Analog to Digital Converter

Abstract

Processing of seismic signals is now completely digital. The analog signals from the seismic sensors must therefore be converted to numbers readable by a computer. This is done with a so-called analog to digital converter (ADC). The continuous analog signal is converted to a series of numbers representing the signal samples at discrete time intervals. The amplitude resolution is typically 1 μV and the time interval is typically 10 ms corresponding to a rate of 100 samples/s. The digitization will introduce errors (quantization errors) into the data since a continuously varying signal is replaced with a limited number of discrete values. Much of the efforts in improving the ADC process are related to minimizing these errors.

An ADC will give an output binary number when we put in a specific voltage. Currently most digitizers give out a number from 0 to 2¹⁶ or 2²⁴ (or ±2¹⁵ and ±2²³) and are correspondingly called 16 and 24 bit converters with dynamic ranges 90 and 138 dB respectively provided the digitizer do not generate noise itself.

In order to obtain the highest dynamic range, converters use the technique of sampling with a high sample rate, low pass filter the signal and resample at a lower rate. The quantization errors of the individual samples in the oversampled trace are averaged over neighbouring samples by the low pass filter and the averaged samples therefore have more accuracy. This is used in the most common converter, the Sigma Delta ADC which digitize with a very low resolution (typically 1 bit or only level 0 and 1) but a very high sampling rate so that successive samples are highly correlated, get an estimate of the signal level, add the quantization error to the input signal, get a new estimate etc. This process will continue forever and the actual value of the input signal is obtained by averaging a large number of estimates and limiting the signal bandwidth with a numeric low-pass filter.

In addition to the quantization error in amplitude, errors can also be introduced due to the discrete steps taken in time which means that signals with a frequency higher that half the sample rate (Nyquist frequency) cannot be resolved. If they are present, the digitized signal will be distorted. This a called aliasing and all digitizers must have a filter to remove signals above the Nyquist frequency.