|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
87.74.29.26
In Reply to: Re: er... posted by Werner on September 14, 2006 at 23:30:45:
Regarding ADC's - I'm surprised to hear you say that they are problematic. I assumed even at 16/44 they have been pretty much 'perfect' in their capture of a signal for many years.Apart from which, most are now 24-bit and much higher sampling. 16/44 masters are created in software from higher resolution recordings.
Follow Ups:
I assumed the same for many years.In reality just about any commercially available ADC chips*, including top models like the PCM1804, do their anti-aliasing filtering with linear-phase half-band FIR filters with relatively short length:
-this is cheap in silicon
-only 6dB attenuation at fs/2
-often only about 80dB of stop-band rejection above fs/2 (i.e. full-band aliasing occurs at levels coincident with the source's ambience and fine detail)
-filter impulse response with pre and post echo, about 1 ms wide (upper 60dB of impulse). You rang, mylord?All of this is very poor. Luckily indeed that many have reverted to 88.2kHz or 176.4kHz recording followed with downsampling in the DAW, hopefully using better filters.
(* There exist pro ADC units that do better, notably dCS and a few others. These do not rely on commercial ADC chips from TI/BB/Crystal/AKM but rather on in-house developed discrete solutions, often with much more rigorous AA filtering)
--
One filter I have made is:
.flat to 15kHz
.-3dB at 18kHz
.-130dB at 20kHz
.600us impulse width (almost twice as fast as a regular CD-audio AA filter)
This is a 'safe' filter in that it allows even for cheap DACs with primitive half-band reconstruction filters to play without imaging.Another filter I have is more like (don't have the details here):
.flat to 16kHz
.-3dB at 24kHz (!)
.-130dB at 28kHz
.400us impulse width
This filter is remarkable in that it deliberately allows aliasing to occur, folding back in the 16-22kHz band. The reasoning is two-fold: many adults (including me) don't hear beyond 15kHz, and above ~12kHz the human ear cannot distinguish pitch, so aliasing occuring above 12kHz is perceived as a brightening, and not as a mixing-in of dissonant tones. dCS follow the same strategy.
I use this filter mainly for recording analogue-master LPs, often with purely acoustic contents and innately smoothly rolling off above 12kHz or so. In short: when the source programme is not rich in treble contents.All work in progress. Present filters are long and computationally intensive, and FIR/linear phase. I want to arrive at much shorter filters, asymmetrical FIR, and somewhere between linear phase and minimum phase.
Very interesting indeed! I can just about follow the concepts, but far be it from me to critique.I whole-heartedly agree that it was very foolish to adopt sampling rates in recording that weren't factors of the defacto 16/44 RBCD, I honestly don't understand why this ever happened.
Regarding our hearing, again I think we have to be careful in differentiating 'band-width', 'frequency-response' and 'time-domain resolution' or 'coherence'.
An anecdote;
my own hearing rolls off over 15 KHz (I'm 46), but I was nonetheless very easily able to hear the difference in Max Townshend's home development/reference system (listenin to DVDA and SACD) when he switched out his 'super-tweeters' - not subtle at all - and these drivers produce no sound pressure even remotely within the range of my hearing.
Actually, more correctly, they produced almost no sound pressure within the range of my hearing - I doubt the crossovers used completely 'brick-wall', but they were probably rolled off to the threshhold of audibility ( at least -60dB or thereabouts) by 16KHz or so, at 'realistic' in-room levels.
This post is made possible by the generous support of people like you and our sponsors: