R&S FSL WLAN TX Measurements (Option K91)
1300.2519.12 2.137 E-11
%
=
=
1
0
2
)(
ˆ
)(
ˆ
)(
)(
N
v
vs
vsvr
vEVM
(23)
is the momentary error signal magnitude normalized by the root mean square value of the reference
signal power.
In [2] a different algorithm is proposed to calculate the error vector magnitude. In a first step the IQ–
offset in the I–branch
{}
%
=
=
1
0
REAL
1
ˆ
N
v
I
r(v)
N
o
(24)
and the IQ–offset of the Q–branch
{}
%
=
=
1
0
IMAG
1
ˆ
N
v
Q
r(v)
N
o
(25)
are estimated separately, where r(v) is the measurement signal which has been corrected with the
estimates of the timing–, frequency– and phase offset, but not with the estimates of the IQ–imbalance
and IQ–offset. With these values the IQ–imbalance of the I–branch
{}
%
=
=
1
0
ˆ
REAL
1
ˆ
N
v
II
or(v)
N
g
(26)
and the IQ–imbalance of the Q–branch
{}
%
=
=
1
0
ˆ
IMAG
1
ˆ
N
v
QQ
or(v)
N
g
(27)
are estimated in a non–linear estimation in a second step. Finally, the mean error vector magnitude
{}
[]
{}
[]
[]
2
22
1
0
2
1
0
2
err
2
1
)(IMAG
2
1
)(REAL
2
1
)(
Q
I
N
QQ
N
II
gg
govrgovr
vV
))
))))
+
+
=
%%
=
=
(28)
can be calculated with a non data aided calculation. The instant error vector magnitude
{}
[]
{}
[]
[]
2
22
2
2
err
ˆˆ
2
1
ˆˆ
)(IMAG
2
1
ˆˆ
)(REAL
2
1
)(
QI
QQII
gg
govrgovr
vV
+
+
=
(29)
is the error signal magnitude normalized by the root mean square value of the estimate of the
measurement signal power. The advantage of this method is that no estimate of the reference signal is
needed, but the IQ–offset and IQ–imbalance values are not estimated in a joint estimation procedure.
Therefore, each estimation parameter is disturbing the estimation of the other parameter and the
accuracy of the estimates is lower than the accuracy of the estimations achieved by equation (17). If the
EVM value is dominated by Gaussian noise this method yields similar results as equation (18).
To Next Page
Loading...
