INA701: INA701 current sensing larger error than expected

Hi Mitch,

Sorry for the delay, been very busy here.

Some clarification: The DMM is in a fixed range, so range switching won't be causing this. The DMM is calibrated. I just retested with an extra DMM in series and their readouts are well within a 3% margin (in fact below 0.1%).

Indeed the 3% is in between the 100mA and 1A range, that does not mean that it is only in that range. Below 100mA it becomes significantly larger, this might be down to resolution. I did not measure beyond 1A as the customer is not interested in this. 

I've taken 100 samples from the INA702 (see screenshot below for the setup of the EZShunt app), downloaded the CSV and averaged this in a spreadhseet. DMM shows an average well over 100 samples. 

I'm pretty confident that the DMM is not the problem (since another one in series indicates the same). I'm willing to believe that the INA701 accuracy specs are genuine, yet there is a large discrepancy between the DMM readouts and the EZShunt app readout. 

I think this could be caused either by:

- Some error in the setup (which would be interesting since current can't go anywhere in a closed loop)
- Some error in the EZShunt app and/or incorrect settings. See screenshot below for the settings used during the measurements.

How can we identify what's going wrong here? 

Kind Regards,

Martin Koster
Sofrides IO

Hello Martin,

Can you try again with the conversion time and averaging settings on the INA701 set to max? Also, make sure there is a good ground between the INA and the supply, DMM, and Load.

Regards,
Mitch

Hi Mitch,

Will do, but could you explain what you mean with good ground between DMM and the others? The DMM is part of the current supply on the high side, if I connect one of the sides of the DMM to GND we'd have a shortcut from supply to GND...

Kind Regards,

Martin Koster
Sofrides IO

Hey Martin,

Ya, sorry, I shouldn't have included the DMMs in that list. The important part is to have the load and supply grounds to be the same as the INA ground. It's a common mistake for people to not share the ground of the load with the INA. 

Regards,

Mitch

Hi Mitch,

Yes, load and supply grounds should be the same. I already found out in earlier INA projects :-) I guess finite CMRR and infinite CM can lead to interesting problems...

Setup was connected as you suggested with good GND between load and source. But...

Meanwhile I've found another clue that might be relevant. So far I've been using the "Collect data" function in the EZCurrent tool, with 100 samples, exported these to CSV using the button and had Excel to calculate the average over said 100 samples. This results in the over 3% discrepancy. 

Today I ignored that entire graph bit, and read values directly from the register map and converting it, assuming 190uA / LSB conversion factor. Now suddenly the discrepancy very decent, dropping below 0.1% getting towards the 1A range. 

Looks to me there is a bug in the EZCurrent tool for the current calculation. Bus voltage seemed perfectly OK. If I look at the graphs that the tool shows, I get the same impression. E.g. for a current set at 50 mA, most datapoints shown in the graph are above 50mA, almost none are below. 

Any suggestion how to confirm this? I don't have source code of EZCurrent of course, so can't check there. Should be relatively easy for you to repeat the test though.

Another note: INA701 datasheet august 2025 shows two different conversion factors for the current register. In 6.6.1.5 it specifies 480μA/LSB, table 7-4 however specifies 190µA/LSB. The latter seems correct. Is this an error in the datasheet?

Please advize, and thanks again.

Kind Regards, 

Martin Koster
Sofrides IO

You're welcome!