488 Chapter 15
A powerful alternative test method uses a sinusoidal waveform and checks the amplitude
probability density function. Missing, or misplaced codes, and nonmonotonic behavior
can show up readily in such a plot.
Linearity can be assessed overall by using a sinusoidal test signal since the output signal
will contain harmonics. The performance of an ADC can be carried out in the digital
domain by direct use of the discrete fast Fourier transform. The DAC can be checked by
driving it with a computer-generated sequence and observing the output in the analogue
domain. The trouble with using the simple harmonic distortion test is that it is not easy to
check the dynamic performance over the last decade of bandwidth and for this reason the
CCIR twin tone intermodulation distortion is much preferred.
Differential nonlinearity is the random unevenness of each quantization level. This defect
can be assessed by measuring the noise fl oor in the presence of a signal. In a good DAC the
rms noise fl oor should be approximately 95 dB below the maximum rms level (assuming a
modest margin for dither). The output buffer amplifi er will contribute some noise but this
should be at a fi xed level and not dependent on the DAC input sample values.
The basic ADC element simply provides an output dependent on the value of the digital
input. During the period while a fresh sample is settling, its output can be indeterminate.
Thus the output will usually be captured by a sample and hold circuit as soon as the DAC
has stabilized. The sample and hold circuit is a zero order hold circuit that imposes its
own frequency response on the output signal ( Figure 15.28 ); correction for which can be
accommodated within the overall reconstruction fi lter. The fi nal fi lter is used to remove
the higher components and harmonics of the zero order hold.
15.10.3 Other Forms of ADC and DAC
Flash converters ( Figure 15.29 ) function by using an array of comparators, each set
to trigger at a particular quantizing threshold. The output is available directly. These
days the technique is most commonly employed directly as shown in digitizing video
waveforms. However, there is a use for the technique in high-quality oversampling
converters for audio signals.
One great benefi t of operating with digital signals is their robustness; they are, after
all, construed as either 1 or 0 irrespective of the cable or optical fi ber down which they
travel. Their disadvantage is that the digital signals do cover a wide bandwidth. Since
bandwidth is a valuable resource there has been much effort expended in devising ways in