cliveb Wrote: > I don't know of a program off-hand that will crop LSBs and redither, but > last night I did some casual tests using an audio editor. I had a > recording of a vinyl record in a 16bit 44.1kHz WAV file on disk, and > extracted a short section from it to play with. What I did was apply an > amplitude reduction (in order to truncate LSBs), then re-normalised it > to bring the level back up and thus pull the quantisation noise up to a > much higher level. This in effect truncates the LSBs *without > re-dithering*, so any effects should be even more audible. The results > were: > > 1. Reduce then increase amplitude by 36dB, effectively truncating the > bottom 6 bits. This is the "60dB test". The raised quantisation noise > was very clearly audible, and completely unacceptable, which suggests > to me that, although the dynamic range of the LP is probably less than > 60dB, one cannot afford to be cavalier and chuck away 6 bits of > resolution when recording. > > 2. Reduce then increase amplitude by 24dB, effectively truncating the > bottom 4 bits. It was very difficult to detect a significant difference > after applying this change. I felt that maybe the noise was changed > slightly, but never detected that characteristic "zzzzy-ness" of > digital quantisation noise. Perhaps better headphones would have > revealed a more obvious difference. > > 3. Reduce then increase by 18dB, effectively truncating the bottom 3 > bits. No audible difference. This was no surprise: the surface noise on > a vinyl LP is so high that it effectively "self-dithers" the recording > at around the 12-13 bit level. So if you have a decent soundcard with a > noise floor below -90dB, it's safe to record at peak levels as low as > -12dB then normalise. > > Bear in mind this was a casual, sighted test. > The equipment used was: > Linn Sondek LP12/Lingo/Ittok/Karma > Naim 42.5K used as phono preamp > M-Audio Audiophile 2496 soundcard > Home built headphone amp > Sennheiser HD535 headphones
Thanks for your tests! I'm slightly suspicious of using headphones for evaluating recording fidelity. BTW, I read recently that LP's "dynamic range" (really: S/N ratio -which is not the same thing) varies with frequency. S/N above 3 kHz was around 90dB if I remember correctly! Add at least 10 dB of signal recovery into the uncorrelated noise, and you have 100 dB. I'll do some tests using your bit reduction method! P.S: some people claim "infinite" dynamic range for dithered CD, in which case LP too must have infinite dynamic range.. It is true that signal can be distingiushed below LSB on a dithered system, but I haven't investigated what qualities (or lack thereof) this sub-noise signal has. -- P Floding ------------------------------------------------------------------------ P Floding's Profile: http://forums.slimdevices.com/member.php?userid=2932 View this thread: http://forums.slimdevices.com/showthread.php?t=21836 _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
