by Richard Elen,
August 9, 2000
Originally published on AudioRevolution.com
could reduce the perceived quality of DVD-A to somewhere between a good MiniDisc
and a below-average CD," says a leading classical recording engineer.
The record industry's search for a "watermarking" system that would make itpossible to trace the origin of digital audio recordings despite their processing through internet audio compression techniques such as MP3 (MPEG I Layer 3) and the copying of high-density digital media such as DVD-Audio discs, has run into another major problem.
In British tests, leading record producers were astonished to find that they could clearly hear the supposedly "inaudible" digital watermark, during replay demonstrations of the Verance watermarking system chosen by the record industry's Secure Digital Music Initiative (SDMI). The watermark must be robust enough to survive MP3 compression and similar Internet distribution techniques, while remaining inaudible to users of high-end DVD-Audio discs.
A growing number of industry pundits are coming to regard the goal of a robust, inaudible watermark as being impossible to achieve. Internet distribution techniques rely on "lossy" compression to minimize the file sizes and thus the download time for subscribers. They do this by using one of a number of psychoacoustically-optimized algorithms to determine which sounds in a recording are audible and which are not. Inaudible sounds may be masked by other sounds, or exist at frequencies to which the ear is insensitive. MP3 is the most popular such technique, but there are several others. The techniques are referred to as "lossy" because they throw "inaudible" data away - unlike the MLP (Meridian Lossless Packing) compression technique used on DVD-Audio discs, where all the data is meticulously preserved while still achieving significant data compression.
If a compression algorithm removes inaudible sounds, it can be argued that a watermark will be removed by such techniques - if the watermark is inaudible. If this is the case, it serves no purpose. However, if the watermark is not removed by lossy compression, it will be, by definition, audible. When this evident syllogism was raised by a subscriber to the "surround" internet surround-sound mailing list recently, Robert Stuart, head of leading British professional and consumer audio manufacturer Meridian and co-developer of the MLP compression system, replied, "This is indeed the core of the problem."
The watermarking system currently under consideration would affect equally both the DVD-Audio system (which uses high-sample-rate PCM - pulse code modulation, the most common digitization method used in digital audio systems for the past thirty years) and the competing Sony/Philips Super Audio CD system (which utilizes a bitstream approach called DSD - Direct Stream Digital). However the DVD camp has been considerably more insistent on watermarking than SACD licensees, with a consultant close to Philips noting recently that "Our stated position [on watermarking] as of the AES in Paris [in April this year] was that we would include it only if forced to by content owners."
The SDMI tests in Britain appear to have been very disappointing. The following comments by Tony Faulkner, one of the UK's top classical recording engineers, were posted on the Pro-Audio mailing list in the wake of the UK SDMI tests:
"...The watermark listening sessions themselves were pretty disappointing in my judgment. Poor unfamiliar dull source material, unfamiliar monitoring, limited value A/B/X test procedure...The only consistently usable track for me was (ironically) a 96k/24 transfer off an old analogue Petrushkha, because the differences were easier to identify...
"...with Petrushkha I scored 75% in identifying the watermarking - on two separate runs. It sounded like medium distance buzzing bees (high frequency ones) with a clogged stereo image when the Petrushkha got louder and more complex...
"...I have no doubts in my own mind now that the Verance watermark is clever enough and effectively unobtrusive enough for non-critical low-to mid-fi,... i.e. up to but excluding DVD-A, SACD and high-quality CD. The bad news is that it was audible on poor quality bandwidth limited archive analogue material to a 49 year old engineer with a cold and "747-Ears" the day after a 12 hour intercontinental flight. For audiophiles paying extra money for a new player and for new discs, judging by what I heard yesterday the watermark could reduce the perceived quality of DVD-A to somewhere between a good MiniDisc and a below average CD.
"The myth about the watermarking being optional is becoming very tiresome too. How will it be optional for listeners to major label output ? How is it optional for performers ? How will it be optional for producers and engineers generating regular releases for major international participating labels ? How will it be optional for DVD-A players and recorders manufacturers to choose not to build in and to pay for the technology ?
"I believe that the strategy of watermarking high-quality material on high-quality carriers is fundamentally flawed if the watermarking is audible on high-quality systems. Further I believe that the testing so far has been inadequate in terms of sample size and quality of test material and methods. If it is audible now with a 2bit copy management payload, how will it sound with a 72bit full identifier payload ?"
Faulkner also alludes to criticism of US tests of the system as representing too small of a sample to be statistically valid. Following a demonstration in Nashville, some concern was expressed by at least one attendee, engineer Chuck Ainlay, who said, according to an article in WebNoize "...the test was conducted impartially, but the only test music provided was a new recording by electric guitarist Mark Knopfler, not orchestral, jazz or other acoustic music with a wider dynamic range and more detail than pop music." However there are some apparent errors in the article, notably the spelling of Ainlay's name, and the failure to point out that Ainlay was the engineer on the Knopfler recording, so the quote may be misattributed. ("Audiophile Label, Engineers Question Verance Watermark", at http://news.webnoize.com/item.rs?ID=9648.)
Even so, it is evident that as it stands today, digital watermarking is in big trouble. After the UK demonstration, according to British science magazine New Scientist (July 22), an SDMI representative admitted, "We are starting all over again."
The significance here is that while DVD-audio players are now being shipped, the always hot topic of copyright protection is going to very possibly continue to delay DVD-Audio for the consumer. If handled improperly, this situation could result in poor sounding DVD-audio and or SACD formats. With billions of dollars in sales waiting for both the software and audio hard ware manufacturers, you can count on the industry doing every thing they can to create a more "inaudible" watermarking to avoid quality concerns much like those of CDs versus Vinyl in the early 1980's.