Actions
Forum Thread Details
  • State Verified Answer
  • Replies 6 replies
  • Subscribers 42 subscribers
  • Views 430 views
  • Users 0 members are here
Related

How to interpret MAX11203 AD-converter self calibration bitvalue?

Engztrom

Hi,

What is the exact relationship between the self calibration gain bitvalue and the actual scaling it does?
E.g.what bitvalue would be measured with a saturated input (maximum voltage measured) and
the self calibration gain set to, say, 0xBFFFFF and the self calibration offset set to 0x000000?
The datasheet is very vague about that. This is all it says:
:
"SCGC: Self-Calibration Gain Register

The self-calibration gain register is a 24-bit read/write register. The data written/read to/from this register is clocked
in/out MSB first. This register holds the self-calibration gain calibration value. The format is always in two’s complement
binary format. A write to the self-calibration register is allowed. The written value remains valid until it is either rewritten
or until an on-demand self-calibration operation is performed, which overwrites the user-supplied value. Any attempt
to write to this register during an active calibration operation is ignored.
The self-calibration gain value is used to scale the self-calibration offset corrected conversion result before the system
offset and gain calibration values have been applied, provided the NOSCG bit in the CTRL3 register is set to 0. The
self-calibration gain value scales the self-calibration offset corrected conversion result by up to 2x or can correct a gain
error of approximately -50%. The gain is corrected to within 2 LSB."

And it also mention SCGC=0xBFD345 and SCOC=0x00007E as an example of values you can get when the self calibration is carried out.

Thanks for any help!

  • 0

    Hi,

    Just adding an answer for now unless you find a better one.. I don't know the relationship you request, but I'm curious, is it important? If it is not in the datasheet, it may vary slightly from part to part. I have not read the datasheet, but it seems to me that if you're not doing auto-calibration and are instead doing self-calibration, then any value could be written and then adjusted/trimmed up or down until the output is at the expected value. If it takes too long to trim then an algorithm could be coded to home in to a value within (say) a couple of dozen value attempts worst-case.

  • 0

    I had a look at the data sheet yesterday and it didn't make too much sense to me either !

    I think your best bet is to get the part connected up with an adjustable DC input and some means of tweeking the registers and reading the ADC  - then have  a play with it to see what it does,

    The thing that confused me about the desctiption was the refence to the register values being 2s compliment - which implies signed and then the example gain settings are negative. My guess is that the offset is 2s compliment but that the gain is unsigned - but you'll need to test it.

    MK

  • +1 in reply to shabaz

    It is important because (whatever the datasheet says) it seems to be necessary to have some kind of calibration values set.
    In theory (as we interpret the very fuzzy datasheet) it should work fine skipping both self calibration and system calibration (this is actually the default setting) and settle with the 60% scale that will then be given (presumably to have margins for component variations), but... we can't get that to work at all!
    Without any calibration factors set the output from the AD-converter becomes very non-linear (why??)
    And with the ratiometric connection we use it's not possible to carry out any of the built-in calibration methods either.
    But it seems to work fine when writing down the hardcoded calibration values 0xBFFFFF and 0x000000 (or something close to that) and then do our own factory calibration in software. But what exactly does these written down bitvalues represent?
    And will it always work with these values, with respect to component variations?
    (I.e. what are the margins?)
    .
    .

  • 0 in reply to michaelkellett

    Yes, the datasheets are very crude, confusing and (possibly) erroneous!
    In order to implement self- or system calibration we would have to tweak the hardware so ratiometric connection is not used during calibration (that is Plan B...)

  • +1 in reply to Engztrom

    Hi,

    Regarding "But what exactly does these written down bitvalues represent?"

    That's my point - why does it matter? _If_ you deem that you have to do calibration anyway (whether it is self calibration, or auto calibration), then that can be achieved with zero knowledge of what the values represent, although as Michael says, you could just try it and see if you can identify the pattern as you adjust the value.

    Regarding "And will it always work with these values, with respect to component variations?" that would be an assumption that I do not think one can make unless it was explicitly in the datasheet. Are you trying to assume that it might be similar within a batch or something? I think it is expected to be a different calibration value per part.

  • 0 in reply to shabaz

    Well, if it only worked as the datasheet suggests, then a 60% scale should give sufficient margins without calibration, which would be sufficient for us.
    But if calibration values to get 60% must be written down (despite what the datasheet says), then I would have to know what values to use for 60%.
    Else we would have to launch "Plan B"...