Choosing one of today’s PC-based data acquisition system (DAS) boards can be confusing. You are faced with an ever-growing number of vendors, models and features. And deceiving advertising claims add to this confusion.
Product comparisons are often based on sampling rates, the number of input channels, the number of converter bits and cost, overlooking the true metric of performance–accuracy. Ads that do tout accuracy often make these claims under conditions that make the product appear to perform better than it will in a real-world application.
The selection process can be greatly simplified if you understand the factors that affect accuracy and the parameters that describe these factors. Choosing a board that provides a full set of guaranteed specifications will provide you with the performance you expect.
Specifying Performance as a System
Every component that appears in the analog signal path affects accuracy. In a typical DAS board, these components include an input signal multiplexer, an instrumentation amplifier and the A/D converter. Remember, system accuracy will NEVER be any better than the weakest link in the signal path.
Three general classes of accuracy specifications exist:
o DC accuracy specifications.
o Temperature stability specifications.
o Dynamic accuracy specifications.
DC accuracy characterizes board performance with a steady-state input signal on a single input channel. It may be expressed as a percentage or as a fraction of the least-significant bit of converter resolution.
In a system of N bits, the weight of the least significant bit (LSB) is 1/2 N. An ideal A/D converter always measures its input with a maximum error of 1/2 LSB. In an ideal 12-bit system, this translates to an error of 1/8,192 (0.012%).
Real-world A/D converters are usually less than ideal. A/D converters are available in several accuracy grades. The lowest–and least expensive–accuracy grades offer significantly degraded performance. Many low-cost 12-bit converters are usable only to 11 bits, and many 16-bit converters are usable only to 14 bits. Beware of DAS board claims based on the number of converter bits without an accuracy specification, as they are meaningless.
Many DAS boards include a programmable gain amplifier before the converter input. This architecture permits the board to scale a smaller range of input voltages to the full A/D converter input range.
To prevent system accuracy from being degraded, the gain accuracy and linearity of this amplifier must be better than that of the A/D converter. These parameters are generally expressed as a percentage and should be specified for each gain. In general, these parameters degrade as gain increases.
Many low-cost DAS boards use inexpensive programmable gain instrumentation amplifiers. Gain accuracy of these parts may be approximately 0.2% (or 1 part in 500), with linearity errors in the range of 0.01%. This translates to a gain accuracy of about 9 bits, and linearity of about 12 bits. This is surprising, since these amplifiers often appear on boards that claim to be “16 bits.”
If all input channels in your application require the same gain, it may be possible to recalibrate the A/D converter to negate this gain error. However, if at least one channel requires a different gain, this mismatching reduces the accuracy of the acquired data to the gain accuracy of the amplifier. Many DAS boards use precision resistor networks to set amplifier gain, maintaining full converter accuracy regardless of the gain selected.
DC accuracy specifications should include temperature coefficients that describe stability as a function of temperature. These specifications should include A/D converter zero and full-scale drift, amplifier gain drift and amplifier offset voltage drift. It is often useful to consider the effect of a 10oC to 20oC change in temperature (within the PC and variations in ambient) on DC accuracy that may have occurred since initial calibration.
In the case of amplifier offset drift, keep in mind that the temperature coefficient is multiplied by the gain of the amplifier. An input offset drift of 50 uV/oC at a gain of 100 translates to an output offset voltage drift of 5 mV/oC. As a result, the accuracy of a 12-bit system employing such a device could easily be reduced to 10 bits due to normal room-temperature variations.
DAS board specifications related to dynamic accuracy describe the maximum sampling and multiplexing rates that may be used to acquire data with the same accuracy as in the steady-state case. They are often exaggerated by citing the maximum sampling rates that may be attained by the A/D converter.
The performance of the instrumentation amplifier, the input multiplexer and the associated control circuits ultimately determine how well a DAS board performs in this respect. To be meaningful, dynamic accuracy specifications must include the effects of this circuitry operating together as a system.
To maintain full accuracy, the instrumentation amplifier output must settle to within an error band determined by the magnitude of 1/2 LSB before the next conversion. It is more difficult to meet this requirement in a multiplexed environment because the required change in output voltage can greatly exceed that required by a single input channel over the same time period.
When an amplifier is optimized for shorter settling times, it becomes more difficult to maintain basic DC accuracy. Designs that meet both the accuracy and settling time criteria are more expensive, and low-cost DAS boards often employ amplifiers that fail to meet them.
Settling time also increases as the amplifier gain increases. Dynamic accuracy claims often mask these effects by considering a single input at the lowest possible gain. A good dynamic accuracy specification should always include the effects of multiplexing and gain.
Settling time must encompass the effects of all changes that occur during the multiplexing process. Many DAS boards permit the gain of the instrumentation amplifier to change on a channel-by-channel basis during multiplexed operation. This feature can significantly increase amplifier settling time. Without a guaranteed specification related to gain changes, you can never be sure of the sampling rate required to maintain full accuracy in this mode of operation.
The circuitry that controls input multiplexing and the A/D converter also plays an important role in dynamic accuracy. Timing of internal events must be tightly controlled to ensure adequate amplifier settling time before initiating the next conversion. When triggering or clocking a DAS board from an external source, it is also important to control delays to correlate the acquired data to the
trigger or clock event.
Many low-cost boards use software to control channel and/or gain selection, conversion initiation and triggering. Since delays introduced by DOS and Windows-based software applications are nondeterministic, correlation between the timing of individual events is lost. The maximum attainable accuracy or the sampling rate suffers as a result.
To accurately control the timing under all conditions, these events must be coordinated by hardware. Low-cost DAS boards often use one shots to control timing, but the accuracy of these devices permits timing to vary as much as 50%. Control logic based on synchronous design techniques assures you optimal and consistent performance under all conditions.
Real-World Sources Affecting Performance
All DAS board specifications assume that inputs are driven by ideal voltage sources. Real-world sources often have non-zero output impedance and/or limited current source capabilities. In a multiplexed environment, these can degrade accuracy of any DAS board to less than its published specifications.
When the multiplexer switches to the next input, the instrumentation amplifier input (and other parasitic) capacitance is charged to the voltage that was present on the previous channel. The output resistance of the next input source forms an RC circuit with this capacitance. The time constant of this circuit is now included in the system settling time and may result in reduced accuracy.
To determine if this effect is present in your application, compare measurements made on each input channel with and without multiplexing. If a significant amount of degradation is observed, it may be necessary to provide a buffer between the signal and the DAS board input.
To fully appreciate the difference between various DAS board designs, consider more than the number of bits, the sampling rates and cost. An objective evaluation of performance cannot be made without a full set of specifications backed by a vendor whose integrity you can trust. The true value of any product must be judged by its ability to deliver the performance you expect in your application.
About the Author
Vincent P. Wasnewsky is the Senior Staff Engineer at Keithley MetraByte. He has more than 15 years of experience in analog and digital hardware design, and holds B.S.E.E. and M.S.E.E. degrees from Worcester Polytechnic Institute. Keithley Instruments, Inc., MetraByte Division, 440 Myles Standish Blvd., Taunton, MA 02780, (508) 880-3000.
Copyright 1995 Nelson Publishing Inc.