The main purpose of this article is to equip the reader with the basic knowledge required in order to qualify and quantify a suspected authoring error in a DVD's audio track. Identification of such problems may provide the impetus towards returning the disc to the store and to seek out an alt.-region release not subject to the same defect, or, in more egregious cases it may provide the basis for a call for the manufacturer to recall the defective product. It is my (perhaps deluded) conviction that better equipped and informed consumers will eventually lead to better products, as more DVD companies begin to realize that they can no longer get away with their reprehensible quality assurance & control practices. Some very elementary principles of digital signal analysis will be reviewed, and some related techniques are directly applied to the analysis of the audio track of some recent DVD releases.

Figure 1 shows a sine wave with a frequency of 440 Hertz. In music, this is referred to as the A above middle C. 440 Hz simply means that the wave goes through 440 full cycles each second, resulting in a wave period of 1/440 seconds, i.e., about 0.002 seconds, which can be seen in Figure 1 as the time needed for the sine wave to complete one cycle. Click here to listen to the tone (mp3). Figure 1 represents the wave in the so-called "time domain," where the horizontal axis represents time, and the vertical axis represents amplitude. It is however often more constructive to look at a signal in the "frequency domain." Figure 2 shows the exact same signal, but this time in the frequency domain. In this representation, we have the frequency (Hz) along the horizontal axis, and the "intensity" of each frequency contained within the signal along the vertical axis. In this case, we see a single peak at the frequency 440 Hz (= 0.440 kHz). This informs us that our signal contains, of course, a strong component of 440 Hz and nothing much else. The sinusoidal waveform is the most basic repetitive pattern in existence. A sine wave consists of only one single frequency, without any harmonics, or (in musicians' terms) "overtones."

But what if the frequency content of a signal changes with time? To study the evolution of a signal's frequency content, we need a sonogram. This is essentially a series of frequency domain plots (as shown in Figure 2) stacked side by side, one for each instant in time. Figure 3 shows a sonogram for the above 440 Hz tone. Time runs along the horizontal axis, frequency along the vertical axis, and "intensity" (the 3rd dimension, coming out of the plane of your computer monitor) is now represented by the use of colors. We notice, quite as expected, the presence of a pure 440 Hz (0.440 kHz) tone for the duration of the 2-second audio clip.

Let's look at a more complex signal, such as Arvo Pärt's beautiful Ode I (mp3, excerpt). Figure 4 shows, along the top, the corresponding sonogram, and below it we see a zoomed-in view of the portion selected by the light blue box at the top. The horizontal traces in the sonogram indicate sustained voices at various frequencies — a signature we indeed expected to see, based on what we just heard in the audio clip. Now let's do the unthinkable and pollute the audio clip using this annoying sound — a 10 kHz sine wave. The modified music piece sounds like this, with Figure 5 showing the resulting sonogram. The sonogram is identical to that of Figure 4, except for the addition of an unwelcome 10 kHz frequency component which fades in about 2 seconds into the clip.

Such a 10 kHz noise is easily spotted and eliminated by audio engineers, and would thus not be present in the final product: the DVD or CD we buy in the store. But what if a somewhat higher frequency noise accidentally found its way into the system somewhere along the way? If the audio engineer is not aware of its presence, it may very well end up on the final digital master. The average range of hearing for children is from a low of 20 Hz to a high of 20 kHz. By the time a person hits the age of 20, years of attending loud concerts may already have killed off the 'high tones'. Studies show that the high range for an average young adult is about 16,000 Hertz (16 kHz). In another 10 years, a 30 year old is often down to a mere 12,000 Hertz (12 kHz). Such a person would simply not be able to detect the presence of, say, a 14 kHz squeal in an audio track. Unfortunately for those of us who have somewhat above-average hearing, many CD and DVD studios appear to be hiring such average (in terms of hearing) personnel. Yours truly, age 40 and perfectly able to hear well past 17,000 Hz, has, along with a not insignificant fraction of the population, for decades had to put up with such high-pitched squeals — and there is generally no sympathy to be received from family, friends and DVD/CD store managers. They are simply unable to hear what is emanating from the loudspeakers and promptly attribute your complaint to a bad case of tinnitus. Classical music CDs have traditionally been the worst offender, but we have noticed a worrisome trend in that similar problems are increasingly starting to crop up on DVD audio tracks. Let us use some very simple digital signal processing techniques to study this phenomenon and some other noise-related artifacts. By way of example, we will take a closer look at some recent popular DVD releases.

This is a brilliant piece of cinema, and it is presented here in a truly excellent video transfer. However, what the numerous rave reviews scattered across the Internet neglect to point out is that the original English soundtrack suffers from a serious defect. Listen, for example, to this audio sample, taken from 01:47:27 into the movie: [ mp3 | aiff ]. Those of you with an acute sense of hearing will most likely get an instant headache from the high-pitched squeal ("ringing"), even on fairly low-end loudspeakers and headphones. The corresponding plot of amplitude as a function of time is shown in Figure 6. It looks pretty normal. But let us take a closer look at, for example, the particular moment in time we have marked with a vertical yellow cursor. Looking at this moment in the frequency domain, we see what is shown in Figure 7. Notice especially the prominent peak just below 15 kHz. The exact frequency is 13800 Hz. A sonogram of the entire clip is shown in Figure 8. Note the 13.8 kHz trace, more or less present throughout the clip. This incredibly high-pitched squeal reappears at frequent intervals throughout the movie, regrettably making watching the DVD a very unpleasant experience. Here is yet another example, from 01:04:50 into the film: [ mp3 | aiff ]. The amplitude plot is shown in Figure 9, and the frequency domain plot from the window in time marked in red is shown in Figure 10. Again, note the presence of an unwanted, highly distracting prominence just below 15 kHz. The sonogram is shown in Figure 11. You may find it instructive to play the presented aiff files in a piece of free software called Frequency.

Another highly anticipated DVD release, another immense disappointment. This Seville release, a direct port of the Wellspring release, suffers from a very similar problem to that of the above discussed disc. A high pitched ringing starts only a few minutes into the film, riding on top of the voice of the "invisible filmmaker" character. Starting at 00:46:11 into the film there is an otherwise quiet moment, allowing us to have a closer listen at the interfering component: [ mp3 | aiff ]. The ringing gradually starts to become noticeable about halfway into the clip, and by the end of the clip it is almost unbearable, threatening to shatter one's wine glasses in the cupboard. Figure 12 shows the clip's amplitude plot, wherein nothing abnormal is evident. The frequency domain representation of the signal within the portion marked with a red band is shown in Figure 13. Note the prominence at 15000 Hz. Figure 14 shows the frequency content a few seconds later, within the time slot marked in Figure 15. Notice how the interfering 15000 Hz component has increased in power from time 00:46:17 to 00:46:22. The sonogram shows this increase quite clearly, in Figure 16. Again, this kind of plot is proof positive that the disc is suffering from a serious defect.

This disc, while not exactly a masterpiece of Cinema, serves to illustrate yet another kind of audio problem which can be studied, quantified, and brought to the attention of the manufacturer. In terms of musical performance, pure inspiration, raw energy, and unpolished, intimate, footage, U2 Go Home: Live from Slane Castle is the by far best U2 concert DVD to date, and perhaps one of the best live concert DVDs out there. But the PCM Stereo track — which is what this author likes to listen to on headphones while traveling — is hopelessly messed up. Listen to this audio clip (this is the left channel), starting at 00:49:25 into the film: [ mp3 | aiff ]. You will immediately become aware of periodic injections of white/pink noise into the audio channel. The clip's sonogram shows this very clearly, in Figure 17. This defect appears in the left channel at frequent intervals throughout the film, and often during exceptionally intense and inspired musical moments. DVD reviews found on the Internet praise the PCM Stereo track as "excellent." We beg to differ.

Several people have called to our attention a "cyclic noise" on the audio track of this Criterion release of Robert Bresson's 1945 film. The following is an audio clip from early on in the film, where the protagonist is riding the elevator up to her apartment: [ mp3 | aiff ]. The "whirring" sound is normal, it is the sound of the elevator, but the larger-scale "pulsations" are not — they continue after the elevator stops. The sonogram for this clip is shown in Figure 18, wherein the "pulsations" are clearly evident. The pulse width is 0.6 seconds, and the pulsation period is 0.7 seconds. The noise clearly originates with the inferior source materials used by Criterion. Any attempt at digitally cleaning up the audio track would presumably have done more damage than good in this case. While it is very easy to remove a discrete-frequency signal, such as the 15 kHz ringing described above, by applying clever retouching in the frequency domain before transforming back into the time domain, it is very hard to remove the kind of broadband noise exhibited by this disc.

Les Dames du Bois de Boulogne was made in 1945, and Stranger Than Paradise was made almost 40 years later. Yet, the latter exhibits a pulsating noise remarkably similar to that of the former (discussed above). Here is a sample audio clip, starting at 00:07:25 into the film: [ mp3 | aiff ]. The corresponding sonogram is shown in Figure 19 (contrast enhanced for clarity). The top shows the entire 80-second audio clip. The three bursts towards the end is the phone ringing three times, each time followed by Eva telling Willy that the phone is ringing. At 22 s we see horizontal structure corresponding to car tires screeching outside the window. Other well-defined vertical (broadband) features are environmental noises generated by Eva herself. The bottom view shows the first (selected) portion in close-up. We call your attention to two artifacts. Firstly, the lack of all frequencies in the important 1.4–1.6 kHz regime, contributing to a rather "hollow" sound. It should not surprise us in the least if it turns out that the observed frequency dropout is due to an attempt by MGM to filter away some interfering signal component in that range, in the process throwing the baby out with the bath water. Secondly, a low-frequency pulsating noise (similar to that exhibited by the Les Dames disc), which is very audible and distracting, manifesting itself as periodic excursions as indicated. To the viewer this sounds like a constant "whoosh-whoosh-whoosh" at a 1-second periodicity. This is heard during large portions of the first half of the film. This author does not recall hearing this effect in the theatre. The pulsating noise heard on this release, as well as on the Les Dames disc, is presumably due to actual physical damage to the optical soundtrack running up the side of the film. Damage occurring to a wound reel might explain the cyclical nature of the problem. In both cases the DVD companies may be faulted for choosing to base their release upon obviously damaged source elements.