TMCS1123: Noisy output waveform recommendations

Erika, 

I see a few issues here. The device looks like it is seeing some systematic distortion, but the customer is also operating the device well down in terms of the full scale range. Can we see the schematic how the customer has this implemented in the system?

An issue the customer is likely seeing here that is complicating the issue is simply noise on the output of the device, so throttling of the device BW here will help. Hall effect sensors are typically not super precise in terms of noise, and are therefore not utilized for such small signals as the customer is attempting to measure here.

For all variants of the TMCS1126, the noise density is 150uA/sqrt(Hz). At full bandwidth, assuming a brick wall analysis, 150uA * sqrt(250kHz) = 75mArms referred to the input

I'm not sure which variant the customer is using here, but assuming they are using the TMCS1123x5 (which is 150mV/A based on the 1A reference), then the ideal output would be 150mVpp, but from the above, referred to the output 75mArms * 6 * 150mV/A = ~70mA referred to the output. So, with this in mind, each sample on the scope could have up to 70mV noise coupled here as well. 

Please let me know if we can look at the schematic, or understand how they are generating this waveform in their system. 

Hi Carolus,

Here is the schematic. The off-page signal goes directly to the microcontroller pin. 

Erika, 

Nothing in the schematic jumps out as problematic. Is the voltage signal generated at the inputs of the TMCS1123 also sinusoidal, or is it square wave? I was also hoping to see what the input structure of the system looks like here. 

The input is mostly sinusoidal; it’s an inherently distorted 60 Hz line fundamental due to the motor’s behavior. The current clamp on channel 2 from the previous capture confirms that the distortion isn’t from the TMCS. As you mentioned- their biggest issue is that they don’t need the 20.6 Ap-p range of even this highest sensitivity device even for the overcurrent condition that this is supposed to help catch.

They set up the EVM to measure/compare the zero-input noise levels using just the sensor (scope shot 1) and then adding the 720 Hz filter with the 10k + 0.022 uF (scope shot 2). These measurements were taken with a spring hook ground rather than the wire clip on our prototype board last time which could've also added to the noise from the longer ground loop. It seems like this cut-down bandwidth reins in the noise and is also good for the SAR ADC in the MCU sampling this. In the end, they will be computing RMS current for input to the logic.

Do you have any other recommendations to prevent additional noise? Or is there any additional information needed?