Noise whitening in a WLAN receiver
View Patent ↗In a method of whitening noise in a signal received by a receiving device from a transmitting device via a multi input multi output (MIMO) communication channel, a data unit is received at the receiving device via the MIMO communication channel. A channel estimate H corresponding to the MIMO communication channel is determined based on the training signals. A noise scaling factor s is determined. The noise scaling factor s is a ratio of variance of noise at the receiving device to a variance of noise at the transmitting device. A noise whitening matrix W is determined using the channel estimate H and the noise scaling factor s. The noise whitening matrix W is applied to (i) the channel estimate H to generate an effective channel matrix H eff and (ii) to the data symbols to generate whitened data symbols.
1. A method of whitening noise in a signal received by a receiving device from a transmitting device via a multi input multi output (MIMO) communication channel, the method comprising:
receiving, at the receiving device, data from a data unit via the MIMO communication channel, wherein the data includes at least training symbols and data symbols;
determining, with the receiving device and based on the training symbols, a channel estimate H corresponding to the MIMO communication channel;
determining a noise scaling factor s, wherein the noise scaling factor s is a ratio of variance of noise at the receiving device to a variance of noise at the transmitting device;
determining, using the channel estimate H and the noise scaling factor s, a noise whitening matrix W; and
applying the noise whitening matrix W to (i) the channel estimate H to generate an effective channel matrix H eff and (ii) to the data symbols to generate whitened data symbols.
2. The method of claim 1 , wherein determining the noise whitening matrix W comprises:
constructing a matrix B, wherein
B
=
[
H
H
s
I
]
,
wherein H H is a Hermitian transpose of the channel estimate H, and I is an identity matrix of dimensionality N r ×N r , wherein N r corresponds to a number of receive antennas via which the data is received at the receiving device;
performing QR decomposition of the matrix B to generate a unitary matrix Q B and an upper triangular matrix R B ; and
extracting the whitening matrix W from the unitary matrix Q B .
3. The method of claim 2 , wherein extracting the whitening matrix W from the unitary matrix Q B further comprises:
partitioning the matrix Q B into a matrix Q B1 and a matrix Q B2 , wherein
Q
B
=
[
Q
B
1
Q
B
2
]
,
wherein the matrix Q B1 is a matrix of dimensionality N ss ×N r , and wherein the matrix Q B2 is the matrix of dimensionality N r ×N r , wherein N ss corresponds to a number of spatial streams over which the data is received by the receiving device; and
obtaining the whitening matrix W by scaling the matrix Q B2 by the noise scaling factor s.
4. The method of claim 1 , further comprising, after applying the noise whitening matrix W to the channel estimate H and to the data symbols, equalizing the data symbols using a MIMO equalizer.
5. The method of claim 4 , wherein the MIMO equalizer is a zero forcing equalizer.
6. The method of claim 4 , wherein the MIMO equalizer is a maximum likelihood (ML) equalizer.
7. The method of claim 4 , wherein equalizing the data symbols comprises equalizing the data symbols based on the effective channel estimate H eff .
8. The method of claim 1 , wherein the training symbols comprise a plurality of training signal fields corresponding to a plurality of spatial streams over which the data is received, including at least a first training field corresponding to a first spatial stream and a second training field corresponding to a second spatial stream, wherein determining the whitening matrix W includes beginning computation of the whitening matrix W prior to receiving the second training field.
9. The method of claim 1 , wherein determining the noise scaling factor s comprises selecting the noise scaling factor s from a plurality of stored noise scaling factors s.
10. The method of claim 1 , wherein determining the noise scaling factor s comprises determining the noise scaling factor s dynamically based on one or more fields of the data unit.
11. The method of claim 10 , wherein determining the noise scaling factor s dynamically comprises determining the noise scaling factor s based on at least i) a received symbol of the one or more fields, ii) a decoded symbol corresponding the received symbol and iii) the channel estimation H.
12. An apparatus, comprising:
a network interface configured to receive data from a data unit via a MIMO communication channel, wherein the data includes at least training symbols and data symbols,
wherein the network interface includes:
a channel estimation unit configured to determine, based on the training symbols, a channel estimate H corresponding to the MIMO communication channel;
a noise scaling factor generator configured to determine a noise scaling factor s, wherein the noise scaling factor s is a ratio of variance of receiver noise to a variance of transmitter noise, and
a noise whitening unit configured to determine, using the channel estimate H and the noise scaling factor s, a noise whitening matrix W, and
wherein the network interface is further configured to apply the noise whitening matrix W to (i) the channel estimate H to generate an effective channel matrix H eff and (ii) the data symbols to generate whitened data symbols.
13. The apparatus of claim 12 , wherein the noise whitening unit is configured to:
construct a matrix B, wherein
B
=
[
H
H
s
I
]
,
wherein H H corresponds to a Hermitian transpose of the channel estimate H, and I is an identity matrix of dimensionality N r ×N r , wherein N r corresponds to a number of receive antennas via which the data is received;
perform QR decomposition of the matrix B to generate a unitary matrix Q B and an upper triangular matrix R B ; and
extract the whitening matrix W from the unitary matrix Q B .
14. The apparatus of claim 13 , wherein the noise whitening unit is further configured to:
partition the matrix Q B into a matrix Q B1 and a matrix Q B2 , wherein
Q
B
=
[
Q
B
1
Q
B
2
]
,
wherein the matrix Q B1 is a matrix of dimensionality N SS ×N r , and wherein the matrix Q B2 is the matrix of dimensionality N r ×N r , wherein N ss corresponds to a number of spatial streams over which the data is received; and
obtain the whitening matrix W by scaling the matrix Q B2 by the noise scaling factor s.
15. The apparatus of claim 12 , wherein the network interface further includes a MIMO equalizer configured to equalize the whitened data symbols.
16. The apparatus of claim 15 , wherein the MIMO equalizer is a zero forcing equalizer.
17. The apparatus of claim 15 , wherein the MIMO equalizer is a maximum likelihood (ML) equalizer.
18. The apparatus of claim 15 , wherein the MIMO equalizer equalizes the whitened data symbols based on the effective channel estimate H eff .
19. The apparatus of claim 12 , wherein the training symbols comprise a plurality of training signal fields corresponding to a plurality of spatial streams over which the data is received, including at least a first training field corresponding to a first spatial stream and a second training field corresponding to a second spatial stream, wherein the noise whitening unit is configured to begin computation of the whitening matrix W before the network interface second training field is received.
20. The apparatus of claim 12 , wherein the noise scaling factor generator is configured to select the noise scaling factor s from a plurality of stored noise scaling factors s.
21. The apparatus of claim 12 , wherein the noise scaling factor generator is configured to determine the noise scaling factor s dynamically based on the data unit.
22. The apparatus of claim 12 , wherein the noise scaling factor generator is configured to determine the noise scaling factor s dynamically based on at least i) a received symbol of the data unit, ii) a decoded symbol corresponding the received symbol and iii) the channel estimation.