HDCD - Mix Magazine Article
Pacific Microsonics' HDCD - Looking Beyond The Compact Disc
Reprinted from Mix Magazine, June 1999
This scenario poses an interesting dilemma for companies such as Pacific Microsonics of Union City, Calif. Because the CD is the only viable consumer digital delivery format available, the company's efforts to promote its High Definition Compatible Digital technology have until recently tended to focus on HDCD's benefits for that particular format, and its success to date has been built largely on its ability to provide 20-bit performance in a 16-bit medium. But with the introduction of new formats offering better resolution without special processing, Pacific Microsonics must now address questions of HDCD's future relevance, even as it successfully moves the process from the audiophile fringes to the mainstream of consumer audio electronics.
HDCD is a multifaceted process that is not specifically tied either to the CD or to the format's underlying Red Book specification, which calls for linear PCM audio at a 16-bit word length and 44.1kHz sample rate. "HDCD is really a brand name for a family of synergetic and interrelated technologies," says Michael Ritter, Pacific Microsonics co-founder and VP of professional audio. "All of the technologies are aimed at providing lower distortion and higher resolution through any form of digital audio recording and playback. Some parts of the system apply through all release formats, while subsets of the technology only apply to certain release formats." To help ensure a role for HDCD in new PCM formats such as DVD, the company is introducing this month at the Tokyo AES the Model Two HDCD processor, which will work not only at the 44.1/88.2kHz sample rates, but also at 48/96/176.4/192 kHz.
Mastering the Mass Market
"The continued and growing acceptance of HDCD on the pro side," Ritter says, "has translated into our current list of well over 2,000 HDCD CD titles released so far." According to the company, that's double the number of HDCD titles available since the end of 1997. And with more than 125 million HDCD-encoded CDs sold so far, it's hard to argue that HDCD is simply an obscure phenomenon. Many of the encoded CDs are very high-profile projects, including recent releases from artists such as Bruce Springsteen, Garth Brooks and Jewel.
"Look at the Billboard Top 200," Ritter says. "At any given time you'll find between 20 and 30 HDCD titles. So it's not just a lot of titles, but titles from important artists. That has really caught the notice of some of the larger companies in the consumer electronics arena. In the past, the consumer hardware manufacturers who have licensed HDCD-more than 90 companies-were relatively high-end. But this last year we started to see products that are at prices that will really move us into the mass market." As evidence of this trend, Ritter cites recent HDCD licensing announcements, including a $399 DVD player from Toshiba, a $400 CD changer from Harman-Kardon, a five-disc changer from Denon at the $299 price point, a Marantz consumer CD-R recorder (HDCD decoding only), and upcoming products from Kenwood and TEAC.
Although these recent announcements indicate that HDCD is gaining momentum in the consumer hardware market, the current installed base of CD players and receivers equipped with HDCD decoding covers perhaps 1% of the total installed base of players. That means most of the people who buy HDCD releases, at least the many Gold and Platinum titles, are not actually listening to them with HDCD decoding. Yet there's been no groundswell among fans to demand that record companies sell the same recordings in non-encoded versions. By itself, this does not prove Pacific Microsonics' claim that HDCD encoding makes music sound better even if it is not decoded. But it does add credibility to the word "compatible" in the product's name. And the fact that many top facilities are using the process despite the limited installed base of decoders indicates that mastering experts believe that some of HDCD's conversion and encoding processes improve the translation of audio signals into the digital domain, and that these improvements are evident in many playback situations even without decoding. All of this ties in neatly with Pacific Microsonics' argument that its future is assured even if the CD is dethroned.
The HDCD Process
The methodology of the HDCD development team involved a thorough review-conducted intermittently over the first several years-of every component in the signal chain in order to isolate and either eliminate or minimize negative effects. Johnson and Pflaumer were guided not simply by test measurements but also by extensive listening tests. "Frequently," they say in their paper, "changing a design parameter produced a clear perceptual correlate that eluded quantification by measurement." They refer to the system they came up with as a "conjugate system of encoding and decoding," but they believe (as noted above) that some benefits are derived from improvements in encoding, particularly in the area of filtering, that do not require complementary decoding to be realized upon playback.
According to Ritter, HDCD's sonic improvements begin with the Model One's unique A/D conversion process. "The actual A/D converter in the Model One runs at 24 bits and 176.4 kHz currently; the Model Two will also convert at 192 kHz. We improve the linearity of our conversion with a high-amplitude broadband dither signal that we mix in with the program in the analog domain. The dither appears to be random, but the system knows at any given instant precisely what the amplitude of that dither signal is. And because we use our own custom, discrete, full-ladder converter with excellent amplitude and phase accuracy, we are able to apply an 'anti-dither' signal, exactly out-of-phase and matched in time, in the digital domain after conversion. That nulls the dither noise out of the signal."
The next steps in the process are related to filtering, and they depend on the sample rate of the destination (release) format. "If it's going to be a 176.4 or 192kHz DVD-Audio release, then we will not decimate that signal; we use a proprietary filter [non-oversampled] optimized to that sample rate. If it's going to be 88.2/96 kHz, we use 2:1 decimation, and once again we use a filter optimized to that frequency. But in both high-resolution settings, the Nyquist frequency is high enough that we don't use the 'dynamic decimation' process that becomes necessary when we go down to 44.1 or 48 kHz."
Dynamic-decimation filtering is HDCD's response to the well-known problems inherent in filter design for digital conversion systems where the Nyquist frequency is only slightly above the range of human hearing. "A filter designer who has to make a 'brick wall' filter at 22 kHz is confronted with conflicting requirements," Ritter explains. "You want to have flat frequency response out to at least 20 kHz, but you can't have any energy above 22 kHz or you will get alias distortion. This requires a very sharp multipole filter with a very steep transition between the passband and the stopband, which has a number of distortive effects on the signal. It smears transients and causes significant ripples in the passband. If you try to simplify the filter, then to avoid totally unacceptable aliasing you have to start rolling off at 13 to 15 kHz, and even then the signal will not be completely cut off by 22 kHz."
Ritter describes the HDCD approach to this problem: "We slightly delay the 88.2kHz signal, not enough to cause any sync problems but enough that we can do a continuous Fast Fourier Transform. The resultant information is digitally analyzed in real time by an algorithm that determines, based upon a model of the mechanics of hearing and psychoacoustics, what is perceptually dominant in the signal from instant to instant. And that information is used to optimize the decimation filter. One moment you might have a sudden sharp transient, so it uses a filter with minimum time dispersion to pass the transient cleanly. The next instant, there might be a cymbal crash, so it uses a filter that minimizes alias distortion. All the filters are the same length, so you are not getting a phase shift as this is going on."
Another element in the process of downconverting for CD is word-length reduction to 16 bits. "We never simply truncate," Ritter says. "And with the introduction of Version 2.0 of the Model One at the end of 1998, available as a flash-ROM upgrade to existing units, we now have a palette of four 16-bit dither options." The dither and the dynamic decimation together, Ritter believes, add up to a big improvement over typical CD sound. "The reduced distortion- sharper transient response and reduced aliasing-becomes part of the digital recording and will be heard on any player, whether it has HDCD decoding or not," he says. Nonetheless, the optimal playback setting is one in which the playback filters are matched to those used in recording. To achieve this, the Model One hides control information in the signal that tells the HDCD decoder which filter to use. This data is encoded as a pattern in the dither used for word-length reduction. It occurs only 1% to 2% of the time, and the company says that extensive testing has shown that it is inaudible.
In addition to dynamic decimation filtering, HDCD uses the control data to fit a 20-bit dynamic range into a 16-bit signal. Two types of complementary amplitude encoding/decoding are available; the use of either is optional. At the high end of the dynamic range, "peak extend" allows the user to boost gain by up to 6 dB. For quiet signals, "low level range extend" may be used to add up to 7 dB of gain. With both dynamic processes, the control data allows the decoder to restore the dynamics of the original signal. "When you have control over the whole signal path from start to end, you can optimize the record side for what you know will be happening on the play side," Ritter says. "You just can't take it to that degree of optimization if you have no control over playback."
In sum, then, HDCD is a multi-tiered process, with some aspects involved at all supported sample rates and word lengths, and others used only in certain contexts (see page 95). "Our proprietary A/D conversion, D/A conversion and filtering processes are used at all times," Ritter says, "though the filters used at the 176.4/192 kHz sample rates are different from those used at 88.2/96 kHz. When you get down to the 44.1/48kHz sample rates, then you have 'dynamic decimation filtering' going on. And when you use 16-bit resolution, you also have the option of using 'amplitude encode/decode' processing."
HDCD for DVD and DTS
Another area where Pacific Microsonics is working to broaden HDCD applications is multichannel delivery. In January, the company announced an agreement with DTS Entertainment (Digital Theater Systems Inc.) to "work together to enable producers, engineers and record labels to use the HDCD process as the front-end recording technology for DTS Digital Surround 5.1 music CD releases." Ritter explains that "in the DTS header there is now a place for a flag that has been identified as an HDCD flag. That feature is implemented now in DSPs from companies such as Analog Devices that handle decoding of both DTS and HDCD. If a player has both DTS and HDCD in it, then if a DTS recording was made with HDCD, the HDCD playback filter will be selected. But there will be no amplitude processing going on." Ritter also notes that with the Model Two, HDCD encoding will be possible not only for DTS CDs, but also for DTS tracks on DVD-Video discs.
The multichannel processing required for implementing the DTS capability is achieved by using three HDCD processors together. "For surround," Ritter says, "you put one of the units into master mode and the other two into slave mode. There is an AC-coupled precision word clock that comes out of the master that you daisy-chain through the slaves, and then you terminate the last unit. That gives you a conversion clock that is synchronous throughout all the converters."
The use of multiple processors will also allow HDCD to be used to create high-resolution surround material for the DVD-Audio format. One potential complication there is the Meridian Lossless Packing scheme used to fit high-resolution multichannel into the DVD transfer rate. But Ritter says MLP should have no effect. "Any system that is bit-for-bit lossless," he says, "will have no impact on the HDCD signal."
Granting that HDCD will work for DVD-Audio, the question arises again as to why you would want to use it in a format that already has such high resolution. Ritter's answer is that better resolution actually increases the need for optimizing all links in the recording/playback chain. "Let's take a camera analogy," he says, "where HDCD is like a lens with very high resolution and very low chromatic aberration. That quality can actually be better appreciated if you are making a 4x5 plate than a 35mm image. So the truth is that the advantages of all the work we do to achieve extremely low distortion and extremely high resolution are even more evident in a high-resolution format like DVD-Audio than with the CD."
Beyond the purely technical considerations, Ritter adds that "CDs will continue to be the dominant format for a long time to come, so player manufacturers will want to include HDCD in their DVD players for playing back CDs. And once that capability is there, it doesn't cost any more to make it available for DVD playback, as well. That means that in addition to the advantages in performance, there are also no obstacles in terms of cost or marketing. So I think we are going to do very well as far as getting HDCD into future generations of audio players."