Switch on pop-ups!
There are numerous scientific measuring instruments which generate smoothly varying analogue signals that represent some physical quantity. The following table lists some common examples:
|Signal Transducer||Measurement||Output generated|
|pH electrode||proton concentration||voltage|
|oxygen electrode||oxygen concentration||current|
|strain gauge||force / displacement||voltage|
|polarograph||hplc column eluates||current|
|flame ionisation detector||gas chromatograms||current|
It is normally immaterial whether the measuring transducer inherently produces a voltage or a current output, since the two types of electrical signal are readily interconverted and scaled using an operational amplifier. Analogue signals must however be converted into digital form before the information can be processed by a computer.
High quality analogue circuits are expensive to build and maintain. Ceaseless attention to detail is necessary to eliminate noise and electrical artefacts and preserve all the information contained in the original signal. Digital information in contrast is much more resilient. It can be readily transmitted, stored and analysed without losing any of the original data.
The normal strategy with scientific measurements is to include the minimum of analogue circuitry very close to the transducer to boost the signal to a convenient level, and then to convert the information into digital format as soon as possible after the original measurement. This operation is performed by an analogue to digital converter (ADC). These devices generate a digital output proportional to their analogue input. Although they can operate "free standing" they are more commonly used under computer control. Some analogue pre-amplification and filtering [also called "signal conditioning"] is usually necessary between the transducer and the ADC. The digital connections to the ADC commonly radiate some low-level high-frequency analogue noise. It is difficult to convert a very weak analogue signal into digital format without this noise causing interference.
A huge variety of specialised conversion circuits have been used for particular applications, but a popular "general purpose" arrangement is illustrated below.
The ADC accepts a smoothly varying analogue input in the range -5 volts to + 5 volts, and generates a 12-bit digital output in the range 0 - 4095. For example:
|analogue input||-5 volts||-2.5 volts||0 volts||+2.5 volts||+5 volts|
An output range restricted to 4096 steps means that signal changes of less than 0.024% cannot be recorded by the system. Such "graininess" is an inherent feature of digital storage. At first sight this precision seems more than adequate, but greater accuracy may be required in order to extract a small varying signal in the face of a much larger fixed voltage that is always present. More expensive ADCs offer greater precision [16 - 20 bits] but take longer to perform the conversion. There is a trade-off between price, accuracy and conversion speed. This may be an issue with rapidly varying signals.
In the present example a negative analogue signal is meaningless [you cannot have a negative amount of light] so that whole of the negative half scale from -5 volts to 0 volts input is unused [0 - 2047 digital output]. The smallest detectable light intensity corresponds to an analogue input signal of +2.44 millivolts and a digital output of 2048.
The digital output from the ADC is usually expressed in arbitrary units. In order to convert the raw data into a more meaningful form, it is usual to subtract off the "zero" reading [2047 in the present case] from every data point, then divide by the full scale digital span [2048 in our case] and finally multiply by the corresponding analogue full scale reading [100 nanoamps for our equipment]. The result is thereby expressed in more conventional units.
This process is analogous to temperature conversion from Farenheit to Celsius: first subtract off the freezing point of water (32°F) so that our new scale starts from zero, then multiply by 5/9 to allow for the different scale widths (180F between freezing and boiling corresponds to 100C).
Note that the optical system has no intrinsic calibration. The actual amount of light producing a particular analogue signal depends on the photomultiplier dynode voltages and the amplifier settings, and would in any case change from one phototube specimen to the next. If necessary the system could be calibrated against a light source with known intensity. Very often, however, we are only interested in relative changes and the absolute calibration does not matter.
On the other hand, it is often very important that the response should be monotonic, linear and unbiased. Phototube and ADC manufacturers put a lot of effort into this. The photomultiplier is likely to perform well if the supporting circuits are well designed [dynamic range over 1 million fold variation in light intensity] but the converter chips are far from perfect and cheap ones may introduce significant errors.
Here is another example: a medium light intensity might generate a photomultiplier output of 30nA, which after analogue signal conditioning would produce a signal of +1.5 volts. From this the ADC would generate a digital result of 2662.
Bit 11 is the most significant bit, and bit 0 is the least significant. Note how the analogue scale runs from -5 volts to +5 volts, so that an input signal of +1.5 volts is 6500mV above the arbitrary reference point of -5 volts.
analogue input: 6500mV = 5000 + 1250 + 156.2 + 78.1 + 9.8 + 4.9 (approximately)
digital output: 2662 = 2048 + 512 + 64 + 32 + 4 + 2 (exactly)
The brightest light that can be recorded with this system would produce an analogue input signal of +5 volts, and a digital output of 4095 [all 12 ADC bits set to "1"].
ADCs commonly perform their conversions on a trial and error basis, one bit at a time. These devices include a digital to analogue converter (DAC) which generates an analogue signal from a digital word. The control logic switches the most significant bit [in this case bit 11] to a binary "1" on a trial basis. If the result [5000mV in this case] is less than the input signal then the bit is left switched on, otherwise it is reset to binary "0". The ADC then tests the succeeding bits in order, finishing with the least significant, which is worth only 2.44mV in the present example. At each stage the converter achieves a better and better approximation to the input signal, so these devices are also known as "successive approximation registers".
There are many other types of ADC and these may have advantages for particular applications. The raw digital information can be stored on disc or tape, or input to a computer for further processing. Advanced data acquisition systems may incorporate analogue multiplexers to follow several inputs at the same time, and each channel may have auto-ranging and zero offset facilities to achieve a better match between the input data and the ADC dynamic range.