The Vinylizer
On Tue, 10 Aug 2010 18:01:57 -0700, Arny Krueger wrote
(in article ):
"Audio Empire" wrote in message
On Mon, 9 Aug 2010 17:05:12 -0700, Arny Krueger wrote
(in article ):
"Audio Empire" wrote in
message
On the playback end, it was D/A converters that were not
able to do a full 16-bits linearly (early Philips
players (Magnavox) didn't even try. They used 14-bit D/A
converters and the little Magnavox FD-1000 sounded MUCH
better than the Japanese 16-bit units of the day).
The above account ignores the fact that oversampling was
used to obtain 16 bit performance from 14 bit parts.
For all practical purposes, the converters were 16 bit.
No, the D/A converters were 14-bit.
They were in an oversapling configuration. This is well known. The objective
of the oversampling was a trade off of speed which was in abundance, for
linearity which was costly.
They used 14-bit
converters because Philips believed (and rightly so) that
the then current 16-bit DACs weren't very linear.
In 1972 (ten years earlier) I worked with 16 bit, 200 KHz DACs that had 1/2
bit linearity and monotonicity. The only problem with 16 bit DACs was their
price before the CD player market ramped up production.
Yes, so the ones used by many CD manufacturers weren't very linear, and those
which were were more expensive than mass-market manufacturers wanted to
spend. In the early days, numerous things were tried to get around this
problem, lower bit D/As, over sampling, single bit D/As that used the same
bit for everything (insuring the steps were absolutely the same, and
therefore linear) etc. Eventually, the D/As got better (laser trimming,
etc.) and the sound of CD players improved. Today, they're pretty close to
"perfect".
The
fact that they used 4X oversampling to achieve 16-bit
resolution is irrelevant to my statement.
Your statement was false because of the false claims that it included
including "..the little Magnavox FD-1000 sounded MUCH better than the
Japanese 16-bit units of the day). In fact they both were sonically
transparent or very nearly so to the extent that they absolutely blew away
the analog equipment of the day, given proper source material to play which
was readily available from the onset.
My experience tells me otherwise. Sorry about that.
The claim that there was a signficant and large audible
difference has been investigated with DBTs and found to
be yet another audiophile myth.
Sorry. I had both the Sony CDP-101 and The
Philips-Maganvox FD-1000, and I beg to differ. The Sony
sounded awful (still does) and the little Maggie was much
more listenable (and still is). I ended-up giving the
Sony to a friend - he didn't like it either.
I don't believe that we have ever been treated to your technical
measurements or the results of proper statistically-analyzed, time-synched
level, matched comparisons of them. The extant well-controlled listennig
tests involving them tell a different story - both units were eminantely
listenable given that they were in good working order.
Nor have we been treated to your test results and technical measurements or
the results of proper statistically-analyzed, time-synched level, matched
comparisons of them, either.
They
also had really crude multi-pole anti-alaising filters
and produced, what would be considered today,
unacceptable levels of quantization error.
As a rule there are no anti-aliasing filters in playback
devices. Aliasing is only possible in ADCs and
resamplers.
Nyquist requires that the upper frequency response limit
of the reconstructed waveform (the Nyquist frequency) be
half of the sampling rate and the signal at the sampling
rate must not have sufficient amplitude to be
quantifiable. This means that the reconstruction filter
must be very steep to avoid there being significant
signal at 44.1 Khz.
Now you've had a chance to review the relevant technical material and change
your story. The filters are now properly identified as "reconstruction
filters". Yet you present this all like its a correction to my statement
which was correct all along.
I was just using the standard parlance as I explained above (and you
"conveniently" snipped the part where I SAID THAT YOU WERE RIGHT, but that
these reconstruction filters are commonly called anti-alaising filters, even
though that term is not strictly correct. I'm not just addressing you in this
thread, you know? And if you're going to debate with me, I'd appreciate it if
you would try to be a little more honest in your snippage, OK?).
Meaning that above the Nyquist
frequency (in this case 22.05KHz) cutoff needs to be as
absolute as possible leading to designs of filters with
as many as six poles (before the advent of cheap digital
filtering, that is).
If you think that the origional CD players had 6 pole filters, then you are
again not telling it like it was. If memory serves there were about 15
inductors and 15 capacitors per channel in the reconstruction filters of the
CDP 101. This was pretty typical. Any second year engineering student knows
that filters like these have about 30 poles (in pairs).
Even worse. I had forgotten and memory "didn't serve". It's been a long time,
so what?
|