AD7730: Increase in resolution

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

I am currently using your AD7730 parts to interface directly to a 2mV/V strain
gauge bridge with an excitation of 5V.  We would like to use this type of part
for many reasons however,
it is proving that the resolution achievable is not high enough.  I am idealy
looking for 20 bits at around
5Hz and maybe 18 at 100Hz.  I am happy to consider another device and perhaps
some external analog amplification before conversion if you feel this is the
only option.


First thing to do is compare the expected resolution with measured resolution.

2mV/V bridge with a 5V supply gives a 10mV full-scale. With the PGA set for
gain of 128, CHOP enabled and the output update rate set to it’s minimum value
of 50Hz, the noise is 40nVrms. The effective resolution is 10mV/40nV = 250k
codes or 17.9bits. The peak to peak resolution is 10mV/(40nV x 6.6) = 37.8k
codes or 15.2bits.

The AD7730 and AD7730L remain state of the art for weigh-scale and bridge
signal conditioning.

A couple of options for improving resolution are:

· Quickest and easiest solution is  to increase the excitation voltage.

· Use a digital post filter. The minimum output update rate of the AD7730 is
50Hz, using a digital filter (a simple moving average filter implemented in a
DSP or micro) to decimate to 5Hz and low pass filter can increase the
resolution by up to 2.5bits. Doubling the sample rate improves signal to noise
ration by 3dB or 0.5bits.

· You could look at building a low noise pre-amp. But you need to ensure that
the output referred noise of the preamp is approx 3x less that the noise
contribution of the AD7730 in the gain range to be used.  For example, using
the AD7730 in gain of 16, noise is 115nVrms. You need to build a preamp with
output referred noise less than ~40nVrms. For a gain of 4, you need input
referred noise of 10nVrms or 66nVpeak to peak. The AD797 with typical  low
frequency noise of 50nVpp could be used but you would need to compensate for
the offset, input bias current, temperature drift of this opamp. You also need
to remember that a 1k resistor at room temperature has noise of 4nV/root(Hz).
There’s a lot of work in designing a low noise preamp.