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.

TMS320F28388D: Quantification Noise ADC

Part Number: TMS320F28388D
Other Parts Discussed in Thread: C2000WARE

Hello.

We are rtrying to measure quantification noise on the ADC by connecting ADC intput to an RC to ground.

However, when the system is connected we only see a 0 value (both at 12 and 16 bit setup).

However, we do see some quantification noise when we input a constant level at 2v (for example).

Why is this?

Thank you

  • Hi,

    Connecting the ADC input through an RC to ground is probably not effective way to measure quantization noise.  The ADC/channel will have offset errors that may be slightly below reflo/ground so any conversion value that is produced with the channel connected to ground through an RC will always produce a zero result,  You will only see noise/error/variation if you apply a clean DC signal that is above reflo.  Taking a histogram of conversions on a signal that is midpoint of FSR would be one method to measure the variation (ADC quantization error would affect the variation but also note that stability and accuracy of the VREF and input signals would influence the variation).

    Regards,

    Joseph

  • Hello, from my point of view using the Histogram measurement is not valid since there are different noise sources: Vref, ADC linearity, Vdc, ADC quantification, etc

    What is the ENOB (Efffective Number of Bits) for theADcs of the  F2838X family?

    How could we detect the offset error and compensate?

    Regards

  • Hi,

    We are able to attain below typical ENOB on the F2838x family in our lab setup:

         - 16-bit differential:      14.65 LSBs

         - 16-bit single-ended:  13.5  LSBs

         - 12-bit single-ended:  11.2 LSBs

    For determining the offset, you can slowly ramp up a DC signal staring at a slightly negative value (~ -7LSB) and do a conversion.  The point where you see the digital value switch from '0' to '1' would be where the offset error is.

    Regards,

    joseph

  • Thank you, we will use this table as a reference.

    About your method for calculating the offset, you would need a really precise source to detect the offset level. I don't think it could be used on a "field" setup.

    How do ADC Calibration functions work then?

    Regards.

  • During factory testing, the 12-bit and 16-bit offsets for the ADC are stored in the device.  If you are using the C2000Ware driverlib routine to set up the ADC mode (12-bit or 16-bit) with function ADC_setMode(), it would populate the ADC offset depending on the resolution that is chosen so you would not have to manually do this routine.

    How I described the offset calibration (slow ramp with precision source) is how offset is calibrated during factory testing. 

    Regards,

    Joseph

  • Yes, I am calling that functions.

    That is why I did not expect to get a pure 0 when reading quantification noise. Shouldn't the calibration call remove this offset?

    Thank you

  • Hi,

    My guess is you would probably see code 0 all the time when you run the conversions as you described with a grounding resistor and cap filter.  Any input value less than LSB/2 would convert to code 0.  The code might start switching if you bump up the input value close to LSB/2 which is hard to produce in a field environment (especially with 16-bit mode).  Even if you are able to produce that source, not sure how you are planning to use the results to determine the quantization error.

    The AC characterization done on the F28388D ADCI believe is comprehensive.  In the AC characterization, we  capture a precision sine wave and run FFTs and determine the SNR, THD, SINAD and ENOB of the ADC for all the modes (these are available in the datasheet).  The amplitude is close to VREFLO and VREFHI so coverage is close to full scale.  This dynamic conversion test should cover quantization noise across the ADC range and should be part of SNR measurement.

    Regards,

    Joseph