As long as memory is not infinite, you will need always to round somehow (in some cases not 2+2), or add more memory for the precision you need, but you won't be able to get the exact result of e.g. square root of 2 as it requires infinite space/precision/bits/bytes, etc to represent.The Unsigned wrote:I understand that theoretically one DAW might round 1.225222222 to 1.23, etc and that there might be minute differences.Kemmler wrote:you're right but you're not 100% right, a floating-point "calculator" does not always give absolutely correct answers, so there is some potential for variability - THEORETICALLY. However when you consider the size of the potential error to the SNR of any one of your pieces of gear, it's laughable. So, yes, you're right, but in theory there is some error introduced when summing.
However, from the tests I've seen there is no audible difference when you sum in two different DAWs. Just the VI debate for starters; they summed 12 tracks in Pro Tools and Logic and the results nullify. Then there are the many other null tests posted in this thread that prove these small calculation differences are inaudible in the rendered audio.
So from this I conclude these differences aren't worth mentioning. If they can't be heard then why should we care? To me they aren't differences until they affect the output.
I leave you this link about audio myths, I don't know if it was posted here, nice to watch: