• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 613 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

PGA970: Temperature compensation for 3-wire and 4-wire connections

Alessandro Cominelli
Intellectual 375 points
Part Number: PGA970

Dear Mr or Mrs,
I am working on a project involving PGA970. The goal of our electronics is to permite the use of 4-wires LVDS or, alternatively, of 3-wires LVDTs (hald bridge).
In our project we want to use the PGA970 to drive two LVDTs at the same time, one demodulated by S1 and one by S2 channels, so we monitor the secondary of one LVDT on S1 input and the secondary of the other one on S2 input, but we never monitor the excitation.

We are in an advanced phase of the project and we are now facing issues with temperature drifts. We found in software user's guide (application note SLDU024 PGA970 Software User's Guide (ti.com)) an example of temperature compensation firmware (page 18 of SLDU024) and we also found the same example in the reference firmware provided for the ARM M0 microcontroller integrated in the PGA970. As far as I understand, the example is intended to work for 5-wire connection, that is connecting one secondary of the LVDT to one demodulator input (S1) and the other one to the other demodulator (S2) and then calculating the measure as PADC = (S1-S2)/(S1+S2).

Can you confirm this?

Unfortunately this is not the connection we are using, so we require an alternative algorithm, able to teke in account the temperature drift of both waveform output and demodulator input.

Is there an alternative algorythm I can use to compensate the measurement in case I use 4-wire connection or 3-wire connection (only reading the LVDT secondary coil)?

I saw that the temperature-compensation formula involves many coefficients saved in FRAM (h0EE, h1EE, h2EE, g0EE...n2EE), which I imagine have different values from chip to chip.

What do those coefficint stand for? Maybe we could  use those coefficint also for our purpose, but we require to understand better what they represent.

Do you have any suggestion?

Thank you in advance.

Best regards,

Alessandro

  • 0
    TI__Mastermind 18900 points

    Hi Alessandro,

    We have no other firmware or algorithms to offer. Please see this FAQ with more information: https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/802433/what-is-the-e2e-support-model-for-the-pga900-and-pga970 I have heard of customers trying to measure a 4-wire or 3-wire instead, so it at least has been attempted. We're not experts in firmware, which is where most of the effort lies.

    Section 6.1 of the software user guide talks about the example compensation algorithm you are referring to. You should be using AINx inputs for an external temperature sensor (the performance of the internal sensor is explained in the datasheet and the feedback has said that it is not as good as expected). Doing calibration over multiple temperature points should be good enough to calculate reasonable coefficients for most customers' use cases. 

    Best,

    -Cole

  • 0 in reply to Cole Macias
    Intellectual 375 points

    Dear Cole,

    thank ou for your answer.

    At this point I think we will do a characterization of our board at various temperaturs.
    I have a further question: do you think that the behavior of PGA970 at varying temperature strongly differs from chip to chip? Or is it almos the same for all components? This is an important point for us, beause a similar behavior between different PGA970 would permit us to avoid calibrating any PCB as a function of temperature.

    Thank you in advance.

    Best regards,

    Alessandro

  • 0 in reply to Alessandro Cominelli
    TI__Mastermind 18900 points

    Hi Alessandro,

    All the drift specifications are in the datasheet. You're looking at 100s of ppm/C for the dominant sources of error. You can do an error analysis with root mean sum of squares to quantify the error. Remember, I wouldn't find the worst case error. Typical worst error is much better of a calculation. The theory is discussed in this TI Precision labs video: https://www.ti.com/video/series/analog-to-digital-converters-adcs.html Titled: Statistics behind error analysis of ADC system

    I've definitely heard of customers doing 1T calibrations so others would seem to agree with you. It all depends on what you can or cannot accept.

    Best,

    -Cole

    edit: added some extra info related to a different device

  • 0 in reply to Cole Macias
    Intellectual 375 points

    Dear Cole,
    We tried measuring the actual temperature drift due to the PGA970, using a mini temperature chamber stabilized at +/- 0.1°C and we made a readout of ADC3 connected to the internal temperature sensor at different temperatures.
    To get data in °C from raw data we exploited the formula contained in the PGA970 datasheet (chapter 7.3.1.7 Internal Temperature Sensor) and we found very different values from different PGA970 samples. In particular, we used 3 identical PCBs, all inserted in the same chamber and we interrogated the 3 PGAs. We found that the raw data are extremely similar between PGAs (less than 2% difference in the values registered from different chips); nevertheless, we found that calibration coefficients are different, so at the end we got very high differences in measured temperatures. For instance, in a measurement with chamber fixed at 20°C, we registered 28°C on one PGA970 and 51°C on another one, that is almost double! The third PGA970 measured 35°C, showing still a big difference with respect to the others.
    I double checked calculations and I could not find any error there.
    Do you expect such a big difference in temperature measurement?
    Thank you in advance.
    Best regards,
    Alessandro

  • 0 in reply to Alessandro Cominelli
    TI__Mastermind 18900 points

    Hi Alessandro,

    Alessandro Cominelli said:
    Do you expect such a big difference in temperature measurement?

    In short, no. Most of our customers can get <1% accuracy from what I've seen.

    Alessandro Cominelli said:
    We found that the raw data are extremely similar between PGAs (less than 2% difference in the values registered from different chips);

    I was going to ask for this. If they're very similar then I don't have much else I can do for you. It sounds like you have some bug in your calibration routine or firmware implementation. As I previously commented, I cannot help with those types of inquiries. 

    Best,

    -Cole