Quote:
Originally Posted by ckramsky  i've been thinking alot about this lately, and maybe someone can help answer my question...
if you are working at 24bit (or even more so, 32 bit floating decimal point), there is obviously a limit to the amount of increments that the least significant bit can plot on the Y-axis (obviously, 2^24). but this number is an EXTREMELY specific number.
my question is, that with a number this specific and detailed, it raises the question in my mind that the least significant bit could almost be as detailed as the amount of magnetic flux that an electron can pass to tape.
or to put it another way, how big is the difference between the magnetic flux of electrons on tape to that of electrons sent for the least significant bit in binary code (in 24bit, or 32bit floating decimal)
i know it really doesn't matter in the long run and it wont help you mic up a guitar any better, but i was just curious if anyone knew the answer.
-charlie |
This is a bit more complex than it seems to be honest. Conceptually speaking, the smallest increment of signal one can impose on tape is basically defined as it's signal to noise ratio - exactly as in digital signals - but for vastly different reasons.
Tape's magnetic performance is produced by 'domains' of magnetic influence. These not only have a variance in size - but they also have a tendency to act in unison with themselves and exhibit hysteresis (which means that they are reluctant to be magnetised and equally reluctant to be de-magnetised - and they are highly variable from domain to domain as well). So on it's own tape is a pretty poor recording medium - you can think of in terms of a 'granular' effect - which is a great many times worse than the electrical noise due to electrons and temperature in the recording process.
To get over the distorting effects of this magnetic hysteresis and domain effect the tape is 'biassed' - which basically means that we hit it with fairly high levels of supersonic signal to 'shake' this behaviour around as we record the program - to overcome much of this hysteresis and variation and reduce the distortions this causes.
Even so - the domain and hysteresis behaviour causes significant noise in terms of variation in magnetic effect as the tape passes over the head (random behaviour). This is something that cannot ever be completely lost - and ends up giving us a brute signal to noise ratio of about 55 - 65dB depending on how hard we record signal to the tape and how much distortion we can stand from thrashing it.
Although the freq content of this tape noise isn't flat - in approximate terms it corresponds to a digital signal of around 10 - 11bits.
As far as the notion of harmonic 'detail' is concerned - a digital signal that is properly dithered can essentially have almost infinite purity in terms of distortion - with solely the noise figure being determined by the number of bits we have to transfer the signal. So in reality an 11bit digital signal is many times more pure than tape could ever be - even though it will have a similar noise figure.
I hope this helps and I have explained it well enough :-)