are having a problem with the gain calibration. The part is being used in
bipolar mode. At Vin- a voltage of 1.25V is applied, which is derived from the
internal reference. The voltage at V+ varies between 0V and 2.5V.
If a gain calibration is carried out, the ones density is not 62.5% as quoted
in the datasheet but 75%
In the uncalibrated case, the sensor has an amplification factor error (gain
error?) of ca.9%, which cannot be corrected with the gain calibration.
It's difficult to say whether the results which you are seeing are out of
specification, because the AD7720 is specified with an ideal digital filter at
the output. In fact the datasheet only mentions the transfer function of the
AD7720 (analog in, single-bit digital bitstream out) on page 13 in the "Offset
and Gain Calibration" section. Ideally, the ones density should range from
37.5% for a zero scale input (-Vref/2 in bipolar mode, 0 in unipolar mode)
62.5% for a full scale input (+Vref/2 in bipolar mode, +Vref in unipolar mode)
I would be interested to know the uncalibrated transfer function of the AD7720
which you are using. The "calibration functions" on the AD7720 are not very
The MZERO pin forces the input to zero scale in unipolar mode or mid scale in
bipolar mode, allowing you to measure the error and compensate for it
externally. Likewise, the GC pin only forces the input to full scale, allowing
you to measure the error and compensate externally.
The way you use this data depends on your application. Usually you would carry
out a zero scale calibration first, compensate for the offset error and then
carry out a full scale calibration to compensate for the gain error. Is this
what you have done? If not, how are you calibrating the AD7720?
There are some basic considerations to follow in order to get accurate results
from a sigma delta modulator.
1) Input drive circuitry must be able to drive the dynamic input impedance
correctly. This either requires an op amp with a very high slew rate or a
circuit such as fig 23, which uses the anti-aliasing circuit to isolate the
dynamic nature of the input. If the dynamic input impedance is not correctly
driven (i.e. if the source has a high output impedance), this usually results
in a gain error (gain too small). Could this be causing the problem?
2) If you are using the internal reference, you must take into account its
errors. If the internal reference is not accurate enough, you should use an
external reference such as the AD780.
3) Antialiasing circuitry. You have to consider the possibility of high
frequency components being aliased into the passband. Do you have adequate
antialiasing circuitry at the input of the AD7720?
4) Applying an input signal that is too large will cause the modulator to be
overloaded and result to be inaccurate. In extreme cases, the modulator goes
unstable. It takes some time for the output to recover from instability, during
which time DVAL goes low. Overloading can occur with low frequency baseband
signal but it can also be caused by high frequency signals beyond the Nyquist
frequency being aliased down into the baseband.
5) General grounding, decoupling and layout precuations. Have you followed the
recommendations given in the datasheet on page 13?