Archimago wrote: > > So the dude measures distortion levels in a 16-bit format at -65dB. Very > low level! So he's basically only using the lowest 6-bits! Is that what > they did? And given that CD players / DACs in 1985 were incapable of > full 16-bit accuracy, it looks bad with at best 36dB dynamic range. And > lots of noise, etc... And says this is is a flaw of the digital > system!? > > Indeed. These guys should try recording some -65dB signal below the peak > allowable amplitude on a piece of vinyl and tell us how it sounds - if > it's even audible above the surface noise at all!
Digital systems have this kind of distortion in the absence of dithering when sampling a continuous-time signal and assigning integer values to the samples via the nearest integer function (int_value = floor(float_value + 0.5)). The quantization "noise" of an undithered conversion of this type is not random, but correlated with the signal. If you use some software like Matlab to generate a sine wave of amplitude -65 dBfs in a 16-bit system (amplitude = 10(-65/20) * 32767), then quantize each sample value to the nearest integer and do an FFT on the resulting sequence, you see heavy harmonic distortion which indeed gets worse as the amplitude of the continuous-time sine wave goes down. If instead you add random noise having a triangular probability density function of 2 LSBs peak-to-peak to each sample before rounding to an integer, the harmonic distortion completely goes away. The quantization noise is randomized in this case. The math behind all this is covered in '_Wannamaker's_Ph.D_thesis_' (http://www.robertwannamaker.com/writings/rw_phd.pdf), which he did under Stanley Lipshitz. So if the test signals used were generated by simple rounding of the floating-point values without dithering, you'll see this gross distortion at low signal levels. But the distortion is not in the system in this case, but in the test signal! ------------------------------------------------------------------------ andy_c's Profile: http://forums.slimdevices.com/member.php?userid=3128 View this thread: http://forums.slimdevices.com/showthread.php?t=105507 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/audiophiles
