conversions have some chance of being outside this range.
In addition, the differential linearity error of each code and
the quantization performed by the converter result in histo-
grams which can deviate from the ideal. Figure 14 shows a
histogram of 5,000 conversions from the ADS7813.
Keep in mind that the time period when the comparator is
most sensitive to noise is fairly small. Also, the peak portion
of the noise “event” produced by a digital transition is fairly
brief as most digital signals transition in a few nanoseconds.
The subsequent noise may last for a period of time longer
than this and may induce further effects which require a
longer settling time. However, in general, the event is over
within a few tens of nanoseconds.
AVERAGING
The noise of the converter can be reduced by averaging
conversion results. The noise will be reduced by a factor of
1/√n, where ‘n’ is the number of averages. For example,
averaging four conversions will reduce transition noise by
half, to 0.3LSBs. Averaging should only be used for low-
frequency signals.
For the ADS7813, error correction is done when the tenth bit
is decided. During this bit decision, it is possible to correct
limited errors that may have occurred during previous bit
decisions. However, after the tenth bit, no such correction is
possible. Note that for the timing diagrams shown in Figures
2, 5, 6, 7, and 8, all external digital signals should remain
static from 8µs after the start of a conversion until BUSY
rises. The tenth bit is decided approximately 10µs to 11µs
into the conversion.
For higher frequency signals, a digital filter can be used to
reduce noise. This works in a similar manner to averaging:
for every reduction in the signal bandwidth by two, the
signal-to-noise ratio will improve by 3dB.
QSPI INTERFACING
APPLICATIONS INFORMATION
Figure 15 shows a simple interface between the ADS7813
and any queued serial peripheral interface (QSPI) equipped
microcontroller (available on several Motorola devices).
This interface assumes that the convert pulse does not
originate from the microcontroller and that the ADS7813 is
the only serial peripheral.
TRANSITION NOISE
If a low-noise DC input is applied to the ADS7813 and 1,000
conversions are performed, the digital output of the con-
verter will vary slightly in output codes. This is true for all
16-bit SAR converters. The transition noise specification
found in the Specifications section is a statistical figure
which represents the one sigma limit of these output codes.
Using a histogram to plot the number of occurances of each
output code, the distribution should appear bell-shaped with
the peak of the curve representing the nominal output code
for the given input voltage. The ±1σ, ±2σ, and ±3σ limits
around this nominal code should contain 68.3%, 95.5%, and
99.7%, respectively, of the conversion results. As a rough
approximation, multiplying transition noise by 6 (±3σ) will
yield the number of unique output codes which should be
present in 1,000 conversions.
Convert Pulse
QSPI
ADS7813
CONV
PCS0/SS
MOSI
BUSY
DATA
The ADS7813 has a transition noise figure of 0.6LSB,
yielding approximately 4 different output codes for 1,000
conversions. However, since ±3σ is only 99.7%, up to three
SCK
DATACLK
CS
EXT/INT
3291
CPOL = 0 (Inactive State is LOW)
CPHA = 1 (Data valid on falling edge)
QSPI port is in slave mode.
FIGURE 15. QSPI Interface to the ADS7813.
Before enabling the QSPI interface, the microcontroller
must be configured to monitor the slave select (SS) line.
When a LOW to HIGH transition occurs (indicating the end
of a conversion), the port can be enabled. If this is not done,
the microcontroller and A/D converter may not be properly
synchronized. (The slave select line simply enables commu-
nication—it does not indicate the start or end of a serial
transfer.)
832
821
33
0001h 0002h
23
FFFDh FFFEh FFFFh 0000h
0
0
0003h
FIGURE 14. Histogram of 5,000 Conversions with Input
Grounded.
®
ADS7813
16
BB [ BURR-BROWN CORPORATION ]
