Sonic Frontiers SFD-2 D/A processor The PMD100 HDCD Decoder

Sidebar 6: The PMD100 HDCD Decoder

The PMD100 HDCD decoder chip not only decodes HDCD-encoded sources, but also has within it an 8x-oversampling digital filter. The filter, which is integral to the HDCD process, is used whether or not the source signal has been HDCD-encoded. Consequently, the PMD100 replaces the NPC filter used in most other processors.

The filter section within the PMD100 was designed with a number of design criteria that were derived from Pacific Microsonics' psychoacoustic research—criteria reportedly not applied to existing digital filters. Pacific Microsonics [now owned by Microsoft—Ed.] isn't saying much about what's going on in the filter, but they did provide a few details:

First, the filter coefficients are specified to 27-bit precision—a longer word length that results in greater computational precision. The filter's stopband rejection (the amount of attenuation above half the sampling rate, called the Nyquist frequency) is a whopping 120dB, which should completely eliminate foldback of high-frequency energy into the audioband.

The filter can also be programmed by the digital-processor designer to optimize the performance of the particular DAC used in the processor. Eight dither options are built into the chip, six of which are increasing levels of shaped high-frequency dither specially suited for multi-bit DACs. To further optimize DAC performance, the timing of the PMD100's de-glitch signal can be adjusted to suit the DAC (if the DAC has this provision). UltraAnalog DACs, as used in the SFD-2 Mk.II, can take advantage of this option.

Because HDCD-encoded discs have a higher crest factor or peak:average ratio than conventional CDs, they sound quieter (footnote 1). This difference in subjective volume is compensated for by attenuating non-HDCD signals by 6dB within the processor. This gain matching, which can be performed in the filter or the analog output stage, produces a subjectively similar playback level on all discs.

There's so much more going on in the HDCD process than can been described here. Pacific Microsonics plans a full exposition of the technology in the form of papers to be presented at the October 1995 Audio Engineering Society Convention in New York.—Robert Harley



Footnote 1: The ear uses a signal's average level to determine loudness. A signal that has a high peak:mean ratio will therefore have a lower loudness than one that reaches identical peak levels but has a higher average level. Compression is almost always applied to commercial recordings to increase their apparent loudness. If you've ever wondered why TV ads sound so much louder than the programs that surround them, this is why.—John Atkinson

X