Chapter 3
Device Overview and Theory of Operation
©
National Instruments Corporation
3-7
Because the ADC samples at 64 or 128 times the data rate, frequency
components above one-half of the oversampling rate—32 or 64 times the
data rate—can alias. The digital filter rejects most of the frequency range
over which aliasing can occur. However, the filter can do nothing about
components that lie close to integer multiples of the oversampling rate—
64 (for
f
s
> 51.2 kS/s), 128, and 256 times the data rate, and so on—because
it cannot distinguish these components from components in the baseband
(0 Hz to the Nyquist frequency). If, for instance, the sample rate is 50 kS/s
and a signal component lies within 25 kHz of 6.4 MHz (128
×
50 kHz), this
signal is aliased into the passband region of the digital filter and is not
attenuated. The purpose of the analog filter is to remove these higher
frequency components near multiples of the oversampling rate before they
get to the sampler and the digital filter.
While the frequency response of the digital filter scales in proportion to the
sample rate, the frequency response of the analog filter remains fixed. The
response of the filter is optimized to produce good high-frequency alias
rejection while having a flat in-band frequency response. Because this filter
is second-order, its roll-off is rather slow. The filter has good alias rejection
at high sample rates, but as a result of its slow roll-off, does not filter aliases
as well at lower sample rates. The alias rejection near 64 or 128 times the
sample rate versus sample rate for the NI 4472 is illustrated in Figure 3-6.
For frequencies not near multiples of the oversample rate, the rejection is
better than 110 dB.