9.4.2
Time base settings for FFTs
When doing FFTs, there are two topics that affect the acquisition:
1
It makes life easier when the final FFT yields spectral lines with a distance
∆
f
that is a "nice" value. Otherwise stated: the FFT bin size should
preferably be a nice value. Sometimes, this is also called the "frequency
resolution". The bin size is determined by the actual frame size or frame
length:
bin size = 1 / T
in which T is the total frame size time. E.g. a one
-
second frame size will result in a 1 Hz bin size, a 0.5 second frame size
results in a 2 Hz bin size.
2
The frame size in samples should be equal to a power of two.
Fundamentally, most FFT algorithms work on data sets with a length of
2^N.
The binary clock base of the internal time base in combination with the division
factors allow for a broad range of values that meet both requirements. In the
table below (see "Examples of FFT bin sizes" on page 141), various sample
rates are given, as well as the corresponding division factor (divisor). The table
shows the bin sizes that result from these sample rates in combination with
various sweep lengths.
Example: from the table (see "Examples of FFT bin sizes" on page 141) you
can read that a sample rate of 40.960 kHz and a frame length of 8192 samples
result in a 5 Hz bin size, i.e. the spectral lines are 5 Hz from each other.
"Nice" values are considered to be "minor" values that easily fit in "major" values
for (grid) display purposes.
GEN3i
140
I3763-3.1 en HBM: public
