On Nov 15, 6:29 pm, codifus wrote:
On Nov 14, 6:11 pm, "Arny Krueger" wrote:
"DJ" wrote in message
It's supposed to upsample CDs to SACD. Has anyone heard
about this player, and better yet, auditioned or own one?
Reference -
http://www.emmlabs.com/html/audio/cdsa/cdsa.html
The basic premise is ludicrously flawed. No mechanical or electrical
process can accurately recreate music that isn't already present in the
recording.
"MDAT is unlike anything the industry has seen, or heard, before. Here's
why: Rather than address the digital signal as a series of sine waves--as is
standard convention
This just isn't true. Standard convention is to address the digital signal
as a series of samples.
"--the MDAT-equipped CDSA SE processes (and upsamples CD audio to DSD for
conversion to analog) by dynamically adapting to the transient nature of the
musical signal.
In fact the basic nature of musical signals is exactly what they just said
they don't do. Musical signals are composed of a series of sine waves. Every
musical signal can be accurately analyzed and represented as a collection of
sine and cosine waves. CD players don't do that, but FFTs do. The human
ear, being largely composed of a collection of narrow-band filters, can also
be characterized as addressing the musical sound as being composed of a
series of sine waves.
In this way, the CDSA SE is utterly unique and singularly able to
preserve the phase, frequency and
dynamic integrity of the original signal.
In fact the best way to preserve the phase, frequency and dynamic integrity
of the original signal is to treat it as a series of samples, which is what
they already said that their product does not do.
Once you've heard this level of improvement in terms of resolution, nuance
and dynamic shading, there's no going back.
So where's their reliable bias-controlled lisetening test data that supports
this claim?
Doesn't all this assume perfect behavior of a D/A system?
Given the known fundamental resolution of the human
auditory periphery, "perfect" is simply irrelevant. "Practically
perfect" is achievable.
The main
reason that oversampling came about is to deal with the limitations or
flaws in the digital filtering process.
Wrong.
The reason why oversampling was implemented (and it was
implemented long before a lot of people here seem to think
it was), was to be able to move the anti-imaging process
out of the analog domain, where the implementations were
not so much "flawedd" in some vague sense, but expensive
and difficult to implement in any repeatable fashion using
conventional analog topologies, into the digital domain where
a number of rather significant constraints were relaxed.
Things like smearing and phase
issues. By oversampling, you're not re-creating, but rather improving
the phasing and smearing issue.
Huh?
It is well known that part of the
reason that 44.1/16 was "flawed" because the filter digital filter
that needs to be applied should have so steep a curve which tends to
cause unwanted, audibly unpleasant artifacts.
"It is well known" by whom? Yes, a lot of things are "well
known" in the high-end audio realm, and many of those
"well-known" things are wrong.
Let's please set the record straight. An oversampling
reconstruction/anti-imaging filter in a D/A converter
MUST have a brick-wall low-pass cutoff at below half
the original sample rate, whether it's implemented
as a pile of expensive resistors, inductors and capacitors
or whether it's implemented as an oversampled filter.
The cutoff MUST be below 22 kHz and it MUST be essentially
a brick-wall filter.
What an oversampled filter lets you do is push the majority
of that filtering to the digital domain, where you have many
more degrees of freedom in your design.
Oversampled filters work thusly: Take you incoming
stream, at 44.1 kHz. By itself, it contains the base-band
audio from 0-22 Khz, an image from 44 to 22 kHz,
an image form 44 to 66 kHz and so on. You HAVE to
get rid of all of those images, thus the requirement
for the brick wall filter.
When you oversample, let's say by 8x (to make the math
easy), now you have your original 0-22 kHz base band
signal in a new base band from 0-176 kHz, an image
384-176, another image from 384-528 kHz and so on.
Now, instead of trying to implement some wildly difficult
analog filter at 20 kHz, you can implement a nice, really-
steep, near brick-wall, linear phase (if you want), low
delay (if you want) or whatever, completely in the digital
domain: your cutoiff frequency is tree octaves below the
Nyquist point, so your artifacts are miminal, and all you
have to do when your done is have an external, gentle,
simple (and, thus, cheap) analog filter sufficient to
remove artifacts at 176 kHz and above.
And, you should note, the MAJOR portion of the cost
of implementing a brick-wall filter in the analog domain
is in the cost of the parts and assembly, thus substantially
raising the per-unit cost of players. The per-unit cost
of an oversampled filter is essentially zero: you probably
already have all the silicon you need anyway.
With this recently introduced Consonance Linear 120 player, it boasts
no over-sampling and no digital filter. It's well received by several
reviewers. Here's a link to the theory behind the digital filterless
DAC;
http://www.sakurasystems.com/articles/Kusunoki.html
This was soundly rejected by the rest of the signal
processing world decades ago. Only in high-end audio
does this sort of patent nonsense not only survive but thrive.
I find this player very fascinating because it goes a whole new way
about extracting digital audio data. My guess is that to go the route
of making a digital filterless DAC, you have to build all the
associated components, the opamps and clocks and ICs to a
fantastically high, and expensive, standard.
Nope, what you have to do is spend a lot of money
on replacement tweeters and output devices, because
ALL of those images are being sent raw out to your
amplifier and tweter.
Such designs are the result of one of two possibilities:
1. Technical ignorance and incompetence on the part of
the product designer,
2. The hope on the part of the product designer of technical
ignorance on the part of the consumer base
In other words, to deal
with imperfect components in the DAC chain, they got rid of the
digital filter and made them remaining components to much more
stringent standards.
No, it's simply a lack of fundamental understanding of
the most basic principles of signals and circuits.
This comes at a price, of course. If there ever
comes a time when gold plated, silver deposited,
1 u meter ICs became cheap, this technology may
find its way to the lower end consumer
audio market like that $50 Walmart CD player.
You wanna take the bet?
I'll bet good money that in 5 years, this "technology", if
you can call it that, will not make it at all out of the
boutique high-end audio-as-jewelry marketplace. In
fact, I'll bet that it will die the type of obscure ignoble
death uniquely reserved for this sort of gross technical
incompetence and negligence .