ADS1262: Charging/Discharging Behaviour on Measured Signal

Hello Chris,

Thanks for your reply!  

No, I don't have any components on the inputs (other than the described).  No diodes or other.

I wanted to use the PGA in its bypassed setting in order to have maximal temperature range.  I will redo the experiment tonight with it set to 1 to verify this hypothesis.  Could you please direct me to the section that describes the input bias current relationship?

To me, it really seems as if there is some kind of power/decoupling/resonant or feedback mode that has been setup.  The waveform looks like a classic charging/discharging response to some step stimulus.  Not sure what would generate a stimulus/step threshold every hour or so though...  I initially suspected that the temperature of the resistors was causing some of this drift but given the specified drift values this seems like a stretch...

No joy.  Ran similar experiments where I set the gain to 1 but enabled the PGA and a second set where I set the gain to 4 (PGA enabled obviously).  It seems to have changed the time constant but, perhaps surprisingly, not the peak 2 peak magnitude of the oscillation between the gain=1 PGA enabled vs. disabled case.  When gain is set to 4 the peak 2 peak oscillation magnitude decreased slightly.  To me, this really seems like some kind of feedback that is being setup.  Perhaps a supply or supply decoupling issue.  I should mention that I have not performed a calibration yet as I assumed that the calibration is simply a zeroth and first order adjustment and wouldn't effect transient, periodic effects such as what we're seeing.  Could I send you my layout/circuit diagram perhaps?  I am hoping that there is some kind of silly mistake there...

Hi Neal,

For input bias current specifications, take a look at the "ADC1 ANALOG INPUTS" section on page 8, as well as figures 25 through 30, starting on page 16. You'll notice that figures 25 and 26 show much larger bias currents when the PGA is bypassed.

Thanks for trying the test again with the PGA enabled! I had figured that the input bias current would be the mostly likely cause of the 20 uV offset, but if you still see it with the PGA enabled then I'd be inclined to look for other external causes. I'd also agree with you that calibration wouldn't change this behavior, it would only help to remove the initial offset.

I'd be glad to look at your layout, feel free to send it to me via pa_deltasigma_apps@ti.com.

A few other questions I have: 

• Have you tried cleaning your boards? We've seen solder flux residue cause strange offset drifts before:

• Is your resistor bridge on the same board as the ADC, or is it connected to the ADC via long wires? Long wires can act as antennas and pickup ambient noise or signals.
   
• What is providing your 5V reference? Is it stable? Since you've configured your measurement ratiometrically, I would expect that your circuit would be fairly immune to any changes in the reference voltage. However, it is difficult to get 100% matching between the signal and reference inputs, so a change in the reference voltage may still have a small effect on the measurement.

A few other troubleshooting other ideas...

• As a sanity check, you could short the inputs together (connect both inputs to one side of the bridge) and measure the offset over time. If the offset is stable, then you know that the periodic offset is a real signal that is external to the ADC. 
  
  
• You might also try removing the X2Y caps just to rule them out as a possible source of leakage current. Perhaps they could have been damaged during soldering.
  
  
• If you have a large ESD bag, you could try placing your circuit in the bag and closing it up (as best you can). The bag can isolate the circuit from air currents, and if you ground the bag it will also act like a Faraday cage.

Best regards,
Chris

Hello Chris,

Thank you very much for your insightful suggestions.  I have sent you my schematic and board layout, please let me know if you need anything else.  Regarding your suggestions:

- We scrubbed the boards with alcohol after soldering and they look very clean under the microscope.  I will give it another go just to be certain.

- The entire ADC circuit is on the same PCB with the same ground plane.

- The 5V reference is questionable.  I assumed since it was a ratiometric configuration that I would gain some immunity but perhaps not.  I was wondering about this myself so I ran the following experiment: as  mentioned before, the sensor with the fixed resistors is on the same SPI bus as two other sensors (identical except they have thermistors in lieu of fixed precision resistors. BTW, they are also exhibiting nearly identical fluctuations, just much larger magnitude.)   I configured these extra sensors so that one read the analog voltage and the other read the digital voltage (relative each sensor's internal reference).  I have attached the figures generated.  Please note that this test is run simultaneously.  There is definitely some correlation between the oscillation measured on Sensor 1 and the voltage drift measured on the analog and digital voltages of sensors 2 and 3 respectively but it is hard for me to know how much source drift is acceptable....

- Shorting the inputs seems like an excellent test.  How should I best do this?  Is there a way to do this internally to the chip by setting the MUX perhaps?

- Good idea.  Was planning to start removing caps in general and retest.  Will do and send data.

- Good idea.  Will do.