A.2 Method

In order to perform a code spread analysis, it is important to know how much contribution there is from noise internal to the ADC. In order to quantify this, sample VREFLO, the low reference voltage, through ADC channel B5 and introduce an offset of approximately 30 LSBs. To set this internal connection write a 1 to the VREFLOCONV bit in the ADCCTL1 register. Then, calculate the standard deviation of the conversions; this will be the internal amount of deviation and should be taken as the baseline value.

Once there is a known baseline of code spread, incrementally add components to the signal path. For each, measure the amount of standard deviation from the resultant conversions. Subtract the acquired value from the baseline; this is how much deviation each of the added components contributes to the overall deviation. The reason this is compartmentalized is to potentially isolate different parts of the signal path and quantify each of their contributions. The goal is to try to minimize this value as much as possible so that there is little to no added system deviation from that of the baseline (internal code deviation). Verifying this ensures that the signal path does not introduce noise prior to the ADC conversion and thus lead to higher efficiency.

Possible ways to reduce code spread are ensuring the system has a low noise level and the input signal is as clean and reproducible as possible. Adding a low pass filter to the signal path will help remove high frequency noise and interference. The time constant of the filter depends on the sampling speed and the application requirements. Another best practice is adding buffering amplifiers between stages in the signal chain to ensure impedance matching.