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.

ADS1256: ADC calibration advice

Part Number: ADS1256
Other Parts Discussed in Thread: OPA189, OPA2189, OPA350, REF70, REF6025, REF7025

I am designing a calibration board for use with the ADS1256 and a 12-bit MCU SAR and I would appreciate some advice:

If I understand correctly, the ADS1256 SELFCAL can be used for any PGA setting, but is it based on the same 2.5V reference for all PGAs? In other words, does the ADS1256 have internal division of the reference voltage?

---

Is it reasonable to have e.g. a 1-year calibration interval, meaning that we can have a separate post-production/calibration device that is used once a year? Or, should the product itself incorporate a buffered resistor divider for each internal reference so that it can self-calibrate every time the product powers up? (This would of course still rely on the built-in 2.5V reference.)

---

Due to tolerances, if the precision reference base for the calibration board were an exact 2.5V, there would be both a difference compared to the less exact 2.5V reference inside the product and inaccuracy for each of the resistor divider-generated voltages. In some 50% of the cases this would result in overranging the ADS1256. Is such a slight overranging (probably below some 0.2%) a practical problem in terms of accuracy? The 'ideal FSC' has some margin before 'binary saturation' (or whatever it might be called).

Or, should we make sure that the voltage levels used for calibration are guaranteed to be below the less exact product-internal 2.5V reference? (This is perhaps more of a pedagogical task for the staff measuring and adjusting the calibration board: "Don't ask why, just adjust the calibration board to be 0.5% less than nominal values."...)

---

Is it a good idea to design such a calibration board with a 2.5V precision reference that feeds a number of resistor dividers with OPAMP unity gain buffer so that full-scale differential and single-ended measurements can be made of known full-scale input voltages for each PGA setting? This should enable not SYSGCAL but measurement data for adjustment of SELFGCAL so that the practical result would be equivalent to SYSGCAL?

(And another with a 5.0V or 4.096V precision reference for unbuffered calibration.)

Or would connecting the calibration board precision 2.5V reference to each product connector/interface in turn enable SYSGCAL without a need for all these gradually lower voltages?

---

Could you recommend a suitable OPAMP unity-gain buffer for the resistor dividers? If I understand correctly, such OPAMPs must be designed to work with high input impedance, which rules out some zero-drift, low-noise OPAMPs?

I am currently looking at OPA189/OPA2189, but are there other ones better suited for this?

---

Should you always incorporate a feedback resistor for the unity-gain buffer? Would 100 to 330 ohm be good as a rule of thumb or should it match the input impedance from the resistor divider?

---

Should there be an RC network at the output of each unity-gain buffer?

---

For the product's 2.5V reference, it is sometimes a good idea to use an OPAMP buffer. The OPA350 is commonly used as an example, e.g. in the ADS1256 EVM, but are there newer, higher-performance OPAMPs, that are currently recommended?

---

An unrelated question:

In this recent post Bryan Lizon86 says "you should not be measuring ground-referenced signals with the buffer enabled.".
e2e.ti.com/.../ads1256-differential-inputs-and-single-ended-inputs

Is this because the ADS1256 buffer is not precise near GND?
"With the buffer enabled, the voltage on the analog inputs with respect to ground [...] must remain between AGND and AVDD − 2.0V." (datasheet p15)

Best regards

Niclas

  • Hi Niclas,

    One point of clarification here: the ADS1256 does not have an internal reference. You mentioned "less exact 2.5V reference inside the product", and I was not sure if "product" meant the ADS1256 or the board you are designing, so I just wanted to be clear on that fact.

    There is a lot of information here, but I am a bit confused as to what your end goal actually is. Can you summarize quickly what you want to accomplish with the ADS1256? Are you trying to calibrate the ADS1256 specifically, or the entire signal chain? What will the ADS1256 be measuring in the final system, or is the ADS1256 only used for calibration?

    In general, I would say the more components you add to your system, the more error you are going to have to calibrate. Once you start adding resistor dividers, buffers, additional references, this might overshadow the error caused by the ADC. Do you have an error budget for your system? What other components do you have on your board (other than the ones you are considering for the calibration)? Since it is not really clear to me what you are trying to do, I am not sure how much calibration is really necessary or if it is worth the trouble to add all of this functionality to your board.

    You also asked a question about how often to calibrate, to which I do not have a good answer. Again, this would depend on what your error budget is over time and temperature, as well as the other components in your system.

    Some points about the ADS1256 calibration scheme:

    • The SELFGCAL uses the external reference voltage as the input source, and this voltage is internally scaled according to the PGA gain selected.
    • The SYSGCAL requires you to apply full-scale voltages to the input of your system. These voltages would change depending on the VREF voltage you choose, the PGA gain, and the buffer setting. If the buffer was off, VREF = 2.5V, and PGA = 1, then you would need to apply a 5V input. If the buffer was off, VREF = 2.5V, and PGA = 32, you would need to apply a 156mV signal.
    • The voltages applied for the SYSGCAL need to be ≤full-scale and >0.8*full-scale.
    • You do not have to perform the SELFCAL if you are going to perform SYSCAL as well, since the latter supersedes the former
    • The SYSOCAL requires you to manually short the inputs together.

    You can use a buffered reference such as the REF6025 with the ADS1256. We also have a new series of high performance references coming out called the REF70 series: https://www.ti.com/product/REF70. Once I better understand what the end goal is I can help you identify any required buffers.

    Finally, the ADS1256's internal buffer requires some additional headroom and therefore limits the minimum input signal to AGND. I would not recommend measuring a single-ended input with the buffer enabled since there is no tolerance on the input. Any slight variation might cause this input to dip below AGND, and then you would start operating in the buffer's nonlinear region, leading to bad data.

    -Bryan

  • Hi Bryan,

    Thank you for quick response and clarifying some things.

    I am aware that the ADS1256 does not have an internal voltage reference.

    The product in question is a handheld instrument. It is currently equipped with a MAX6126 2.5V reference directly feeding the ADS1256.

    (The REF7025 is good news. It will give MAX6325 stability at single-LiIon-cell supply voltage. This is AFAIK best in class so I'll replace the MAX6126 with REF7025 once it becomes available.)

    The product also has analog MUXes for two of the ADS1256 inputs. (The other six ADS1256 inputs are direct.) Several but not all of the various signal paths go via (different) connectors to cable-attached external sensors. In other words, I expect differences in terms of series resistance and offsets. So, SYSGCAL seems wise for calibration of the entire signal chain.

    ---

    The standalone (lab-located) calibration board is planned to contain MAX6325 (2.5V) and MAX6341 (4.096V) references. They are slightly better than MAX6126 and can be trimmed.

    It's the calibration board that would contain a number of buffered resistor dividers. I will follow your advice and set these voltages between 0.8*full-scale and 1.0*full-scale. (No overranging during calibration, so some 0.95*full-scale.)

    ---

    The calibration board signal paths is planned to consist of MAX14661 MUXes from 16 voltage levels via 2 internal 'lines' to 16 output pins. The product (=handheld instrument) will be connected to these 16 output pins when being calibrated. Shorting the inputs together is planned for certain calibration steps. Operation of the MUXes and a few relays for routing GND to the appropriate product connector will be automated.

    ---

    I'm making a new version of an existing product, originally due to the 8051-based MCU becoming obsolete, but there are lots of things that really should be done differently. E.g. transition from only doing single-ended measurement to mostly differential. Some of these S-E measurements are near-GND... This requires a lot of work because it also means redesigning the original subsystems. There is no error budget as of now. Just a great potential for improvement.

    Just to be clear: using the buffer for S-E measurements is OK so long as you stay some tens or a hundred of mV above AGND? It's near-zero S-E measurements that are unreliable, especially with buffer?

    Thanks

    Niclas

  • Hi Niclas,

    Thanks for the additional information, this helps clarify what you are trying to do.

    Regarding single-ended measurements: I would consider any input that is not referenced to ground, but to some other fixed potential, to be pseudo-differential. A single-ended measurement therefore is exclusively a ground-referenced input, so maybe this is just some confusion over nomenclature.

    But yes, if your AINP or AINN voltage was at 100mV above AGND, then this is allowed if the buffer is enabled.

    Since we have exchanged a lot of information in the preceding three posts, let me know if there is anything that I missed and anything that you still need support with. Thanks!

    -Bryan

  • Hi Bryan,

    Thanks. I think these questions are still open:

    Could you recommend a suitable OPAMP unity-gain buffer for the resistor dividers? If I understand correctly, such OPAMPs must be designed to work with high input impedance, which rules out some zero-drift, low-noise OPAMPs?

    I am currently looking at OPA189/OPA2189, but are there other ones better suited for this?

    Should you always incorporate a feedback resistor for the unity-gain buffer? Would 100 to 330 ohm be good as a rule of thumb or should it match the input impedance from the resistor divider?

    Should there be an RC network at the output of each unity-gain buffer?

    For the product's 2.5V reference, it is sometimes a good idea to use an OPAMP buffer. The OPA350 is commonly used as an example, e.g. in the ADS1256 EVM, but are there newer, higher-performance OPAMPs, that are currently recommended?

    Niclas

  • Hi Niclas,

    Since these questions are unrelated to the ADS1256 or data converters in general, can I ask that you post them in the amplifiers forum where one of our product experts there will assist you? I could offer you some advice from my experiences, but they would be able to get you the best answers to these questions.

    I would also suggest including a picture of what you are trying to do, including expected voltages in the system. For example, they will need to know what supplies are available, are bipolar inputs required, the resistances you intend to use, etc., in order to provide the best support.

    Link: https://e2e.ti.com/support/amplifiers/f/amplifiers-forum

    -Bryan

  • I received an automated e-mail asking me to respond to the suggested answer.

    As I see it, half of my quite complex question has been well answered, but the other half not at all. In the amplifiers forum I got answers to most of my OPAMP questions.

    To put things into perspective, in my recent experience, Texas Instruments has the best design support that is accessible to small and mid-size customers. I have previously received first-class support in this forum and also had a highly valuable dialog in the USB forum. Maybe my expectations were set too high now.

    It's perfectly acceptable to provide design support that only responds to detail questions about specific parts, but I would argue that it's good business to go the extra mile and also engage in design discussions with customers and use these discussions as triggers for production of application notes.

    The really important and interesting question is still open: Does TI have anything to say about HW design of a (system) calibration board for its AD converters? It is most likely related to amplifiers but it has everything to do with data converters.

    Some of the questions are:

    * Is a basic precision reference-to-resistor divider OPAMP buffer an adequate source of a stable calibration voltage? Do bells and whistles improve? If so, which ones?

    * Recommendations on grounding? Solid plane for stability, star grounding, or perhaps minimal copper for reduced heat transfer? Should a closed enclosure mean that heat equilibrium will be reached and hence heat transfer in a solid plane becomes a non-issue?

    * What's the value of an ambitious power supply design? Is there a point in temporarily switching to battery power when using the board?

    * Is solid-state signal MUX switching good enough or are dual-coil mechanical relays better?

    I would say this topic is a good candidate for an application note.

    Best regards

    Niclas

  • Hi Niclas,

    Thanks for the feedback. We always appreciate hearing that our support is valued, while always striving to improve where possible.

    Some of your questions do have answers (more like guidance, since the true "answer" will likely be different on a system by system basis). For example, this FAQ summarizes our recommendations when it comes to layout and grounding: https://e2e.ti.com/support/data-converters/f/data-converters-forum/755516/faq-pcb-layout-guidelines-and-grounding-recommendations-for-high-resolution-adcs

    Or, for your question about a closed system, this is generally not recommended because it will likely increase the temperature inside the enclosure, providing a positive feedback loop for self-heating and therefore higher temperature drift errors, higher leakage currents in ESD diodes, and many other unwanted effects. But if you needed a hermetically sealed enclosure because your system was exposed to a very dirty environment for example, then you would have to deal with these effects. In that latter case, you would probably steer the conversation toward minimizing power consumption, using higher accuracy components, and/or calibracting over temp and time.

    Some of your questions are a little more open-ended. For example, you asked the question "Is a basic precision reference-to-resistor divider OPAMP buffer an adequate source of a stable calibration voltage?" My question to you would be "how do you define 'adequate' "? Do you need 1000ppm stability? 100ppm? 1ppm? There really is no answer to this question without knowing what the goal is, and I believe you mentioned that you did not have an error budget yet. This is not wrong or bad, it just limits the help we can offer. We could suggest just using the best components available e.g. 0.001% resistors, etc., but that is probably not feasible for most systems and is not really helpful to you.

    As a result, our focus on this forum tends to be troubleshooting, with a secondary goal of limited system design support. Of course we try to help as much as possible, but you can imagine that without knowing all of the system constraints - timelines, power budget, error budget, size limitations, cost, etc. - it can be challenging to offer specific, meaningful advice at a system level (and many engineers are hesitant to share all of this information, which is understandable). To your point about questions turning into content, we do generally convert commonly-asked questions into FAQs on this forum, while more in-depth questions do get turned into application notes or other collateral. For example, about two months ago we released an app note related to RTD wire break detection that resulted from engineers asking questions about this topic (feel free to check it out if you are interested in learning more: https://www.ti.com/lit/an/sbaa483/sbaa483.pdf). I would also point you to our Precision Labs content, which is ever-expanding and includes a vast curriculum of fundamental and design knowledge about ADCs, DACs, amplifiers, and many other components: https://training.ti.com/ti-precision-labs-adcs. Much of this content began as discussion topics with other engineers, whether on this forum or via direct contact.

    I hope this helps clear up what our intent is with the support on this forum, as well as some of its limitations. Please let me know if you have any additional questions about this topic. If not, I will close this thread out, and if you have any new questions in the future, you can always start a new thread.

    -Bryan

  • Hi Bryan,

    Thank you. Let's close this thread now.

    Regards

    Niclas