the MAX195 finishes the old conversion, allows four
clock (CLK) cycles for input acquisition, then begins
the new conversion.
_____________Dynamic Performance
High-speed sampling capability, 85ksps throughput,
and wide dynamic range make the MAX195 ideal for
AC applications and signal processing. To support
these and other related applications, Fast Fourier
Transform (FFT) test techniques are used to guarantee
the ADC’s dynamic frequency response, distortion, and
noise at the rated throughput. Specifically, this involves
applying a low-distortion sine wave to the ADC input
and recording the digital conversion results for a
specified time. The data is then analyzed using an FFT
algorithm, which determines its spectral content.
Conversion errors are then seen as spectral elements
other than the fundamental input frequency.
Signal-to-Noise Ratio and
Effective Number of Bits
Signal-to-Noise Ratio (SNR) is the ratio between the
RMS amplitude of the fundamental input frequency to
the RMS amplitude of all other ADC output signals. The
output band is limited to frequencies above DC and
below one-half the ADC sample rate. This usually (but
not always) includes distortion as well as noise compo-
nents. For this reason, the ratio is sometimes referred to
as Signal-to-Noise + Distortion (SINAD).
The theoretical minimum ADC noise is caused by quan-
tization error and is a direct result of the ADC’s resolu-
tion: SNR = (6.02N + 1.76)dB, where N is the number
of bits of resolution. A perfect 16-bit ADC can, there-
fore, do no better than 98dB. An FFT plot of the output
shows the output level in various spectral bands. Figure
25 shows the result of sampling a pure 1kHz sinusoid at
85ksps with the MAX195.
By transposing the equation that converts resolution to
SNR, we can, from the measured SNR, determine the
effective resolution or the “effective number of bits” the
ADC provides: N = (SNR - 1.76) / 6.02. Substituting
SINAD for SNR in this formula results in a better mea-
sure of the ADC’s usefulness. Figure 26 shows the
effective number of bits as a function of the MAX195’s
input frequency calculated from the SINAD.
If your intended sample rate is much lower than the
MAX195’s maximum of 85ksps, you can improve your
noise performance by taking more samples than neces-
sary (oversampling) and averaging them in software.
Figure 27 is a histogram showing 16,384 samples for
the MAX195 without averaging, with an ideal “noiseless
conversion,” and with a running average of five sam-
ples. The standard deviation is 0.621LSB without aver-
aging and 0.382LSB with the running average. If fewer
data points are needed, normal averaging (e.g., five
data points averaged to produce one data point) can be
used instead of a running average, with similar results.
Figure 22. Supply Bypassing and Grounding
Figure 23. Power Dissipation vs. Conversions/sec When
Shutting the MAX195 Down Between Conversions
