17
the AWG’s 10 MHz low-pass
filter (middle trace). The
TDS 744A FFTspectra for the
two signals are overlaid below
the time domain waveforms. The
salient characteristic of the
unfiltered noise spectrum is that
it rolls off with a (sin x)/x func-
tion with the first null at the
32.768 MHz clock frequency and
subsequent nulls at multiples of
the clock rate. If the goal is to
add this noise waveform to the
10.7MHz FM carrier, then noise
density is required only in the
vicinity of 10.7 MHz. The
filtered noise signal is a suitable
bandwidth-limited source. Thus,
when using the AWG noise
function, one consideration is to
account for the clock rate depen-
dent roll-off.
Maximizing “Randomness”
The second property to consider
when using the AWG noise
waveform is to observe that the
noise waveform itself is a pre-
calculated series of points that
will repeat at each period of the
record length. The period of the
32K point noise waveform at a
32.768 MHz sample rate is 1ms
and the exact noise waveform
repeats at a rate of 1kHz. This
periodicity translates into the
resulting noise spectrum. The
ideal noise waveform would
exhibit no periodicity (i.e., no
repetition). While this is not an
option with pre-calculated AWG
waveforms, the effect of the peri-
odicity can be reduced by
increasing the period of the
noise waveform relative to the
corresponding signal waveform.
Figure 18 shows how the AWG’s
sequence editor converts the
32K point FM waveform into a
256K point waveform which is
simply 8 concatenated copies of
the same waveform. Thus, if the
same clock waveform of
32.768MHz is used, the resulting
signal waveform is identical.
However, if a 256K noise wave-
form is generated, then the
period of the noise waveform is
increased by a factor of 8.
Figure 18. The 32K point FM waveform can be
converted to a 256K point waveform by simply
sequencing or concatenating 8 copies of the
original 32K waveform. This expansion means
that a 256K noise waveform can be added to the
FM waveform instead of a 32K noise waveform.