I have said that there are 6 major audio issues that stand between listeners and a level of believability and engagement that parallels or improves upon the live experience. The Berkeley Alpha DAC Reference addresses one of these major issues, that of a-musical digital distortions, more effectively than any DAC I have heard.

Yes, It Is Expensive

The Berkeley Alpha DAC Reference Series 3 is priced at $28,000 in the U.S. That’s an expensive DAC and if you are of the world view that nothing that expensive can be “worth it” because it is outside your budget, I can understand. I will also say that if it passes the “show me better for less” test, then it could be a reasonable purchase for well-heeled audiophiles. I don’t know whether it passes that test, but I want to find out and will be reviewing other contenders as soon as I can. I wanted to start this project by looking for DACs that are more than a hair’s breadth better than others. But, beyond helping with your shopping, there are other items of interest raised by the Berkeley Alpha DAC that make it a benchmark of sorts.

Here Are Two Sentences About The Alpha DAC, If You’re In A Hurry

Some readers want to get to some kind of point as fast as possible and then get on to other things. For those viewers, I invite you read Robert Harley’s review of the Alpha DAC Reference Series 3 on The Absolute Sound website. It doesn’t address the issues I cover here, but it does explain why this DAC is held in high regard.

If that is too much time and effort, Robert summarizes with:

“The Reference DAC exquisitely reveals musical detail, but in a way that doesn’t call attention to itself. Rather, the presentation is densely textured and infused with a wealth of the finest microstructure of instrumental dynamic, timbral, and spatial cues.”

The things I have that add to and unpack Robert’s review require some context and detail.

What Problem Are We Trying To Solve?

Here is where the conversation gets interesting for listeners willing to think about how we really improve the audio experience. My assertion is that most audio progress is organized to yield small, incremental advances in parameters where products are already high performing. This, I suspect, is because engineering is organized around problems that we know how to work on…and so we do. Over a period of ten years or so, the resulting incremental progress adds up, and you can buy a much better DAC or speakers or preamp than you could a decade ago. But in this model that I’m outlining, I suggest that after 50-75 years of work, we have passed a threshold with standard distortions wherein incremental advances don’t make for a much better sounding stereo. And yet there remain six big problems that are sitting there damping the believability of the musical experience, but which don’t fit inside the standard engineering model and thus aren’t progressing very quickly. This, I think, is why many listeners can hear a $300,000 stereo and a $30,000 stereo and ask whether the $300k rig is “that much better”.

By way of automotive analogy, incremental advances on the standard dimensions get you a better car, not an airplane. We need “airplanes” to address the issues of believability and engagement.

I said that there are six major (above threshold) problems with music reproduction that stand in the way of believability. These are:

• The problem of recording standards
• The problem of visual images
• The problem of spatial imaging
• The problem of bass in real rooms
• The problem of dynamics
• The problem of a-musical digital distortions

These are tough issues because they don’t seem to lend themselves to traditional work on frequency response or distortion levels or noise at the component level. Don’t misunderstand, those are all worthwhile endeavors, and I certainly wouldn’t have predicted 15 years ago how much there was to be gained by working on noise. So, I hope that such work continues. And there is a certain joy for listeners to making small audio progress on a regular basis. But we should understand the kinds of goals we are pursuing and the kind we are not and set our expectations accordingly.

In the interest of getting closer to believable music reproduction, a different kind of project that I invite you to follow is a search for DACs that address problem #6: the above-threshold a-musical distortions of digital. Of course, we have want to understand how to get better source digital data. But confining ourselves to DACs, a problem has been around for a while (since 1982, really) encourages considering DACs that are and aren’t brand new. The Alpha DAC for example dates from 2019. But if Moore’s law and newness alone could solve the problem of digital distortions, then it wouldn’t be on the list above. As it happens, the Alpha DAC convinced me that it was possible to make progress on a-musical digital distortions and thus make more than incremental progress on believability. And so, our journey begins.

A-Musical Digital Distortions

Digital signals have many advantages. But I will assert that the continued interest in vinyl playback for high-end audio suggests there is still something wrong with much of the digital playback we experience. I’ve asked a lot of experienced listeners try to explain what they like about vinyl or what disturbs them about digital and they generally have to point vaguely at the problem. I may do no better, but as a first point I want to say that when knowledgeable music lovers and audiophiles prefer a medium (vinyl) that has massively inferior distortion, crosstalk, frequency response and noise levels, an open-minded person would see that as a data point suggesting some kind of digital error is prevalent. And meaningfully problematic.

Digital processing can be handicapped by artifacts that are unlike those we are accustomed to in the analog world. Digital processing is very complex and is done in a variety of different complex ways, so I only offer examples of the kinds of artifacts that can occur – this is not a comprehensive list nor even necessarily a list of the more problematic artifacts. My point here is that digital artifacts can occur, even though viewing them through analog-style metrics shows that the digital signals crush the performance of pure analog.

Examples of digital artifacts would include:

• aliasing difference errors, where high frequency sounds recorded at lower sampling frequencies can, with downstream nonlinearities, generate signals in the upper midrange or lower treble not in the original music
• pre-ringing, wherein partial musical sounds during playback occur before they occur in the main recording due to A/D errors
• side tones, where certain types of input signals trigger output sounds not in the original signal due to math errors in the D/A process
• clipping due to inter-sample overloads

Again, those are examples of where digital might go wrong. Some of these errors require specific inputs and thus can be hard to measure with some kinds of test signals.

More important for listeners is that a somewhat consistent aspect of known digital distortions is their a-musical character. To summarize, if digital processing adds a signal (we can call this noise if you like, but note that this isn’t typical analog noise), and if that noise is correlated to the music (related to the music in level and timing), it will tend to make the music sound odd. Or wrong. If the noise is not correlated to the music, the ear will have an easier time ignoring the result. Ideally, we’d have neither uncorrelated noise nor correlated noise, but correlated noise is a real problem.

I want to say it again: correlated noise can be difficult to hear explicitly because it is embedded in the music signal.  But if you know what real music sounds like, you know that reproduced music with correlated noise is wrong. The noise bothers you. The noise is a distraction. The noise reduces the sense of believability.

Examples

I notice this uncorrelated noise, not explicitly, but in the frequently awful high frequency sounds on digital recordings. Cymbal strikes are often problematic, where I regularly hear a harsh, crashy, piercing tone even when the cymbal isn’t struck violently. Something like:

“CHrrhhhrsscHrchrscrh”

Violin often doesn’t sound right either when rendered by digital. With violins, the tone with digital tends to be monochromatic, unlike the real deal. Some female vocals can be problematic as well, as can some piano. The telltale elements are thin, imbalanced harmonics. But this analytical language may be misleading: the big deal is that music sounds wrong and that is distracting, encouraging a focus on the audio not on the music.

As you may have guessed, I’m taking you through all this because the Berkeley Alpha DAC Reference has these distortions to a much lower degree than other DACs I have had in my reference system. This isn’t to say the Alpha DAC eliminates such distortions. With multiple artifact types and the possibility of distortions introduced in the recording, this is a multi-dimensional problem.

One example of the Berkeley’s distortion reduction comes from the Black Pumas first album. On the song “Colors” there is extensive cymbal work along with upper range choral singing. I listened to this track on an embedded AKM DAC chip (4493), a Topping D70 Pro with ESS 9038 DAC chips, the dCS Lina with Master Clock and the Alpha DAC. I tried a variety of filter and mapper settings when these were available. All the standalone DACs were fed by an Aurender N200 streamer via USB (which meant the Alpha DAC had the outboard Berkeley Alpha USB Series 2).

All the DACs aside from the Alpha DAC have some noticeable distortion of the difficult passages on the Black Pumas album (remember this is just one example from among many). This leads to a sharpness of treble tones and makes cymbals and choral sounds more homogenized than they would be in real life. The other DACs didn’t all sound the same by any means, the Lina being maybe halfway between the chip-based DACs and the Berkeley. But they all, save the Berkeley, had distortions of the type I am describing.

I want to expand on this a bit to help you understand what we’re talking about but from the perspective of how digital sounds when done “better”, meaning more believably. With the Alpha DAC, these treble sounds seem to be “unpacked” in the sense that what sounds like a single transient tone on other DACs is revealed by the Alpha DAC to be multiple tones and harmonics. The Alpha DAC presents these slightly spread out in time so that each can be heard, as is the case with real instruments. With the Alpha DAC fundamentals, harmonics, and body resonances all have their own place in the mix.

I heard this Alpha DAC strength clearly on Hilary Hahn’s 2023 hi-res recording of the Ysaye Sonatas for violin. I could clearly hear the fundamentals, overtones and violin body as a rich harmonic structure rather than a stripped-down sequence of tones. As luck would have it, just before “DAC overload week”, I had invited a violinist friend into my new listening room to hear what the real thing sounds like, and a standout quality was this time-dimensionality of the different notes she was playing.

Jon Valin, one of our reviewers, has a complimentary perspective about normal digital, which when inverted applies to the Alpha DAC: he says tone colors should be developed with a lifelike duration, yielding timbral density and maintaining dimensionality at a 3-D level. The Berkeley does this on many recordings.

Expanding on the idea, I observe that the other DACs I used for this review can seem to be brighter than the Alpha DAC. I think this is not due to a frequency response issue, it is due to the addition of distortion components that attract extra attention from the ear, especially in the treble. The Alpha DAC doesn’t generate that distortion, and it reveals the timing of the underlying components. It has as much total treble musical energy, but the energy is slightly spread out in time and doesn’t seem as over-energized. A big point here is that you won’t solve these digital distortion problems with tone shaping. You may ameliorate them, but you won’t solve them.

With the Alpha DAC, I got a lesson in treble balance vs. treble distortion. Listening to Limits of Language by Field Music, for example, I noted that I felt the mixing and mastering was on the bright side. But moments later I noted that with the Alpha DAC this treble balance wasn’t bothersome. I realized I had learned to instinctively recoil from tipped up treble balance because I associated this with amplified digital artifacts that distract and annoy. With the Alpha DAC, a different treble balance is more like changing your seating position in a concert hall, where closer positions tend to have more treble, yet you don’t live in fear of Row E vs Row J. They just sound a bit different.

These richer tone colors aren’t confined to the treble. Midrange seems richer too, although you might notice this only by comparison. And depth is improved, in some cases quite substantially.

Oddly, bass definition on the Alpha DAC also seems slightly better than I am used to. That seems weird to me, but I know that many reviewers have reported that super tweeters can have a primary effect in the bass, so I try to retain an open mind. I wouldn’t say that the bass definition benefits of the Alpha DAC are significant enough to be anything more than icing on the cake, but many people like frosting. I do too, so who am I to judge?

Technical Background

If you’ve made it this far, you may want some idea of what is causing these results. Such things generally can’t be fully explained, but here are some nuggets.

Berkeley Audio Design is a long-established company, offering its first DAC in 2008 but with a digital processing history extending back to Pacific Microsonics and the development of HDCD in the 1990s. They are rather tight-lipped about the technology in the Reference DAC. They do say that it is a multi-bit design (as are the vast majority of modern DACs). They seem very concerned about clock accuracy, but they also emphasize that many design decisions require unusual approaches to get timing and noise to the lowest possible levels. An example of this is using assembly language instead of a high-level language for key on-board software. Isolation of signals from noise sources and careful management of the thermal environment (the Alpha DAC weighs 30 lb largely to stabilize the interior thermal environment) are also examples of attention to detail. Berkeley also places great emphasis on listening tests, and in my discussions with them they repeatedly brought up examples of design goals that had been established as far back as the mid-1990’s based on revelations in listening experiments. They seem quite clear that the human ear and brain can perceive things that you might not expect and that these perceptual capabilities must be the guiding force behind design work.

Summary

The Berkeley Alpha DAC Reference Series 3 is the first DAC I’ve heard that crosses the threshold where a-musical digital distortions are significantly reduced as an important distraction. This is a big achievement in my view, as explained here. I will continue a search for other DACs that hit this milestone. This process hopefully will find some successful lower-priced offerings, which may include Berkeley’s less than half-priced Alpha DAC Series 3. And I hope, even if this necessitates higher prices, to find some DACs that make further meaningful progress.