National Instruments AT E Series - Manual (Page 36)

Chapter 3 Hardware Overview
AT E Series User Manual 3-10 ni.com
Considerations for Selecting Input Ranges
Which input polarity and range you select depends on the expected range
of the incoming signal. A large input range can accommodate a large signal
variation but reduces the voltage resolution. Choosing a smaller input range
improves the voltage resolution but may result in the input signal going out
of range. For best results, you should match the input range as closely as
possible to the expected range of the input signal. For example, if you are
certain the input signal is not negative (below 0 V), unipolar input polarity
is best. However, if the signal is negative or equal to zero, inaccurate
readings occur if you use unipolar input polarity.
Dither
When you enable dither, you add approximately 0.5 LSB
rms
of white
Gaussian noise to the signal to be converted by the ADC. This addition
is useful for applications involving averaging to increase the resolution
of the AT E Series device, as in calibration or spectral analysis. In such
applications, noise modulation is decreased and differential linearity is
improved by the addition of the dither. When taking DC measurements,
such as when checking the device calibration, you should enable dither and
average about 1,000 points to take a single reading. This process removes
the effects of quantization and reduces measurement noise, resulting in
improved resolution. For high-speed applications not involving averaging
or spectral analysis, you may want to disable the dither to reduce noise. You
enable and disable the dither circuitry through software.
Figure 3-7 illustrates the effect of dither on signal acquisition. Figure 3-7a
shows a small (±4 LSB) sine wave acquired with dither off. The
quantization of the ADC is clearly visible. Figure 3-7b shows what happens
when 50 such acquisitions are averaged together; quantization is still
plainly visible. In Figure 3-7c, the sine wave is acquired with dither on.
There is a considerable amount of noise visible. But averaging about 50
such acquisitions, as shown in Figure 3-7d, eliminates both the added noise
and the effects of quantization. Dither has the effect of forcing quantization
noise to become a zero-mean random variable rather than a deterministic
function of the input signal.