When you build an A/D converter
into a circuit, as part of a system, the first thing to do is to measure its
performance. The reason is that A/D converters are very layout and
bypass sensitive; the A/D specifications provided by the manufacturer were
measured using an idealized test fixture; probably your implementation
isn't quite so ideal, and requires a little refinement. Therefore, for any new
A/D circuit design you will always need to measure the performance of your A/D
to assure yourself that its performance in your circuit at least nears
the part's rated specifications.
More than likely, your first design will need refinement, and you will end up
repeatedly measuring and improving the performance of the A/D converter in your
circuit and system. With hard work and persistence, you will eventually measure
something close to the manufacturer's A/D specifications.
The following uses ScopeDSP™ to analyze a data
sample taken from a high-performance 14-bit, 10 MHz A/D converter made by a
leading vendor of digital and analog devices. The performance of this part
(really a chip set) is so high that achieving the part's specified performance
in an actual application circuit (not just a test fixture) was a major
challenge. In fact, just testing the part is a major challenge because it is
extremely difficult to generate analog signals which have less distortion than
this A/D converter produces. The captured signal we will analyze is the end
product of literally months of iteration to achieve the converter's specified
performance in its application circuit. For those of you who'd like to play
along at home, the A/D signal we will analyze is captured in the file
twomhz.bin. This is binary file whose data samples are
in IEEE 32-bit format (corresponding to ScopeDSP's "Binary:float"
data file format). It contains 8192 real
points of a 1.953125 MHz sine signal which was sampled at a 5.0 MHz rate.
Below is a plot of the entire signal:
This picture isn't much to look at because the sine is at a high frequency
relative to the sampling rate and because the sample size is large compared to
the number of horizontal pixels. Let's zoom in:
This plot is a little more meaningful, but still doesn't look much like
a sine wave except to the most highly trained eye. Next,
let's transform it to the frequency domain...
© 1997-2005 Iowegian
International Corporation. All rights reserved.
"ScopeDSP" is a trademark of Iowegian International Corporation