _{/ 6ij07} (23)

|0045] If (/ mod 4) = 3 , a pilot position at the start of the symbol is interpolated as:

P _{1 3} = 0.661P _{1 0} + O.333P, , (24)

[0046] Initially, the estimated channel for a preceding symbol is not available but may be obtained using a number of different techniques. For example, in one technique, the receiver waits until it receives and saves four symbols for each of the pilots (l-l,i), (l-l,i+3), (l-l,i+6) and (l-l, i+9) locations. Once, the initialization is carried out, the above interpolations, as shown in equations (18)-(22), are performed to estimate the channel.

[0047] Figure 7 is a block diagram of a pilot interpolation block 600 disposed in a receiver and adapted to interpolate pilot symbols in accordance with any of the techniques described above. Pilot interpolation block 600 is shown as including, in part, a memory 610 and a processor 620. Memory 610 is adapted to store data representative of pilot symbols that are used to perform interpolation. Processor 620 retrieves the data stored in memory 610 to perform pilot interpolation and channel estimation in accordance with any of the expressions described above. Processor 600 may be a general central processing unit (CPU) commercially available or may be a dedicated logic block adapted to perform the interpolations described above.

[0048] The above embodiments of the present invention are illustrative and not limiting. Various alternatives and equivalents are possible. Other additions, subtractions or modifications are obvious in view of the present disclosure and are intended to fall within the scope of the appended claims.