Tektronix AWG 2021 Brochure page 29

Figure 36 compares the original
and filtered data patterns. The
upper two traces are the unfil-
tered data pattern and its spec-
trum. The lower two traces are
the filtered data pattern and its
spectrum. Note how the spec-
trum of the filtered version rolls
off more quickly. The spectrum
of a modulated carrier shows the
same results. Figure 37 shows
the filtered baseband pattern
0
-10
-20
-30
-40
-50
-60
-70
-80
10500 10550 10600
modulating (BPSK) the
10.7 MHz carrier, as in Figure 27.
Figure 38 shows the difference
in their spectra.
The convolution operator can be
applied to multi-level patterns.
Figure 39 shows Gaussian
filtered I and Q baseband
patterns for the 16-QAM signal
in Figure 32. (The unfiltered I
pattern is shown in Figure 31.)
The falling edge of the data
10650 10700 10750 10800 10850
Frequency (kHz
clock output defines the center
of the symbol period. Using the
marker output as a data clock
provides a convenient reference
when characterizing the perfor-
mance of symbol timing recov-
ery circuits. Careful attention
was given to wrapping data at
the ends of the data patterns so
that the convolution result
would be continuous across the
seams.
Figure 38. Spectrum analyzer plots of unfiltered
10900
(upper) and BT=0.5 Gaussian filtered (lower)
BPSK carriers at 10.7 MHz. The data rate is
40 kbaud. Compare the roll-off to the baseband
roll-off in Figure 35.
AWG data clock output
I baseband pattern
Q baseband pattern
Figure 39. Gaussian filtered multi-level baseband
modulation is shown. The AWG generated a data
clock output on one of its marker outputs. The
bottom trace is the other AWG marker output
generating a once per pattern pulse for scope
triggering.
29