why would you think the 8 least significant bits would be zeros rather than the most significant?
Hi Bob,
The "least significant bit," as you may realize, corresponds to the smallest resolution increment, while the "most significant bit" corresponds to the largest.
For example, if the maximum possible value ("full scale") at the analog output is 2 volts, on the digital (SPDIF) output a logic "1" on the msb would indicate that the corresponding analog output is greater than 1 volt. The next most significant bit would have a weight of 1/2 that amount, so a 1 on the two most significant bits would indicate a value of greater than 1.5 volts. Etc. The least significant bit in a 24 bit word would have a weight of 2volts/2^24 (two volts divided by 2 to the 24th power), which is 0.000000119 volts.
So setting the 8 least significant bits to 0 would introduce very miniscule inaccuracy, while setting the msb's to 0 simply would not work.
For further confirmation of this, see the section in the middle of this page defining the time slots in the AES/EBU and SPDIF subframes:
http://en.wikipedia.org/wiki/AES/EBURegards,
-- Al