DVD Sound Quality
Dan and Priscilla Lavry. Atop their Bösendorfer sits the AD122-96MKII Gold.
Dan Lavry: Truth in Tone Coloring
His invention makes it possible to have DVDs with audio fidelity
by Stefan Zucker
With digital sound, the choice of an analog-to-digital converter is of major importance. Bel Canto Society only began to issue CDs in 2000 because before then all converters truncated overtones and added digital artifacts, such as hardness, glassiness or glare, to an unacceptable degree. The audio used by nearly every DVD company, including those that put out operas, is deplorable. Although DVDs are in vogue, we abstained from issuing them because we refused to be party to a travesty of the sound.
DVD audio typically consists of Dolby Digital at 192 KHz, a compressed format that retains 12.5 percent of the original sound and is designed to trick the ear into not realizing something is missing. But if you compare the Dolby Digital conversion to the original, the degradation is unmistakable. The Dolby Digital sound leaves out punch, brilliance and nuance. Attacks are dulled. Orchestrations sound thinner. Voices lose some of the color that gives each one its unique timbre. Many overtones are missing.
Another obstacle: The technician who authors a DVD with Dolby Digital must encode a squashed dynamic range into the program. With dialog, this sometimes may help intelligibility. But with music you want to maintain the original dynamics, without soft passages pumped up loud or loud ones attenuated. Some DVD players allow the consumer to bypass the squashing (“dynamic-range compression”). But most consumers don’t know enough to seek out players with the bypass option, or don’t realize they own a player with it or forget to turn it on or off. (Is it possible to bypass dynamic-range compression on your player? If so, you’ll find out in the menus. Different manufacturers call this feature by different names. If you ask a salesman, be prepared for a bewildered stare.)
An alternative to Dolby Digital is Pulse Code Modulation (PCM), an uncompressed format, used for CDs. Encoding and authoring with PCM audio is more tedious and time-consuming. PCM also is more expensive because it eats up more disc space; programs that otherwise would fit on a DVD5 have to go on a DVD9 or, indeed, two DVDs. (In 2000, I hoped Direct Stream Digital-not PCM-would be the future, but recent DSD converters are wretched. In any event, DSD is not compatible with DVD.)*
We chose PCM audio for our DVDs. The analog-to-digital audio converters built into DVD encoders truncate overtones and add digital artifacts, so we decided to acquire an outboard converter compatible with DVD encoding (which the converters we owned were not). We directly compared a number of high-end converters-the Prism AD-2 and others, including the Lavry AD122-96MKII Gold. It was truest to the tone colors on the analog tapes we used as sources. I sought out the inventor, Dan Lavry, to find out why.
He plays an accordion and a Bösendorfer piano and edited analog tape. Before becoming preoccupied with digital audio, he designed digital converters for MRI and CAT scans and developed high-speed integrated-circuit test systems used by IBM and Toshiba. His father, Marc Lavry, was a classical music composer, his sister, Efrat, plays the harp in the Haifa orchestra-and his wife, Priscilla, trained to be an opera singer although she did not attempt to make a career. Lavry’s background is astonishing in a field where inventors typically speak of technology and nothing else. Perhaps because of musical sensitivity, his choices in audio design are more discerning than other engineers’.
Dan Lavry speaks his mind
I feel I am the guardian of the original sounds of the music. Transparency is my most important goal. I don’t want to add to or subtract from the sound of the music. There are many roadblocks to transparency, and my job is to surmount them.
I make my Gold Series converters from the ground up, not from integrated circuits, which are ready-made. Others spend $20-30 for the parts for a two-channel converter. These companies have to fit everything on to an IC. They cannot choose many parts available to me, such as film capacitors, inductors, special resistors and transistors. I’m not restricted as they are by space or heat, and I can use higher voltage.
Other designers pick an analog-to-digital or digital-to-analog integrated circuit, add a front-end analog circuit, often just a copy of what the IC maker suggests in its application data, make a printed circuit board including some other ICs (such as receivers and transmitters) and put the result in a chassis with a power supply. The quality of the analog front end, clock jitter, power-supply cleanliness and good layout will make a difference.
One always could go an extra step and bypass some of the internal workings of the converter ICs, such as substituting the up-sampling or down-sampling inside the IC, making one’s own DC removal and more. But to do things really well, one must be ready to outdo the IC makers, and by a large margin. It may seem like an impossible task for a single designer, to beat large companies with hundreds of millions in sales. It is more than do-able and here’s why:
IC makers often try to cram a whole bunch of stuff into a single IC solution. The whole thing must be limited to a space not much larger than a pin head, and it runs “boiling hot” at two watts or so. Why cram a whole function into a single IC? Mass production and selling a complete cost-effective solution.
IC makers must produce results while being limited by their own technology. There are a lot of things they simply cannot do inside the IC. I can use a whole range of resistor, capacitor and inductor values, and I also can choose these parts on the basis of the material they are made of. IC designers are playing the game with a lot less flexibility. They do not have any inductors inside the IC. They have space inside only for very tiny-value capacitors, of certain limited materials. For the most part, they cannot handle high voltage or current or much power.
Some IC makers offer advantages equipment designers are longing for, such as keeping all parts close together and at almost the same temperature, so that they track well, along with laser trimming for great matching of parts and so on. (Lasers are used to shave resistors to very precise values.) Yet the restrictions inside an IC are serious. The technology used by an IC maker for one portion of a converter may not be ideal for another portion. The goal of fitting everything into an IC often stands in the way of optimizing a design for the greatest performance.
The most serious limitations in an IC-based converter are real estate, heat, narrow selection of part types and restricted ranges of values. Many innovative ideas simply are put aside because they won’t fit in the IC or will make too much heat. The main goals are mass production and low cost. It costs a huge amount to make a new IC, and the IC makers need sales volume.
So far I’ve only mentioned technology, but there is also marketing: I was sad to see the industry sell analog-to-digital and digital-to-analog converters that sample at 192 KHz. They know 192 KHz is wrong from a science standpoint. Often it is the little guy who has to fight those King Kong¬?size companies to send them in the right direction. They made not one technical objection to anything I said against 192 K. Yes, they tried, but it took very little to show their arguments to be wrong technically.
My two digital-to-analog converters, the DA924 [for the professional studio] and the DA2002 [for the home], are the only ones anyone makes from scratch. They utilize custom-made resistor networks. Resistors change on account of temperature, so I tie my resistors to a heating element with a thermostat. Components, including resistors, change over time. To overcome aging, the design builds in automatic calibration-the unit readjusts itself when it is turned on. Mine are the only converters with self calibration.
Reducing clock jitter is a critical issue in digital audio. Both the DA924 and the DA2002 provide a superior timing clock: my invention CrystalLock (TM) is a processor-based approach that substantially outperforms the usual phase-lock, loop-based circuitry.
Traditional PCM converters are least accurate near total silence (digital black). These inaccuracies are very noticeable, for example, when a piano note fades away. The DA924 and DA2002 are designed with over 100,000 extra codes for more faithful reproduction at near silence.
My converters utilize a processor-based upsampler, followed by a DMOS deglitcher for clean transitions and a discrete output stage.
The AD MKII is obviously better than any IC converter on the market. My first model provided significantly better performance in 1995 than any IC converter today. It offered a true 120 dB dynamic range (the MKII offers 127 dB), while the best of the ICs cannot get near 120 dB without “A weighting”-a way to get a few extra “marketing dBs.” The MKII yields 130 dB under A-weighting tests. The A-weighting curve is used to measure analog audio gear. The curve puts less weight on distortions and noise in frequencies where the ear is less sensitive, so it does nothing at, say, 1-3 KHz. The idea would work if the curve were right and everyone used it all the time. Some people use A weighting to make their specs look better. I do not use A weighting. I think it’s overused and I’m not sure it’s a good curve.
And then there are proprietary issues I do not want to describe.
For Dan Lavry’s white papers, “Sampling, Oversampling, Imaging and Aliasing: A Basic Tutorial,” “Sampling Theory for Digital Audio,” “Acoustic Bit Correction,” “20 Bit Equipment for 16 Bit Work?,” “Understanding FIR (Finite Impulse Response) Filters: An Intuitive Approach,” “Understanding IIR (Infinite Impulse Response) Filters:An Intuitive Approach,” and “On Jitter,” visit www.lavryengineering.com and click on “Support.”
*We do use Windows Media, a compressed format, for our Webcasts. In my opinion it sounds better than Dolby Digital. And the Webcasts are free.–SZ