|'); } // End -->|
In Reply to: Re: Upsampling software posted by Tom Dawson on April 18, 2005 at 01:35:38:
Some of the modern crop of redbook CDs do give high res a run for the money.... the linked CD is a great example of what is acheiveable with 16/44.1 (and its some great music too).
There is nothing wrong with 44.1 part (although on most playback systems the filtering is inadequate to do it justice, higher sample rates do not provide anything beyond what was already there - they just relax the real-time filtering requirement), 16 bits may not seem like enough but it seems more modern techniques of downsampling a master gets around the 16 bit problem quite well.
I'm glad you're pleased with Redbook - I never was and its best still falls dismally short of SACD, DVDA and good old analog, both on paper, and in the listening room, in my experience.
Possibly the most significant, but not the only problem with sampling barely above the 20khz Nyquist Limit is that intra-aural delay differences between channels are obscured, severely limiting imaging capability. Keep in mind that studies have shown that the human ear-brain mechanism can process delay differences between the ears of as low as 2 microseconds, or stated another way, around 1 degree laterally while Redbook is more than an order of magnitude worse. The most obvious result of this is smeared, imprecise imaging.
The other major problem with this 20khz brickwall bandwidth truncation whose justification has long been obsoleted is the sonic thumbprint it imposes on all material. It is not necessary to be able to hear sine waves above 20khz, or even 5 khz to be capable of detecting the effect this has on musical material since the difference tones produced by middle ear nonlinearities are often prominent right through the ear's midband region of highest acuity.
It makes little sense to be too enraptured by the (seldom utilized & proprietary) noise shaping juggling which is but a partial band aid to this problem and which has the undesirable side effect of producing constant relatively high levels of hf noise which under certain playback conditions can damage hf drivers.
Then, of course, there is the item that you refer to that such brickwall filtering close to the audio band makes it very difficult to avoid coloring the resulting audio with aliasing artifacts due to 'imperfect' filtering.
Knowledge of these fairly well known hearing mechanisms was common among well informed people until Sony's 'Perfect Sound Forever' advertising campaign caused its most deplorable and long lasting effect on the intellectual curiosity of the suitably impressionable.
You make some interesting points... but I do have a couple of comments:
1) SACD is absolutely full of ultrasonic noise, noise shaping is the only reason it works at all.
2) As long as both left and right channels are band limited and the sampling frequency is sufficient (ie at least double the largest frequency) there is no reason one channel can't be a 2 microsecond (or 2 nanosecond) delayed version of the other no matter what the sample period is as long as it meets the above criteria. _It has nothing to do with sample period_. The limiting factor would be the number of bits since you would be approximating continous sample magnitudes. This would be true even if your highest signal frequency were 5 Hz and your sample rate was 10 Hz.
I think the extra bits definitely help a lot though (they make the sample magnitudes more "continuous"), and the higher sample rates generally buy you less filter problems but unless you believe ultrasonic frequencies are necessary (and SACD is pretty noisy there) then it doesn't actually buy you any extra information, just an easier way of getting it out.
You're correct regarding SACD's limitations & wrt that, some have described some equipment as being 'SACD Friendly or unfriendly', depending on how much SACD's ultrasonic component excites their nonlinearities.
To a certain extent resolution compensates for sampling rate when attempting to recreate interchannel phase information, but when the information is not at a very high level in the first place with precisely localized sound sources, then not that much can be achieved with only 16 bit linear in the first place which I regard as extremely nonrobust on its own merits.
Sorry if I was a bit abrupt regarding Redbook. However I blame the facile adoption of it and similar professional standards, in practice, for a drastic deterioration of a generation of recorded audio for no justifiable reason at all.
This post is made possible by the generous support of people like you and our sponsors: