When I reading some books related to ADC, I find that while they were calculating the allowed minimum error referred to input of an ADC, some of them said that the error should be less than 1/2 LSB, which is the quantization noise, yet another declaimed that the error should be less than 2-N or 1 LSB, which is the quantization noise floor. I wonder whether the two statements are valid or equivalent. If the 2nd is valid, does it assume that the FSR is 1 so that 2-N is 1 LSB?