Editor's Note: As part of his twice-monthly column here at High-Def Digest, from time to time Josh Zyber answers frequently asked questions related to High-Definition and both Blu-ray on HD DVD. This week: Josh tackles the subjects of whether Uncompressed audio is better than Lossless, and what Dialogue Normalization really does to an audio signal.
By Joshua Zyber
A couple of months ago, I wrote a column called Blu-ray and HD DVD Audio Explained that spelled out the basic functions, features, and differences among the various audio formats available on both High-Def disc types. In it, I explained that uncompressed PCM audio (as found on many Blu-rays) is an exact replication of the studio master, encoded on disc without compression, and that the lossless audio formats Dolby TrueHD and DTS-HD Master Audio are also bit-for-bit identical to the studio master once decoded. Doing the math, that should mean that a lossless track is also identical to an uncompressed track. Indeed, that is the case. However, some confusion remains as to whether an uncompressed track is actually better than a lossless one.
Now that both High-Def formats have been available for over a year, and each has built up a catalog of hundreds of titles, we have several cases where two high-resolution audio tracks (one lossless and one uncompressed) can be directly compared for the same movie. Examples include Warner's dual-format releases of 'The Departed' and 'Troy: Director's Cut', which feature lossless TrueHD on HD DVD and uncompressed PCM on Blu-ray, or Sony's Blu-ray release of 'Ghost Rider' with both PCM and TrueHD on the same disc. Theoretically speaking, playing the same movie's soundtrack in both lossless and uncompressed encodings should sound absolutely identical, shouldn't it? Well, yes, except that sometimes there are extenuating circumstances that come into play, and indeed some listeners have tried comparing the soundtracks and claim to hear a difference between them.
So what would cause a lossless track to not be identical to an uncompressed track? To get to the bottom of this, let's first take a look at the ways in which each audio format is encoded.
Isn't All Compression Bad?
(Note: Please keep in mind that the following examples have been simplified for explanatory purposes. The actual mathematical workings of lossless compression are more complicated than I can explain here, but this should serve to illustrate the basic concept.)
Let's begin by looking at how uncompressed audio works. A PCM track is an uncompressed digital format that is 100% bit-for-bit identical to the studio master. If the studio master is:
Then the PCM track pressed onto the disc is:
Pretty straightforward, right? The problem when it comes to High-Def discs is that, since the PCM file is totally uncompressed, an entire movie soundtrack takes up a huge amount of disc space. With their greater storage capacity, Blu-ray discs may often have enough room for this, but space is generally more cramped on HD DVD. Even on Blu-ray, some studios prefer to use that extra space for other purposes.
On the other hand, Dolby TrueHD and DTS-HD Master Audio are "lossless" compression formats. Although they're compressed to take up less disc space than a PCM track, once decoded they're also bit-for-bit identical to the studio master. Think of this like a ZIP file that holds a PCM track. Once you unZIP the file, you get a 100% identical copy of the original PCM, without compromising any sound quality. What these formats do is drop certain data, and instead use flags to indicate that the empty spaces in the stream are meant to be filled with that data when decoded. As an example, let's pretend that we have a movie that's half sound and half complete silence. A PCM track would look like this:
As you can see, all those 0s at the end are needlessly taking up space on the disc, literally for nothing but complete silence. To losslessly compress this audio, a TrueHD or Master Audio track would instead look like this:
By dropping the 0s, the lossless version takes up vastly less room, but when decoded those missing 0s are filled in and it looks like this again:
Voila! A perfect reproduction of the studio master at less than half the disc space.
Standard Dolby Digital, DTS, and (to a lesser extent) Dolby Digital Plus and DTS-HD High Resolution are all "lossy" compression formats. In the above scenario, they'd not only drop the 0s, but also drop some of the 1s that are deemed less critical to human hearing. The decoder chips for these formats will then have to make intelligent guesses about what those missing bits were in the first place, under the belief that most people won't be able to hear the difference. The higher the bit rate, the less guessing they have to do. DD+ and DTS-HD HR are not only higher bit rate than old DD and DTS, but have smarter decoders that are better at guessing the missing bits. Still, they're not a perfect replication of the studio master the way that the PCM or TrueHD and Master Audio formats are.
Comparing Apples to Apples
Now that we've seen how lossless compression works, before we can legitimately compare a lossless track to an uncompressed track, we have to be sure that we're actually comparing the same thing. Over the past year, I've read countless discussion forum postings (and a few editorials from people who ought to know better) in which viewers have tried to compare the soundtracks of different movies to prove a point about one audio format being superior to another. The reasoning usually goes something like this: "The PCM track on 'Black Hawk Down' sounds better than the TrueHD track on 'Batman Begins', therefore PCM must be better than TrueHD."
Unfortunately, this entire argument is based on a huge logical fallacy. You can't compare the soundtracks of completely different movies and draw conclusions about the audio formats used on their video discs. Maybe 'Black Hawk Down' just has a better sound mix than 'Batman Begins'? By the same token, I could argue that the TrueHD track on 'Batman Begins' sounds a lot better than the PCM track on 'The Benchwarmers', so have I just proven that TrueHD is inherently superior to PCM, even though someone else just proved the opposite by picking different titles to compare? Of course not. The entire train of thought is hopelessly flawed.
If a person likes apples better than oranges, does that mean that the crate used to ship the apples is superior to the crate used to ship the oranges? For that matter, does this opinion really mean that apples are better than oranges, or just that this one person happens to prefer them? Likewise, is the 'Black Hawk Down' soundtrack actually superior to the one for 'Batman Begins', or is it just that someone likes that mix better?
To further complicate matters, even when you're trying to compare the same movie's soundtrack in its different formats, you may still not quite be comparing apples to apples if the two tracks aren't encoded at the same bit depth. While both Blu-ray and HD DVD are capable of utilizing lossless and/or uncompressed audio up to 24-bit resolution, studios may choose to encode at 16-bit resolution instead, depending on the bit depth of the original source or concerns about conserving bandwidth. For example, on that copy of 'Ghost Rider' with both TrueHD and PCM on the same disc, the TrueHD track is encoded at 20-bit resolution, while the PCM track is encoded at 16-bits. Even though it's the same movie soundtrack, and technically both audio formats are "bit-for-bit identical" to their respective sources, in this case the studio chose to use a downsampled source for the PCM option, which may affect the final audible outcome in TrueHD's favor.
When making conclusive claims about the technical merits of one audio format over another, it's critical to accurately take all these factors into account.
Dialogue Normalization – Benefit or Menace?
So let's say we pick a single movie with its soundtrack available at the same bit depth resolution in both uncompressed and lossless formats, like the 'Troy: Director's Cut'. Now we should finally have a case where playing the Blu-ray's PCM track and the HD DVD's TrueHD track back-to-back should sound instantly identical, right? Well, almost.
Now there's a new wrinkle to consider. Many Dolby audio tracks are encoded with a function called Dialnorm, which is short for Dialogue Normalization, a feature Dolby offers to set the default playback levels. The idea is to avoid having some discs start very loudly and others start very quietly when a receiver is set for the exact same volume. To do this, Dialnorm sets a default center of the soundtrack at a common average, using dialogue as a baseline. Therefore, the relative loudness of movie dialogue should be the same from one Dialnorm-encoded disc to another without a viewer needing to change the receiver volume from normal preferences.
There's been a certain level of hysteria about Dialnorm from members of the audiophile community, who misunderstand its purpose and functioning, and believe that it fundamentally alters the soundtracks encoded with it. In actuality, Dialnorm does not affect a movie soundtrack any more than raising or lowering the Volume setting on your receiver does. Contrary to common misconception, Dialnorm does not "boost" the dialogue relative to the rest of the sound mix, or in any way alter the track's dynamic range. A Dialnorm-encoded soundtrack has the exact same peaks and valleys as a soundtrack without Dialnorm; it's just that the Dialnorm track will contain an extra flag in the metadata telling the receiver to either increase or decrease its entire volume scale globally before playback, so that all movies start on the same scale. And it only does this once at the start of the movie; it does not cause fluctuations after the movie begins.
At any given volume setting on your receiver, a movie like 'Gosford Park' will deliver dialogue crisply and clearly, but the soundtrack won't get much louder, because that film is practically all dialogue. Switching to 'Jurassic Park' at the same setting, dialogue will come through just the same as it did for the last picture, until the dinosaur roars shake your house to pieces, because that movie has a lot of sound effects that are much louder relative to the dialogue. Dialnorm will not make 'Gosford Park' a house-shaking experience, or make 'Jurassic Park' any less of an auditory powerhouse. It just sets them both so that their dialogue is at the same loudness as one another.
This is relevant to our discussion because a Dolby TrueHD track encoded with Dialnorm may begin at a higher or lower starting volume than a PCM track without this feature, even though it's the same movie's soundtrack and the receiver is left at the same setting. There's a well-known principle in auditory research that has shown that listeners typically perceive a recording played back at a louder volume as better in quality than the same recording at a lower volume. That's because the louder the playback, the more pressure generated by its sound waves. At a difference of just a few decibels, the listener may not necessarily be able to tell that one track is being played louder than the other, but subtle sounds in the recording will suddenly start to vibrate their eardrums more forcefully. The result will be that the louder track seems to have more clarity, breadth, and "impact," when in fact the only real difference is that it's being played a little louder.
In order to properly compare the same soundtrack on two different audio formats, they must first be matched to the exact same volume, and this will require a sound level meter to measure precisely. Once that's been accomplished, the audible differences between an uncompressed encoding and a lossless one vanish. Being set for different starting volumes doesn't make one track better or worse in actual quality than another; they just need different volume settings on your receiver.
Does the Hardware Affect the Results?
One last factor to take into consideration: A lossless audio track is really only bit-for-bit identical to its source if it's been decoded and processed correctly. In my review of the Toshiba HD-XA2 HD DVD player, I noted a bug in its audio section that causes bass management for the multi-channel analog outputs to be applied inaccurately when the "Digital Out SPDIF" control is set for Bitstream rather than PCM. That player also seems to apply Dynamic Range Compression whether you want it or not unless all speakers are set to a Small size. Without the required workaround settings (SPDIF at "PCM" and all speakers Small) all movie soundtracks seem to be lacking bass over those audio connections.
If a viewer weren't aware of this problem, a first inclination might be to assume that the Dolby Digital Plus and Dolby TrueHD audio formats used on HD DVD were poor quality. However, this is actually just a defect in one specific player, and not at all indicative of the audio formats themselves.
Similarly, although Fox Home Entertainment prefers to use DTS-HD Master Audio on its Blu-ray releases, at the present time there isn't much hardware that can decode the full lossless extension to the codec. Most currently-available Blu-ray disc players and A/V receivers instead extract the lossy DTS "core," so the majority of listeners aren't hearing the format to its fullest potential. That's not a knock against Master Audio, but rather a limitation imposed by the playback hardware.
What It Boils Down To
The number of new audio formats on Blu-ray and HD DVD have caused a great deal of consumer confusion, especially with three separate formats (PCM, Dolby TrueHD, and DTS-HD Master Audio) all designed to accomplish the exact same goal -- a perfect replication of the movie's audio master. Apprehensions about lossless compression being inferior to an uncompressed version of the same soundtrack are not borne out by the facts. One methodology may have technical advantages over the other in terms of space savings, but the end result is the same whether the disc you buy has an uncompressed soundtrack or a lossless one. They're both equally good, so sit back and enjoy.
Källa: High-Def Digest
------------------------------------------------------------------Jag lägger till en fråga om Dialogue Normalization vara eller icke vara, så svara gärna efter att ni läst artikeln.
Det här inlägget har ändrats av Addex 08 december 2007 - 11:25