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.

LM5066I: Telemetry linearity issues

Part Number: LM5066I
Other Parts Discussed in Thread: LM5066

Tool/software:

Hello,

we have a design where an LM5066I is used to switch 48V. In our application, power switching requirements are pretty low, usually well below 40W.

To get some acceptable power measurement accuracy, we use an Rsense value of 47mR.

So for testing what I did is connect an electronic load to the circuit. That load can be configured to draw a constant current (and I verified with a multimeter that the current values are indeed correct), and I did that and increased the current in 50mA steps, each time measuring the voltage across Rsense and reading out the raw Iin ADC measurement values from the device (register 0xD1/0x89) (CL set to 50mV).

I noticed a few things:

  • Values are not really consistant. If I keep the current at the same value and do 100 readouts of the register, about 3 out of 10 measurement values are totally off. Meaning the value is not +/- 10% off, but is instead zero or double of what it should be. At first I thought maybe the I2C is faulty and that is why I am reading invalid values, but I can for example read out the MFR_MODEL register 10000x in a row without any errors. Also, the I2C signals look fine on a scope. So this seems not to be a problem with the I2C signal, the device just gives invalid data sometimes.
  • If I plot the raw register value from the READ_IIN register against the current, what I get is not a linear plot, but something with steps in it. There are ranges of Rsense voltages that give me the same ADC value. Looks something like this (looks the same across boards and multiple measurements):

  • Also, when I adapt the calculation coefficients according to Rsense = 47mR (table 48) and then do the conversion of these ADC values to actual current values, what I get is mostly off the expected value by quite a lot (at least a lot more than the expected +/- 3.5% accuracy according to the data sheet). With a plot like the one above a linear fit will not help much to increase accuracy.

So I do not really understand what is going on here. Is this a layout issue, is it a device issue, am I misunderstanding how the device works and this is expected behaviour?

Here is an image of the circuit we're using. Note that originally, we were using the LM5066 and then later switched to the LM5066I. But except different calculation coefficients and a different Rpwr value this should not have any impact.