I'm enjoying the Mastering Overview in Logic 401 (so far, highly recommended). One technical issue has me confused. 24-bit is supposed to give more headroom (or "tailroom" as the video suggests) than 16-bit. The graphs in the video confirm this as 24-bit is able to capture a broader range of amplitude; however, my fairly uninformed impression is that bit-depth is more a question of accuracy of the actual signal value (quantization). For example, if a signal is between -10 and 0 (I'm just choosing numbers, the units are irrelevant for my question), 16-bit will set a "grid" of 65,536 (picture horizontal lines spanning the vertical axis) and the signal value is assigned to the closest grid value. With 24-bit it would be a much *finer* grid (16 million+) to this range (-10 to 0) and hence the grid value will always be closer to the true signal. The video seems to suggest something else; it suggests that the 24-bit signal will use the 16 million grid values to span a wider range; e.g. -20 to 0. Is this true? In other words, we have approx. 8 million grid values for -10 to 0 and another 8 million for -20 to -10. Any comments would be appreciated! As would any references (even some of the actual math would help!). Thanks and looking forward to the rest of Logic 401. -m

This is a good question. I've actually heard some differing opinions on this from different people while i was studying digital audio at uni. My understanding at this point is that what Steve H is saying is correct. When you're recording in 16 bit, chances are you're not actually using all 16 bits, as you'll be giving yourself a significant amount of headroom so as not to clip the converter. So while 16 bit gives you a theoretical 96dB, in practice you probably only have about 70-80dB of dynamic range on average. Factor in cumulative noise throughout the system and you lose even more dynamic range as the noise floor impinges on the total useable range. The maximum value in 16 bit is 16 values of 1, (ie 1111111111111111), which will be equal to 0dB FS. The maximum value in 24 bit will also be 0dB FS, just 24 values of 1. 24 bit essentially gives you more space to work with by pushing the noise floor much lower and farther from the dynamic range that you will be recording in. I should mention that this is just my understanding of the subject, i'm not sure if it's exactly technically correct but hopefully it helps. Would love to hear some other opinions on the subject.