The FieldDAQ device uses a combination of analog and digital filtering to provide an accurate representation of in-band signals
and reject out-of-band signals. The filters discriminate between signals based on the frequency range, or bandwidth, of the
signal. How the filter discriminates signals based on their frequency is known as frequency response. In general, the frequency
response of a filter is described by a signal attenuation (magnitude response) and an input delay (phase response) for every
input frequency.
-
Magnitude Response—The three important frequency ranges, or bandwidths, to consider for magnitude response are passband, transition band, and
stopband:
-
Passband—The range of frequencies at which the filter attempts to pass a signal without modifying it. The small amount of
variation in magnitude at these frequencies is called passband flatness. This is the frequency range of signals that you want
to measure.
-
Transition band—The range of frequencies in which the filter magnitude response has started to roll-off such that it attenuates
signals by some amount, but has not reached the full attenuation amount. The shape of the transition band has an impact on
the alias rejection and how signals are represented in the time domain (for example, step response).
-
Stopband—The range of frequencies at which the filter attenuates input signals to its maximum attenuation level. Ideally,
you want to choose a filter with a stopband that covers frequencies of noise sources that you do not want in your measurements.
-
Phase Response (Input Delay)—Filters delay the input signal by some amount when processing data. In some cases, the delay is a function of the input signal
frequency; when this is the case, the phase response plot is useful for knowing the exact delay at different input frequencies
and the maximum variation between signals of different frequencies within the passband. Call the AI.FilterDelay property in
NI-DAQmx to read the input delay.
Each FieldDAQ filter has a different frequency response to serve different applications:
-
Brickwall—Has a passband that extends to approximately 0.4 times the sampled data rate, then drops with a very sharp transition band
to reach the stopband and properly reject any alias components or noise that may exist. The brickwall filter is designed to
provide full alias protection from any signal component of more than one-half the sampled data rate (Nyquist frequency).
-
Butterworth—Has a passband independent of the sampling rate (as opposed to the brickwall and comb filters), which offers more flexibility
when filtering out noise that is below one-half of the sample rate. However, depending on your settings, you may see alias
components of higher frequency signals in your measurement that extend beyond one-half of the sample rate due to the larger
transition band.
-
Comb—Has a smaller passband because its transition band starts early in the frequency range. The comb filter has shorter group
delay than other filters and better representation of signals in the time domain (step response). The passband and transition
band notches track the data sample rate similar to the brickwall filter. The comb filter's transition band features equally-spaced
notches at different frequencies. It is common to use the comb filter with a specific sample rate to align the notches of
the transition band thereby removing a specific noise-source frequency from measurements.
The FieldDAQ filters vary between input delay across signals in the passband:
-
Brickwall—All input frequencies have the same amount of delay when going through the filter. Choose this filter for applications where
linear phase is required or applications where data correlation between different devices and configurations is required.
-
Butterworth—Delays signals by a variable amount depending on their frequency.
-
Comb—All input frequencies have the same amount of delay when going through the filter. The comb filter has a shorter delay than
the brickwall filter. Choose this filter for applications where linear phase, short delay, or data correlation of different
devices and configurations is required.
Refer to the FD-116xx Specifications for details on the amount of variation in the passband gain and signal delay you can expect for different input frequency
ranges.