[FAQ] INA241 vs INA296: Selecting Devices with or without PWM Rejection

Let me first describe the problem PWM rejection is trying to address. With standard current sense amplifiers such as the INA296, common mode voltage transients on the inputs disturb the output voltage. This causes high frequency, 200mV ripple on the output for each high-speed edge of the input common mode voltage. Here is the figure from the INA296 datasheet illustrating this:

Now let’s compare this output response to the INA241:

You can see that the output maintains ~50mV of voltage coupled from input to output. Here is an actual scope capture from lab where channel 4 is the AC coupled output of the INA241:

In practice, the INA241 samples the output at some point during the high frequency transient and holds it for approximately 1µs. The output also “steps to the correct output value towards the end of the 1µs holding window. You may also notice the HF transients still getting picked up on the probe. For testing the INA241, it is recommended to use the “pigtail” method for GND loop minimization:

As described in the INA241 datasheet, “when INA241x-Q1 senses the large common-mode ΔV/Δt transients on the input pins, the device holds the output for 1μs, thereby preventing the common-mode disturbance from propagating to the output.” The ΔV/Δt where the PWM rejection activates in the INA241 is for common mode transients greater than 100V/µs. If the ΔV/Δt is less than 100V/µs, then the performance is the same as the INA296 in terms of common mode coupling from input to output.

You will also notice that at the top of the datasheet for the INA241, the maximum supported switching frequency is 125kHz, corresponding to an 8µs total period. It is also assumed to be used in an application with 50% duty cycle for the voltage waveform (4µs on and 4µs off). This is where the following statement in the datasheet comes from: “The enhanced PWM rejection is achieved up to a PWM frequency of 125kHz or if common-mode transient edges are separated by a 3μs interval or more.” It is recommended to operate at or below the 125kHz switching frequency for common mode voltages, corresponding to a 1µs hold time with 3µs recovery.

What is the tradeoff of this feature? As shown in the previously provided scope captures, there is a 1µs period where the output is not in a valid state if the common mode voltage rise time slope is more than 100V/µs. It will not accurately track the input waveform for these fast transients for 1µs. This becomes potentially more critical in PWM applications where the switching frequency is approaching 125kHz. In these faster PWM applications, the phase accuracy may be more critical than common mode coupled transient spikes on the output. In this case, the INA296 is the preferred device.

Can you minimize common mode transients and have phase accuracy in higher speed PWM applications? Yes, but nothing is free. You will need to implement passive components to create a differential filter across the inputs of the INA296 and add common mode capacitors on each input as shown in this FAQ: (+) [FAQ] INA181: How do I properly design an input filter for my INA18x-Q1 device? - Amplifiers forum - Amplifiers - TI E2E support forums

To conclude, we offer current sense amplifiers with and without PWM rejection. PWM rejection reduces the coupling on the output from input common mode voltage transients greater than 100V/µs with the tradeoff being phase accuracy. With common mode voltage transients with less than 100V/µs, devices with and without PWM rejection perform the same. If phase accuracy is crucial in your high-speed application with faster than 100V/µs common mode voltage transients, we suggest using a current sense amplifier without PWM rejection. By increasing physical solution size with passives and BOM cost, you can also implement an input filter as shown above.