 |
Web Hosting by Netfirms | Free Domain Names by Netfirms |
|
|
|
Creating a noise-free environment for test and measurement is seldom practical, but there are ways to control the effects of noise in PC-based data acquisition while increasing measurement accuracy.
Noise pickup resulting from your DA system's configuration and location in the test environment introduces measurement inaccuracies. Data acquisition systems come in a variety of configurations--PC expansion-slot boards, PCMCIA cards, and separately enclosed instruments--and each variation incorporates its own noise-reduction measures (see Photo 1). Expansion-slot A/D conversion (ADC) boards, such as IOtech's ISA-bus-based DaqBoard, are designed with multilayer printed circuits to resist the noisy environment inside a PC. PCMCIA cards have metal jackets to shield their printed circuitry. Separately enclosed instruments, such as IOtech's parallel-port-based DaqBooks, are shielded by their enclosures and by the fact that they are situated outside the PC's noisy interior.
Regardless of the data acquisition system configuration, users should test their system's noise level by connecting the ADC input to the analog common line and observing fluctuations in the ADC output. Connecting the ADC input to the analog common line isolates the cause of the noise to the ADC itself. Users may need to perform more careful diagnostics when using an external voltage source because noise can arise from the external source or from the input leads.
Most laboratory and industrial environments have several sources of electrical noise. For example, an AC power line generates 50- or 60-Hz noise. Users can easily reduce this noise by installing RF line filters and transient suppressors or using a DC power source.
Heavy equipment is also a source of noise, particularly when turned on and off. Local radio stations are a source of high-frequency noise, and computers and other electronic equipment can create noise in all frequency ranges. If at all possible, users should isolate their test equipment from such external noise sources. When this isn't unfeasible, users may be able to reduce noise interference by using filters.
Current flow through one or more unintended paths, known as ground loops, creates measurement errors. Each ground has a unique voltage potential, which may differ from other grounds in a data acquisition system. Under general operating conditions, these potential differences, or common-mode voltages, range from millivolts to tens of volts. The frequency of the voltages generally is 50 or 60 Hz, depending on the local power frequency. Under unusual circumstances such as a power line ground fault or a local lightning strike, voltage differences are higher and can damage electrical systems. One way to minimize the large differences in ground potential is to use the same grounded AC outlet to power the computer, data acquisition system, and system under test.
However, even when all the data acquisition equipment operates from the same AC outlet, ground loops still occur. For example, a user who connects a data acquisition system without galvanic isolation to a nonisolated (grounded) device can create unwanted ground loops. Users can employ differential connections to reduce the effect of the common-mode voltages frequently caused by such unwanted ground loops. In situations where a common-mode voltage exceeds 10-15 volts, users should consider employing isolated signal conditioning equipment. Generally, data acquisition systems that are isolated or that are floating electrically prevent ground loops.
When installing shielded leads, the shielding should never be grounded at both ends. Any potential difference-causing current flow through the shield (with the capacitive proximity of the shield to the center conductor) can result in a noisier connection. Grounding the leads at both ends also simultaneously couples other noise into the signal leads, which negates the advantage of using shielded leads.
Input leads made of twisted pairs often prevent noise as effectively as shielding. Twisted pairs are made by twisting two loose wires into a single spiral pair. The tighter the spiral, the more effective the twisted pair. Twisted pairs also improve most measurements taken in reverse proportion to the signal levels.
Some 60-, 120-, and 180-Hz residual noise emitted from AC equipment is virtually impossible to eliminate. Depending on the frequency of the signal being measured, users can employ either low- or high-pass filtering. A filter is an analog circuit element that attenuates an incoming signal according to its frequency. A low-pass filter attenuates frequencies higher than the cutoff frequency, and a high-pass filter attenuates frequencies lower than the cutoff frequency. As frequency increases beyond the cutoff point, the attenuation of a single-pole, low-pass filter increases slowly. Multipole filters provide greater attenuation.
Users can easily make simple low- and high-pass single-pole filters using a resistor and capacitor combination. The theoretical cutoff frequency of these filters is 1/(2piRC). To benefit from a single-pole filter, the frequency of the signal being measured should be significantly above or below the noise frequency. If a signal's frequency is within the frequency range of a noise, users can construct a bandpass filter by placing high- and low-pass filters in series.
Sometimes a simple filtering approach at the signal source helps. For example, a small capacitor (range 0.001 µF to 0.1 µF) across a signal source removes much of the high-frequency noise. In the case of a user measuring the voltage or current of a brush-type DC motor, a single capacitor across the motor's power terminals will considerably reduce RF generated by the motor's brushes. The same technique works well with strain gage outputs and other low-level, low-frequency sensors.
In most cases, users can increase data accuracy by making differential voltage measurements (see Figure 3). This is accomplished by attaching the channel input leads directly to the voltage point and the most appropriate reference point. Differential connections work well because the noise on the high analog input lead closely approximates the noise on the low analog input lead. This noise voltage can be considerably higher than the signal that the user is attempting to measure.
One of the most common mistakes made when setting up a differential measurement is failing to establish a biasing path between their differential inputs. All differential measurement applications require a biasing path. Although specific applications may already have an established biasing path, others must have one established or the data will be invalid. Users can easily establish a biasing path by connecting a resistor (10 K Ohms to 1 M Ohms) from the low input line to the analog common line. In an application with multiple measurement points referenced to the same analog common line, users need only one connection to the data acquisition system's analog common line.
The common-mode rejection ratio (CMRR) of an instrument is typically specified in decibels (dB). A ratio between two voltages of 10 corresponds to 20 dB. A CMRR of 80 dB means that a 10-mV signal can be measured in the presence of a 1-V common-mode voltage with a common mode error of 0.1 mV. It is important to note that CMMR is a frequency-dependent parameter, which drops rapidly as a function of frequency. At DC to 60 Hz, the CMMR will be higher than at 10 kHz; thus higher frequency common-mode noise is less easily rejected with differential measurement.
Many users may not consider averaging their data for fear of losing valuable information on signal variations. However, when the signal is steady, averaging provides more accurate data by reducing noise via the square root of a number of averaged data samples. Data collected for dynamic signals should not be averaged because it tends to become distorted.
Before a user decides to average the data for their steady signal measurements, they should consider several important factors. Averaging generally eliminates only random noise; it cannot eliminate many types of system noise (e.g. noise that occurs with the same delay after a triggering event). Averaging is useful only to the extent that the noise component of a signal averages to zero. Noise in measurements decreases only as the square root of the number of measurements. Therefore, in certain applications, reducing the RMS noise to a single count by averaging would require far too many samples.
The type of ADC used also makes a difference in data accuracy achieved through averaging. Slower, integrating-type ADCs tend to be more accurate than high-speed ADCs (e.g., a successive approximation converter register) because the integrating ADC has more time to take a reading from the actual voltage and the conversion has an averaging effect. High-speed ADCs convert temporarily captured voltages on a preceding sample-and-hold stage. Therefore, any single reading has a degree of uncertainty because it includes the value of noise at the instant of a sample-and-hold capture. Any individual reading can differ from the ideal by the system reading uncertainty. If a user is scanning a group of signals, the values from each channel can be sorted and averaged.
Several techniques are available for reducing noise errors in analog measurements. By choosing the best test environment, careful grounding the system, and correctly wiring the input leads the entry of noise into the measurement can be minimized. Differential measurement, filtering, and averaging all reduce the impact of noise. But to obtain maximum accuracy, most applications need a coordinated mix of these methods.