ADS1120: Differential ADC Recommendation

Hi tedex45,

The ADCs mentioned make a differential measurement based on one input relative to the other so it is possible to measure +/-5V even if the absolute input voltage is between 0V and 5V.  If the AINP input is 5V and the AINN input is 0V you will read out a result of + full scale (+5V).  If the AINP input is 0V and AINN is 5V, you will read out a result of - full scale (-5V).  However, we need to be clear on the actual voltage you desire to measure relative to a common ground point. 

The ADS1118 cannot accept an input voltage below ground as the supply is unipolar.  In the case of the ADS1120, the analog supply operation can be either unipolar or bipolar.  In the unipolar supply case, the input cannot go below ground as AVSS would be connected to ground.  In the bipolar supply case you could connect the ADS1120 to +/-2.5V . 

In either the bipolar or unipolar supply case while using the ADS1120, the full-scale range would be +/-5V for a 5V reference.  However, you cannot measure beyond the supply rails, so the absolute input voltage will be limited to +/-2.5V relative to ground for the bipolar supply operation.  If the sensor output is +/-5V relative to ground, then some method must be used, such as a voltage divider, to bring the voltage into the correct input range of the ADC.

So the first question is, what is the true sensor output?

Best regards,

Bob B

Hi Ed,

+/- 6.144V is simply the full-scale range of the ADC which determines the value of 1 code or count.  It does not mean you can apply an analog input voltage greater than VDD + 0.3V as this will damage the device.  So to measure +/-5V with the ADS1118, you will need to choose the +/- 6.144V setting and use 5V (or maximum of 5.5V) for VDD.  So the internal scaling does not necessarily mean you can actually measure the full-scale range.  The same would also be true for the +/- 4.096 V setting using a VDD of 3.3V.  Even though the scaling or the output code value relates to the full-scale range of +/-4.096 V, you cannot apply a voltage greater than VDD + 0.3V.  See the note under Table 7 on page 26 of the ADS1118 datasheet.

As far as anti-aliasing filter, you should have one.  There will be the potential of some aliasing at the data output rate, but in most cases designing a filter for the output data rate is a bit overkill.  The actual device sampling rate is the modulator frequency (250 kHz) and this should be your primary consideration.  See section 10.1.3 of the datasheet for more detail.

As the ADS1118 uses a switched capacitor input sampling you want to limit your series resistance to prevent error of the filter resistor voltage drop.  I would try to limit my filter resistance to 1k ohm maximum (for each input  resulting in 2k for the filter cutoff calculation).  Also keep in mind the effects of using large capacitors as the caps used my have significant temperature coefficient and DC bias effects.  I recommend using C0G/NPO caps if possible.  Obviously using small resistances and small values of capacitance will limit the cutoff frequency.

Another filter consideration relates to power line cycle noise (50/60Hz) as the ADS1118 does not attenuate these frequencies well and have the potential for aliasing.

Best regards,

Bob B

Hi Ed,

This is kind of a tough question to answer as I know nothing about your sensor, the data rate or environmental conditions.  From an engineering perspective there can be over engineering if the theoretical limits are used but the conditions of the system do not warrant it.  For example, 96dB might be the ideal at fMOD/2, but how large of a signal is anticipated at that frequency in the first place?  So you really need to have a full understanding of your system.  It may be required that due to external noise sources you need a higher order filter.  It is possible to create an effective 2nd order RC filter for the ADS1118.

If you are concerned about 50/60 Hz rejection, then you may want to reconsider using the ADS1120 which does have this filter capability at 20sps.  If 50/60 Hz is a problem, then you will need to adjust your filter accordingly. 

Best regards,

Bob B

Hi Bob,

My signal passband that I want to measure is 1kHz. That means I would need to use an ADC with a data rate (f_DR) of at least 2kHz right. So the ADS1120 would be right for me as the ADS1118 has only a max. of 860SPS. 

Placing my anti-aliasing RC filter with cut-off frequency at e.g. 5kHz should be enough to not attenuate my passband but still enough to attenuate unwanted signals at 125kHz (f_mod/2). I guess that the attenuation at 125kHz will now be lower since the cut-off frequency is raised, but since I am not expecting large signals in that frequency range anyways, I should be fine. In applications, where these frequency ranges do matter I guess one would need an more complicated active filter to per se achieve the -96dB attenuation. Does this sound right?

With this configuration, I should be able to accurately measure signals up to 1kHz with no aliasing?

Thanks again,

Ed

Hi Ed,

I believe you have the right concepts.  However, keep in mind that the FIR digital filter response for the ADS1120 is not flat-pass but closer to a SINC filter response so you will start to see some roll-off even at 1kHz (see figure 59 in the ADS1120 datasheet.) 

Best regards,

Bob B