MM912J637 Vsense input current causes ~40mV offset error

[igw]

Hello all,

We are using the MM913J637 device.

The data sheet specifies a 2.2kohm resistor on the Vsense input (for over voltage protection).

The data sheet does *not* specify any input leakage current or resistance for the Vsense input. (The data sheet mentions nothing about this.)

With a 12V input I measure about 40 mV drop across the 2k2 resistor (so about 16 uA). This varies slightly but linearly with voltage. At Vin=10V it is about 30mV, at 18V it is about 54mV. I have not measured how it varies with temperature.

This seems like a *large* offset error given the other analog specs are much tighter than this.  There is no mention of this in the data sheet nor of the need to apply any sort of scaling correction.

I assume it is related to the fixed 1/28 voltage divider on the Vsense input but there is no info on what the input looks like. It would appear to be equivalent to about 735 kohm.

What is the architecture of the Vsense input?

Is there any information about how the input current will vary with temperature?

Can the value of the protection resistor be lowered to reduce the offset error?

Thank you for any comments,

Ian

[alexis]

Hello Ian,

Please find below my comments:

What is the architecture of the Vsense input?

[AA] VSENSE and VOPT are inputs with an internal resistive voltage divider.

There is an ESD structure at the pin to protect against ESD/EMC events. This structure has to be protected against surge current the external resistor is there to insure this protection.

Is there any information about how the input current will vary with temperature?

[AA] Yes, the part was tested at Freescale End of Line by adding into the measurement path a  perfect 2.2kOhm (by computation). So the leakage current induced into the resistor is bring into consideration into the VSENSE and VOPT gain and offset compensation values. These values are also available across temperature (delta value to room are available into the compensation repository).

There, it is recommended to customer to run calibration of the voltage at room temp to cover the variation of the resistor to its production process. The temperature compensation change will be handled by running a calibration using a linear interpolation of the channel compensation versus temperature.

There you can use the calibration request interrupt that is driven by internal temperature measurement.

If the temperature is also impacting the value of the external resistor, you can run a compensation over temperature or use a computation law that will minimize the error due to the resistor variation.

All of this can be handled through compensation look up table located into the flash and constructed during the initial compensation of the channel at your end of line.

Can the value of the protection resistor be lowered to reduce the offset error?

[AA] The resistor is there into the application schematic to protect the product against ESD and EMC (protecting the ESD structure against surge current). So yes, but this might lower the sustainability of the product to ESD and EMC events.

[igw]

Hello Alexis,

Thank you for reply.

So, I think you are saying that the effect of the 2.2k resistor should be calibrated out automatically if I have applied the calibration process within the data sheet correctly.  Assuming I have an accurate 2k2 resistor with low temperature coefficient then I should not have to worry at all about any extra calibration or correction in software.

Is this the case?

The problem is then, I am fairly sure I have applied the calibration correctly. We have created a temperature calibration lookup table and this is being applied when the temperature changes by more than +/- 5 degrees (in my case that is the temperature cal window we have programmed).  I will have to investigate more at our end if you can assure me that the documented MM912J637 cal process accounts for the external 2k2 resistor.

Many thanks,

Ian

[alexis]

Hello Ian,

You are right the 2.2k resistor is calibrated automatically, assuming the resistor has low temperature coefficient and is accurate. On this latest point (I mean accuracy of the resistor), a calibration at room temperature after your end of line will set an higher accuracy.

Best regards,

Alexis