 |
 |
 |
 |
 |
 |
Upsampling: the magic bullet ?
For some reason it is very often the case that while two parties fight
to win a battle, a third one, unknown by the two others, nicks the prize.
This might just as well be the case with the battle that’s going
on between SACD and DVD-Audio.The technology offering what seems to be
the perfect solution to any demand of new format is called“upsampling”.
It is said, as shown by the several interview at the bottom, that upsampling
would be the ‘magic bullet’, or the perfect go-between.
The technology: what does it do, and how does it work ?
Upsampling is said to releave the digital signal contained in standard
CD’s of its disadvantages, and bring it very close to the sonic
qualities that SACD has to offer. It is said to smoothen the sound, take
away the cold, analytical and ‘digital’ feel that conventional
CD is bothered with.
So how does it do that ?
Upsamling works by increasing the resolution of conventional CD to that
of a DVD. The technique that is used to do just that is quite complicated
but can easily be understood with the use of some intuitive grasp of basic
mathematics. We will start by explaining the very basic principles of
the PCM theorema once more, which is the best way to allow everyone to
understand the idea of upsampling. In the process of explaining the working
of the ‘upsampling’ process, we will in the meantime make
our point that upsampling is in fact no more than hyped ‘oversampling’,
a 20-year-old technique that is around since the very first days of Compact
Disc.
To get back to the basics of PCM-digital, and it’s obvious drawbacks,
there is a finite amount of information contained in any linear (PCM)
digital recording.
At best, assuming perfect conditions throughout the recording chain, this
quantity of information is dictated by the sampling rate (how many samples
of the original analog waveform are taken per second) and the sampling
depth (the precision, measured in bits, with which these samples are measured).
The standard rate and depth for the Compact Disc is 44,100 samples per
second and 16 bits per sample.
Sonic improvements can be realized by increasing either the sample rate
or the sample depth, as long as there is no significant limiting factor
in the analog path before the signal is converted to digital. However,
once you have captured everything that a given sample rate and depth can
"hold," any leftover information is lost forever.
On playback, the trick is to preserve every iota of the information contained
within the digital recording in order to recreate as much of the original
performance as the recording managed to capture.
When Sony and Philips first introduced Compact Disc in 1982, the partners
themselves disagreed about how to best accomplish this goal of preserving
the information on playback.
Given the limitations of 1982 technology, Sony felt it best to use 44.1
kHz, 16-bit digital to analog converters. After all, these would exactly
reverse the process that had been conducted to get the recording onto
the CD in the first place.
Philips elected to use 14-bit converters running at a higher frequency–specifically,
176.4 kHz, which is four times faster than 44.1 kHz. In order to do this,
Philips used a technique called "oversampling," whereby three
additional samples are interpolated between each original sample from
the recording itself. This process is done using various mathematical
techniques, but creates intermediate values between the "real"
or original samples. The result is similar to a connect-the-dots child’s
drawing that now has more dots, spaced more closely together. Specifically,
it contained four times as many dots and became known as "4x oversampling."
Mathematically, both the Sony and the Philips approaches in 1982 were
identical. Each contained exactly the same amount of information–Philips
had simply re-packaged the information into a different form, using more
samples with less depth per sample. The advantage Philips had was that
their approach allowed them to use better-sounding analog circuitry after
the signal had been converted back to analog.
This takes a moment to explain. When digital audio is converted back to
analog, it does not recreate the continuous, smooth waveform that was
originally sampled. Instead, the waveform is only a reasonable facsimile
of the original, but with small "staircase steps" all along
the length of the curve. These sharp edges create large bands of distortion
centered on multiples of the conversion frequency, with a bandwidth equal
to plus-and-minus the range of frequencies recorded. In 44.1 kHz audio
recordings, these bands extend out ± 20,000 Hz around 44,100 Hz
and its multiples (88.2 kHz, 132.3 kHz, 176.4kHz, etc.)
Note that the first of these bands of distortion reaches as low as 24.1
kHz (44.1 kHz minus the 20kHz bandwidth of audio). This is precariously
close to what we hear, and it needs to be filtered out effectively in
order to avoid serious sonic compromises like IM distortion, overloaded
tweeters, and power amps having fits over large-signal ultrasonic information
they were never designed to amplify.
For this reason, Sony had to include what are called "brick-wall
filters" in their first-generation CD players: ninth- to eleventh-order
analog filters that added complexity, cost and significant audio compromises
( phase shift, in-band ripple, and other problems).
By contrast, the Philips approach was designed to fool the digital to
analog converter into thinking that it was working with a 176.4 kHz signal.
Even though there was no additional information contained in the signal,
the fact that it was being converted at 176.4 kHz meant that the bands
of severe distortion were centered on multiples of 176.4 kHz (and they
were still only ±20 kHz wide). Thus the lowest edge of this band
of distortion reached down only to 156.4 kHz–nowhere near the audio
band, and easily handled by a simpler, better-sounding analog filter.
In the long term, the Philips approach won out. Every CD player since
approximately 1984 has used oversampling to ensure that a simpler, better-sounding
output filter could be used.
Of course, in relatively short order, DACs (Digital-Analog-Converters)
capable of reproducing all 16 bits at higher frequencies (4x, 8x, even
16x 44.1 kHz) became available. Then 18-bit DACs, 20-bit DACs, and even
24-bit DACs became available, all capable of running at extremely high
rates with excellent accuracy. Even the cheapest DACs today will outshine
the finest technology had to offer in 1982, and by a wide margin.
The point is here that "upconverting" digital audio to higher
rates for conversion has been around, and in fact has been the overwhelmingly
dominant way of doing things, since somewhere between 1982—1984,
depending on whom you ask. Historically, it has been called "oversampling."
But whether you call it "oversampling," "upconversion,"
"upsampling," or anything else, it remains the same. It is a
method by which the originally sampled data is converted to a higher rate
for the purpose of better-quality conversion to analog.
But why then, do these upsampling machines make an ordinary CD sound
nearly as good as SACD ? The principle here at work is basically the same
as the one we saw happening in the late 80’s, when the first ‘outboard
DAC’s’ were being introduced. An outboard Digital-Analog-Converter
appears to do nothing more than the DAC inside the cd-player. But in fact
in has a great advantage over inboard DAC’s, because it is housed
in a separate chassis, with a dedicated power supply. More than one manufacturer
(eg Cambridge Audio)has shown empirically that the performance of the
DAC can be significantly improved by simply taking it out of the cd-player’s
box, housing it into its own chassis, and feeding it with a dedicated
power supply. “It’s not what you use, it’s how you use
it’, is a very-well appreciated saying in the realms of audio engineering.
Upsamling does in fact do the exact same thing, it just takes the Digital-Analog
conversion out of the cd-player’s box one stage earlier than usual
standalone DAC’s do, and then sends it to the DAC. This way of working
ensures that every step of the conversion from the digital data on the
CD to an analog signal going to the amplifier can make use of its own
casing (and therefore is better protected from all sorts of noise-inducing
phenomena such as magnetic radiation from power lines, standard electro-magnetic
interference (EMI) and radio-frequency interference (RFI)), can be fed
through its own power line and power supply, and therefore give far better
performance than the same circuitry would have done if left inside the
cd-player. So yes, upsampling càn sound better then standard oversampling,
but it is not a new technique, more a marketing hype that is used to make
CD survive in the battle between SACD and DVD-Audio. Will it prove to
be ‘the magic bullet’ that will beat the two new competing
formats ? Well, only time can tell, but we have gathered a few expert-opinions
on the matter to see which way it might go.
The first interview of which the most relevant excerpts will be quoted
here is one with Sam Tellig, one of the world’s most respected audio
reviewers, giving his opinion on upsampling in Stereophile.
“ The big advantage of upsampling is that digital distortion and
noise are pushed well out of the way – above 30kHz. So the digital
and analog filters can be shoved out of band, too, minimizing phase shift
and group delay (In simple terms, phase shift wrecks spatial reproduction,
while group delay ruins harmonic reproduction.). Phase shift and group
delay are why conventional CDs often sound airless and harsh. “…”with
upsampling, I hear more space, more air and more low-level resolution,
and even, subjectively, wider dynamic range. The sound opened up and bloomed
– almost the way it does with SACD. I heard harmonic delicacy –
an extension, and sweetness to the treble that reminded me of analog at
its best, the way that SACD does. If you could take a six-buck CD sound
almost like a 20-buck SACD, that would be some trick, wouldn’t it
? That’s exactly what upsampling did for me. Upsampled sound also
seems more delicate, harmonically. It’s not just an easing of the
hardness and edginess so long associated with CD. With upsampling there’s
a rightness about the harmonic presentation. So natural was the sound
that I stopped trying to analyze the performance of the processor and
simply enjoyed the music. It doesn’t give you absolute SACD-like
quality, but it’s very near. Besides that, it also has the advantage
that you don’t have to buy a new player, ánd you don’t
have to buy your entire CD-collection all over again.”