Rohde & Schwarz R&S FSL3 - Page 156

R&S FSL WLAN TX Measurements (Option K91)
1300.2519.12 2.131 E-11
In case of an additive white Gaussian noise (AWGN) channel the FFT is described by [4], [5]
kl
phasephasej
klkl
neHgaKr
kl
common
l
kl
,
(
,mod
)timing(
,
)(
,
+××××=
+
(10)
with
the modulation–dependant normalization factor
mod
K ,
the symbol
kl
a
,
of sub–carrier k at symbol l ,
the gain
l
g at the symbol l in relation to the reference gain
1
=
g
at the long symbol (LS),
the channel frequency response
k
H at the long symbol (LS),
the common phase drift
)common(
l
phase of all sub–carriers at symbol l (see below),
the phase
)timing(
,kl
phase of sub–carrier k at symbol l caused by the timing drift (see below),
the independent Gaussian distributed noise samples
kl
n
,
.
The common phase drift in equation (10) is given by
lrests
common
l
dylTfNNphase +×××= /2
)(
 (11)
with
80=
s
N being the number of Nyquist samples of the symbol period,
64
=
N being the number of Nyquist samples 64
=
N of the useful part of the symbol,
rest
f being the (not yet compensated) frequency deviation,
l
d # being the phase jitter at the symbol l .
In general, the coarse frequency estimate
coarse
f
ˆ

(see figure 1) is not error–free. Therefore the remaining
frequency error
rest
f represents the not yet compensated frequency deviation in
kl
r
,
. Consequently the
overall frequency deviation of the device under test (DUT) is calculated by
restcoarse
fff +=
ˆ
. Remark:
The only motivation for dividing the common phase drift in equation (11) into two parts is to be able to
calculate the overall frequency deviation of the DUT.
The reason for the phase jitter
l
d # in equation (11) may be different. The nonlinear part of the phase
jitter may be caused by the phase noise of the DUT oscillator. Another reason for nonlinear phase jitter
may be the increase of the DUT amplifier temperature at the beginning of the burst. Please note that
besides the nonlinear part the phase jitter
l
d # also contains a constant part. This constant part is
caused by the not yet compensated frequency deviation
rest
f . To understand this, please keep in mind
that the measurement of the phase starts at the first symbol
1
=
l of the payload. In contrast the channel
frequency response
k
H in equation (10) represents the channel at the long symbol of the preamble.
Consequently the not yet compensated frequency deviation
rest
f produces a phase drift between the
long symbol and the first symbol of the payload. Therefore this phase drift appears as a constant value
("DC value'') in
l
d # .
Referring to the IEEE 802.11a measurement standard Chapter 17.3.9.7 "Transmit modulation accuracy test''
[6], the common phase drift
)common(
l
phase must be estimated and compensated from the pilots. Therefore this
"symbol wise phase tracking'' (Tracking Phase) is activated as the default setting of the R&S FSL–K91/K91n.
Furthermore the timing drift in equation (10) is given by
lkNNphase
s
kl
××××=  /2
)timing(
,
(12)