Logo

Begin typing your search above and press return to search. Press Esc to cancel.

The Future of Audio: Integrated Systems?

This article is intended to be a bit different from most of what we publish in TAS. I want to make the argument that we could be executing a slow but powerfully important transition to a new system architecture for high-end audio. This new architecture is necessary to take significant further steps toward musical enjoyment. Those steps may be big enough that the new architecture becomes dominant. Or, the new architecture may simply create an alternative that works for some listeners and not for others. But, either way, I think this new architecture could be hugely significant. 

When I say “architecture,” I’m referring to the conceptual layout of the things we buy to reproduce music. One way to think about this is to observe that high-end audio has, since its inception in the 1950s, had one basic physical architecture: recording:source>preamp>amp>speakers/room:listener. Cables were used at each point marked by “>.” There was no feedback around the entire chain or between the elements. Sources evolved from turntables to compact disc players to computers. With that evolution, the recording medium went from vinyl to CD to hard disc and streaming. That is, while technologies changed (e.g. tubes to transistors or CDs to hard discs and streaming), the architecture really didn’t.

The physical architecture, though, is secondary to the logic behind it. The essential, almost completely dominant, logical architecture for high-end audio, from about 1955 to 2015, was the idea of modular, component-based systems. You could swap preamps or amps or speakers from any of dozens of manufacturers in a system, because the building blocks and interfaces were standardized. At the same time, modularity meant that the budget consumed for each improvement was limited to the cost of a new component, not the cost of an entire system. This approach worked well, to a point. The hoped-for benefit was that evolving technological refinement at the component level would lead the user up a stepwise path toward higher and higher musical performance. 

The standard high-end audio architecture was chosen to a large degree on practical grounds, which helps in understanding why there might be substantial performance left on the table. Especially since a crucial rule is this: structure begets performance. It takes little more than a glance at a Dachshund, a Bassett Hound, a Corgi, a Greyhound, a Saluki, and a Dalmatian to imagine that the latter are much faster than the former. The structure of long legs and relatively high power:weight are faster than the structure of short legs and relatively low power:weight. 

Because structure begets performance, audiophiles mainly have derived their performance goals from the architecture. Not directly or knowingly, perhaps, but the performance goals of audiophiles and the industry serving them naturally were set and constrained by the goals that were achievable within the constraints of the architecture. Another way of saying this is that audiophiles and the audio industry, for the most part, didn’t lay out goals first (e.g. musical realism) and then pursue the architectures that would deliver them. So, we really aren’t pursuing a process that is designed to achieve the end of musical realism. We want musical realism, but we have an architecture that is restricted to only partial success.

To understand the impact of the prevailing architecture, observe that the traditional modular-component architecture has adopted the main performance goal of what we might call “intra-component high definition.” High definition can be seen in other concepts with which audiophiles will be familiar: resolution and depth. Related to this is the idea of “low distortion.” Importantly, these ideas are generally applied at the component level (hence the “intra-component”). 

These concepts—intra-component high definition and low distortion—are good and valuable parameters of audio performance. And the industry has made real progress on these fronts. It can be hard to remember that audio amplifiers were invented well after the automobile (1912 vs. 1885). And important basic concepts like negative feedback (1934) and bipolar transistors (1950) are also comparatively recent—the internal combustion engine dates from 1879 or earlier. As a mark of continued progress on the high-definition and low-distortion fronts, almost no one would trade a good 2019 system for a good 1989 system.

At the same time, intra-stage high definition and low distortion are not the only audio parameters that matter. The standard architecture assumes that errors introduced between, say, the performer and the mastertape are minimal, consistent, and irrelevant. But are they? The standard architecture assumes that errors introduced at each stage in the reproduction chain are minimal and not compounding. But are they? The standard architecture assumes that errors introduced by equipment placement and room characteristics are minor and irrelevant. But are they? 

Those who have been around audio a long time might observe that the goals that actually are addressed by the standard architecture are, to a degree, byproducts of history, driven by the analog-signal era, where signal degradation between stages of amplification was a difficult and important issue. This issue doesn’t really disappear in the so-called digital era, but over the course of decades it can, and has, come to dominate the thought process about where the audio frontier lies. Intra-stage distortion was a problem, we internalized that it was a problem, and as we reached the asymptotic part of the progress curve we kept going back to low intra-stage distortion as the goal because we didn’t know what else to do—or we did, but the architecture worked against doing something. This is now a problem. The problem, I would say.

 

So, the narrow focus of audio goals is not a nefarious plot; it is the natural byproduct of sensible decisions taken 50 or more years ago. Many people—audiophiles and audio companies—are aware that “something” is amiss. Older audiophiles commonly remark that “something” is unfortunately different now than it was back in the 1970s and 1980s when progress seemed to be a quarterly event. TAS founder Harry Pearson periodically observed that, despite decades of progress, high-end audio systems still sound vastly different from live music (“the absolute sound”). He wished to point out that “something” was needed to go farther toward the absolute sound. The problem is that the “something” isn’t explicitly clear as long as you assume the standard architecture. 

I wish to argue that the reason macro-progress toward the absolute sound is limited is because of the constrained micro-performance focus that we have inherited from the modular component systems architecture and its focus on intra-component distortion. The assumption in focusing on intra-component distortion is that if the recording is decoded, amplified, and delivered to the listener faithfully at each step in the reproduction chain, then all is well. This is a strategy to deliver low distortion in some portions of the chain, but not necessarily a good strategy to achieve low overall distortion from performer all the way to the brain. 

To propose an explicit alternative that I think actually would help, the strategy I want to point to is focused on low total distortion based on feedback loops encompassing the entire chain from mind to recording venue. I think we can call this latter approach the “integrated systems architecture.” 

So, my hypothesis is twofold. First, I want to suggest that high definition and low distortion are parameters on which we have reached an asymptotic level of performance advance. Which is to say, things are pretty good as far as these parameters are concerned and they naturally aren’t getting and can’t get better very fast (also, such performance advances tend to be expensive). Second, I suggest that integrated systems, taken as a concept, provide a significantly richer wealth of opportunities to develop audio products and services that advance the state of the art much farther toward musical realism. In part, that is because, in the integrated systems era, audiophiles and the audio industry would first lay out fundamental goals and then pursue the architectures that would deliver them. Not the obverse.

These two related hypotheses don’t present a forced choice. We can advance the cause of low intra-stage distortion and at the same time pursue the advances allowed by integrated systems. But we have to recognize and pursue the latter as avidly and intensely as we can, and—resources being limited—that probably means less progress on intra-stage distortion than we would get with the industry structured as is. There will be resistance to such a direction because audiophile mindsets are focused on the traditional architecture and because the industry is not arranged to deliver integrated systems. This sort of problem happens in industries all the time—witness the current transition of automotive technology from internal combustion to electric powertrains with its attendant haters and reshuffling of power (or previously from voice communications from land lines to smartphones or…). 

This is all pretty high-level, abstract thinking. To bring it closer to our understanding, we need to define the alternative logic of integrated systems. 

The integrated systems architecture is, first and foremost, a logical construct that is aimed at six goals supporting the overall goal of musical realism. Four of these goals are primary:

  • Environmental integration. The logic of integrated systems is that the largest errors imposed between the source signal and the listener are due to differences between the assumed environment and the actual environment (e.g. room shape and materials; loudspeaker and listener placement). An integrated system, then, is one that works to achieve results within its environment. It is integrated with the environment. 
  • Listener integration. We can easily imagine that different listeners process music differently along the ear-brain signal-processing system. Said more simply, different listeners care about different things. As a result, even if “the absolute sound” is one fixed thing, in a real world of systems that have distortions, different listeners practically may need to receive different musical wave fronts to experience musical accuracy in their heads. In addition, a hi-fi system is a virtual reality machine, and it is easily imaginable that technical “inaccuracies” might be introduced that lead to greater virtual reality. These are, possibly, heretical remarks in the logical paradigm of component systems, but integrated systems could, and probably would, abandon these assumptions to make it easier for listeners to tune systems to work well.
  • Sub-system integration. The component model revolves around standard building blocks with standard interfaces. Integrated systems have the freedom to abandon both the building blocks and the standards. One likely benefit of this is that sub-systems can be optimized for each other. One small example of this is the use of active crossovers. These have significant potential benefits in reducing inter-driver interference, but generally such crossovers can’t be used because the component building block model is based on a power amp to passive loudspeaker architecture.
  • Recording integration. We have tended to view the recording as some sort of absolute. But realistically, recordings are made with different venues and different equipment and different monitoring and mixing preferences. The likelihood is that significant errors are introduced before the recording data even gets to a home audio system. True integration would take these errors into account.

The secondary integrated systems goals are:

  • Packaging integration. Audio systems are generally used in homes, and the way systems physically fit in the home is potentially important to domestic acceptability and to placement. Similarly, the user interface of the audio system may be integral to its enjoyment and usage. Finally, packaging for easy setup may affect the tuning of the system and the frequency with which technological upgrades are implemented. One-box systems, like the Naim Mu-so or the Dynaudio Music series, may typify packaging integration when compared to a standard component system, but the concept is more general than that, as we can see in the Kii Three or the Gayle Sanders Eikon. 
  • Cost integration. We can imagine that cost reductions are achievable with integration. The standard idea here is exemplified by the integrated amp, which uses one chassis rather than two (preamp and amp) and one power supply and one shipping container rather than two. But we can also imagine that a designer, with a wider range of system elements to optimize, might be able to produce a target level of performance at lower cost than designers bound to optimize their individual components. A core idea here may turn out to be that if we can push development from hardware into software, per-unit costs may be reduced. 
 

I have labeled these latter two goals secondary to make it clear that some of the obvious example candidates of musical systems, like the Naim Mu-so and other good one-box systems, are mainly examples of pursuing these secondary goals. The primary idea behind the integrated systems movement, were it to advance, would be the first goals described above. That is, the integrated systems philosophy is primarily a performance-enhancing philosophy, not a convenience or cost-driven philosophy. In my conversations with people about integrated systems, this is rarely understood, because the examples we have come so often from the edges of the audiophile realm. 

Before this gets completely out of hand, let me summarize the story up to this point. The basic observation behind integrated systems is that the recording venue and the recording system and the room in which a playback system exists, and the array of listeners within that room, and the signal processing of those listeners have a huge impact on the perceived accuracy of sound reproduction. This seems rather jejune, until you realize that the modular-component architecture does essentially nothing to address the issue of recording anomalies, room-by-room performance, listener-by-listener signal processing, and cross-component optimization. An end-end system can exhibit ±15dB variations in output at different frequencies in different rooms and the traditional system architecture provides almost no structural way to address these issues. And we haven’t even arrived at the listener’s brain nor have we addressed the problems imposed by system interactions. You can swap as many power amps or power cords as you like, and you aren’t going to fix these problems. The magnitude of the problems is too big, and the problems are too complex. You need a different system architecture based on end-end results.

KEF LS50W and MartinLogan Dynamo 800X

If you’ve read my rant/treatise on integrated systems, I hope you found it interesting, but it is mainly an assertion that there are severe limitations to how well a modular-component system can perform. While I suppose it might be possible to lay out a full theory of integrated systems to show why they will do things that modular-component systems can’t, I doubt this would work to persuade consumers. I, and most of our readers, don’t have the technical knowledge to know whether such arguments are correct. Thankfully, since integrated systems are emerging, it is possible to take a more empirical approach to seeing if an integrated systems architecture really helps. One sample isn’t proof, but my experience with two KEF and MartinLogan products suggests to me that there is a strong possibility that integrated systems are a useful path to progress. 

My plan for this test was simple: Try an integrated system of sorts and see what observations arise. If it doesn’t work well, we don’t know anything, but if it does, we might learn something. 

Beyond that, for practical reasons I had to choose something to try. I chose the KEF LS50W for this test for a pretty simple reason: KEF also makes a traditional passive version of the speaker that is the basis of the LS50W, so it was, to a degree, possible to test the two architectures I’m discussing against each other. 

The LS50W is philosophically aligned with this test to a degree, but only to a degree. First off, it has built-in power amplifiers and an active crossover. The power amps are interesting in that the low-frequency section is a Class D amp, while the high-frequency section is a Class AB amp. The LF amp is much more powerful (200W vs. 50W). These seem in part to be packaging-related and cost-related choices, but KEF makes it clear that they are also performance-driven. I think it is interesting that these engineering optimizations are ones that consumers probably wouldn’t make or perhaps wouldn’t even know how to make. The active crossover is another decision that consumers could in theory make, but probably wouldn’t. First, it is likely that many consumers are unaware of the effects of crossover slopes on inter-driver interference. If they are, they might not feel confident choosing slopes and crossover points. And even then, they would possibly recoil at the budget needed to execute a multi-amp, active crossover system—where those electronics might cost more than the entire system under review here. 

This brings us to my first observations about this system. The total cost of the integrated system under test is $3000, assuming you have a computer or CD drive as a source. Now $3000 isn’t chump change, but in standard audiophile terms it isn’t that expensive. If we started with the passive LS50 ($1500/pair) and the Dynamo 800X ($800) subwoofer, we could add an integrated amp/DAC from the likes of NAD, Yamaha, Music Hall, or many others for about $1000. We’d get 80 to 150 watts per channel. With some cables, we’d be right around the $3000 that this integrated system costs. 

Two related points come from this. First, we’re talking about an integrated system that is comparable in budget to a conventional system with $1k mini-monitors and a mid-power integrated amp/DAC. Expectations should be calibrated accordingly. Second, it is unlikely at this budget level that a user could assemble an electronic crossover and multiple amp setup. We don’t know yet if that architecture is valuable, but by choosing an integrated system, we can see Goal #6: Cost Integration at work. 

I added the MartinLogan Dynamo 800X subwoofer midway through my listening because I found that the KEF isn’t really a full-range speaker and because the Dynamo series employs Anthem Room Correction—an application of the integrated systems idea that fits the brief of this test. 

To get down to performance, the integrated LS50W and Dynamo 800X sounded obviously better than the passive, non-integrated version of the system (which I drove with a Devialet D-Premier to reduce arguments about whether the particular mid-priced integrated amp used was the issue). The midrange of the integrated system is smoother, which I found to be important to the sense of realism on offer. A big effect here comes with being able to hear each ensemble as a naturally balanced whole. And instruments sounded more transparently rendered, with a greater sense of space.

Maybe just as big, KEF allows fine-tuning of the treble level of the LS50W with an app. That tuning allowed me to dial in the HF balance of the system in a way that I couldn’t do with the passive LS50. And in a way that I found very important, even though my adjustment was 1dB in magnitude. This tweak transformed the treble balance of the system from distracting to quite natural. I would go for a little more adjustability in a perfect world, but the KEF app gave me 75% of what I needed.

 

Bass with the integrated system was more balanced too, and, dare I say, integrated. My experience with subwoofers is that their addition to a system is often fraught with problems which cause bass to sound unnatural. This is partly because low frequencies are so affected by room dimensions and subwoofer placement that you simply face a choice of various lumpy LF regions. MartinLogan addresses this with DSP equalization (Anthem Room Correction), and my measurements show that ARC and the personal tuning adjustments allowed (e.g. changes in overall level and output below 30Hz) make a huge difference. Subs are also problematic because it is hard to set the phase correctly. Here I cheated a bit and used my measurement system to adjust subwoofer phase. Using my ear I was close, but things still weren’t right. The measurement system, and the fact that MartinLogan allows continuous phase adjustment (360 degrees), allowed me to change phase by about 20 degrees from my original setup and that adjustment yielded important results. Ideally, if adhering to the integrated systems manifesto, KEF or MartinLogan would have provided a way to do this automatically. In the end, bass on this system was deep, controlled, and less distracting than on many systems. I had a greater sense than is usual that bass was just part of the music, not something generated by an audio system.

Analytically, the LS50W and Dynamo 800X are very good, and I would say substantially better than the system you might build around passive LS50s. But I wouldn’t want you to assume that they sit at the edge-of-the-art in performance thanks to the transformation wrought by partial integration of various pieces of the system. A few examples may help. The first is in the domain of dynamics. I had these systems in a smallish room and even with this attempt to not ask too much in terms of output, there are many more dynamic systems on the market. Output isn’t really the issue; this particular system is just consistently a bit soft. Another limitation is in the lower treble where I thought the LS50W was slightly dry sounding (a characteristic shared by the LS50 in my experience). 

But, from a big picture perspective, the LS50W/Dynamo 800X system did something that I am coming to view as a valuable hallmark of good audio systems: it sounded quite good on almost everything I listened to. To give a sample of my playlist: Mahler 1, 5 and 7, Los Lobos, Brandi Carlile, Rachmaninov Piano Concerto No. 2, Little Feat, Kate Rusby, Wagner Flying Dutchman, Pretenders, Fairport Convention, Messiaen Organ Works, Radiohead, Vaughan Williams Sea Symphony, Art Blakey. In my lengthy experience of high-end audio, this ability to engage the listener with just about all the music you might want to play suggests a fundamental rightness to the important things the system is doing. The more you just want to listen to more music, and the more you are well-rewarded for that, and the fewer distractions on offer, the better. 

I think we often can be enticed down the path of seeking an amazing sonic experience, only to find that this outstanding result comes at the price of many recordings sounding just okay or worse. Reviewers are tempted to write this off as “the system was very revealing of flaws in the recordings I played.” Yes, perhaps, but perhaps the system has distortions that work well on occasion but not very often. You have to decide if that is what you want, but I find I am more and more interested in enjoying the music with the limited time I can devote to it. I found the LS50W/Dynamo 800X particularly good at working in that context. 

I should also mention the ease with which these results were achieved. I spent less time setting up this system than I do on average by about 75%. I did my placement calculations, set the speakers there, put the sub where MartinLogan said to put it, wired things up (five wires!), adjusted the apps and that’s about it. I did measure for sub phase and level. And I did move my listening chair by 3″ later on and made one adjustment each to sub and treble level—all based on listening. For the right listener, there is a freedom derived from the limited ability to tweak via component substitution. Or, one might say, there is a psychological refocusing on the music that comes from such a simple system. 

The question these observations about performance and usage raise is whether this consistent high level of listener engagement is a byproduct of the integrated systems element of the system. We won’t know, of course, until we have many more examples. But, in reflecting on what was intriguing about this system, I came to the conclusion that the musical engagement I experienced was a direct byproduct of the integrated approach. My listening notes say “90% of state of the art on everything—100% on nothing.” I think this is, at this price point, an indication that integration has lowered overall system distortion significantly. If the development of integrated systems can thereby lead to substantially high listener engagement, I think that is a big deal for the future of audio. 

Read Next From Blog

See all
BLOG

Q&A with Andy Kerr of Bowers & Wilkins

What ignited your interest in high-end audio? Did it come […]

BLOG

When Prince Was King

When Sign o’ the Times came out in 1987, Prince […]

BLOG

How to Build Your Own Low-Cost, High-Performance Streamer

There is no product category with a wider range of […]

BLOG

Q&A with Martijn Mensink of Dutch & Dutch

Did your interest in the high end come from the […]

Sign Up To Our Newsletter