Post Go back to editing

AD2S1210 Input voltage scaling

Hi all,

I plan to use the AD2S1210 to implement a generic resolver measurement solution.

My customer recommends support of a wide range of transformation ratios and exitation voltages.

Thus, I an in need of an input voltage scaling solution in case the signal from the resolver is larger than the AD2S1210 is capable to deal with.

I found an article that suggests the following solution:

What do you think about using a (dual) digital potentiometer as Ra?

Regards, Niels



Adding tags for internal tracking purposes - SK 7-Oct-19
[edited by: @skowalik at 1:37 PM (GMT 0) on 7 Oct 2019]
Parents
  • Niels,

    Can you please post the original circuit article in this thread?   I would like to read through the text to make sure I'm not misinterpreting the circuit diagram.

    The one thing I would caution is that depending on the resolution/performance requirements for your system is that you are aware of both the tolerance and drift specifications for your selected components.  Mismatched components can introduce error terms in your output which are position dependent and could complicate calibration/manufacturing.   You may be better off having a range of standard components that you populate based on the resolver selection.   By selecting components with a similar footprint you can choose the absolute, tolerance and drift values that best meet the application need.  

    Hope that helps.

    Sean

  • Hello Sean,

    here it is:

    PDF

    It's the "Analog Dialogue 48-03", March (2014).

    You can also download it here:
    www.analog.com/.../precision-rtdc-measures-angular-position-and-velocity.pdf

    Regards, Niels

Reply Children
  • Neils,

    I reviewed the document and my understanding of the diagram was correct.   I do however stand by my recommendation to use standard values for your production runs and to adjust the attenuation factor to suit the particular resolver transformation ratio.  

    Sean

  • Sean,

    I found dual potentiometers with a 0.1% resistance match between its two channels (e.g. AD5262).
    At a first sight, this sounds good. Fixed resistors also have 1% or 0.1% tolerance.
    Can you please explain your reasons for refusing the potentiometer approach?

    Populating different resistor values to adapt the AD2S1210 to the resolver output voltage is not an option. The end-product will be used with several different resolvers, resulting in a wide range of possible input voltages on the SIN and COS pins. It is essential that it can be software-controlled adapted to 4-34Vpp at the SIN and COS inputs. This means: no soldering. :-)

    What kind of signal conditioning do you suggest for such application?

    Niels

  • Niels,

    My primary concern is gain mismatch between the SIN and COS channels and the potential for that mismatch to change versus temperature.  However, I had not considered using a dual potentiometer such as what you've described.  With the trim resistors in the same form factor they are likely going to match fairly well from both an absolute and temperature perspective so you are probably going to be okay.  Also if there is a sufficient number of taps in the digipot you can probably also take advantage of the digital interface in your calibration.   Just be mindful that mismatches between the attenuation stages of the SIN and COS channels will introduce positional errors which vary at twice the angular displacement, thus you'll have relative maximum errors at 45 degree steps.

    Sean

  • Sean,

    according to equation (9) of the above posted document:

    I came to the conculsion that a gain mismatch of 0.1% between the SIN and COS inputs will result in an error of app. 5LSB when running the AD2S1210 in 16 bit resolution.

    Do you agree?

    Niels

  • Sean,

    I wonder if there is any other kind of signal conditioning you would suggest for RDC applications that require software adjustable, wide range, voltage level adaption?

    Regards,
    Niels

  • Niels,

    Apologies for not getting back to you sooner.   I think the most obvious thing to be to be concerned with any time you are dealing with voltages outside the rails of an active component is the potential for an over voltage stress to cause damage to your device.   However, this has been addressed fairly well in the circuit note and would not require software to adjust.

    The only other option I can think of is the output excitation gain.   In your set of resolvers the likelihood of encountering not only variable transformation ratios but also different excitation requirements will likely mean you need to change your excitation output gain as well to obtain optimal performance. I would take a look at circuit note CN0276 for the output buffer design.

    Sean

  • Sean,

    thank you for your help.

    Let me resume our talk to help others having the same question:

    Using the signal conditioning approach shown in "Analog Dialogue 48-03", March (2014) with a dual potentiometer as RA is a possible way to achieve software adjustable input voltage ranges for SIN and COS inputs of the AD2S1210.

    It is critical that the dual potentiometer channel-to-channel resistance matching is as good as possible. A 0.1% missmatch would lead to an error of app. 5LSB when running the AD2S1210 with 16 bit resolution.

    Be aware that the AD2S1210 data sheet states an angular accuracy of +/-10 LSB  for the device (A, C grade).

    By the way, offset errors are more disastrous than gain errors. 0.1% Offset error (4mV for an Full-Scale input signal into the AD2S1210) leads to app. 10 LSB positional error.

    In addition, you suggest to implement a software adjustable excitation voltage swing, since resolvers can have very different excitation voltage swing requirements. CN0276 shows the output buffer design for the EXC outputs of the AD2S1210. Due to the differential nature of the excitation output, it  is necessary to change the gain of two OP-Amps. Again, the gain matching between EXC+ and EXC- is important. A dual potentiometer with good channel matching is a feasible approach.

    Regards, Niels