ADS8694: Accuracy reacheable based on SBAA243

Hello Benjamin, We are looking into your query and will get to you soon

Hello Benjamin,

The accuracy achievable is dependent on a number of factors including the accuracy of  the shunt resistor, the gain and offset errors of the current sense amplifier, the gain and offset errors of the ADC, the accuracy of the reference, and the common-mode voltage errors of the current sense amplifier.

Offset errors tend to dominate when the input current is close to zero, and gain errors dominate when the signal is close to full-scale. In most applications,  the user is able to perform offset calibration by performing a null measurement, and therefore in most cases the dominant error may be the gain error of the current sense amplifier (+/-0.20% maximum).

The typical total gain error of the system is calculated by performing the square root-sum-of-square (RSS) of the typical gain error specifications of the different components in the signal chain.  The worst case gain error is the addition of the max gain error specifications of the devices in the signal chain.  The worst case error is a very conservative estimate.

The INA240 voltage supply is VS=5.5V maximum; therefore is output is limited to +VS-0.2V; and VGND=0.10.  The ADS8694 optimal scale setting would be 0V-5.12V.

The design shown on SBAA249 was able to achieve a gain error 0.056% typical  (based on the RSS of typical gain error datasheet specs  of the devices) and 0.281% worst case error (based on the addition of the maximum gain errors of the components) after performing a 1 point offset null calibration.  Below is an example of a typical/max gain % error calculation of the design of SBAA243.

http://www.ti.com/lit/an/sbaa243/sbaa243.pdf

Thank you,

Kind Regards,

Luis Chioye  

Thank you for your quick reply,

Our requierements concerning the current are :
- 0.5% Accuracy on converter from 400mA to 10A.

I think the upper analyse is done for 10A setpoint, hence my question about the ADC reference change by an ADS8694 to benefit of PGA.
With PGA, we can increase the accuracy when we measure a low current.

BR,

HI Benjamin,

The ADS8694 has an integrated PGA that drives the 18-bit SAR ADC. Since the ADS8694 operates from an analog +5V supply, the PGA main purpose is to scale and level shift the single ended signal and convert it into a fully-differential signal, while driving and charging the internal S/H capacitor of the SAR ADC. This simplifies the design, since the user does not have to design a front-end amplifier to drive the ADC. The scales available are: +/-10.24, +/-5.12. +/-2.56, 0-10.24V and 0-5.12V.

Since the shunt amplifier (INA240) output swing is limited to 0-5V (5V supply), you would choose the 0-5.12V scale setting. The ADS8694 will provides 18-bit resolution at 0-5.12V scale; where the Least Significant Bit weight is 19.53uV. The ADS8694 provides SNR=90.5dB at the 0-5.12V scale setting; this translates to an equivalent noise of 54.03uVRMS at the ADC input.

In general, the dominant noise contributor limiting the resolution in the system comes from the shunt amplifier (not from the ADC). Looking at the noise simulation shown on p.5 SBAA240, the noise contribution from the shunt amplifier (INA240) is 103.2uVRMS at its output (G=20V/V).

Since the ADS8694 (or ADS8910) resolution performance is higher than the shunt amplifier, the resolution of the measurement is limited by the shunt amplifier noise performance, and increasing gain will not improve resolution. Averaging samples may help reduce noise, increasing resolution. Also, performing a null or offset calibration will increase accuracy.

Thank you,

Best Regards,
Luis Chioye

Hi Luis,

Thank you for these precisions.

I understand that ADC has a better resolution than shunt resistor, PGA can't counter this problem.
For me, PGA permits to reach full scale when we measure a low amplitude signal, we increase noise and useful signal.
With an average filter, we could increase accuracy on low amlitude signal.

I think your resultats (typical and worst case) is based on 10A measure and not on all range (correct me if I wrong).
However, as I precised on my last message, we are looking for a solution to reach an accuracy of 0.1% on a current mesurement from 400mA to 10A.

BR,

Benjamin