I've been experimenting w/16bit-44.1KHz, 24-96, and 24-192 files in CEP2.1 (some processed with 32 bit float) and have found some of my preferences. Cool Edit/Adobe offers some choices for bit depth and noise shaping and quality of filtering and I'm curious what people have found to be best. What other programs have you tried for this (Ozone, Wavelab or other) and have you compared. Some hard and fast rules apply (as it is written) i.e., don't dither more than once, do it on the down sample and of course, the settings are dependent on the type of music but not all is as clear cut. This can be very time consuming if you compare samples especially if settings are checked with music of various amplitude. In any case what have you found that you stick with? Some say don't dither if there is going to be further processing but what in the case of converting from 32bit/96KHz samples to 24/192KHz for DVD-A authoring - this would be a lower word length with a higher sampling rate and with further processing to come. Lots to chew on, Any bites?
Dithering is a way of hiding the Quantization distortions that are a part & parcel of PCM encoding by introducing deliberate noise to mask the distortions. It is inly really necessary - and some would argue even not then - when going from 24-bit fixed to 16-bit fixed. From 32-bit float to 24 bit it is simply pointless & irrelevant as you will never hear any errors at -144dB. Another reason is that 32-bit float is actually 24-bit fixed with a shifting mantissa - the additional "8 bits" are exponents, and the core audio is still at 24-bit. When encoding for DTS-CD or Dolby Digital, feed the encoders with 24 bit files - they perform better. SurCode will accept 32-bit float, but to be brutally honest I doubt anyone will ever hear the difference unless working on seriously expensive gear. The real issue is for 16-bit audio from 24-bit. 32 bit float gives you almost limitless headroom - around 1500dB or so - but if you are even anywhere near the ballpark (never mind actually in it) then your levels are far too hot anyway so any distortions will be the audio clipping at the master output where it all reverts back to 24-bit fixed. CEP is again problematic as it's default 32-bit float is actually 16-bit fixed with a 16-bit mantissa. Avoid this where possible. To try & answer your points & questions: 32/96 to 24/192 is pointless. All you are doing is padding with zeroes. If you go from 32/96 to 24/192 and them do any firther processing it can be argued that as the processing will be carried out at 192 it is worthwhile. Do not - ever - dither anything if you need to do further work. And only ever when going down - never up.
Your explanation is quite helpful. It is also as I have read and found in my experience. For the sake of brevity I didn't add to many details in the question. For converting to 24bit/192 (for DVD-A 2 channel files) I have found that it isn't necessary to dither (just as you wrote). My confusion is that I have files from different sources and didn't specify which and what purpose. Vinyl transfer, as I expected, did greatly benefit from the use of dither although it is difficult to compare various settings without trial and error and quite time costly. The end result with these files (in CEP2.1) has been excellent. Digitally sourced files seem better left w/o dither when downsampled to final resolution. Coincidentally, the few DTS and Dolby Digital files that I tried did sound better with 24 bit processing but not in any higher resolution conversion.
Cool Edit Pro is (was?) a fine application and greatly under-rated at the time IMHO. It was one of those apps that was loved by those who knew about it in a world that was dominated by Soundforge - even WaveLab was not that well-known back then. How things have changed! We still use the Adobe Audition incarnation of CEP - AA2.0 & 1.5 - for running scripts that take a 24/96 (or whatever) "stereo" SQ Quadraphonic matrix and splitting it out into 4 separate WAV files. Works even better than the Tate II decoders do. Just try not ti use the floating point versions as IMHO it's not all that worthwhile in 16/16 mode. 24/8 is a different matter - it is just that I personally find 16 bit audio to be so damned fragile. Dither is - as you say - one of those things that is hotly debated on a regular basis. Some swear by it, others swear at it. With Vinyl transfer you need to take considerable care, as despite what the vinyl buffs will tell you the resolution of vinyl is actually very poor - you are lucky if you have much more than around 40dB dynamic range. Every time you play the record, the resolution drops too. What I believe people are loving about the sound of vinyl is the fact that it has dynamics. Most modern CD is dreadful sounding because of excessive use of compression & brickwall limiters & maximizers like Waves L2. The default for "High Resolution CD Master" is -5dB, fer chrissakes. That is -5dB off the transient peaks, and +5dB on the overall level. That is -10dB dynamic range gone right out of the window there. This is far from being "High Resolution", it is a volume gain at the expense of the mix dynamics. You cannot do this with vinyl matering, as even if you could cut it without burning out the cutting heads or going through the walls of the groove, the stylus would never track. The entire mess came about with engineers adding a compressor into the 2-buss, emulating what Radio compression does. Trouble is that clients liked this, so left it in - not realising that thre radio will still recompress it all over again. Then the CD changer came in and it was all over. Before the CD changer, the ears played tricks on us in a good way. By the time you had unloaded the CD, got the next one in & hit play you had forgotten how apparently loud the volume was. With the CD changer, it loads in a secoind or so and the difference is apparent. Nobody wanted the quietest CD in the changer - despite the fact that louder is not better - so we got the mess we have now where things are mixed for iPOD earbuds. It's a disaster for quality. MP3 should carry a health warning, it really should, and those bloody earbuds are destroying the subtleties of mixing. Aargh - I am ranting again. Sorry....
No, rant all you like on this 'cuz this loudness thing is a curse. I mean everyone, EVERYONE has a volume button/switch/pot/attenuator/control and we know how to use it. This compression and brickwall limiter misuse just kills the dynamics of the music to the point where I just won't listen to it. If people knew the difference they wouldn't take it. Even a song that only marginally grabs me, if done right, I will listen to over and over. I don't want to knock ipod nation, as it dates me but the market has perpetuated the mindset of convenience and abundance at the expense of quality. Always thought that quality sells, of course that was before disposable appliances and walmart moving good jobs out the neighborhood. OK, I'll keep this PC. Quality still sells, it's just harder to come by. Quick observation on Vinyl. Average dynamic range might be 30 to 35 db but it can and does reach 70+ db. To me it seems to reproduce the HF better - less harsh, cold or sterile as some claim digital does. LF sounds more life-like also. That's real bass and percussion rather than some studio produced protooled foolery. In fairness, digital can sound great but it's best just doesn't match the best vinyl. Nostalgia is tough to overcome. Now well done vinyl transferred to digital and in high-resolution can be, well, steak with steak sauce.
