CDBCapture+

6

DS327F1

Figure 6 shows the results of a Histogram Analysis.

This analysis was obtained by grounding the signal

input to the evaluation board and acquiring data.

The results of this analysis are useful in determin-

ing how much noise and offset error is in the sys-

tem and A/D converter. In an ideal system with the

input grounded, all readings should be zero, but due

to noise and offsets, there will be some variation in

the readings. A DC offset will result in the mean

value of the readings being non-zero, and the

MEAN value displayed to the left of the graph is an

indication of the offset, in lsb’s. If the histogram

fits the normal Gaussian distribution, then the Stan-

dard Deviation (STD_DEV) will be a measure of

the RMS noise in the system. The overall peak-to-

peak variation in A/D counts is available by view-

ing the MINIMUM COUNT and MAXIMUM

COUNT values shown to the left of the histogram.

The statistical peak-to-peak noise is defined as the

interval which contains six standard deviations, or

99.6% of all count occurrences, so multiplying the

STD_DEV value to the left of the histogram by six

will yield the peak to peak noise.

Figure 7 shows the analysis screen for a calculation

of a converter’s Differential Nonlinearity (DNL).

To test for DNL, samples are taken using a linear

over-ranged triangle wave for the signal. If a large

number of samples are taken, then the number of

occurrences of each code should be the same if the

converter has perfect differential nonlinearity.

Codes that occur less than average will have a neg-

ative DNL and codes that occur more often than av-

erage will have a DNL greater than zero. The graph

shown in Figure 6 shows the variation in DNL for

each count of the converter. A mouse-controlled

cursor allows the operator to select a particular

code and read out its specific DNL below the

graph. To the left of the graph the maximum and

minimum DNL values for all the codes are dis-

played. If a code has a DNL of -1, then that is a

missing code, it never occurred during the sam-

pling.

Figure 5. Grounded Input FFT