TMCS1133EVM: Inline current measurement gain

Hello,

Thank you for your post. 

I would like to confirm that you are sensing ±3.2A?

Here is an excel calculator that I made where you can adjust to parameters to calculate the expected error in your system.

I pre-filled what you provided and, at this low level of input current, the expected error without calibration is about 3.6%. 

Copy of TMCS1133 Error Calculator with Lifetime Corrected.xlsx

Can you provide scope capture of the common mode voltage rise and fall times?

For measuring the output, are you probing directly at the device pins on the EVM or at a test point elsewhere?

Best Regards, 

Joe

Hello,
I confirm that I have connected a probe (x1 gain) directly to the attachments on the eval board.

Voltage and current  amplitude at 500kHz, current is+2,4A to -1,8A (asymmetry come from the amplifier driver at 500kHz)

Voltage (2 probes to do differential) = grey
current clamp ELDITEST CP6000 1V/A = green

Current comparison :
scope current clamp ELDITEST CP6000 1V/A = green,
TMCS1133 = red


Common mode voltage
floating voltage probe blue = + side of the load where the TMCS1133 or INA149  is connected.
I measure 118V/us
yellow is the other side of the load used to trace the grey differential curve above.

Thank you.

Hello, 

Thank you for the additional information. 

Have you used the calculator to find your expected error at your input current?

Can you also confirm that the output clipping for the negative edge is expected?

Thanks, 

Joe

Hello,

Using the calculator I find 3,56% RSS total error.
The error come mainly from the offset (3,25%) and from the sensitivity error (1,4%)
As I measure in AC RMS or with the FFT the DC offset error should be reduced, am I correct ?
Also I am at 500kHz and according to the datasheet Figure 6-7 above I expect 0,5dB loss wich add -5,93% error ((10^(-0,5/20)-1)*100)

So I have +-3,56% RSS error and -5,93% due to the gain.
From the FFT above
3,14dB for the current probe = 1,43A RMS
4,01dB for the output of the hall effect sensor 1,59A RMS

That give me +11% difference with the probe as reference and I expect having -2,37% to -9,49% (RSS+ gain)

I confirm that the clipping in the negative is expected and due to our amplifier command.

For the INA149 I will create a dedicated topic in the correct forum.

Regards.

Hello, 

Unfortunately, at this low of input current, even when taking AC RMS, your error will be dominated by offset error. 

Can you provide scope captures of the levels of noise on the following pins of the TMCS1133:

• VS pin to GND pin
• GND pin to GND test point somewhere else on the LV circuitry
• VOUT pin to GND pin

Please use an oscilloscope for this measurement with a low inductive GND loop on the probe:

Best Regards,

Joe

Hello,

Here are the noise measurement. I did it with no input current and using the ground spring tip as you have specified.
VS pin to GND pin


GND pin to GND test points

VOUT pin to GND pin


Thank you

Hello,

Thank you for providing these. 

It looks like these signals are clean from a noise perspective. 

I am starting to agree with you that this may be caused by the switching CM input voltage. 

The faster the switching common mode, the worse your issue seems to be in terms of gain error. 

One suggestion I have is to try a targeting input capacitor across the inputs to potentially help with the dV/dt. 

Can you place an 0603 or 0402 100pF targeting capacitor across the input polygons for IN+ and IN-?

My hope is that this will reduce the impacts of the dV/dt.

Best Regards, 

Joe

Hello,

Thank you for the suggestion.

I was able to solder a 0603 100pF C0G 50V 5% (würth PN 885012006057) directly across the pin of the sensor as I have no footprint for it.

Red: TMCS1133C3A
green: ELDITEST CP6000 1V/A
300kHz the amplitude is now 0,3dB lower compared to the current clamp (0,1dB difference before)

500kHz still have +0,8dB compared to the current clamp

Regarding the measurement compared to the first one I made I think that adding the capacitor did not make such a difference.
I think that trying a bigger value would add error for low frequency.

Regards

Hello Erwan,

Due to the U.S. holiday we are currently out of the office but someone will reply to you on Tuesday.

Regards,

Javier

Hello Erwan, 

We are attempting to internally build up the capabilities for this level of high dV/dt but this will take time.

In the meantime I would suggest trying to solder multiple caps to target different frequencies. 

Do you think you can try soldering a 47pF and 180pF capacitor on top of this 100pF?

This should target 500MHz, 640MHz, and 1GHz frequencies for fast transients. 

Best Regards, 

Joe

Hello,

I have added both 47pF and 180pF on top of the 100pF. All are C0G 50V from wurth.
I think it make the measurement worth. See the results bellow at the same frequency as above to compare.
At 300kHz I have now 0,69dB more on the TMCS1133C3A compared to my current clamp vs 0,3dB before.
Red: TMCS1133C3A
green: ELDITEST CP6000 1V/A

At 500kHz: 1,57dB more on the TMCS1133C3A compared to my current clamp.
I have also noticed that I have 0,4A variation on the peak to peak waveform that correspond to 32mV on the input of my scope.
I have few mA variation on the current clamp as reference.


I measure up to 60mV variation using low inductance ground tip. I don't know if this is related to my issue and where this come from.
500kHz same config as above but without the TMCS1133C3A gain applied on the scope.

Thank you,
Erwan

Hi Erwan, 

Thank you for running that experiment. 

My next question is how are you sampling the output? For our datasheets we oversample the output to remove the inherent output noise. 

Best Regards, 

Joe

Hello,

For now I am using approx 30MSPS 16bits or 125MSPS lower resolution with my scope but our final application will use a slower ADC like 8MSPS. We will be able to do oversampling because we want 500KhzBW and could decimate 8MSPS down to 1MSPS.

The picoscope have the hability to decimate if you think it will improve my measurement I can try it. It will work as a low pass filter and because I am using the FFT to get my measurements, I don't expect it to have a huge benefit here.

Regards,
Erwan

Hi Erwan, 

I would suggest that you try the decimate feature on the scope. 

I know this is an older application note we had to LPF the output of the TMCS110x to reduce the noise floor: https://www.ti.com/lit/an/sboa518/sboa518.pdf 

So I would recommend oversampling as much as your application permits. 

I hope this helps, 

Joe

Hello,

I have done some more test and I can't see much difference decimating or adding the capacitors.
I think the fact that we measure using the FFT already help with the noise. We will probably go for this solution and try to improve the precision on the final layout. We may be able to do some sort of calibration/compensation if the error is consistent through all components.
I am waiting for your feedback testing with higher dv/dt to confirm if the error come from this.

Regards.