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.

TIDA-00913: settling time for INA240

Part Number: TIDA-00913
Other Parts Discussed in Thread: INA240, TMS320F28069M


The following picture is in the TIDA-00913 reference design:

I have some questions about that:

1-)What we can do for avoidance of reading those transient events in the output of INA240? 

2-)The settling time for INA240 is 9.6-9.8 micro second. Is this settling time just for edges of PWM signals (changing from low to high or high to low);in other word, if we read the current in the middle of PWM signal, should we consider this settling time (acquisition window for 9.8 micro second) or we can immediately read the current (minimum acquisition window))?



  • Dear Mohammad,

    thank you for your interest in TIDA-00913 and INA240. 

    1) To avoid sampling during the transients with a 3-phase inverter, the three (or two) phase currents are typically sampled center aligned to the PWM, when there's no switching (either 001 or 111 vector). This is a typical approach in drives. 

    If - at higher PWM frequencies - the duty cycle should be smaller than the phase current settling time (per item 2), an option is to avoid reading that specific phase and take the readings form the other two phases, which have a higher duty cycle. Then calculate the third phase from the other two, assuming the sum of the 3-phases =0.

    2) The INA240 settling time you refer to is with respect to a transient common mode voltage (AC common mode rejection), after which the INA240 settles to an accuracy 0.5% to final value. If you sample earlier than the signal may not have settled to 0.5%, but a slightly higher. Whether this is tolerable in your design, you'd need to validate yourself.

    In the TIDA-00913 design guide you can see an example response in figure 50 during PWM switching with a 0 to 48V common mode voltage transient. 


    Martin Staebler

  • Thanks for response,
    I have three other questions,

    1-) As you mentioned, for correct reading of current, we have to use 3 inline sensor and sample 2 of those at center aligned to the PWM, but this is exactly the same thing we do in low side current sensing (the only difference is in the low side sensing, we can do sensing just in 000 vector but in inline sensing we can do sensing in either 001 or 111 vector) . so what is the benefit of inline sensing if in the software is like the low side sensing?
    2-)If we sample at center aligned to the PWM, according to the picture in my first post, there is no transients in the INA240 output and the output of INA240 has been settled to the final value , So can we use the minimum acquisition window in MCU?
    3-)Can the RF filter in the output of INA240 remove (or decrease) the transients that exists in the picture of my first post?

  • Dear Mohammad,

    please see below.

    1) A key difference between inline sensing and low-side sensing is that the phase current with inline sensing is continuous, despite more or less impulse noise or settling time after common mode transients. This allows for example oversampling over e.g. one PWM period and/or dual PWM update of the FOC current controller, which is not possible with a low-side shunt topology. The impulse noise depends also on the DC-link voltage and of course the switched voltage slew rate, which is quite higher with GaN FETs compared to Si-FET. By slowing down the slew rate it is also possible to reduce the impulse noise contribution.

    2) The TMS320F28069M sample acquisition window can be as small possible as the as long the 2.2nF capacitor can charge the internals 1.6pF S/H capacitor to the expected accuracy.

    3) The noise spikes you refer to on the scope plot are attenuated, but also measured in system with the TMS320F28069M ADC. Please refer to figure 50 in the TIDA-00913 design guide, which shows an interleaved measurement of the INA240 transient response. Note: The x-axis and y-axis titles are swapped on fig 50, the x-axis should be time [us], the y-axis is phase current [A]. The impulse noise you can see is likely due to X-talk and GND bounces between the TIDA-00913 and the F28069M LaunchPad, which are connected through the standard headers. I don't think increasing the output RC filter will help in that case, but rather increase the F28069M ADC's sampling time or apply oversampling w/ median filtering to remove impulse noise.

    Martin Staebler