On July 10, 2014, a new documentary, titled The Distortion of Sound, premiered at the Grammy Museum in Los Angeles, California. It promised to deliver “an eye-opening exposé of the current state of sound”, enlightening music fans as to what they have been missing out on by listening to the lossy audio formats that now dominate today’s music industry. Following the premiere, Harman, the company that produced the film, made it available for free on its own dedicated YouTube Channel.
When I first saw the trailer for The Distortion of Sound, I was excited at what appeared to be a polished and well-researched documentary examining the decline in sound quality at the hands of convenience. Unfortunately, after watching the 22-minute film I was left feeling confused and angry.
The Distortion of Sound is chock-full of misinformation and deliberately misleading audio examples. The filmmakers misuse and confuse the term ‘compression’ so many times that even the president of the Audio Engineering Society, Dr. Sean Olive, who is quoted extensively in the film, would have difficulty pinpointing just what type of ‘compression’ any given line of dialogue refers to.
As SonicScoop readers know, the term “compression” has two main uses in the audio field. There is:
1) “Dynamic Range Compression”, a reduction of fluctuations in level (eg what happens when you send a bass guitar through a LA-2A) and:
2) “Data Compression”, a reduction in the size of a file (often resulting from the use perceptual encoding, e.g. MP3s).
Having spent the good part of the past 10 years teaching audio engineering to undergrads, I am used to students confusing the two terms at first, and using them interchangeably. It’s an understandable mistake for students and laypeople to make.
It is however, a completely unacceptable mistake for an established audio company such as Harman to make—and to make again and again.
Their film consistently weaves between both senses of the term, referring to loudness and dynamic range compression in one sentence, then MP3s and data compression in the next without missing a beat.Interview any musician about the state of sound today, and they are likely to have something to say about either or both of these very different types compression. But it is up to the director to discern which type of compression they are talking about and to make sure that remains clear.
The film also features some astoundingly misleading audio examples that insult even the casual listener’s intelligence. At 11:20, after discussing how MP3s are ‘dangerous’ as they ‘are not what is faithful to the original recording’ because ‘portions of the original signal just aren’t there anymore’, the filmmakers play an example of what this apparently sounds like in order to illustrate the problem. In their audio example, a clean-sounding violin is transformed into a distorted, band-limited, modulated version of itself. The resulting sound is absolutely nothing like the result of perceptual encoding — not even close. The worst MP3 preset on iTunes sounds light-years better than the example presented.
This is one of several cases in which the producers have deliberately manipulated the sound to make it painfully obvious that the ‘before’ sounds much better than the ‘after’. The big and recurring problem with the film is that the kind of processing used to achieve this result is not the kind processing that the segment is actually discussing.
Harman have made a film that argues that by listening to MP3s, people are missing out on the ‘original experience’ as lossy formats sound much worse than the original. To demonstrate this, they could have played the original clip, and then the MP3 version, and let the difference in sound speak for itself. But they didn’t. Instead, they manipulated the sound clips to support their desired conclusion.
It is easy to understand why they would do this. The biggest problem facing the filmmakers is that if they had taken the honest approach, the actual difference just isn’t that obvious to the average viewer. (And certainly not when you’re watching the film on YouTube, which is another issue entirely).
It is likely that the producers of the film likely tried a legitimate ‘before and after’ MP3 test, but were not convinced that listeners would be able to hear a difference. (In many cases, an audible difference is there, but it is far from night-and-day for any layperson, and perhaps many audio engineers).
So instead, the filmmakers set out to manipulate and completely change the original sound of the file through the aggressive use of signal processing effects, resulting in a sound that is nothing like the original, but also nothing like the sound of perceptual encoding. This is much like attempting to demonstrate the difference in performance of a Ferrari with Fuel A versus Fuel B, but before testing with Fuel B, taking a moment to stop and slash the tires.
This kind of dishonest presentation pervades the film. At 12:05, the filmmakers present a drum recording, and then switch to the alleged MP3 version, which they have treated to be so far beyond distorted as to be unlistenable. The film suggests that this effect was actually the result of perceptual encoding.
Assuming that the original audio was 16-bit 48kHz and was then converted to a MP3, this would never result in such brutal distortion. I recorded the original drums from the film and tried various different MP3 conversions, (well aware that the audio is already data compressed). The result? Even a 40kbps MP3 doesn’t sound as trashed at the example presented, and this is less than 1/6th the size of a standard iTunes track today. A discerning listener will also likely conclude that some form of dynamic compression has been applied to the drums in the film—It clearly sounds like they have been hit hard with a limiter and a bucket-load of makeup gain. (The dead giveaway is the significant increase in the level of the room sound in the clip.)
The drums may very well have been converted to a MP3 after this (or before it), but all bets are off by this stage. You can’t rightly slam a drum recording through a limiter to the point where it’s falling apart at the seams, and then point to perceptual encoding as the villain. The film not only routinely confuses the dynamic range compression with data compression, but it baldly presents the sound of one as the sound of the other.