Specifications and Accuracy

Computer-Based Instruments and Calibration
Computer-based instruments are measurement systems that place the instrument's intelligence and measurement circuitry inside a computer. Therefore, a computer-based instrument is a computer system that contains a plug-in digitizer and application-specific software.

A graphical representation of a computer-based instrument is shown in Figure 1. In this figure, the digitizer is represented by the Acquisition module and the application-specific software is represented as the Analysis and Presentation sections.

Figure 1 - Elements of Instrumentation

Understanding the Computer Environment
Because computer-based instruments exist inside a computer chassis, there are often concerns as to how the computer environment can affect measurements. In order to understand how computer-based instruments can operate inside a computer, you must first understand the computer environment itself. This environment includes temperature effects, electromagnetic interference (EMI), and power supply variations.

Electromagnetic Interference
EMI is an issue to consider when dealing with any electronic device. In the case of computer-based instruments, emissions from other devices can affect signal integrity. This interference results in instruments that perform poorly or do not meet specifications. You must consider two types of EMI when designing a computer-based instrument-radiated EMI and conducted EMI. Radiated EMI refers to a device that radiates an interfering signal through the air, much like a broadcasting antenna. Conductive EMI, on the other hand, uses the systems ground or power planes to transfer interference to other devices. In the case of a computer, the major contributors to these EMI problems are the video system and power supply.

Power Supplies
While the computer power supply also acts as a source of radiated emission, its overall contribution is small compared to other computer components. The power supply contributes to emission problems by adding switching noise to the power supply of the computer. This means that you must condition the system power before using it with a computer-based instrument.

Environment Effects
Temperature ranges and variations always pose a concern for computers users. As more peripherals are added to the computer chassis the overall airflow is reduced, leading to temperature gradients in the computer. From a design point of view, this means that the computer-based instrument must withstand above ambient temperatures and subtle temperature variations without affecting measurement capabilities.

Designing for the Computer Environment
You must understand the issues of the computer environment to design computer-based instruments. Without knowing the possible hurdles an instrument faces, you cannot create an instrument that meets its published specifications inside a computer.

Designing for Electromagnetic Interference
In order to deal with EMI, computer-based instrument designers use a variety of techniques, including shielding and advanced circuit board layout techniques. Of these two techniques, proper layout is by far the most critical.

The first step of the layout is to minimize current loops. This means placing components as close together as possible. If less metal exist between the components, there is less area for induced currents. A good layout also uses small traces and limits the amount of metal at connections to high impedance components such as operational amplifiers. By limiting the areas available for current to circulate, good circuit layout limits the effects EMI can have on a circuit.

If proper layout does not provide enough immunity to EMI, you may add a shield to the instrument. A shield is simply a metallic partition between the circuitry and the rest of the computer environment, and is used to keep electromagnetic radiation from entering sensitive portions of the circuitry. Shields reduce the amount and intensity of any interference that reaches the computer-based instrument, thus reducing the effects of EMI.

Resolving Power Supply Issues
To resolve the issues associated with conducted interference in computer power, a computer-based instrument must contain its own power supply conditioning. The conditioning power supply design must be robust enough to manage voltage fluctuations, or else the measurement accuracy of the instrument suffers. To solve this problem, instrument designers use voltage regulators, boost converters, and filtering to create a stable power supply.

Boost converters and voltage regulators act as the front end of the computer-based instrument power supply. These components regulate the instruments' power supply to provide a stiff source that is "noise free" and immune to power fluctuations present on the computers' power bus.

Once you have a good power supply, you can use bulk storage capacitors and bypass capacitors at strategic locations on the instrument. These capacitors remove any signal noise that the instrument might inject on its own power supply. This prevents noise in non-critical sections of the design from affecting critical portions.

Dealing with the Environment
Temperature and other environmental changes can affect sensitive measurements because circuit components drift, imposing measurement error, with environmental changes. To account for this drift, you can use internal calibration circuitry and onboard signal references. This circuitry allows the computer-based instrument to self calibrate and compensate for environmental changes.

The simplest form of calibration circuitry is a digital-to-analog converter (DAC). When connected to the instrument's measurement circuitry, the DAC acts as a controllable voltage source that can add voltage to correct for gain and offset error. Internal calibration also is the point where the software and hardware resources for the instrument begin to come together. During internal calibration, the software driver programmatically controls the DAC to generate the appropriate correction values for the selected measurement mode.

Before any adjustment is performed, the instrument must know how much the measurement circuitry may have drifted. The use of onboard signal references resolves this problem. The onboard reference represents a very stable source that has a high resistance to temperature change and low drift over time. Knowing that the reference voltage is very stable with respect to temperature, the instrument measures its value and attributes any noticeable error to environmental affects in the measurement circuitry. Using a signal reference, you can calculate the offset and gain error corrections for the measurement circuitry. These corrections are then applied via the calibration circuitry. Because of this, the instrument can correct for environmental variations at any time.

Interpreting Accuracy Specifications
National Instruments provides absolute and relative accuracy specifications for many of its products. These specifications help you determine if the product's performance meets your measurement requirements.

The specification tables often break down into two parts: absolute accuracy specifications and relative accuracy specifications. Of these two, absolute accuracy is used to define the overall uncertainties of a measurement. The table shown below is the accuracy specification table for the NI6070 multifunction I/O product.

NI6070 Accuracy Chart

Relative Accuracy
Relative accuracy compares the difference between two or more measurements. That is, it indicates the degree to which two or more measurements distinguish from one another. The two major contributors to relative accuracy are the resolution of the device's analog- to-digital converter (ADC) and the system noise. The accuracy tables provide the theoretical relative accuracy, which excludes noise effects, as well as the averaged relative accuracy, which includes noise effects.

As an example, assume you monitor a voltage once per second using the +/-10 V range of a NI PCI-6070 board and then averaging 100 points for each measurement. From the accuracy table shown above, we find:

Averaged Relative Accuracy = 1.11mV This means that each set of measurements will have to be 1.11 mV greater or less than previous measurements in order for the device to detect a difference in the input voltage.

Keep in mind that because relative accuracy is due to the board's analog-to-digital converter, it will not be affected by calibration.

Absolute Accuracy
Absolute accuracy is the specification you use to determine the overall uncertainty of your measurement. Absolute accuracy specifications apply only to a successfully calibrated board device. So, you can use these specifications to verify the operation of the board device before and after adjustments are made to the measurement circuitry.

Absolute accuracy is made of includes four separate components. By combining these components you can determine the uncertainty associated with the board's device's measurements.

Percent of Reading

Offset

Noise & Quantization

Drift

Combining these components, the formula for calculating the uncertainty of a measurement is:

Measurement Uncertainty = +/-((Input Voltage * % of Reading)/100 + Offset + Noise and Quantization + Drift) However, drift is already accounted for in the table unless your ambient temperature is outside the range of +15 to +35 degrees C. For instance, if your ambient temperature is 45° C, you must account for 10 degrees of drift. This is calculated by:

Drift = Temperature Difference * Percent of Drift per °C * Input Voltage Below is an example for calculating uncertainty for the PCI-6070 one year after the last calibration. The example uses a voltage range of +/-10 V, averaging of 100 samples and an input voltage of 9.5V. In this example, the ambient temperature is considered to be between +15 and +35 °C. Hence, we can ignore the temperature drift component. Combining the components from the above table, we have:

Measurement Uncertainty = +/-((9500. mV * 0.0714)/100 + 6.38 mV + 0.846 mV) = 14.0 mV If you average a different number of points than 100, the value for Noise and Quantization changes. Use the following equation to account for a different number of averages:

Noise and Quantization for x averages = (Averaged Noise and Quantization from table) * sqrt (100/x) where x is the new number of averages.