Introdution:
Noise is a variation in a measurement of a process variable that does not reflect real changes in the process variable.
A signal from a sensor can have many components. This signal will always have as one of its components the process value that we are measuring, but it may also contain noise. Noise is generally a result of the technology used to sense the process variable. Electrical signals used to transmit instrument measurements are susceptible to having noise induced form other electrical devices. Noise can also be caused by wear and tear on mechanical elements of a sensor.
Noise may also be uncontrolled random variations in the process itself. Whatever the source, noise distorts the measurement signal.
Effects of Noise:
Noise reduces the accuracy and precision of process measurements. Somewhere in the noise is the true measurement, but where? Noise introduces more uncertainty into the measurement.
Noise also introduces errors in control systems. To a controller fluctuation in the process variable from noise are indistinguishable from fluctuations caused by real disturbances. Noise in a process variable will be reflected in the output of the controller.
Eliminating Noise:
The most effective means of eliminating noise is to remove the source. Reduce electrically induced noise by following proper grounding techniques; using shielded cabling and physical separation of signal cabling form other electrical wiring. If worn mechanical elements in the sensor are causing noise repair or replace the sensor.
When these steps have been taken and excessive noise is still a problem in the process variable a low pass filter may be used.
Low Pass Filters:
Smart instruments and most controllers have noise dampening features built in. Most of these noise dampeners are actually low pass filters.
A low-pass filter allows the low frequency components of a signal to pass while attenuating the higher frequency components.
Fortunately for us, noise tends to fall into the higher end of the frequency spectrum while the underlying process value tends to lie in the lower end.
Selecting a Filter by Cut-off Frequency:
Attenuation of a signal is a reduction in its strength, or amplitude. Attenuation is measured in
decibels (dB).
dB of attenuation = 20 log10(Amplitude In/Amplitude Out)
For example: let’s say we have an amplitude ratio of 0.95 (the value of the signal out is 95% of
value of the signal in), the dB of attenuation would be:
20 log10 (.95) = −0.45
An attenuation of 0 dB would mean the signal would pass with no reduction in amplitude while a large negative dB would indicate a very small amplitude ratio (at -10 dB of attenuation we would have an amplitude ratio of 0.32).
The ideal filter would be designed to pass all signals with 0 dB of attenuation below a cut-off frequency and completely attenuate all frequency components above the cut-off frequency. This ideal filter does not exist in the real world. -3 dB of signal attenuation has been established as the cut-off frequency in filter selection. Figure 3-8 illustrates the effect of a filter with a 3 Hz cut-off frequency on a noisy 1.2 Hz signal. Where a filter is selected by choosing cut-off frequency, select that is above the frequency of your process value.
Noise is a variation in a measurement of a process variable that does not reflect real changes in the process variable.
A signal from a sensor can have many components. This signal will always have as one of its components the process value that we are measuring, but it may also contain noise. Noise is generally a result of the technology used to sense the process variable. Electrical signals used to transmit instrument measurements are susceptible to having noise induced form other electrical devices. Noise can also be caused by wear and tear on mechanical elements of a sensor.
Noise may also be uncontrolled random variations in the process itself. Whatever the source, noise distorts the measurement signal.
Effects of Noise:
Noise reduces the accuracy and precision of process measurements. Somewhere in the noise is the true measurement, but where? Noise introduces more uncertainty into the measurement.
Noise also introduces errors in control systems. To a controller fluctuation in the process variable from noise are indistinguishable from fluctuations caused by real disturbances. Noise in a process variable will be reflected in the output of the controller.
Eliminating Noise:
The most effective means of eliminating noise is to remove the source. Reduce electrically induced noise by following proper grounding techniques; using shielded cabling and physical separation of signal cabling form other electrical wiring. If worn mechanical elements in the sensor are causing noise repair or replace the sensor.
When these steps have been taken and excessive noise is still a problem in the process variable a low pass filter may be used.
Low Pass Filters:
Smart instruments and most controllers have noise dampening features built in. Most of these noise dampeners are actually low pass filters.
A low-pass filter allows the low frequency components of a signal to pass while attenuating the higher frequency components.
Fortunately for us, noise tends to fall into the higher end of the frequency spectrum while the underlying process value tends to lie in the lower end.
Selecting a Filter by Cut-off Frequency:
Attenuation of a signal is a reduction in its strength, or amplitude. Attenuation is measured in
decibels (dB).
dB of attenuation = 20 log10(Amplitude In/Amplitude Out)
value of the signal in), the dB of attenuation would be:
20 log10 (.95) = −0.45
An attenuation of 0 dB would mean the signal would pass with no reduction in amplitude while a large negative dB would indicate a very small amplitude ratio (at -10 dB of attenuation we would have an amplitude ratio of 0.32).
The ideal filter would be designed to pass all signals with 0 dB of attenuation below a cut-off frequency and completely attenuate all frequency components above the cut-off frequency. This ideal filter does not exist in the real world. -3 dB of signal attenuation has been established as the cut-off frequency in filter selection. Figure 3-8 illustrates the effect of a filter with a 3 Hz cut-off frequency on a noisy 1.2 Hz signal. Where a filter is selected by choosing cut-off frequency, select that is above the frequency of your process value.
Selecting a Filter by Time Constant
The effect of a low pass filter is to introduce a first order lag in the process variable response.
Low pass filters may sometimes be referred to as first order lag filters.
Some filters are configured by selecting a time constant for the lag response of the filter. The relationship between the cut-off frequency and the time constant of a low pass is approximately given by:
Cut - Off Frequency ≈ .1/ 5 Time Constants
For example, to configure a filter for a cut-off frequency of 60 Hz specify a filter time constant of seconds
Time Constant ≈ 1/5 Cut - Off Frequency =1/5*60 =0.0033
Bellow Figure shows the step response of the 3 Hz filter illustrating the filter time constant of 0.068 seconds.
No comments:
Post a Comment