MP4245 reported current inaccuracy

DC-DC Power Converters
1

Hello. I am using an MP4245 in a design to generate a 12V power rail from a battery. So far everything has worked fine for us, with the exception of the current reported by the regulator via I2C. I am using Virtual Bench (v4.0.45.1034) to monitor the output current while using an electronic load to test the output current capabilities. Unfortunately, the current reported by the MP4245 is consistently lower than the actual current. It is about 11% under the actual value when delivering about 25% of the Iout Limit value (4.8A), and only as good as 4% when around 75% of the Limit value. I understand that the expected range should be +/-5%, so I’m mostly concerned about the error at the lower end.
I have the EV4245-VE-00A evaluation board, which I programmed via I2C to output 12V and matched my other configurations from the MP4245. When I do the same exact tests, I get -5% error at 25% of the Limit value, and around -1% at the 75%+ range. I modeled my MP4245 layout and design off the EV4245, because the application was quite similar. Interestingly, I measure the exact same voltage across the sense resistors on both boards (20mΩ, which results in 20mV for 1A, 40mV for 2A), so this seems like an issue with the device incorrectly reading the voltage as opposed to the current bypassing the sense resistors.
What could be causing this discrepancy? I made Kelvin connections from the sense resistors as was done on the EV4245, but even if I didn’t I’d expect the current to be higher than the actual, not lower. But the fact that I measure the same voltage across the ISENSE pins of both the EV4245 and MP4245 is odd, since that only leaves voltage sensing circuit inside the IC. The only difference between the EV4245 and the MP4245 circuits is that we used a different resistor divider on the FB pin, which results in a different Vout Scale Loop value.
Below is a table comparing the MP4245 and EV4245 actual and reported current readings (*100% the value is the closest I could get to ILIM without causing OC).

Device 25% Current (A) Reported Current %Error 50% Current (A) Reported Current %Error 75% Current (A) Reported Current %Error 100%* Current (A) Reported Current %Error
MP4245 1.2 1.078 10.17 2.4 2.266 5.583 3.6 3.469 3.64 4.72 4.625 2.01
EV4245 1.2 1.141 4.92 2.4 2.391 0.38 3.6 3.594 0.17 4.5 4.578 1.73
2

Did you build your board identical to the EV board? What accuracy do you expect? I can’t see an accuracy spec anywhere in the datasheet, although I didn’t look too hard. Usually current measurements errors are dominated by offset which is why your low current readings are so bad. The nice thing about digital stuff, is you can perform a calibration, slope and offset and this is good so long as the numbers aren’t terribly temperature depandent.

3

The board isn’t exactly identical in terms of layout, but I did follow it in broad strokes (used power planes, as much of the BOM as was possible, similar component proximity to the EV layout, etc.). I saw on the spec sheet that the ISENSE OC threshold was about 5% with a 20mΩ sense resistor, which isn’t exactly a direct spec for the IOUT readings, but it did match with the behavior I saw on the EV board. I am mainly wanting to check if there is something obvious I missed, before I start applying software calibrations. The discrepancy seemed significant, especially when the higher current readings are very close to the real on the EV board, and my layout never gets better than 4% and is as bad as 11%.