• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 63 subscribers
  • Views 445 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.

UCD9240 out of calibration

EDGAR BROWN
Intellectual 450 points
Other Parts Discussed in Thread: UCD9240, UCD9248

We just completed a design with the UCD9240 (80-pin version) and we are in the testing stages. The design is as close as being 'by the book' as it gets with the caveats that the documentation is not consistent with respect to the connection of some pins (e.g., leave floating, or connect a resistor/capacitor to them). We have used development board schematics (from TI and XILINX) to resolve those inconsistencies. To minimize any risk, we went as far as using TI recommended parts for nearly all design decisions.

The modules are running off a regulated 5V power bus (which comes from yet another TI module).

The same exact issues were seen in three separate test boards (all three prototypes tested).

Among the many issues we have had:

- All measurements (inputs and outputs) are consistently 10% below the programmed value. We have had to 'lie' to the fusion interface to get them to the proper set points.

- The 3.3V regulator has a 3.1V output. We had to reduce the bias resistor to get it to 3.3V, and this node is overly sensitive to the resistor value. And yes, we followed all the equations in the design guides, the value we had to use is below the minimum given by the equations and nearly an order of magnitude below the maximum acceptable value.

- The BPCAP node, measured with a high-impedance voltmeter, is at 1.65V (instead of the nominal 1.8V). Which could explain the rest of the problems.

These parts were supposed to be factory-calibrated to better than 5%, and I don't see a calibration constant that would remove such gain/scale error. What is going on here?

  • 0
    Intellectual 450 points

    After talking to TI this seems to be a known (albeit undocumented) problem with the UCD9240. So for the sake of other designers out there:

    The problem originates from insufficient bypassing of the analog reference section. If the bypass capacitors are not close enough to the analog supply/gnd pins, or the traces are slightly too long or with one too many vias, the reference section settles to a wrong value (1.65V in my case), which affects all other IC variables.

    The solution is to place the capacitors as close as possible to the analog supply pins, and perhaps add a capacitor or two just to be sure. I have another design that works correctly, in which I added a separate ferrite bead and bypass capacitors just for the analog section.

  • 0 in reply to EDGAR BROWN
    Intellectual 450 points

    We added 0.1µF X7R capacitors across the analog supply pins on the original boards and this brought the reference output to better than ~2.7% from spec (instead of the ~10% we had before).

    After bringing the reference to the right level, the UCD9240 PWM frequency got to 980KHz (from the programmed 1MHz), previously it was in the 750KHz range.

    This fix was verified in four different boards with very similar results.

  • 0 in reply to EDGAR BROWN
    TI__Intellectual 1755 points

    All,
    We have found that the V33A power pin on the UCD92xx parts needs good, low ESR decoupling to ensure an accurate internal clock and reference.  This has been born out by Edgar's application, where he is using the 80-pin UCD9248 and by other customers using the 64-pin and 48-pin versions of the device.  We see good results if there is a 0.1 uF cap directly across the V33A and Agnd2 pins.  As is always good layout practice, there should be no vias between the cap terminals and the power/ground pins on the controller. 

    For existing designs where the decoupling is not as "tight" as it could be, we have found that increasing the capacitance on this pin also helps.  In at least one case increasing the cap on V33A from 0.1u to 1.0, improved the internal voltage reference noticeably.  In this case the sigma of the monitored temperature was reduced by ~4X with the larger decoupling cap.