UCD90120A: UCD90120A Current measurement inaccuracies?

Hello

Which MON pins are used for IOUT measurement/

Please be noted that MON10-13 can not measure any voltage below 200mV as described in the data sheet.

This may impact your accuracy on the low current.

Thanks

Yihe

I am using MON1-MON9 for current measurements.

Hello

There are many factors to contribute accuracy. Do you have the probe to check the voltage on the MON pin. there could be some ripples which are not measured by the current meter.

• Reference voltage inaccuracy due to V33A ripple
• Noise on MON pin
• Grounding noise
• High input impedance to MON pin.  There is a leakage current in / out of the input.  In this case, it can be as high as +/- 5uA.  If you have a source impedance on the input of 1Kohm, this can contribute another 5mV of error.  .

Regards

Yihe

Thank you for the suggestions Yihe. I do not believe any of the factors you identified explain the inaccuracies I have seen.

1. The UCD90120 is powered by a low noise linear regulator (TPS7A8300), through a low pass filter (2 ohm, 4.7 uF) to the V33A pin of the UCD90120A, so I do not think noise on the reference is an problem.

2. I also do not think noise on the MON pin is the problem, as I currently have nothing running on this board except the UCD90120A and UCD9246E and several switching regulators that these devices control. All regulators are at no load except the particular rail I am testing.

3. Similar to above, I don't think ground noise is an issue here. I have 10 contiguous ground planes, and again, nothing else currently operating on this board except the UCD90120A and UCD9246E.

4. High input impedance on the MON pin is definitely not a problem, as I am driving them directly from an INA214 current sense amplifier.

I'd again like to emphasize that the theoretical measurement resolution on the particular rail that I provided data on is 0.12 mA, but I am seeing a resolution of 15 or 16 mA. This is 2 orders of magnitude worse that it should be. I think that either the UCD90120A firmware, or the Fusion Digital Software is dropping least significant bits, or truncating somewhere.

I'd also like to note that I am seeing similar current measurement inaccuracies on output currents measured with the UCD9246E.

Finally, I'd like to share the results of an experiment I did that may shed some light on this problem. I experimentally changed the "Iout Cal Gain" from 5000 to 5. The reported values were 1000 times greater than expected, of course, but after dividing the reported values by 1000, the accuracy was MUCH better as shown below.

This could be a viable work around, except that at with only 3 digits to the left of the decimal point reported by the Digital Fusion software, this would limit the current I could measure to 999 / 1000 = 0.999 amp.

Hello

Thank you for the further test and update.

If you have a current sensing at 100x gain, the IOUT CAL Gain shall be set to 10 instead of 5000 or 5.

with a IOUT_CAL_GAIN =10mohms, the measured current by UCD9012A is = 1000/10 * the voltage on the MON pin.

I checked the design, The reason is in order to save memory to provide large range detection, only 6 bit are used for fraction which is equal to 1/64 = 15.625mA. The IOUT resolution is 15.624mA. You test result also matches this.

Regards

Yihe

Thanks again Yihe.

I have a 50 milli-ohm sense resistor and current sense amplifier with a gain of 100 (on this particular channel), so the gain should be 5000 milli-ohm. This was validated by the data in the first table I sent. With a load (as measured by external equipment) of 0.498 A, and the gain set to 5000 [milli-ohm], the Fusion Digital Power software reported a value of 0.499 [A].

This software "design", using only 6 bits for the fraction, is quite problematic for me.  The part has a 12 bit ADC, and I was counting on that resolution, but the software is limiting me to only 6 bits. Is there any way we might request a software update to correct this?

Thanks again,

Tom 

Hello Tom

As long as the value reported from GUI matches the result, the gain is good.

We do have a 12bit ADC, but we only have 16 bits to represent the result, excluding the signed bit, only 15 bits can be used.

For the current sensing, most customers need large current(couple hundreds), that's why the resolution is not so important for large current.

At this point, the workaround you mentioned in the previous post would be the best solution.

Regards

Yihe

Hello Yihe,

This has certainly NOT been solved. The UCD90120A is advertised to have 12 bit resolution, and I was counting on that performance in my application. But due to the software implementation, I am only getting 6 bit resolution. Frankly, I don't think this is acceptable and am still looking to TI to provide a solution.

Thank you,

Tom