|
OPTIMIZED FOR
AM-RADIO?
Let's start with the officially remastered
Hound Dog and the reason any sane
person would want to compare that to all other released
versions.
There is little point in complaining over CD releases of
Elvis 50's masters prior to the 50's Box in 1992. Apart from
tapes of different sources being used, commercial digital
audio was new territory that apparently took the industry
about 10 years to fully understand and in the meantime
development brought technology to expand the limitations of
the chosen format. But by 1992 it should have been possible
provide consumers good CD transfers of vintage music as long
as the source tapes were of high quality and most of Elvis'
1950's master tapes had indeed been found a year prior to
the release of the "Complete 50's Masters". This box set was
the turning point and all later releases so far seem to be
from the same digital masters, including the restoration
work done then. In a way these recordings sounded good, but
on the other hand, what was there to compare with? Many
fans, including myself, probably chose to believe the
official statements that these were the definite RCA 1950's
masters. The Sun masters were a different story and for a
few of the RCA masters it seemed that the good tape sources
couldn't be located. One of the few RCA masters that really
sounded bad was Hound Dog. But then, what's good and
what's bad? To be a little more specific, there are at least
3 completely different flaws that can be identified by
listening in any normal system.:
1. Too much dynamic compression, resulting in
distortion.
2. Too much poor noise reduction, resulting in
a total lack of ambience.
3. Poor equalization, resulting a frequency
response close to that of AM radio (~10 kHz).
A closer investigation reveals possible pitch problems and a
very overdone echo that's different on Left and Right
channels. So the song is not even presented in mono.
Although a shame, we'd just have to accept this as a fact if
this was best source around. However, Elvis and audio expert
Juan Luis González Brugal had found a much better
Hound Dog released on the 1989 CD
Billboard Top
Rock'n'Roll Hits-1956 (Rhino R2 70599). After just listening to it in late 1999 I
had my doubts at first, but when comparing it to the BMG
version, it was obvious I had forgotten just how bad BMG's
master is. There is absolutely no comparison; the Rhino
Hound Dog is so much better and much
closer to the original source. Any consumer could tell that.
The first MP3
sample has the intro from the Rhino version
followed by the same part taken from the BMG version. Fine,
no big deal, now that the information of a better source was
out, the situation would be rectified after which nobody
would ever need to mention it again.
Alas no. Instead, in the first published official reply
(Elvis Unlimited #6/2000) we were told that such claims of
the original tape being sent to Rhino "...could lead to a
very unpleasant lawsuit between them (Rhino) and RCA" should it turn
out that Rhino perpetuates this story. The "story", as it
was called, was actually speculation by a couple of
persons on this site and had nothing to do with the people
behind Rhino. The reason for the speculation was us trying
to figure out how the good Hound Dog possible could
have made it to the old Rhino CD and BMG keeps releasing the
worst sounding Hound Dog ever heard. One would have
hoped that the quality of the recording alone ought to be
more important than any "stories", claims or potential
lawsuits.
Then in Paul Dowling's May 2000 Ernst Jorgensen interview,
producer Jorgensen confirmed that BMG's Hound
Dog is mastered from the first generation
master tape or the original album master at least and in a
recent reply in The Man and His Music #49 he finally said
Jehovah: "THE 'BILLBOARD' CD WAS MASTERED AT RCA FROM
"SOME TAPE" IN RCA'S VAULTS." He also suggested that
perhaps they 'doctored' the compression when it was
mastered. 'Doctored the compression'? How do you gate
signals when everything is of 'equal amplitude'? The
compression is far too heavy for it to work. But the comment
about the Billboard CD being mastered at RCA does fit with
analyses and sad revelations of some of the worst known
audio restorations ever done by the music industry. Although
Hound Dog is the worst example, it
turns out that most of the BMG 50's Box is a restoration
disaster. This is easily shown, so let's take a closer look
at the RCA masters as released on the 50's Box, even if it
hurts.
THE ECHO PROBLEM,
THE EQ PROBLEM AND
THE NOISE-REDUCTION PROBLEM
As mentioned before, one of the flaws with
the BMG Hound Dog master was an echo artifact, different on
Left and Right channels. A true mono source should leave a
completely silent difference. In reality, a 2-channel mono
tape would have different subtle dropouts and weaknesses on
each channel. The difference would be a bit noisy with
sudden peaks and scratches as one of the channels drops out
more than the other. A remastered mono song can be made from
the channel that survived best or, if channels are
perfectly synchronized in time, the
average of the two channels (which also would eliminate
random noise that was different on left and right channels).
As our second MP3
sample starts to play we can hear that the
amplified Rhino Hound Dog L/R difference isn't silent. And a
huge dropout or broken peak can be heard after 4 seconds.
Apart from that the difference sounds so consistent that it
is fair to assume that the Rhino CD version is made from one
single recorded track. The difference also implies that
Rhino may have re-equalized whatever source they were using.
Still, there's no big difference between the channels. The
same can't be said of the BMG master that follows. WHAT'S
THAT? Was this remastered in the bathroom? Actually, the L/R
difference reveals a terribly overdone digital
echo or reverb that possibly was applied to both
channels. The point of extracting the difference like this
is that there will always be a random difference with
processing. The left and right channels will not be
processed exactly the same even if all the
settings were the same. So now that we hear the echo
artifact alone an obvious question would be if this echo was
on the originally recorded master? The answer is
no.
Without making it too complicated let's start
with the very first RCA recordings. Complicated because of
the legendary stairwell echo. Any odd echo here could be
explained if it weren't for the fact that we have high
quality mono outtakes that were released on Platinum.
Our third MP3
sample is the Left/Right difference of
I
Got a Woman, first the excellent Platinum outtake, then
the BMG master. As expected by now, the Platinum version has
generated a difference close to that of the Rhino
Hound Dog, but what about the BMG
master? Well, it has the same dreadful echo difference
artifact as the BMG Hound Dog. Why add digital echo when
the original recording wasn't dry? Furthermore, frequency
analyses show that the BMG master has a huge bass
drop from 80 Hz and below. Unfortunately this
kind frequency response is typical of BMG's masters, not
only the 50's masters, but that's another story. By removing
the deep bass and sometimes raising the midrange (even more
evident on the Collectors Gold outtake of I Met Her
Today), recordings do sound clearer, but also
very flat. There is no way to compensate for the lost bass
when so much of the original signal is lost. The Platinum
outtake, with a minimum of lossy restoration, sounds very
good in comparison.
To conclude the echo problem for now, our fourth MP3
sample is a desperate search for life containing
the amplified difference of Don't Be
Cruel, Paralyzed and We're Gonna
Move. They all are infected with that early
1990's digital echo. Even We're Gonna
Move has it which isn't important to anybody who
can recognize a digital echo, but worth knowing just in case
somebody were to say the echo was on the original tapes or
something.
In
1999 BMG Japan released a series of expensive paper sleeve
collectors CD's of original LP's with claims of 24 bit/96
kHz transfers. One of the CD's, "Elvis Christmas Album", was
even said to be taken from a "recently found mastertape".
Still, a careful test of the song Paralyzed from Elvis' second
album reveals that this probably is mastered from the same
physical transfer as the 50's Box version. The difference in
length is about 65 samples (~1.5 ms) after synchronization
and there are no random time fluctuations which rather
suggests multiple transfers from the same DAT source, but
then 65 samples would be a lot. If an old analog tape were
to be transferred physically on two different machines the
error would show. So unless the BMG master was a 24bit/96kHz
transfer in the first place the Japanese claim may be hard
to explain. What they did do was a better normalization of
amplitude and something very interesting; they seem to have
added just a little noise above 16
kHz. This added noise is inaudible and doesn't do any harm
nor any good, but it's interesting nevertheless because it
shows that the Japanese may have been aware of the ambience
problem and probably frustrated because better sources
couldn't be obtained. On one channel they also added a
little bass from 20 Hz and below of about the same
amplitude. Needless to say, this isn't audible either and
even difficult to measure.
THE JAPANESE CHRISTMAS
GIFT
What about the Christmas Album then? This September 1957
session is one of the better sounding on the 50's
Box and still, the Japanese release sounds much
better on a good hi-fi system, mainly because the ambience
is there. Blue Christmas is a fair example for a
review like this as the difference is more evident on silent
songs such as O Little Town of Bethlehem or I'll Be Home for
Christmas. Our fifth
MP3-sample has Blue
Christmas as released by BMG Japan followed by the
regular BMG version. Corresponding spectral plot shows the
similar difference as for the Rhino/BMG Hound
Dog. But considering the Paralyzed trick mentioned above we
need more evidence to prove the Japanese version is as good
as it sounds. A very interesting question here would be if
it would be possible to derive the Japanese version from the
BMG one. It looks impossible at first, but remember that the
BMG versions of this session are the best on the Box so
perhaps it might be possible to "restore" them closer to
original splendor than other songs.
An experiment was prepared like this:
1.
Playback of BMG CD on a normal player (=>D/A)
2. Analog re-equalization of BMG version
3. Re-transfer (A=>D) 48 kHz/20 bit SBM.
4. Digital re-sampling 48 kHz/20 bit => 44.1 kHz/16 bit
of CDDA
5. Digital re-equalization and adding pink noise
digitally
It
turned out it was possible to bring the BMG version closer
to the Japanese, but not all the way. The added pink noise resulted in a
noise-spectrum very similar to that of the Japanese version.
However, the traces of parametric filters could not be
erased and most important of all: It was not possible derive
the true peak-signals above 11 kHz of the Japanese track. So
even if it wasn't easy to show it, there is no way the
Japanese track could have been derived from the BMG
version.
And best of all, on our sixth and final
MP3-sample that has the amplified Left/Right channel
difference of Blue Christmas (sample #5), we finally find
what we've been looking for: A noisy difference with
sudden peaks and scratches as one of the channels drop out
more than the other.
This
Japanese CD is truly mastered from a noisy, old, but genuine
2-channel mono tape. And no CEDAR! One possible analog
generation loss is nothing compared to dreadful digital
restoration. It would be a waste of time and space to write
down what the BMG difference sounds like.
To avoid mistakes
here, there are no signs of any other Japanese paper-sleeve
CD being mastered from anything else but ruined BMG masters,
although I haven't checked all of them.
THE FUTURE
It
is not interesting to find anyone to blame for all this.
That's all in the past, but the following question stands
out as important right now: Did BMG keep the early 1990's
transfers as they were before restoration done
then or did they just keep the restored
disasters? (This question applies to SUN masters as well and
especially You're a Heartbreaker that has a
restoration job close to that of Hound Dog and that should be
compared to the Bear Family ditto, mastered from the same 78
RPM transfer, as released on their first Sun singles
Box.)
I have never heard such a destructive restoration job by any
major record company before. There is not much a consumer
can do, other than not to buy the products or trying to put
pressure on the company to go back to the untouched digital
transfers or to the analog tapes if they weren't kept. The
latter would probably be expensive, but could perhaps be
justified when they realize that their PCM-transfers won't
do as digital archive format for the future anyway. (That's
another subject, if the tapes will make it to a transparent
digital format before falling apart and whether that format
is SACD or not.)
Hopefully
they did keep the basic digital transfers, or the "years of
work, research and technical proficiency" may be the end of
this most valuable of all contemporary music legacies.
S.J.,
November 2000
|
The Real
Thing:
Billboard Top
Rock'n'Roll Hits 1956
Rhino/Billboard CD
(R2 70599)
This page contains short MP3 samples that
are best played or decoded on a PC with Winamp version
2.22(Fraunhofer reference decoder) or version 2.7 or later
(equally flawless). All samples are encoded in 256
kbps/Joint stereo with Lame MP3 Encoder version 1.05, engine
3.58 for full frequency response and a minimum of
artifacts.
MP3-SAMPLE #1:
HOUND DOG
INTRO
1. The Rhino version :-)
2. The BMG version :-(
Waveform of
Hound
Dog test file. The
left source (Billboard) actually has peaks. The right source
(BMG, Elvis 56) is the usual compressed BMG-version.
X-Axis=Time; Y-Axis=Sample Value (+/- 0-32768)
"Perhaps they 'doctored the compression' or
something when it was mastered to give it more of a punchy
sound ." (Ernst Jorgensen, interviewed in The Man and His
Music#49, September, 2000.)
Spectral plot of
Hound
Dog test file. The
left source (Billboard) actually
has frequency response up to 20 kHz and good signal above 15
kHz. The right source (BMG) is the usual noise-reduced
"digitally remastered, optimized for AM-radio"-version.
X-Axis=Time; Y-Axis=Frequency components (black=silence;
yellow=loud)
MP3-SAMPLE #2:
HOUND DOG: AMPLIFIED
LEFT/RIGHT CHANNEL DIFFERENCE
1. The Rhino version
2. The BMG version
The difference can be obtained simply by
subtracting one channel from the other in any wave editor
with a channel mixer. "Vocal Cut" is a common term for this
since the subtraction will remove anything that's equal in
left and right channels that normally would be lead vocals
on a stereo recording.
If
anybody want to try this at home:
1. Rip track 4 of CD 2 of the BMG 50's Box with any digital
extraction tool.
2. Mix the inversion of Left channel with the right
channel
MP3-SAMPLE #3:
I GOT A WOMAN:
AMPLIFIED LEFT/RIGHT CHANNEL DIFFERENCE
1. The alternate Platinum version
2. The BMG master
The difference was extracted the same way as
before. Again, try this at home.
Frequency and
amplitude components of a 5-second sample of I Got a Woman. The magenta source (BMG master)
shows a huge bass drop while the cyan source (Platinum)
shows a more natural frequency content.
X-Axis=Frequency (Hz); Y-Axis=Relative amplitude (dB)
(FFT size: 65536)
"You
can go in there and you can take the top frequencies out and
boost the middle and you can add reverb, or somehow track
the reverb." (Ernst Jorgensen interviewed by Paul Dowling in
May, 2000.)
MP3-SAMPLE #4:
MORE AMPLIFIED
LEFT/RIGHT CHANNEL DIFFERENCE
1. Don't Be Cruel
2. Paralyzed
3. We're Gonna Move
Spectral plot of Paralyzed test. The above source (BMG) seems to be the
same as source below ("Elvis", BMG, Japan). Watch how they
tried to compensate for the noise reduction.
X-Axis=Time; Y-Axis=Frequency components (black=silence;
yellow=loud)
The Real
Thing:
Elvis Christmas
Album
BMG/Japan CD
(BVCM-37085)
MP3-SAMPLE #5:
BLUE
CHRISTMAS
1. The Japanese version
2. The BMG master
Spectral plot of Blue Christmas sample. The left source (Japanese)
actually
has frequency response up to 20 kHz and good signal up to 17
kHz. The right source (BMG) is the usual noise-reduced,
"digitally remastered" version as it appears on the 50's Box
or the 1994 CD "If Every Day Was Like Christmas".
X-Axis=Time; Y-Axis=Frequency components (black=silence;
yellow=loud)
Spectral plot of Blue Christmas experiment.
Source 1 = The BMG version (left)
Source 2 = Experiment, derived from BMG version (middle)
Source 3 = The Japanese version (right)
Note that the added pink noise of the derived sample has
resulted in a noise-spectrum very close to that of the
Japanese. However, the horizontal, narrow/parametric filters
could not be removed as indicated by the arrow at Note #1 in
the figure (dark horizontal lines). Furthermore, neither
re-equalization in the analog domain nor digital filtering
can raise the true peak-signals (vertical lines, magenta in
color) present on the Japanese track above the arrow denoted
Note #2 (11 kHz). There is no way the Japanese track could
have been derived from the BMG version.
X-Axis=Time; Y-Axis=Frequency components (black=silence;
yellow=loud)
"It's
not much better. It's the same." (Ernst Jorgensen
interviewed by Paul Dowling in May, 2000.)
MP3-SAMPLE #6:
BLUE CHRISTMAS:
AMPLIFIED LEFT/RIGHT CHANNEL DIFFERENCE
1. The Japanese version
2. The BMG master
|
