peterpiper wrote:As far as I understand it is cutting on the other side of the scale. If the sampling source signal goes lower than -72dB in a 12bit A/D it can't be heard cause there is nothing but a noisefloor down there.
To get this effect with a 16bit A/D you'll have to lower the signal -24dB cause the noisefloor of a 16bit A/D is 24dB lower.
But like I said before. There are two laws (curves) of bit representation in A/D converters and they are compression
http://en.wikipedia.org/wiki/%CE%9C-law_algorithm
To be honest. Thats just how I understand the theory but it can be totally wrong. I'm not really an expert on this topic 
peace
hmm yeah, what im making of that chart: the closer to 0 db, the finer the encoding gets (more per db); there is some room for user error (the recorded signals +1 db (or so) is encoded as 0db, the recorded 0db is about 1db in encoded terms); i THINK the companding is there to make optimal use of the available bits (this chart is a convention i think, so they just could have 'spaced the bits more', but there probably is a reason (ease of calculation?) why the bits are closer together near the 0db)
what i'm wondering about: this chart goes to -42 db (in encoded terms, otherwise about 64 or 76 depending on the used law), is 128 values (if i read the code underneath the chart correctly), so 7-bit.
would an 8-bit encoder add more values inbetween these, or add more values so the range would be 48 db?
im guessing the second, that would mean that dropping the 4LSB indeed does not give the same result as sampling in 12 bit (if these laws would be followed in the AD encoding, which im not sure, the mpc 60 emphasises the high end before sampling, no idea whether it's compensated in the DA)
The 60 and the 4000 are not 16 bit....
