Because of the holidays, TI E2E™ design support forum responses will be delayed from Dec. 25 through Jan. 2. Thank you for your patience.

  • State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 45 subscribers
  • Views 898 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

ADS1220: ADS1220 acting like 14 bit

Chris Good
Prodigy 40 points
Part Number: ADS1220

I am using a ADS1220 connected over SPI to an FPGA. I have it set up as follows:

AVss = DVss = GND

AVdd = DVdd = 5V (the data sheet shows better noise performance with 5v vs 3.3v)

AREFP = 3.3V

AREFN = 1.65V

AN0 = my signal, ranging from .1V to 3.2V or so

AN1 = 1.65V

20hz continuous mode, everything else like PGA defaulted. So no PGA, no gain, etc.

My results are pretty much like I expect them. However, through a lot of my range var away from 0v, it seems like the ADC is only really using the most significant 14 bits. then the next 4 don't change as expected in any consistent manner, and then the last 6 bits are noise as expected. This seems repeatable with both a precision potentiometer and a transducer I am using. As you can see below, with a potentiometer I get steady relatively noise free conversions, just missing 4 bits that should be useful data.

This is the potentiometer. You can see it works in discrete steps, and I can't hit anything between

This is my analong out pressure transducer, it shows a similar output just more noise

I don't think it is a SPI issue because those 15-18 bits work fine when AN0 is close to 0. I'm doing no weird byte collection, just shifting in 24 bits. My thoughts are it is something involving a reference difference minimum, but the only one I can find is 0.75V.

  • 0
    TI__Guru** 113765 points

    Hi Chris,

    Welcome to the forum!  Let's review one thing at a time.  You are telling me that your reference is 1.65V (3.3-1.65V).  You also said that the only change from default is that you are using continuous mode.  This means that you have PGA enabled and not disabled.  With PGA enabled (default) this means that at some voltage inputs you violate the input common-mode restrictions.  However, I don't think that is the issue you are seeing.

    Looking specifically at the potentiometer, what is the value/make/and part number of the pot you are using, and how do you have it connected?  The steps seem to be very even for each change based on your plot.  Most likely the pot has discreet intervals of steps and areas in between each step cannot be measured. 

    As to the pressure measurement, there could be several possibilities.  In the pressure plot, what is the red line indicating?  Also, how stable is the pressure you are applying?  And is the pressure being compared to the ambient pressure or some other reference pressure?  Even though the pressure appears to move in steps, there are still areas where there is considerable noise even within areas that appear to plateau.  It may be more beneficial to use the same voltage for both excitation and the ADS1220 reference so that the result is more closely related to a ratiometric measurement.  Any noise on the 3.3V or 1.65V of the reference input will directly impact the measurement of the sensor.

    Instead of using the pressure sensor with an analog output, you may find you can achieve better results using a bridge type output sensor measuring differentially.  You can apply gain at the ADS1220 to achieve better noise results.  Also, the measurement can be made ratiometric by applying the same voltage to the sensor excitation and the ADS1220 reference.  This will improve results with noise and drift.

    Best regards,

    Bob B

  • 0 in reply to Bob Benjamin
    Prodigy 40 points
    That is correct, PGA is enabled, gain of 1. Good catch. I'll try both disabling the PGA and separately setting AVDD to 3.3V, just to see if either makes a difference.

    The pot is this guy: www.digikey.com/.../3534145 a 10 turn 1k pot. One end hooked to gnd, one to 3.3v, and the wiper to AN1. I think I occasionally am seeing the steps of the pot when I'm careful enough turning it, but in that specific image the steps were identical in size to the steps I was getting out of my transducer. Could be a coincidence, but unlikely.

    The red line is just my pot. I have 2 different ADC's hooked up to my fpga, and I had them both plotting to compare results. It's a honeywell HSC series, differential +/- 10 inh2o. In this application one side was atmospheric but blocked with a mostly sealed short piece of tubing, the other side is connected to 70ish CC closed area, so light temperature increases from a solenoid are what is causing the pressure to slowly rise. The same 3.3V from a linear regulator is currently the excitation for the sensor and VREFP. Right now I just have a resistor divider from that to make my 1.65V VREFN. If I settle on that setup I'll be getting a regulator that has a midrange out.
  • 0 in reply to Chris Good
    Prodigy 40 points
    I tried hooking the Pot up with one side hooked to 1.6v and one side to 1.7v, so that I had more accurate control over the voltage. Now I'm more convinced the ADC is working as intended as I have tight control within the previous steps. I now believe the Honeywell sensor, even though it's ratiometric analog out, has discrete step sizes. I don't see any mention of this in the HSC datasheet, but I've seen it elsewhere on other honeywell sensor datasheets. However, that's not a problem with my ADC so I'm going to have to move to honeywell on this one. Looks like they also have a new line, RSC, which has this same line of ADC built in. Interesting.

    Thanks Bob for your help.
  • 0 in reply to Chris Good
    TI__Guru** 113765 points

    Hi Chris,

    I guess I'm a little confused.  You have a resistor divider to create 1.65V that you apply to REFN, and from your first post you stated you have 1.65V connected to AIN1 as well.  Is this the same voltage source from the resistor divider?  If so this could be part of the problem.  What are the resistor values you are using?

    I think there may be a weird combination of current draw related to input voltage and bias current creating a current divider between the analog input and the reference.  I would try simplifying the reference first by either using the supply or the internal reference.  I would also try and use some other source other than a resistor divider for the AIN1 input.  If you use PGA disabled you can either tie AIN1 to AGND or use the mux selection AIN0 to AVSS.

    Another option would be to add a buffer amp between the voltage divider output and the ADS1220 inputs.

    Best regards,

    Bob B

  • 0 in reply to Bob Benjamin
    Prodigy 40 points
    Yes it is the same source currently. The resistors are about 6k. I did this so that the actual voltage I care about can swing above and below in order to get the full range of the transducer. Before when I just had REFN tied to ground, I lose the MSB of my conversion because my positive AN never swings below negative AN. I need at least 17 bits noise free to hit my resolution spec.

    I've had AN1 tied to ground, as well as has REFN tied to ground, and used the internal reference, and I have the similar issue of stepping. However, now that I think about it each time it stepped about the same amount of pressure once I did the math on the conversion, so maybe that's just one more reason to believe it's the transducer. I just didn't expect to have an analog transducer have discrete steps.