Verification of the antenna weights of the base station in a UMTS mobile radio receiver by use of the Viterbi algorithm with fed-back transmission diversity
View Patent ↗A mobile station sends to the base stations a bit vector which contains weighting factors (w 1 , w 2 ) for the amplitudes and/or the phases of the signals to be emitted from the antennas. The bit vector is updated by changing one bit position for each time slot, and by transmitting this from the mobile station to the base station in the form of a bit (e.g., an FBI (feedback information) bit). A trellis diagram can be set up comprising the possible states of the bit vector and the possible state changes (branches) and the Viterbi algorithm can be used for the determination of the weighting factors actually used by the base station, taking account of both the currently transmitted training symbols and the previously transmitted training symbols.
1. A method for verification of a bit vector which is transmitted from a mobile station to a base station that comprises two antennas and weighting factors (w 1 , w 2 ) for amplitudes and/or phases of the signals to be emitted from the antennas, wherein the transmitted bit vector is employed in the base station to determine the weighting factors, comprising:
updating the bit vector in the mobile station by changing only one bit position in the bit vector at a time and transmitting the updated bit vector from the mobile station to the base station in the form of a bit, wherein the updated bit vector is associated with target weighting factors;
transmitting current training symbols from the base station to the mobile station using updated weighting factors based on the updated bit vector received from the mobile station; and
determining in the mobile station whether the updated weighting factors match the target weighting factors by evaluating the transmitted current training symbols from the base station and previously transmitted training symbols therefrom, where determining whether the updated weighting factors match the target weighting factor comprises:
establishing a trellis diagram comprising possible states of the updated bit vector and possible state changes or branches between them;
executing a Viterbi algorithm on the trellis diagram to determine the updated weighting factors being used by the base station; and
comparing the updated weighting factors to the target weighting factors, where executing the Viterbi algorithm comprises:
associating a cumulative metric with respective states;
defining a branch metric as a function of the transmitted training symbols from the base station, wherein the branch metrics Δm p,k b are defined by
Δ m p,k b ={ln({circumflex over (p)}( w 1 , w 2 ))+ln ( p ( w 1 , w 2 ))}
where ( w 1 , w 2 ) is a pair of weighting factors corresponding to the state transition and p (w 1 , w 2 ) is an a-priori probability for the occurrence of the weighting pair ( w 1 , w 2 ), and {circumflex over (p)}(w 1 , w 2 ) is calculated from the transmitted bit associated with the updated bit vector and from a hypothesis for the error probability for the transmission of the bit associated with the updated bit vector, and
ln
(
p
^
(
w
_
1
,
w
_
2
)
)
=
-
(
∑
i
=
1
N
path
h
i
(
d
)
-
γ
(
w
_
1
h
1
,
i
(
p
)
+
w
_
2
h
2
,
i
(
p
)
)
2
σ
i
2
(
1
+
γ
2
(
w
_
1
2
+
w
_
2
2
)
)
)
where h i (d) are channel parameters determined on the basis of a dedicated pilot channel for both antennas, h 1,i (p) and h 2,i (p) are channel parameters determined on the basis of a common pilot channel for both antennas, γ 2 is the ratio between the signal-to-interference-and-noise (SINR) values for a DPCH pilot channel and for a CPICH channel, N path is the number of propagation paths, and σ i 2 is the sum of the noise and interference power on the i-th propagation path;
calculating a branch metric for respective branches of the trellis diagram that lead to a state;
calculating a cumulative metric for respective states;
determining the state with the largest cumulative metric; and
selecting the weighting factors (w 1 , w 2 ) which are associated with the largest cumulative metric as the updated weighting factors, where if a total number of possible states as dictated by the bit vector is N, merely M≦N/2 state variables associated with the trellis diagram are followed such that the method comprises:
producing subsequent states on the basis of an initialization state in the trellis diagram in successive time steps until M state variables have been assigned respective cumulative metric values;
calculating 2M subsequent states with cumulative metrics for the M state variables;
sorting the 2M cumulative metrics on the basis of their magnitude; and
assigning the M largest states to the M state variables and rejecting the remaining states such that they are no longer followed.
2. The method according to claim 1 , wherein the cumulative metric m p,k of a state is calculated by
m p,k = m p,k−1 +max b (Δ m p,k b )
where p is an index for the states and k is an index for the time steps, m p,k−1 is the cumulative metric for the previous state, and max b (Δ p,k b ) is the maximum of the branch metrics Δm p,k b .
3. A method of verifying antenna weights employed in a base station employing multiple antennas for transmission to a mobile station, comprising:
determining optimum antenna weights in the mobile radio station by evaluating training symbols transmitted by the base station to the mobile radio station;
generating a bit sequence based on the determined optimum antenna weights and transmitting the bit sequence from the mobile station to the base station, wherein a bit associated with the bit sequence is transmitted for each time slot;
modifying the antenna weights at the base station based on the bit sequence; and
evaluating current training symbols at the mobile station transmitted by the base station based on the modified antenna weights and training symbols previously received from the base station to determine if the modified antenna weights at the base station match the set of optimum antenna weights and thus verifying a transmission accuracy of the bit vector from the mobile station to the base station, where determining whether the modified weighting factors match the optimum weighting factors comprises:
establishing a trellis diagram comprising possible states of the updated bit vector and possible state changes or branches between them;
executing a Viterbi algorithm on the trellis diagram to determine the updated weighting factors being used by the base station; and
comparing the updated weighting factors to the target weighting factors, where executing the Viterbi algorithm comprises:
associating a cumulative metric with respective states;
defining a branch metric as a function of the transmitted training symbols from the base station, wherein the branch metrics Δm p,k b are defined by
Δ m p,k b ={ln({circumflex over (p)}( w 1 , w 2 ))+ln( p ( w 1 , w 2 ))}
where ( w 1 , w 2 ) is a pair of weighting factors corresponding to the state transition and p (w 1 , w 2 ) is an a-priori probability for the occurrence of the weighting pair ( w 1 , w 2 ), and {circumflex over (p)}(w 1 , w 2 ) is calculated from the transmitted bit associated with the updated bit vector and from a hypothesis for the error probability for the transmission of the bit associated with the updated bit vector, and
ln
(
p
^
(
w
_
1
,
w
_
2
)
)
=
-
(
∑
i
=
1
N
path
h
i
(
d
)
-
γ
(
w
_
1
h
1
,
i
(
p
)
+
w
_
2
h
2
,
i
(
p
)
)
2
σ
i
2
(
1
+
γ
2
(
w
_
1
2
+
w
_
2
2
)
)
)
where h i (d) are channel parameters determined on the basis of a dedicated pilot channel for both antennas, h 1,i (p) and h 2,i (p) are channel parameters determined on the basis of a common pilot channel for both antennas, γ 2 is the ratio between the signal-to- interference-and-noise (SINR) values for a DPCH pilot channel and for a CPICH channel, N path is the number of propagation paths, and σ i 2 is the sum of the noise and interference power on the i-th propagation path;
calculating a branch metric for respective branches of the trellis diagram that lead to a state;
calculating a cumulative metric for respective states;
determining the state with the largest cumulative metric; and
selecting the weighting factors (w 1 , w 2 ) which are associated with the largest cumulative metric as the updated weighting factors, where if a total number of possible states as dictated by the bit vector is N, merely M≦N/2 state variables associated with the trellis diagram are followed such that the method comprises:
producing subsequent states on the basis of an initialization state in the trellis diagram in successive time steps until M state variables have each been assigned a cumulative metric value;
calculating 2M subsequent states with cumulative metrics for the M state variables;
sorting the 2M cumulative metrics on the basis of their magnitude; and
assigning the M largest states to the M state variables and rejecting the remaining states such that they are no longer followed.
4. The method according to claim 3 wherein the cumulative metric m p,k of a state is calculated by
m p,k = m p,k−1 +max b (Δ m p,k b )
where p is an index for the states and k is an index for the time steps, m p,k−1 is the cumulative metric for the previous state, and max b (Δ p,k b ) is the maximum of the branch metrics Δm p,k b .