On Sat, Feb 13, 2010 at 6:00 PM, <fons@xxxxxxxxxxxxxxx> wrote: > On Sat, Feb 13, 2010 at 05:37:50PM -0500, Monty Montgomery wrote: > >> The RIAA preemphasis is not a biorthogonal filter. There will always >> be at a minimum some phase/group delay as the 'perfect' inverse filter >> is unstable. You can only approximate it. Another example of an >> analog horror we no longer need to put up with in the digital era :-) > > The RIAA filter applied (normally) in the cutter is > (in Bode-plot form) > > +6dB/oct up to 50 Hz > flat up to 500 Hz > +6dB/oct up to 2122 Hz > flat above that > > It is the combination of 3 simple first order filters. > > The first can't be inverted completely, as it would > lead to infinit gain at DC. It can be inverted with > any precision you care for within the audio range. > But that one is not the problem here. > > The two others can be inverted exactly. Ah, I am perhaps used to seeing... dirtier implementations. Yes, it's intended to just be an EQ. Do you have links to the specified forward/inverse linear filters? I may not be thinking of a proper implementation then. >> A digital EQ alone isn't even close, as digital EQ is nearly always >> phase-linear (acausal). It's a completely different style of filter. > > Sorry but that is nonsense. The pole/zero at 500 and 2122 Hz can > be inverted by about 2 lines of C code, and the phase response > will match the analog one. As long as you don't approach half > the sample rate the 'simple' digital filters are the same as the > corresponding analog ones. And the highest pole/zero frequency > is this case well below that limit. I didn't say that a digital filter couldn't invert it (at least as well as it can be inverted). I said that the typical digital EQ, the kind you're going to find in any general purpose EQ plugin, is going to be an acausal linear-phase filter. Cheers, Monty _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/listinfo/linux-audio-user
