ADIS16209 Accuracy Problems and Assembly issues

Hi,

we have a program in which there is an ADIS16209 as the mail tilt sensor.

we have tested several units via the evaluation board and they performed to spec.

we had made a PCB design which includes the ADIS16209 on-board (according to app notes and the datasheet rev-c).

in this architecture the ADIS is getting stable and filtered 3.3V from a switched DC/DC converter (ripple is lower than 15mV)

we had made 10 assemblies via automated SMT. from the 10 units only 3 ADIS units were functional!!

we had x-ray tests to see the solder on the pads and the solders were looking good but the units were none functional.

we had opened a new and seals batch of components and made another 7 SMT units built (taking the soldering profile, SMT aliment etc' under great care).

in this batch we had all 7 working but not according to spec or expectations...

we configure the SMPL_PRD to 41 and the AVG_CNT to 8 (the max value) and our testing facility's temperature is 23C-25C:

here are some questions:

  1. in all 10 units were experiencing jitter of about 4-6 LSB in total stable conditions - in the Evaluation units the jitter were about 1-2 LSB
  2. in most units we have a drift in measurements that behaves as if the sensor is measuring a larger angular degree:
    for example: we have a measuring jig which is leveled to 0.1mRad and there we have accurate wedges for 5, 10 and 15 degrees, those wedges are also very accurate. the PCB with the ADIS16209 is placed on a metal bracket with leveled planes which are placed on the testing equipment.
    we see a measurement drift of about 0.05 degree per every angular degree, meaning, for every actual 1 degree our ADIS16209 sensors are measuring 1.05 degree, therefore 5 actual degrees are measured as 5.25, 10 actual degrees are measured as 10.5 and 15 actual degrees are measured as 15.75, etc'
  3. we have larger deviation in the positive side than in the negative side - for example we measure -14.5 and +15.7 at actual 15 degrees
  4. we see that the Y incl axis has significantly higher errors than the X incl axis (the data above represents the Y axis were the X axis shows about 20% less deviation....)

we had gone through all calculations, checked and rechecked the SPI interface and we are currently in a very problematic situation - we have some 100 units which we are not assembling because we don't know why the first 10 are not operating according to expectations....

waiting for any help possible!!!

Thanks

  • 0
    •  Analog Employees 
    on Jun 20, 2014 6:08 AM

    I am sorry to hear that this is not going smoother for you, but I think that we can help.  To start with, I want to share some things that we have observed, with respect to the sample rate and filtering settings on the ADIS16209.  Using a sample rate setting of 0x0041 will result in a sample rate of 62.5 SPS, which does not meet the sample rate criteria for a system that has a single-pole filter with a cut-off frequency of 50Hz.  Can you consider changing this back to the factory-default sample rate setting of 0x0014 (SMPL_PRD) register and try this again?

  • 0
    •  Analog Employees 
    on Jun 20, 2014 6:10 AM

    By the way, the previous post was specific to the elevated noise observations.  For the apparent 5% scale error in the measurements you are reporting, can you start by reading register locations 0x1C and 0x1E?  I would expect them to be zero, but we should start by verifying them.  They provide access to an offset correction on the accelerometers, prior to the angle computation.  There is probably more going on with this, so I will give this some more thought.

  • Hi Mark,

    with regards to SMPL_PRD and the AVG_CNT:

    What is the mathematical relation between the SMPL_PRD and the AVG_CNT ? (please provide an example calculation)

    What are the default values ?

      We would like to use the sensor in static mode for tilt measurement, we would like Narrow LPF – what are the optimal settings for this scenario in terms of the specified registers?


    Thank you

  • 0
    •  Analog Employees 
    on Jun 27, 2014 3:33 AM

    The factory default settings were selected to support the type of application you describe.  What lead you to change them?

    The relationship between SMPL_PRD and the sample rate is illustrated on page 14 of the ADIS16209 datasheet:

    http://www.analog.com/static/imported-files/data_sheets/ADIS16209.pdf#Page=14

    AVG_CNT has no impact on the sample rate but does impact the low-pass filter response associated with the averaging filter.  The relationship between the total number of averages and the AVG_CNT setting is also located on page 14 of the ADIS16209 datasheet:

    http://www.analog.com/static/imported-files/data_sheets/ADIS16209.pdf#Page=14

    Does this help?
    Best,

    NevadaMark

  • Dear Mark,

    We have tested the values of the registers you had asked for:

    We got 0x3F51 in PROD_IDI and we got 0 and 0 in 0x1C and 0x1E.

    We are in a time pressing situation and we are no closer of understanding what's the problem with our assembled ADIS16209.

    PLEASE FIND THE TIME FOR US TO TALK.

    I feel that's the only way you could understand the whole picture fast and accurate to help us overcome this issue.

    Please let me know how and when can I contact you.

    Thank you very much