I'm trying to use the PCA9955B for testing of LED characteristics. The datasheet says the current accuracy is 8% chip-to-chip and 6% per channel on-chip. Are the errors always a constant percent that can be tested and calibrated, then compensated in software, or are they non-linear or temperature dependent?
Hi Sam,
I have asked our product group to comment on this, here is the answer:
The matching characteristics are defined in the datasheet with Rext=1kohm, which gives higher values of current and feedback voltage for the current control circuitry to maintain the current. When Rext is increased, the possible output current and likewise feedback voltage is LESS than when 1kohm was used. The output current sensor that controls the output current has an amplifier with an offset voltage at its input. The offset voltage is a major contributor to the matching error. Thus, anytime the maximum possible output current is used, best matching occurs, since the feedback voltage is a much greater value than the offset voltage. When a value of Rext greater than 1kohm is used, the sensed feedback voltage is now a lower value, thus the offset is a bigger percentage and causes a bigger percentage error. Thus, optimum matching is obtained with minimum Rext. This is a well understood phenomena in sensors and op-amps, and perhaps should have been made more clear in the matching section of the datasheet.
Hope it helps.
Best regards,
Tomas
Thomas,
Thanks for the quick feedback. Would tuning rext with a potentiometer or similar be the best way to resolve this, or would I have to implement external current measurement and compensate through software? Is the quoted 6% value measured with rext 1kohm?
Thanks,
Sam