Mode 1 and 2 are 16 bit and 32 bit respectively. Compressed mode (CO) uses an algorithm that uses the smallest of 8,16 or 24 bits to store the differences of the samples in a given data block. The algorithm is compute intensive and should be avoided when throughput on the das is in the 100s of samples per second. It typically uses a little less space than 16 bit mode. But if a site is noisy, it can easily use more than 16 bits per word. The values for CO mode are estimates only since the size can vary so much. CO (quiet) will be about the smallest amount expected, while CO (noisy) will be closer to the largest expected amount.