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.

INA253: Weird behavior of INA253 when used for phase current measurement in B6 bridge

Part Number: INA253

Dear TI engineers,

The INA 253 seemed very promising when testing it without any common mode voltage at the input. However, when applying it for phase current measurements in a full-bridge inverter with a switching frequency of 800kHz, its output voltage seems to behave weirdly. In the following example, the three inverter legs receive the same gate signal with a duty cycle of 50% and the current sensors are connected after a LC filtering stage. The sensors are supplied with 5V (with decoupling capacitor of 100nF) and connected to an external reference voltage of 2.5V. Since no current can flow through the sensors, their output is expected to be flat at 2.5V.

In a first attempt, the star-point of the filter is isolated from GND. Without any gate signal at the switches, the output looks nice at 2.5V:

But as soon as the gate signal is activated, the output (channel C1 yellow) gets an offset (0.48V -> 1.2A) and oscillates arbitrarily (the INA253 variant used is the one with 400mV/A, a division represents then 0.5 A). Might there be a problem when the voltage at C3 (blue curve in figure below) goes below -4 V?

If the star-point of the filter is connected to ground, the offset disappears, but an unwanted ripple still exists:

Questions:

  • Can then the INA253 be used in such a configuration (even preferably without connection from star to GND)?
  • And what could be the reason for this strong deviations?
  • In the data sheet PWM rejection is mentioned as special feature.
    • How das the PWM rejection works?
    • Might there be a problem when the PWM frequency is at 800 kHz?

Thank you for your consideration

Kind regards

Timon Achtnich

  • Hi Timon,

    INA253 can be used in such a configuration. The PWM rejection is achieved through special circuit techniques, together with unique manufacturing technology. The frequency of 800KHz is too high for INA253 to work properly. You’ll have to dial it down significantly in order to see normal output. If possible please try 100KHz or lower.

    Regards, Guang

  • Dear Guang,

    Thanks a lot for the quick reply. For us it is disappointing that we can not use INA253 in our application as we have already buildt the hardware (after tests with the INA253 on a demo board). Reducing the switching frequency down to 100 kHz is not possible. Is there a specification in the data sheet where we could have seen that this is not possible? If not I recommend to add this limitation.

    Can you recommend another shunt current measurement or different current measurement principle for our application?

    Regards,

    Timon

  • Hi Timon,

    there are several places in the datasheet that make clear, that a switching frequency of 800kHz is way too high.

    On the first page of datasheet you will find this:

    "• High Bandwidth: 350 kHz"

    On page 6 in the electrical characteristics you will find the bandwidth specification of 350kHz. Directly below the bandwidth for a distortion of 2% is given with 100kHz.

    And then figure 11 and figure 32:

    Form all this it should be clear, that for the INA253 a switching frequency of 800kHz is totally out of range.

    Kai

  • Hi Timon,

    Like Kai pointed out, the device bandwidth is simply not enough to support 800KHz switching frequency. Thank you, Kai for your post.

    Is this a brand new design, or a spin from an older one? If similar designs exist, how was current sensing implemented (configuration and corresponding devices?) and what motivated you to switch to INA253?

    We don’t have an integrated current shunt monitor that satisfies the speed requirement, but if it is possible to construct a discrete solution with faster amplifiers. It will be low side instead of inline however, such as this, if you’re interested.

    Regards, Guang

  • Hi Kai and Guang,

    Thank you for your answers. Our switching frequency is 800 kHz, however the required bandwith of sensing is only around 50kHz as we have a second order active low pass filter after the current measurement. Therefore the bandwidth of INA253 might be ok, it is no limitation for the application if the signals above 100 kHz are damped.

    However in figure 32 of the datasheet (THD+N over frequency) I see that high frequency signals deteorate lower frequency signals. Therefore not the bandwith is the problem but the THD.

    We used the INA253 as a replacement for an AMR sensor with a bandwith of 400 kHz. The reason for the change is the end of life of this part. We measured the THD+N for INA253 as well as for the AMR based current sensor up to 200 kHz before implementation. The results of the INA253 were even better in this frequency range compared to the AMR based current sensor. There might be a drastical decrease of performance of the INA253 at higher frequencies or when the input is not  sinusoidal anymore.

    For testing purposes we reduced the switching frequency to 80 kHz and the output of the INA253 doesn't look satisfactory. The measurement is done in the same way as in the first post, there is no load attached, therefore zero current is flowing (green gate signal (inverted), blue phase voltage, yellow sensor output). Might the ringing at the phase voltage be a problem for the INA253?

    Thank you Guang, for pointing out possible solutions for our application. For the sake of simplicity we will most probably switch back to an MR-based current measurement.

    Regards,

    Timon

  • Hi Timon,

    the bandwidth specification is always referring to a sine wave, because we are in the frequency domain then (->Fourier). A square wave, on the other hand, contains extrem far reaching harmonics. So, a 80kHz square wave is still way too fast for the INA253.

    Have a look at the "typical applications" in datasheet to see what applications the INA253 is designed for. Look at figure 42. A solenoid is current sensed. See the time scale: 50ms/div!

    Or see figure 43. It shows a sine wave loadspeaker measurement which goes up to 20kHz. The figure 32 shows a distortion of 0.5% at 20kHz. This is what I would consider as the maximum frequency range for a precise sine wave measurement.

    There are other specifications of the INA253 which must also be considered: The slew rate is 2.4V/µs. And the settling time to 0.5% of final value is 10µs. For comparison: The period of a 80kHz square wave is 12.5µs. And the period of a 800kHz square wave is 1.25µs...

    Kai

  • Hi Kai,

    Thank you for pointing out the typical applications for INA253. We are aware that the INA253 is not the optimal solution for our application.

    Regards,

    Timon
  • Hi Timon,

    I know some other people who also made bad experience with the INA253. They also thought that the INA253 can handle fast signals. Unfortunately, you must carefully read between the lines and have lots of experience to see that the INA253 has problems with too fast signals. Finally, the distortion graph and the settling specification tell the truth, whereas the text on the front page of datasheet makes you believe that the INA253 is really fast.

    In Section 8.3.5 you will find the following phrase:

    "The full amplifier bandwidth is always available for fast overcurrent events at the same time that the lower-frequency signals are amplified at a low distortion level."

    This also makes you believe, that the ONA253 is really fast. But the distortion graph tells a different story. 5% distortion at 200kHz will cause lots of intermodulation distortion which makes it nearly impossible to "amplify lower-frequency signals at low distortion level" then.

    I think the datasheet should be much clearer in this point.

    Kai