Rohde & Schwarz R&S FSL3 - Page 158

R&S FSL WLAN TX Measurements (Option K91)
1300.2519.12 2.133 E-11
is calculated as a function of the trial parameters
l
g
~
and
l
d #
~
. Finally, the trial parameters leading to the
minimum of the log likelihood function are used as estimates
l
g
ˆ
and
l
d #
ˆ
.
This robust algorithm works well even at low signal to noise ratios with the Cramer Rao Bound being
reached.
After estimation of the parameters, the sequence
kl
r
,
is compensated in the compensation blocks.
In the upper analyzing branch the compensation is user–defined i.e. the user determines which of the
parameters are compensated. This is useful in order to extract the influence of these parameters. The
resulting output sequence is described by
k
l
r
,
' .
In the lower compensation branch the full compensation is always performed. This separate
compensation is necessary in order to avoid symbol errors. After the full compensation the secure
estimation of the data symbols
kl
a
,
ˆ
is performed. From equation (10) it is clear that first the channel
transfer function
k
H must be removed. This is achieved by dividing the known coarse channel estimate
)
LS(
ˆ
k
H calculated from the LS. Usually an error free estimation of the data symbols can be assumed.
In the next block a better channel estimate
)PL(
ˆ
k
H of the data and pilot sub–carriers is calculated by using all
symbolsnof _ symbols of the payload (PL). This can be accomplished at this point because the phase is
compensated and the data symbols are known. The long observation interval of
symbolsnof _ symbols ( com pared
to the short interval of 2 symbols for the estimation of
)L(
ˆ
S
k
H ) leads to a nearly error–free channel estimate.
In the following equalizer block
k
l
r
,
' is compensated by the channel estimate. The resulting channel–
compensated sequence is described by
kl
r
,
'' . The user may either choose the coarse channel estimate
)L(
ˆ
S
k
H (from the long symbol) or the nearly error–free channel estimate
)L(
ˆ
S
k
H (from the payload) for
equalization. In case of using the improved estimate
)L(
ˆ
S
k
H a 2 dB reduction of the subsequent EVM
measurement can be expected.
According to the IEEE 802.11a measurement standard [6], the coarse channel estimation
)L(
ˆ
S
k
H (from
the long symbol) has to be used for equalization. Therefore the default setting of the R&S FSL–K91 is
equalization from the coarse channel estimate derived from the long symbol.
In the last block the measurement variables
are calculated. The most important variable is the error
vector magnitude
%
=
×=
symbolsnof
l
klklk
aKr
symbolsnof
EVM
_
1
2
,mod
''
,
_
1
(14)
of the sub–carrier
k of the current packet. Furthermore the packet error vector magnitude
%
&=
=
26
)0(26
2
52
1
kk
k
EVMEVM (15)
is derived by averaging the squared
k
EVM versus k . Finally the average error vector magnitude
%
=
=
packetsnof
counter
counterEVM
packetsnof
EVM
_
1
2
)(
_
1
(16)
is calculated by averaging the packet
EVM
of all
packetsnof _
detected packets. This parameter is
equivalent to the so–called "RMS average of all errors
RMS
Error '' of the IEEE 802.11a measurement
commandment (see [6], Chapter 17.3.9.7).