Adapted from Stereophile article "Wadia Know?" (Dec. 1996 by R. Harley)
Left to right: Steve Huntley, Peter Bohacek, Steve Jeffery, Tim Coughlin (COO),
and Jim Kinne. Front lawn of Digital Corp., River Falls, Wisc. USA (ca.
early 1996)
Digital’s Jim Kimie, and Steve Huntley talk with Robert Harley of Stereophile.
Digital made a grand entrance on the high-end audio scene back in 1989
with their revolutionary
2000 Digital Decoding Computer. The four box
2000 broke new ground in several areas, including a radically different digital
filter, a jitter-attenuating-reclocking circuit, the first consumer use o fa
glass-fiber optical interface, and a chassis machined from a solid block of
aluminum. The 2000 debuted to rave reviews around the world, opening the door
to
’s rise in its present status as one of the most respected companies
in digital audio.
But suffered a few missteps along the way. The company was founded by
telecommunications engineers who knew little about marketing h audio products.
Side projects not related to
’s core business drained money and engineering
talent from creating high-end digital processors. Marketing choices damaged
’s US dealer network, and personnel turnovers put the company in disarray.
To top it off
’s policy for the last five years of not providing review
products to the US audio press virtually erased their once-high profile in
the US marketplace.
But all that’s in the past. has a new CEO, a new Director of Marketing,
and a fresh direction. Steve Jeffery, formerly a principal of PS Audio, has
taken the
helm. Peter Bohacek, who worked at Avalon Acoustics and co-founded
Ayre Acoustics, is handling
’s marketing. Fortunately,
’s principal
engineers — who worked on the original 2000— are still with the company.
With this engineering continuity and the new management and marketing direction,
is poised to renew their presence on the h scene. The company has also
l its ban on providing review samples to magazines (as you’ll see from my review
of the
27 Digital Decoding Computer and
7 CD transport in the October’96
issue and the Follow-Up on the Model 2000 else where in this issue).
I recently visited ’s factory in tiny River Falls, Wisconsin, to find
out more about
’s technology, people; and future. I spoke with Peter Bohacek,
Engineering Manager Jim Kinne, and Technical Services Director Steve Huntley.
Jim Kinne began by describing how
Digital was created.
Jim Kinne: was started in 1988 by a group of engineers who left a company
called Applied Spectrum Technologies. The original
2000 was developed
as sort of a sideline project there. Applied Spectrum was not interested in
the audio market, so Don Moses left that company and founded
.
The first product was the 2000 Digital Decoding Computer, shown at the
1989 Summer Consumer Electronics Show. Since then we’ve delved into other areas
of digital design —A/D converters and profession al audio equipment. Now our
direction is stable, designing de coding computers, CD players, and CD.
Peter Bohacek (PB): It’s important to note that, even from the beginning,
the 2000 was a group effort. For example, ’s first software algorithm,
which is one of the things that set the 2000 apart from other products of its
time, was developed as a design project by Bob Moses, Don Moses’ son.
Kinne: Bob Moses had this idea to do a filter algorithm in the senior year
of his [ Engineering] program. Don’s father, Moses, funded a project
to make it into a reality, which is what eventually became the
2000.
Bob never was actually an employee of
. He is now an independent consultant
in the Seattle area.
Some of the other people involved in the initial 2000 development—including
Jim —are still at the company. The Digimaster algorithm became the property
of . Some of the things from the original 2000 are still a major part
of all the current designs —except the 12.
[$1500 12 D/A is the only
product to use an off the-she chip instead
of a custom DSP filter]
Steve Huntley: The goal of today is — and was at that time—to simply
make better sound from a Compact Disc. That’s never changed. . . We tried to
apply some new techniques to audio because Don Moses and his father were audiophiles.
Don’s son knew CD players did interpolation, and started thinking about how
to figure out the points on a curve between the known points. He came up with
a theoretical design of the filter working with some professors at his university—Stanley
Lipshitz [ the University of Waterloo] had some communication with him—and
searched for a way to implement the filter in real time.
When AT&T produced their DSP chip, we were able to run the algorithm on
two of them per channel. That eventually led to the practical implementation,
which was the 2000.
RH: Is telecommunications engineering applicable to digital audio converters?
Kinne: Most definitely. You can view digital audio as being a transmission
system; you just have a storage medium in the middle called the CD. A sound
is generated at a source and it’s transmitted to the end user. But the whole
idea of moving signals between chips — and between transports, converters,
amplifiers — that’s all transmission. The engineers recognized that early
on.
Many techniques used in the audio industry are techniques that were driven by cost constraints—like the S/PDIF interface — and not from trying to achieve high performance. Some very crude telecommunications products use far superior technology to that used in audio— technology that was available at the time of the CD’s introduction.
RH: What are the tradeoffs between creating your own software-based filter and using an off-the-shelf part?
Kinne: I do not see software and hardware as being different filters. The designer of a filter chip could just as easily have developed it in software —like the HDCD filter.
RH: I was referring to the decision to use DSP chips running your own filter algorithm from code held in a socketed EPROM It’s a much more expensive approach.
Kinne: Any time you roll your own, you can make trade offs that are not necessarily cost-based. A great deal of the off-the-shelf filters are designed on a cost basis. How that hardware is implemented is influenced by the need to make it inexpensive for a mass market.
Those designers are trying to refine traditional brick-wall filter techniques to a higher and quality. We have chosen to optimize the time-domain performance so that any pre-echo or post-echo of transients does not exist. That’s apparent in an impulse-response test or a squarewave test, which show those types of effects. The disadvantage is that we have a slight rolloff in the high frequencies. We think that the tradeoffs we make to improve the time-domain response more than compensate sonically for the slight rolloff see in the measured response.
RH: The downside to that approach is that it introduces some passband rolloff. There was some controversy when the 2000 come out because it had a rolled-off treble. Some people took exception to an $8000 product with a 3dB rolloff at 20kHz.
Kinne: It was 3dB at 22kHz —half the sampling frequency.
PB: You see that in audio frequently—products that look like they have measurement weaknesses on traditional measurements, whether it’s single-ended amplifiers or planar loudspeakers. But that’s the key thing about the Digimaster algorithm; it’s not a mistake. It was something that was carefully chosen. That tradeoff was made with the clear knowledge that it was a worthwhile one. The slight attenuation at 20kHz is worth while in comparison with what you gain. Yes, it measures less well, but it performs better sonically.
Huntley: Since none of us hear flat to 20kHz, it’s musically not as much of an issue as it may seem on a test bench.
The Digimaster filter algorithm has been around for some time now. Are you working on new filters?
Kinne: We continue to re search new ways of doing filters that are not conventional — and not necessarily achievable. We have recently been granted a patent for a neural-network-based filter, which at the present time is too costly to implement in a practical product. But as processing technology continues to develop, we see that some day the neural-net work filter will be in our best products. It’ll probably be in the low-end products eventually. That’s just the way technology progresses with time. Processing power becomes cheaper and you can do new things. We have been limited more by commercial practicality than by ideas on how to do the filter.
ABOVE: A a technician inspects a CD player assembly at
’s factory
in River Falls, Wisconsin.
Harley: How sonically important an element of the playback chain is the filter?
Kinne: It’s certainly one of the most important things. You have to do good design throughout, but the filter makes a very large contribution to the sonic quality. The analog out put section is also important. You can just as easily produce a transistor-radio sound using our filter as you can using a brick-wall filter. We think that the best sound comes by making the tradeoffs that we have chosen to make, along with other well-thought-out design practices that we implement.
Harley: appeared from nowhere with a landmark product, the 2000. It
was full of innovative technology, but it seems
didn’t fully exploit
that technology in the marketplace.
Huntley: That’s true. was the first company to recognize jitter as a
source of audible distortion in CD players. But
is not well known as
the company that — I don’t know if you want to say “discovered” —but made the
correlation between sonic performance and jitter performance. That’s a failure
of the company. The same is true for fiber-optic transmission, chassis design,
and noise-reduction techniques.
RH: has chosen to take a low profile in the US audio press the last
five years. Why?
PB: Actually, has chosen to take a low profile not only in the US press,
but in the US in general.
was started by a group of engineers who had
some great ideas about how to build audio components, but I don’t think they
had a frill understanding of what you had to do to make an audio company successful
in the US market. And they made some mistakes.
At the same time, the Asian market was developing fast. Asian consumers were
so interested in buying US-manufactured products, especially something as unique
as the 2000, that there was an entire shift away from the US market because
it was harder for Wadia to become established in the US than in Asia.
But the US market is the most important — and the largest — audio market in the world. And we’re an American company. We’re refocusing our efforts on the US market.
ABOVE: Another Wadia technician carefully adjusts a CD player’s transport
mechanism in final assembly.
RH: Are you developing new products tailored to the US market, or is this change more of a marketing direction than a product direction?
PB: It will initially be a marketing direction. More of our marketing time will be spent in the US, working with US dealers and educating them about what Wadia is and what we’ve done. It wouldn’t be fair to say the 27 processor was designed for the US market—it wasn’t. But all the products we’ve designed will be effective in both markets. For example, there is far more dollar volume in $3000 CD players sold in the US than there is in $15,000 CD players. By concentrating on the development of those kind of products, we’ll be able to sell them in Japan and they’ll be successful in the US.
Kinne: That’s not to say we’ll give up on developing the high-end units. We’ll continue to develop at both ends of the spectrum. We’ll continue to strive to make reference- quality products and a line of products that address different subsets of the market for people who can’t afford the ultimate of what we can deliver
Harley: What is Wadia’s position on HDCD?
Kinne: We are licensees of Pacific Microsonics. We’ve evaluated their filter with both HDCD discs and non-HDCD discs, playing the discs on converters we designed and on other people’s converters — some that are very highly regarded. It was our opinion that, yes, HDCD does do some things that people like. But it also introduces artifacts that we do not believe outweigh the improvements.
In our discussions with Pacific Microsonics, we came up against a situation in which we could not do anything further with them. They have said they were going to deliver a software-based implementation of the filter for evaluation. The have not done so.
PB: We compared units that were identical in every respect except for the filter. We had a standard NPC-type filter, the Pacific Microsonics filter, and the Digimaster filter. You could hear a clear improvement when you went from the standard NPC to the HDCD filter.
Kinne: Compared to an NPC filter, their filter is very good. But it’s not the equivalent of our filter.
PB: Exactly. The feeling here was that the performance o the Digimaster filter was superior That was true regardless of whether you were using an HDCD-encoded disc or a standard disc.
Kinne: An HDCD disc played back through the Wadia filter has certain sonic characteristics that a lot of people find pleasing. We also implemented HDCD-compatible designs that used the Digimaster algorithm, which we think were even better. However; they were designs that Pacific Microsonics would not allow us to make. That created the parting of the ways.
RH: You made an HDCD decoder running on the Digimaster filter?
Kinne: Right.
PB: If you want to do HDCD right, you have to do it with the Digimaster. I can see where Pacific Microsonics might not want to do that.
Kinne: We tested products with HDCD that were equivalent in price to products using our standard Digimaster algorithm and believe that our product is superior —with HDCD discs or with standard discs.
RH: I had a Wadia 2000 in my system in 1989 and in 1996 I still have a Wadia 2000 in my system. Tell me about the upgrade ability of Wadia products.
PB: All the Wadia products since the 2000 were designed with upgradeability in mind. The less expensive products are oriented more to sound quality for the dollar and are not as easy to upgrade.
Part of what you’re paying for when you buy a Wadia product is a construction that allows the product to be upgraded. The 2000 upgrade is an example of that. We’ve taken our latest thinking — the same ideas that are in the Wadia 27 — and constructed a circuit that can be installed in the original Wadia 2000. You can get the performance out of an eight-year-old unit that approaches the best we can do, regardless of cost. Protecting someone’s investment is a strong philosophy of Wadia’s.
RH: The products I’ve seen today in your factory all use surface mount technology. Even the DACs, output stages, capacitors, resistors, and some regulators are surface mounted, which is unusual in high end design. Is there a sonic advantage to surface-mounting?
Kinne: We believe there is because you are getting rid of the leads on the components. If you broke open the packages of many components, you’d find a surface-mount component inside with leads attached to it. Eliminating those leads results in fewer junctions per component. You’re putting things closer together so you have less parasitics [ capacitance or inductance] and less chance for crosstalk. An IC has the very same silicon die in it whether it’s a surface-mount or not. With surface-mount you have shorter bond-wire lengths, less lead-frame inductance, less radiated energy, and less crosstalk in the package.
If we can’t get the quality of the component we want in a surface-mount package, then we’ll use a through-hole part. We do not do total surface-mount design, but choose the best components for the particular application. Sometimes that’s surface-mount, sometimes it’s through-hole.
RH: Tell me about the design process at Wadia particularly about the relationship between developing the technical and musical aspects f the product.
Kinne: We start a project by jointly defining, as a team, the features the unit. The meetings incorporate marketing, engineering, manufacturing, and anyone else who has input into the process. We then go into the design phase, in which we do a prototype mechanical and electrical design. The prototype undergoes extensive sonic evaluation in multiple systems, with different speakers, amplifiers, and listeners. We make adjustments until we have a consensus that we have achieved the sound we want to get from the product.
During that phase we would also be doing manufacturability analysis to make sure we can produce this design. Then we would run another set of prototypes and evaluate them to make sure that they are the same as what we had sought to achieve. Once that’s verified we start beta and show it to distributors and dealers. Then we start full-scale production.
PB: We’ve achieved a good balance between engineering and listening. You really need engineering expertise, especially with a digital product. You can’t design a digital product by the seat of your pants. At the same time, there are things you can hear that you can’t always predict. I think we’ve developed a good balance between our engineering techniques and working to achieve a sound that we think is pleasing and involving and engaging.
Harley Do you try to maintain a certain sound between products?
Kinne: You’ll hear similarities within different generations of products. The 2000 and 64.4 was one generation. Then we had an interim generation like the X-32 and the Wadia Pro. The Wadia 9, the 15, and then the 16 through the 26, was another generation. You’re not going to see major sound differences within those product groups; they are more consistent among themselves com pared to different generations. Now, with the 27 and the 2000 upgrade, we’ve moved into a whole new type of sonic performance.
RH: Is there an iterative design process, with a feedback mechanism from listening evaluations back to tweaking the design?
PB: It’s even closer than a feedback loop. Jim and Steve and I are all in the listening room at the same time, mixing ideas of our sonic perceptions. It’s not a situation where we take the unit away and then tell Jim what we think about the sound. It’s a much closer communication than that, where Jim is listening, too. Steve and I both have some technical understanding. It’s not to Jim’s level, but we have some insight into what he’s doing with design.
Huntley: The difficult part is finding the right point at which to sell the product, because we’re having so much fun making it sound better!
RH: What new projects is Wadia working on?
PB: We’ve already heard experiments here that sound far beyond what we’re doing now. We know we can improve on what we’ve done, and do it in a way that’s upgradeable as well.
For example, we have a time-base corrector in the Wadia 9. It’s like the Genesis Digital Lens — a very good way of reducing jitter. The Wadia 27 doesn’t have that because we couldn’t include it and still have it come in at the price we wanted it to. That’s at least the beginning of what the Wadia 9 upgrade will be.
ABOVE: Wadia’s listening room features Avalon speakers.
RH: How significant a factor is jitter in the performance of a product?
Kinne: Jitter certainly makes a large contribution to the way things sound. As you get products of higher and higher resolution, the jitter becomes more and more important.
Part of the reason jitter might not have been recognized in the past is that products of six or eight years ago had other problems that masked what jitter was doing. It’s probable that one of the most important things in current designs is correlated jitter. You have to address it.
RH: You emphasize correlated jitter rather than random jitter?
Kinne: Each has a different sonic effect. Correlated jitter has certain psychoacoustic effects that can almost be explain able in our testing. We were listening to something that had a design error, and Steve said it sounded like a peak at 180Hz. It turned out that this particular design error introduced a jitter artifact at 180Hz. It had the effect of making it seem like there was a peak there. Yet you cannot measure a peak or dip in the frequency response of the unit. It’s still not a well-understood process, other than you know it’s a problem and you want to get rid of it.
RH: Do you have your own methods of measuring jitter?
Kinne: Currently, we’re using a Wave Technologies Digital Time Scope. It’s similar to a Hewlett-Packard frequency- domain analyzer or modulation-domain analyzer, where it actually measures the time between edges for a large number of samples that you can program with a very high resolution. It is the standard in the telecommunication industry for measuring jitter and high-speed transmission lengths that are used in telephone applications.
RH: Wadia showed a product called the Power DAC at the 1991 CES— and it was playing music. It took in a d signal and drove loudspeakers at the other end. Why hasn’t it became a product?
PB: It’s still playing music! We think it’s a great idea and real y want to make it a product. There are problems to be solved, but it shows promise. It’s not ready to be released, nor is it close enough that we can say we know when it would be released. It’s an idea we’re interested in and will continue to pursue.
RH: Wadia has incorporated digital volume controls on almost all its products. In your view, do the downsides of digital volume control— giving up resolution and the need to digitize analog sources — compensate for the benefit of removing the preamp from the signal path?
Kinne: We believe that the preamp has a certain sonic signature that does not necessarily improve the sound. Removing it from the loop more than offsets the degradation from doing the volume control digitally, which does not put anything in the signal path.
Obviously, it’s not fair to take a digital volume control that’s turned down to a minimum level, feed that into a pre amp, crank the gain up to a listening level, and compare it to a unit that does not have digital volume control. Yes, the dig ital volume control degrades the sound as you run it down, but the intention is to run the digital volume control at a level in which you only give up one or two bits of D/A-converter resolution at a normal listening level. It still far exceeds the 16 bits of the CD master. You’re still operating in the 18- to 19-bit region at low to normal listening levels. By taking out the electronics and the connections and the cables that are associated with the preamp, you will have a better-sounding system overall.
[Random jitter is sometimes called “white” jitter because of its similarity to white noise. Correlated jitter is jitter with a specific frequency or frequencies related to the musical signal being processed. Random jitter, which produces an increase in noise floor, is relatively benign compared with correlated jitter, which creates discrete artifacts. In the jitter plots, published in Stereophile or Audio, correlated jitter is seen as the at 1kHz and its harmonics when driving the converter under test with a 1kHz-sinewave test signal.]
PB: I’m going to take another crack at that. In a digital system there really is no need to have a preamplifier, because it’s easy to have a CD player that has sufficient out put drive. We originally had preamplifiers because the phono cartridge doesn’t have enough output to drive an amplifier. But now you just don’t need that gain. It’s easy to make a D/A converter that has 5V of output that will drive an amplifier/speaker combination. Why build a wonderful, well-engineered piece of equipment with a great power sup ply that’s not providing any gain between your CD player and amplifier? You start to question what its purpose is because it can’t improve the sound.
The only missing link is the volume control. The digital volume control in Wadia products allows you to have nothing in the signal path. You can run straight from the DAC into an amplifier. That’s the advantage you’ve eliminated a whole chunk of circuitry.
RH: But the tradeoff is you have to digitize analog sources like tuner and phono, along with the resolution loss.
Kinne: Obviously, the only high-resolution source available is phono. Everything else has an inherent resolution where digitizing it with an 18- to 20-bit converter is not a compromise.
If you had a phono system, you’d have to make a choice of using a preamp or not. We think that we can achieve very good results with our Wadia 17 which is a 20-bit A/D converter. With a good phono stage in front of it, the sonic results rival all but the very best preamps.
PB: That’s well said. If someone is a true analog devotee, they’re not going to digitize their phono signal. That’s not the intention. Since the majority of most people’s listening is to CD, we optimize CD performance and still allow you to have performance from the other components.
It’s true that if you compare the sound of a good phono stage digitized through the Wadia 17 and into something like the 27, you need to have an awfully good phono system to outperform it. If you spent the same amount of money on a system that’s totally focused for analog reproduction, you get great sound.
Huntley: Our products are designed specifically to drive an amplifier directly. It’s not comparable to putting a passive attenuator between a CD player and a power amplifier. You may get some level of transparency, but you also get terrible dynamic ability and a loss of drive. Designing a product with a low output impedance and a voltage swing that can drive a power amplifier directly is essential for proper digital volume control.
It’s also important to note that there isn’t any compromise in using a Wadia digital volume control in some circumstances. When you turn the volume up to maximum, there is absolutely no change in the signal.
RH: How much potential for further improvement is there with 44.1kHz, 16-bit digital audio in the playback side?
Kinne: I don’t think that we’ve come anywhere near achieving what we can achieve in CD reproduction. It’s an ongoing learning process. You never do something without learning how to do it better the next time. We certainly have ideas on how to make things better.
== ==
ALSO SEE:
Another Stereophile Wadia review
== == ==