The need to have one analogue comparator for each quantisation level means that the circuit becomes excessively complex and costly with increasing resolution. The high speed of the flash design, however, is bought at the expense of resolution. The National Instruments NI 5112 Digital Oscilloscope card, for instance, uses a flash ADC which supports 100 MHz, 8-bit A/D sampling. Flash ADCs are typically used in frame grabbers for digitising video signals, or in high-speed digital oscilloscopes. This approach allows conversion times in the region of 10 ns and sampling rates as high as 100 MHz. The parallel or flash ADC avoids these constraints by providing a comparator for each quantisation level and making all voltage comparisons simultaneously, in parallel. A typical 12-bit successive approximation ADC tends to have a conversion time in the region of 0.5–2 µs, supporting sampling rates up to 2 MHz. Since one comparison is required for each bit, this is one reason why high-resolution ADCs have longer conversion times. The conversion speed of the successive approximation design is constrained by the number of voltage comparisons that have to be made, each of which requires a short period of time to allow V ref to settle down after a bit is changed. This phenomenon is known as missing codes and a guarantee of ‘no missing codes’ is something one should look for in the specification of a high-quality ADC. Any inaccuracy in these levels produces discontinuities in the ADC voltage response, which can result in some of the binary quantisation levels being missed out at particular input voltages. Accuracy depends upon the reference voltage levels for each bit in the binary word being exact binary multiples of each other. On completion of the procedure, V ref = V in, to within the accuracy of the ADC, and the value of the binary number representing that voltage can be read out by the host computer. This is, in effect, a binary search procedure which forces V ref to converge towards V in with each successive step, as shown in Fig. The process is then repeated with the next lower bit, until all 8 bits have been tested. If this causes V ref to exceed V in, the bit is set back to zero, otherwise it is retained. Starting with the highest bit (7) which generates the greatest voltage, each bit is set to one. The value of V ref is adjusted to match V in by successively comparing the effect of setting each bit in the binary word to one. V ref and V in are fed into a comparator circuit which allows the ADC to determine whether V ref exceeds V in. (d) Integrator charge/discharge cycle during an A/D conversion.
(b) Convergence of reference voltage, V ref, to analogue input voltage, V in, with each successive comparison during an A/D conversion. Conversely, the 24-bit ADCs designed for high-precision work may require 20 ms per conversion.
ADCs intended for the digitisation of video signals can have conversion times of 10 ns, but may be restricted to 8-bit resolution to achieve this speed. Generally speaking, the higher the precision of the ADC, the longer it takes to perform a conversion, thus a 16-bit ADC will tend to have a longer conversion time than an 8-bit one.
The 12-bit ADCs, typically found in the laboratory, have conversion times in the range 1–10 µs, and are thus capable of sampling at rates of 100 kHz to 1 MHz. This conversion time places a limit on the rate at which an analogue signal can be sampled. The A/D conversion process is not instantaneous a certain amount of time is required to measure the analogue voltage and to generate the binary output value. ADCs with higher resolutions, such as 20 and even 24 bit, are available, and are used in applications where very high precision is required, the digitisation of signals from an HPLC (high-performance liquid chromatograph), for example. This – a precision of around 0.025% – is usually sufficient for most purposes. Most of those in common use within the laboratory have at least a 12-bit resolution, yielding 4096 quantisation levels. Thus the smallest voltage difference that the 8-bit ADC, with a ± 5 V range, can measure is 39 mV.ĪDCs are available with resolutions varying from 8 to 24 bits. Where V + and V - are the positive and negative limits of the voltage range.