194.78.209.104
In Reply to: RE: Charles Hansen Please Explain posted by Werner on September 15, 2009 at 00:55:24
Let's have a first look at the replay side:
The theoretically ideal DAC anti-imaging (aka reconstruction) filter is defined as the Sinc(t) function. It is characterised as:
-cut-off: at Fs/2
-slope: infinitely fast => a lot of ringing, no imaging
-phase: linear => pre- and post-ringing
It has a lot of pre- and post-ringing, but this can be proven to be of zero (0) influence, provided the source signal was correctly band-limited prior to sampling. This filter can't be realised, but it can be arbitrarily approximated, especially in software-based oversamplers.
Most commercial DACs and replay SRCs (HW and SW) are a decent approximation of the above:
-cut-off: at Fs/2 (half-band for economy) => a bit of imaging
-slope: fast => a lot of ringing at Fs/2 (22kHz)
-phase: linear => pre- and post-ringing
Now to the recording side:
Contrary to the replay side there exists no ideal model of an anti-aliasing filter. The sampling theorem demands a correctly band-limited input signal but gives no clues as to how to do this.
Consequently any anti-aliasing filter for CD production has to be a compromise. Disregarding analogue AA filters, there are two commonly-used digital filter types:
The one most-often used in single-chip ADCs:
-cut-off: Fs/2 (half-band, again for economy) => some aliasing
-slope: fast => lots of ringing
-phase: linear => pre- and post-ringing
This type of filter embeds pre- and post-ringing at 22.05kHz in the recorded signal. Moreover, as it allows some aliasing and as its attenuation is only 6dB at Fs/2, it excites ringing in the playback DAC too (since this sort of AA filter mildly violates the sampling theorem).
Sometimes used in decent software SRCs:
-cut-off: somewhat below Fs/2, say 20kHz or 21kHz
-slope: fast => lots of ringing, but combined with the lower cut-off less or even no aliasing
-phase: linear.
This type of filter still embeds pre- and post-ringing, but it does not excite the DAC's ringing.
Time for a summary:
1) almost all commercial ADCs or production SRCs embed pre- and post-ringing at 20-22kHz into the recorded signal.
2) many commercial ADCs allow some aliasing. This aliasing triggers the otherwise-innocuous pre- and post-ringing of Sinc(t)-based DACs
3) almost all commercial DACs are half-band Sinc(t) approximations.
At this point the presence of pre-ringing is perceived as a bad thing (although the audibility of this has not been proven!!!), and some remedy is wanted.
More later ...
bring back dynamic range
Follow Ups:
Post a Followup:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2009, Audio Asylum, all rights reserved ]
[ Go Green ]
[ Take5 ]