Data Acquisition Systems

　Data acquisition systems are used to acquire process operating data and store it on，secondary storage devices for later analysis. Many or the data acquisition systems acquire this data at very high speeds and very little computer time is left to carry out any necessary, or desirable, data manipulations or reduction. All the data are stored on secondary storage devices and manipulated subsequently to derive the variables of in-terest. It is very often necessary to design special purpose data acquisition systems and interfaces to acquire the high speed process data. This special purpose design can be an expensive proposition.

　Powerful mini- and mainframe computers are used to combine the data acquisition with other functions such as comparisons between the actual output and the desirable output values, and to then decide on the control action which must be taken to ensure that the output variables lie within preset limits. The computing power required will depend upon the type of process control system implemented. Software requirements for carrying out proportional, ratio or three term control of process variables are relatively trivial, and microcomputers can be used to implement such process control systems. It would not be possible to use many of the currently available microcomputers for the implementation of high speed adaptive control systems which require the use of suitable process models and considerable online manipulation of data.

　Microcomputer based data loggers are used to carry out intermediate functions such as data acquisition at comparatively low speeds, simple mathematical manipulations of raw data and some forms of data reduction. The first generation of data loggers, without any programmable computing facilities, was used simply for slow speed data acquisition from up to one hundred channels. All the acquired data could be punched out on paper tape or printed for subsequent analysis. Such hardwired data loggers are being replaced by the new generation of data loggers which incorporate microcomputers and can be programmed by the user. They offer an extremely good method of collecting the process data, using standardized interfaces, and subsequently performing the necessary manipulations to provide the information of interest to the process operator. The data acquired can be analyzed to establish correlations, if any, between process variables and to develop mathematical models necessary for adaptive and optimal process control.

　The data acquisition function carried out by data loggers varies from one to 9 in system to another. Simple data logging systems acquire data from a few channels while complex systems can receive data from hundreds, or even thousands, of input channels distributed around one or more processes. The rudimentary data loggers scan the selected number of channels, connected to sensors or transducers, in a sequential manner and the data are recorded in a digital format. A data logger can be dedicated in the sense that it can only collect data from particular types of sensors and transducers. It is best to use a nondedicated data logger since any transducer or sensor can be connected to the channels via suitable interface circuitry. This facility requires the use of appropriate signal conditioning modules.

　Microcomputer controlled data acquisition facilitates the scanning of a large number of sensors. The scanning rate depends upon the signal dynamics which means that some channels must be scanned at very high speeds in order to avoid aliasing errors while there is very little loss of information by scanning other channels at slower speeds. In some data logging applications the faster channels require sampling at speeds of up to 100 times per second while slow channels can be sampled once every five minutes. The conventional hardwired, non-programmable data loggers sample all the channels in a sequential manner and the sampling frequency of all the channels must be the same. This procedure results in the accumulation of very large amounts of data, some of which is unnecessary, and also slows down the overall effective sampling frequency. Microcomputer based data loggers can be used to scan some fast channels at a higher frequency than other slow speed channels.

　The vast majority of the user programmable data loggers can be used to scan up to 1000 analog and 1000 digital input channels. A small number of data loggers, with a higher degree of sophistication, are suitable for acquiring data from up to 15, 000 analog and digital channels. The data from digital channels can be in the form of Transistor- Transistor Logic or contact closure signals. Analog data must be converted into digital format before it is recorded and requires the use of suitable analog to digital converters (ADC)．The characteristics of the ADC will define the resolution that can be achieved and the rate at which the various channels can be sampled. An in-crease in the number of bits used in the ADC improves the resolution capability. Successive approximation ADC's are faster than integrating ADC's. Many microcomputer controlled data loggers include a facility to program the channel scanning rates. Typical scanning rates vary from 2 channels per second to 10, 000 channels per second.

　Most data loggers have a resolution capability of ±0.01% or better, It is also pos-sible to achieve a resolution of 1 micro-volt. The resolution capability, in absolute terms, also depends upon the range of input signals, Standard input signal ranges are 0-10 volt, 0-50 volt and 0-100 volt. The lowest measurable signal varies form 1 t, volt to 50, volt. A higher degree of recording accuracy can be achieved by using modules which accept data in small, selectable ranges. An alternative is the auto ranging facil-ity available on some data loggers.

　The accuracy with which the data are acquired and logged-on the appropriate storage device is extremely important. It is therefore necessary that the data acquisi-tion module should be able to reject common mode noise and common mode voltage. Typical common mode noise rejection capabilities lie in the range 110 dB to 150 dB. A decibel (dB) is a tern which defines the ratio of the power levels of two signals. Thus if the reference and actual signals have power levels of N, and Na respectively, they will have a ratio of n decibels, where

　n＝10 Log10（Na /Nr）

　Protection against maximum common mode voltages of 200 to 500 volt is available on typical microcomputer based data loggers.

　The voltage input to an individual data logger channel is measured, scaled and linearised before any further data manipulations or comparisons are carried out.

　In many situations, it becomes necessary to alter the frequency at which particu-lar channels are sampled depending upon the values of data signals received from a particular input sensor. Thus a channel might normally be sampled once every 10 minutes. If, however, the sensor signals approach the alarm limit, then it is obviously desirable to sample that channel once every minute or even faster so that the operators can be informed, thereby avoiding any catastrophes. Microcomputer controlled intel-ligent data loggers may be programmed to alter the sampling frequencies depending upon the values of process signals. Other data loggers include self-scanning modules which can initiate sampling.

　The conventional hardwired data loggers, without any programming facilities, simply record the instantaneous values of transducer outputs at a regular sampling in-terval. This raw data often means very little to the typical user. To be meaningful, this data must be linearised and scaled, using a calibration curve, in order to determine the real value of the variable in appropriate engineering units. Prior to the availability of programmable data loggers, this function was usually carried out in the off-line mode on a mini- or mainframe computer. The raw data values had to be punched out on pa-per tape, in binary or octal code, to be input subsequently to the computer used for analysis purposes and converted to the engineering units. Paper tape punches are slow speed mechanical devices which reduce the speed at which channels can be scanned. An alternative was to print out the raw data values which further reduced the data scanning rate. It was not possible to carry out any limit comparisons or provide any alarm information. Every single value acquired by the data logger had to be recorded even though it might not serve any useful purpose during subsequent analysis; many data values only need recording when they lie outside the pre-set low and high limits.

　If the analog data must be transmitted over any distance, differences in ground potential between the signal source and final location can add noise in the interface design. In order to separate common-mode interference form the signal to be recorded or processed, devices designed for this purpose, such as instrumentation amplifiers, may be used. An instrumentation amplifier is characterized by good common-mode- rejection capability, a high input impedance, low drift, adjustable gain, and greater cost than operational amplifiers. They range from monolithic ICs to potted modules, and larger rack-mounted modules with manual scaling and null adjustments. When a very high common-mode voltage is present or the need for extremely-low com-mon-mode leakage current exists（as in many medical-electronics applications），an isolation amplifier is required. Isolation amplifiers may use optical or transformer isolation.

　Analog function circuits are special-purpose circuits that are used for a variety of signal conditioning operations on signals which are in analog form. When their accu-racy is adequate, they can relieve the microprocessor of time-consuming software and computations. Among the typical operations performed are multiplications, division, powers, roots, nonlinear functions such as for linearizing transducers, rims measure-ments, computing vector sums, integration and differentiation, and current-to-voltage or voltage- to-current conversion. Many of these operations can be purchased in available devices as multiplier/dividers, log/antilog amplifiers, and others.

　When data from a number of independent signal sources must be processed by the same microcomputer or communications channel, a multiplexer is used to channel the input signals into the A/D converter.

　Multiplexers are also used in reverse, as when a converter must distribute analog information to many different channels. The multiplexer is fed by a D／A converter which continually refreshes the output channels with new information.

　In many systems, the analog signal varies during the time that the converter takes to digitize an input signal. The changes in this signal level during the conversion process can result in errors since the conversion period can be completed some time after the conversion command. The final value never represents the data at the instant when the conversion command is transmitted. Sample-hold circuits are used to make an acquisition of the varying analog signal and to hold this signal for the duration of the conversion process. Sample-hold circuits are common in multichannel distribution systems where they allow each channel to receive and hold the signal level.

　In order to get the data in digital form as rapidly and as accurately as possible, we must use an analog/digital (A/D) converter, which might be a shaft encoder, a small module with digital outputs, or a high-resolution, high-speed panel instrument. These devices, which range form IC chips to rack-mounted instruments, convert ana-log input data, usually voltage, into an equivalent digital form. The characteristics of A/D converters include absolute and relative accuracy, linearity, monotonic, resolu-tion, conversion speed, and stability. A choice of input ranges, output codes, and other features are available. The successive-approximation technique is popular for a large number of applications, with the most popular alternatives being the counter-comparator types, and dual-ramp approaches. The dual-ramp has been widely-used in digital voltmeters.

　D/A converters convert a digital format into an equivalent analog representation. The basic converter consists of a circuit of weighted resistance values or ratios, each controlled by a particular level or weight of digital input data, which develops the output voltage or current in accordance with the digital input code. A special class of D/A converter exists which have the capability of handling variable reference sources. These devices are the multiplying DACs. Their output value is the product of the number represented by the digital input code and the analog reference voltage, which may vary form full scale to zero, and in some cases, to negative values.

　Component Selection Criteria

　In the past decade, data-acquisition hardware has changed radically due to ad-vances in semiconductors, and prices have come down too; what have not changed, however, are the fundamental system problems confronting the designer. Signals may be obscured by noise, rfi，ground loops, power-line pickup, and transients coupled into signal lines from machinery. Separating the signals from these effects becomes a matter for concern.

　Data-acquisition systems may be separated into two basic categories:（1）those suited to favorable environments like laboratories -and（2）those required for hostile environments such as factories, vehicles, and military installations. The latter group includes industrial process control systems where temperature information may be gathered by sensors on tanks, boilers, wats, or pipelines that may be spread over miles of facilities. That data may then be sent to a central processor to provide real-time process control. The digital control of steel mills, automated chemical production, and machine tools is carried out in this kind of hostile environment. The vulnerability of the data signals leads to the requirement for isolation and other techniques.

　At the other end of the spectrum-laboratory applications, such as test systems for gathering information on gas chromatographs, mass spectrometers, and other sophis-ticated instruments-the designer's problems are concerned with the performing of sen-sitive measurements under favorable conditions rather than with the problem of pro-tecting the integrity of collected data under hostile conditions.

　Systems in hostile environments might require components for wide tempera-tures, shielding, common-mode noise reduction, conversion at an early stage, redun-dant circuits for critical measurements, and preprocessing of the digital data to test its reliability. Laboratory systems, on the other hand, will have narrower temperature ranges and less ambient noise. But the higher accuracies require sensitive devices, and a major effort may be necessary for the required signal /noise ratios.

　The choice of configuration and components in data-acquisition design depends on consideration of a number of factors:

　1. Resolution and accuracy required in final format.

　2. Number of analog sensors to be monitored.

　3. Sampling rate desired.

　4. Signal-conditioning requirement due to environment and accuracy.

　5. Cost trade-offs.

　Some of the choices for a basic data-acquisition configuration include： 1 ．Single-channel techniques.

　 A. Direct conversion.

　B. Preamplification and direct conversion.

　C. Sample-hold and conversion.

　D. Preamplification, sample-hold, and conversion.

　E. Preamplification, signal-conditioning, and direct conversion.

　F. Preamplification, signal-conditioning, sample-hold, and conversion. 2. Multichannel techniques.

　A. Multiplexing the outputs of single-channel converters.

　B. Multiplexing the outputs of sample-holds.

　C. Multiplexing the inputs of sample-holds.

　D. Multiplexing low-level data.

　E. More than one tier of multiplexers.

　Signal-conditioning may include:

　1. Radiometric conversion techniques.

　B. Range biasing.

　D. Logarithmic compression.

　A. Analog filtering.

　B. Integrating converters.

　C. Digital data processing.

　We shall consider these techniques later, but first we will examine some of the components used in these data-acquisition system configurations.

　Multiplexers

　When more than one channel requires analog-to-digital conversion, it is neces-sary to use time-division multiplexing in order to connect the analog inputs to a single converter, or to provide a converter for each input and then combine the converter outputs by digital multiplexing.

　Analog Multiplexers

　Analog multiplexer circuits allow the timesharing of analog-to-digital converters between a numbers of analog information channels. An analog multiplexer consists of a group of switches arranged with inputs connected to the individual analog channels and outputs connected in common（as shown in Fig. 1）．The switches may be ad-dressed by a digital input code.

　Many alternative analog switches are available in electromechanical and solid-state forms. Electromechanical switch types include relays, stepper switches, cross-bar switches, mercury-wetted switches, and dry-reed relay switches. The best switching speed is provided by reed relays（about 1 ms）．The mechanical switches provide high do isolation resistance, low contact resistance, and the capacity to handle voltages up to 1 KV, and they are usually inexpensive. Multiplexers using mechanical switches are suited to low-speed applications as well as those having high resolution requirements. They interface well with the slower A/D converters, like the integrating dual-slope types. Mechanical switches have a finite life, however, usually expressed in number of operations. A reed relay might have a life of 109 operations, which would allow a 3-year life at 10 operations/second.

　Solid-state switch devices are capable of operation at 30 ns, and they have a life which exceeds most equipment requirements. Field-effect transistors（FETs） are used in most multiplexers. They have superseded bipolar transistors which can introduce large voltage offsets when used as switches.

　FET devices have a leakage from drain to source in the off state and a leakage from gate or substrate to drain and source in both the on and off states. Gate leakage in MOS devices is small compared to other sources of leakage. When the device has a Zener-diode-protected gate, an additional leakage path exists between the gate and source.

　Enhancement-mode MOS-FETs have the advantage that the switch turns off when power is removed from the MUX. Junction-FET multiplexers always turn on with the power off.

　A more recent development, the CMOS-complementary MOS-switch has the advantage of being able to multiplex voltages up to and including the supply voltages. A±10-V signal can be handled with a ±10-V supply.

　Trade-off Considerations for the Designer

　Analog multiplexing has been the favored technique for achieving lowest system cost. The decreasing cost of A/D converters and the availability of low-cost, digital integrated circuits specifically designed for multiplexing provide an alternative with advantages for some applications. A decision on the technique to use for a given sys-tem will hinge on trade-offs between the following factors:

　1. Resolution. The cost of A/D converters rises steeply as the resolution increases due to the cost of precision elements. At the 8-bit level, the per-channel cost of an analog multiplexer may be a considerable proportion of the cost of a converter. At resolutions above 12 bits, the reverse is true, and analog multiplexing tends to be more economical.

　2. Number of channels. This controls the size of the multiplexer required and the amount of wiring and interconnections. Digital multiplexing onto a common data bus reduces wiring to a minimum in many cases. Analog multiplexing is suited for 8 to 256 channels; beyond this number, the tec