• State TI Thinks Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 3 answers
  • Subscribers 46 subscribers
  • Views 865 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.

ADS1278 Gain Drift Excessive Over Narrow Temperature Range

Cory Wilson
Prodigy 50 points
Other Parts Discussed in Thread: ADS1278, ADS1274, THS4531A, THS4531, THS4521, REF2025

We have noticed the gain drift on the ADS1278 to be as high as 1000 ppm/c.  For instance, converting a DC voltage on a channel while the temperature is increasing shows a notch like response in the converted value versus temperature.  To verify it is not the input signal, we also captured the differential input to the ADC with a DMM while sweeping in temperature.  A plot of the converted value vs temperature shows a notch at around 1000 ppm/c at a specific temperature.  Once the notch is passed by continuing to drive temperature up, the ADS1278 works fine.  In fact, it works fine for most of the temperature range, there is a very specific temperature for which the gain seems to drift excessively. We see this on other channels as well but at different specific temperatures.  We have successfully used the ADS1278 in other designs.  My question is, what kinds of things can contribute to such excessive gain drift in the ADC that occur at very narrow temperature ranges while the greater ADC temperature performance is inline with 1.2 ppm/c?

  • 0
    TI__Guru 57276 points
    Hi Cory,

    Sorry for my delay in responding.

    Are you providing the DC input voltage directly to the ADS1278 channel inputs, or are you using driver amplifier as a front-end input stage? I ask because the inputs to the ADS1278 are un-buffered, so there will be a voltage transient present on the input pins with every sample. The settling of this transient will depend on how fast the ADC is sampling and how much bandwidth your input driver has, which will in turn affect your system gain error. The gain error drift of your driver may dominate the system gain error drift over that of the ADS1278.

    Also, in what increments are you increasing the temperature? If you could share your plots, that may also be helpful.

    Best Regards,
  • 0 in reply to Ryan Andrews
    Prodigy 50 points

    Ryan,

    So good to hear from you!  Here is the drive to the A/D:

    AIN6_P and AIN6_N feed the A/D with the addition of a 1.5nF differential capacitor on the input.  

    Here is a picture of the gain drift:

    We accumulate 2109 samples over one second and get the accumulated counts per second as you see above.  Bear in mind, in tandem, I have captured the AIN6_P and AIN6_N in tandem with an external DMM and have the response to be quite flat unlike what is converted by the A/D.  The temperature is incremented continuously at about 15 deg./hr.  We have found that regardless of the temperature slope ramp, the notch like response occurs very consistently at the same temperature.  Also, we see this same phenomenon on other channels.  We have two prototype tools and we see it on each tool with each channel notching at a different temperature and magnitude. Again, we have lifted this design for another product and the results there are good.  We also use the ADS1274 and it works quite well in a similar application.  Any insight would be much appreciated.  We are trying something new everyday in hopes to find the issue.  We have try more reference voltage capacitance, different firmware, different power supply and different voltages.  In fact, we have found that this problem only occurs for positive voltages and seems to work fine for <=0 voltages.

  • 0 in reply to Cory Wilson
    TI__Guru 57276 points

    Hi Corey,

    First, I just want to verify what your plot is showing. On the y-axis, you have "Counts per second." Is this the sum of 2,109 decimal output codes? In other words, around 33 degrees C, your average ADC output was 4.428E9/2,109 = 2.0996M, which is about 625.72mV. If that's correct, I estimated your gain error drift to be roughly 140uV/C.

    I believe the way we specify gain error drift is a bit different than how you are measuring it. At each temperature (usually in increments of ~20degC), we collect and average 2 sets of samples with a DC input, first with a positive full-scale input , then a negative full-scale input. We take the difference of these 2 averages to remove the offset and get the slope at that temperature, thus our gain error.

    The typical gain error drift in the datasheet is the average gain error drift over the entire temperature range, calculated by averaging the gain error over range of min to max temperature. If you look at gain error drift in finer increments, you may see results that vary from the typical spec. I think Figure 33 is trying to show this as the Gain Error is plotted for each channel over temp. While I don't see as large of a dip as what's shown in your graph, this could be responsible for some of it. Since you're not removing the offset by testing at positive and negative full-scale, that may also be a contributing factor.

    Lastly, regarding my earlier comments about ADC input settling, the THS4531A is a rather "slow" FDA with only 36MHz unity-gain bandwidth. You may try rerunning your tests with a slower ADS1278 modulator clock frequency (fMOD) to ensure the inputs have enough time to settle after every sample.

    Best Regards,

  • 0 in reply to Ryan Andrews
    Prodigy 50 points

    Ryan,

    The reference voltage we are using is 1.25V so the voltage is not 625.72mV but rather 317.95mV.  Otherwise, you are correct  And, as you can see from the plot of accumulated codes vs. temperature, there is upwards of 300uV "notch" depth in the conversion output.  Keeping in mind, I have measured the differential input to the A/D with a 5.5 digit DMM spitting out data on RS232 for comparison and I can see no dip.  However, I take your point about the THS4531 and I believe we can experiment with the THS4521 to see how that fares.  We are still at a loss here as we have nearly look at every possible angle including SPI timing.  I have tried more capacitance on the differential reference input, increased the reference voltage from 1.25V to 2.5V, monitored all voltages, and compared datecodes on the A/D.  I have placed the system on the bench and applied heat directly to ONLY the A/D and can reproduce the problem quite readily.  When it cools back down, it does so exactly retracing its steps.  Any other ideas we can try while I wait for the THS4521?

  • 0 in reply to Cory Wilson
    TI__Guru 57276 points
    Hi Corey,

    Thanks for clarifying. Yes, with a 1.25V reference, that "notch" accounts for about a 300uV shift in the ADC output. Again, this may be the combination of offset and gain error drift as well as input settling. Is it possible for you to provide a DC input directly to the ADC inputs, bypassing the front-end driver stage? You might try repeating your tests by providing a DC positive and negative full-scale input to the ADC at every temperature to look at the gain error drift. Also, tying the ADC inputs together and biasing them to mid-supply will allow you to look at just the ADC offset over temperature to see if that's changing drastically.

    Best Regards,
  • 0 in reply to Ryan Andrews
    Prodigy 50 points

    Ryan,

    Right, we are considering driving the ADC input with a supply for further testing.  I have one important thing to add now that we are at this point in the discussion.  If I configure the ADC input voltage such that negative codes are gathered, that is, apply -317.95mV we do NOT see the problem.  Further, for any <=0 input voltage we do not see the problem.  Obviously, for any >0 input voltage we see the problem.  I know this is telling me something, but we are still scratching our heads on that.

  • 0 in reply to Cory Wilson
    TI__Guru 57276 points

    Ok. I'm not sure what that tells me either. How are you setting the input common-mode voltage? Remember that the ADS1278 is using a unipolar supply, so the absolute voltage on each input pin cannot be negative. The THS45xx driver will take care of that for you by driving the VOCM pin (which defaults to mid-supply if floating). If you try bypassing the driver, I would fix one of the ADS1278 inputs to mid-supply, and then swing the other input above/below mid-supply to generate the positive/negative differential inputs.

  • 0 in reply to Ryan Andrews
    Prodigy 50 points
    Ryan,
    Ok, I will adjust in the manner you suggest. Also, I believe the attached picture of our drive amp is shown where the VOCM pin of the THS4531 is driven by a precision reference REF2025 at 2.5 V.