We define a paradigm shift as a radically new idea that causes an exponential change from an old idea; at least 10X improvement.

HDR-A objective improvements are over 100X.

HDR-A conversion sounds more realistic for a number of measurable, objective reasons. Let’s walk through them.

1 Low Level Waveform Purity

Because multi-path topology has nearly two orders of magnitude (100X) lower noise than high-performance single-path topology, perceptual low-level signals remain pure and relatively distortion-free, while single-path signals become visually noisy, jaggy, and ragged. This is a paradigm-shifting improvement in DAC performance. Improvements in deep-image program are immediately heard.

2 Reduced THD

Compare a THD+N plot of mid-level and low-level signals. Multi-path THD+N will be significantly lower than any single-path, often more than 2 orders of magnitude (100X) lower. This is because mid-level signals are processed in a higher (more accurate, less noisy) region of the D-A converter core, which isn’t possible in single-path topology. The difference is audible even on average audio systems. Lower THD+N translates into space and timbre that is more realistic than single-path conversion. This is not only a paradigm-shifting improvement in DAC performance, it frankly renders all other DACs technically obsolete.

To our knowledge, no other DAC maker publishes their lower-level THD+N numbers or FFT plots, and yet every acoustic recording engineer knows that the middle- and lower-levels are where we most perceive and “feel” deep-image atmospherics and tactile space. This is why it’s important to slash lower-level DAC distortion, and the only way to achieve that is via imersiv triple-patented multi-path processing. This is a paradigm-shifting improvement in DAC performance.

The comparison is analogous to HDR photography. In a single-path photo, the shady areas (quiet musical passages) lose all detail. But in the HDR multi-path photo, the low-level areas become vibrant and sharp. This is a perfect analogy to multi-path audio. The lower-level audio becomes clear and precise, evoking comments like “the depth and low-level detail is breathtaking” (Paul Blakemore, multi-Grammy winning chief mastering engineer, Concord Music).

Read more about imersiv quiet-side performance in the F.A.Q. section: “what is low-level signal purity.” 

3 Headroom

Triple-patented multi-path HDR-A architecture allows us to increase headroom levels without raising the DAC’s residual noise, something that’s impossible to do with a single-path design. In a single-path DAC, when you raise the headroom, you raise the noise by the same amount. And visa-versa … when you reduce the noise, you reduce the headroom. It’s an inescapable tug-of-war with single-path design.

The inherent “unlinked” nature of multi-path topology means we’ve had to create some entirely new testing algorithms. We’ve named one of these test parameters DSNR: dynamic signal to noise ratio. Because each path will have its own noise signature, we test SNR at the moment when the low-path and high-path are dynamically cross-faded, hence Dynamic SNR, or DSNR. For a technical deep-dive into this and other new multi-path test parameters, see the Audio Engineering Society Paper No. 21106.

4 Noise

Perhaps multi-path’s greatest breakthrough is dramatically lower self-noise, approaching 40dB (100X) lower noise than today’s best single-path DACs. We believe this is the largest single objective specification improvement in the history of audio delivery  … a paradigm shift.

5 Dynamic Range

Combine dramatically lower noise with unlimited headroom, and theoretical linearity becomes bounded only by the 32nd bit. But there’s a practical limit to how much dynamic headroom is “required” in professional audio applications. The imersiv D-1 offers +23dBu (10V) of ISO-free headroom. We believe this is more than sufficient for virtually all pro applications. Those with unique pro requirements should contact the factory for higher customized headroom levels. We’re considering a true 32-bit vacuum tube DAC (40nV to 160V = 192dB dynamic range), mostly for bragging rights.

6 Linearity

Today’s very best single-path DACs can deliver around 130dB of true linearity. Beyond 130dB, these “super-DACs” quickly go non-linear (i.e., bad). Some of these DACs sell for $100,000, or far more. We know of one DAC priced at $300,000 which cannot achieve 130dB linearity.

Compare this with HDR-A which maintains over 170dB of true linearity. This is another paradigm-shifting improvement in D-to-A conversion performance.

7 Self-Calibration

A multi-path DAC has two internal signal paths, what we call the high-path and low-path. Machine-learning-based calibration algorithms (AI) maintain optimal audio performance of the patented multi-path process, including vanishingly low DC-offset. Most DACs exhibit vastly worse offset, often 1000X or more, which can lead to errors and non-linearities in a downstream device (power amp, etc).

8 Design

Beyond its groundbreaking HDR-A technology, the D-1 DAC represents a lifetime of engineering mastery:

Galvanically-isolated external power supply. Keeps USB and Dante signal paths free from ground-induced errors — or what’s been called “ground jitter” or “ground-induced noise”.

Low-jitter clocking, in the femto-second region. Far below audibility.

Twenty-three individual stages of ultra-low-noise power supply regulation.

Free from ISO’s (Inter-Sample Overs).

Dante networking option. Dante is the professional industry standard audio network protocol, with a 32-bit jitter-free signal path and up to 64 audio channels on one standard RJ45 Ethernet connector. We love Dante.

DSP-based command and control. A wide range of DSP-controlled functions are available, including master level (functions as a preamp), level control velocity, MS widening and centering, path and DC calibration, six digital filters, latency selection, harmonic enhancement, and much more.

Aerospace-grade PCB fabrication and componentry. Millennia has four decades of experience in military-spec QC, QA, and design protocols – hand-crafted and supported in Northern California. Contact us. We talk audio all day, every day.

In multi-path architecture, we say the magic is in the low-path — lower-level musical information. Recording engineers know that mid-level and low-level program is where we are most sensitive to spatial and atmospheric detail. Think about being in a concert hall when a sudden loud section (sforzando) is followed by silence (caesura). This very brief period of decay is when you fully sense the size and envelopment of the hall, and your precise place in acoustic space. You sense this by the myriad subtle cues of decaying reverb tails swirling around you, until the sound fades into the black background of the hall itself. That’s atmospherics.

It’s in these softer musical passages where multi-path topology delivers profoundly improved clarity via dramatically reduced THD and noise, translating into timbre precision and atmospheric realism never experienced in legacy DAC architecture. As one multi-Grammy-awarded engineer put it, “like going from 35mm to IMAX.” Single-path topology (all DACs today) simply can’t reproduce low-level information with the exceptional purity and detail of the HDR-A multi-path process. Not even close. You can read more about this in the FAQ section: “what is low-level signal purity.”

So, do we hear a change in “reduced noise”? Yes! Turn the D-1 level control to full — you will hear zero noise. But it’s less about the “hiss” and far more about the improvement in spatial and timbre realism, resulting from reduced mid-level THD. This is where you will hear an immediate and profound difference when compared to any single-path D-to-A converter. It’s precisely the difference between a standard photograph and an HDR photograph.

Compare Distortion+Noise performance of mid-level and low-level signals. imersiv HDR-A distortion will be significantly lower than single-path, often more than 2 orders of magnitude (100X) lower. This is because imersiv mid-level signals are processed in a higher (more accurate) region of the D-A converter core, which isn’t possible in legacy single-path topology. With deep-field program, the difference is plainly audible even on average audio systems. Significantly lower THD+N translates into space and timbre that is far more realistic than single-path conversion. Frankly, HDR-A renders all other DACs technically obsolete.

Something every acoustic recording engineer knows: it’s in the mid-levels and low-levels where we most perceive and “feel” the sense of atmospherics and tactile reality. This is why it’s important to slash lower-volume DAC distortion, and the only way to achieve this is via multi-path architecture.

Imersiv is a new brand of our 35-year-old parent company, Millennia Music & Media Systems. We named it imersiv because multi-path stereo architecture delivers dramatically improved three-dimensional realism compared with legacy single-path audio architecture. Imersiv high dynamic range audio allows you to hear far deeper into the orchestral stage, with a clearer experience of subtle instrument timbre and precise location in 3D space. Imersiv does this by dramatically reducing THD+N at lower listening levels, opening up an immersive sound field never before experienced in stereo delivery.

The imersiv branding will be applied to all future multi-path products, including headphone amplifiers, streaming centers, power amplifiers, A-to-D converters, and more.

Yes, every audio capture and delivery device can be realized in multi-path HDR-A architecture. Let’s explore each element.

1 Microphone

There are decades of patents for multi-path microphones. Some mobile phones use MEMS microphones in a multi-path configuration. The concept is simple: use an exceptionally low-noise capsule and head-amp for the low-path. For the high-path, use a capsule that can withstand brutal SPLs. Terminate both paths into a multi-path ADC which maintains the total combined SPL of the two transducers, resulting in 170dB or greater dynamic range.

2 ADC / Preamp

To our knowledge, the first commercial multi-path ADC was released by the German company Salzbrenner in the early 1990s, achieving a dynamic range of around 153dB. Many companies today are doing a variation on this theme with what is being called “floating point ADC” – which is effectively a clip-prevention scheme.

May I air a gripe? I have a problem with the name “floating point ADC.” Floating point is a mathematical framework, not an audio topology. I suggest our pro-audio industry find another term for a multi-path ADC. Call it floating, adaptive, ranging, heuristic, or multi-path …. but please do not call it “floating point.”

With the realization of the D-1 multi-path DAC, I believe we’ll see a renewed interest in multi-path ADC’s as well. At imersiv we’re working on a high-performance HDR-A ADC, and invite other professional manufacturers to do the same. It’s time to move the audio industry from single-path to multi-path and reap the benefits of profoundly reduced systemic noise, leading to greatly improved image and timbre accuracy.

3 DAW

The good news is that DAWs (digital audio workstations) do not need or use “multi-path” architecture. Once a multi-path-processed signal leaves an ADC, its complete dynamic range is carried in a single-path digital transfer. Same story on the delivery side – when signal leaves the DAW, its entire dynamic range is carried in a single-path format.

The bad news is that most DAWs are still working in 24-pefect-bits. They may claim “32-bits” but 32-bit-float only guarantees 24-bit-perfect processing. As of this writing, there are only two DAW makers we’re aware of that can support 32-bit-perfect capture, storage, processing, and delivery: Cockos and Steinberg. There may be some other DAWs that offer partial 32-bit-perfect operations, but not full-stack.

4 DAC

The imersiv D-1 DAC is the first known multi-path D-to-A converter – with 100X performance improvement over any other legacy DAC made today. We are working on a number of other HDR-A DAC architectural variations. Sign up for the imersiv newsletter and be first to hear of new developments.

5 Power Amplifier

The imersiv patents outline a roadmap for 100X improvement in power amplifier noise, linearity, and dynamic range. A 6-watt headphone amplifier is planned as the first multi-path power product, followed by higher power mains amplifiers. All multi-path power products will exhibit 170dB of HDR-A dynamic range and linearity, with revolutionarily low quiet-side THD performance.

6 Signal Path

This all adds up to a signal path with each individual element performing at 170dB dynamic range and linearity, a result of advanced HDR-A topology. When we sum the noise of each element, we realize a complete audio signal path, from mic to pwr amp, of around 164dB combined dynamic range, or around 27-bit systemic performance. Compare that to today’s very best systemic single-path performance of around 20-bits, or 120dB (actually far less, considering low-level distortions). Multi-path delivers a 100X paradigm-shifting improvement to the entire audio experience, front to back.

Great question. We asked this same question years ago. The short answer is that many recordings have information below the distortion threshold of today’s best single-path DACs. A multi-path DAC is able to reach below single-path DAC limitations to reveal information previously hidden (i.e., below the noise floor and elevated THD of low-level single-path processing). This is both objectively measurable and empirically perceived. Start by reading the myriad comments of platinum producers and Grammy-winning engineers and composers who used the D-1 in their studios for nearly a year of beta testing.

As HDR-A performance is more widely achieved in other audio functions (microphones, ADCs, power amps, etc.), objective and empirical performance will continue to leapfrog today’s single-path designs. Eventually, all critical audio will be captured and delivered via HDR-A multi-path architecture. It’s the only way forward.

It’s a long conversation. Let’s review the most important points.

1 File Formats

Most file formats today can represent 32-bits. These include FLAC, AIFF, WAV, BWF, m4a, WV, and many others. Older 24-bit file formats are fading away by natural attrition.

2 Hardware Interfaces

All of the newer audio hardware pipelines support 32-bits. These include USB, Dante, Thunderbolt, AVB, and others. Legacy 24-bit formats will still be with us for some time (AES, SPDIF, MADI, AES67, etc.). Most audio products now offer both 24-bit and 32-bit operation, but the trend is moving quickly to 32-bit hardware and networking.

3 Workstations (DAWs)

This is a bottleneck. Most DAWs are still working in 24-bits. They may claim “32-bits” but 32-bit float math only guarantees 24-bit-perfect processing. As of this writing, there are only two DAW makers we’re aware of that can support 32-bit-perfect capture, storage, processing, and delivery: Cockos and Steinberg. There may be some other DAWs that offer partial 32-bit-perfect operations, but not full-stack.

We’ve developed an in-house 32-bit ASIO test app which can toggle 32 discrete DC levels (bits) into a DAW, and then compare those bits at the DAW output. If the comparison is not perfect, the DAW is not 32-bit-accurate.

Probably our most frequently asked question.

Here’s the essential point: imersiv’s 40dB dynamic range improvement is achieved mostly as lower noise and distortion. That extra 40dB is not “40dB louder”. It’s nearly 40dB quieter. This is where the multi-path magic happens — in the quiet parts of music, where our neural processing is most sensitive to atmospherics and imaging. It’s our long-term vision to bring imersiv 170dB noise-free HDR-A performance to every element of the audio signal path, including microphones, preamps, ADC, DAW, DAC, and power amps.

But there’s more to the story. Let’s explore the “real world” of dynamic range. Buckle up.

The human ear can detect sound down to -8dB SPL at mid-range frequencies. That’s the true low-range of human perception. At the high-range, recording engineers commonly work with +155dB SPL signals. Snare drums and trumpets peak at roughly +155dB SPL. We stick microphones on these things every day. Hence, for acoustic recording engineers, the real world of dynamic range is roughly (|-8|) + (155) = 163dB.

Will everyone need this level of dynamic range? No, of course not. I’m simply pointing out that 160-165dB is the real world of professional audio dynamic range. imersiv audio gear is designed for the real world, and will never be a weak-link in the engineer’s tool kit, even in the most extreme quiet or extreme percussive real-world recordings.

But I hear you protesting, “120dB SPL is the loudest safe level for our ears.”

At higher frequencies, yes. But at very low “hip-hop” beat frequencies, the safe listening level rises significantly above 120dB SPL. I don’t know if any research has been done on the onset of ear damage at 30Hz, but we do know that boom cars commonly achieve 155dB SPL at very low “beat” frequencies. A large home entertainment sub-woofer now achieves 148dB peak SPL at very low “movie explosion” frequencies.

Let’s use home entertainment as an example. Say we’re mixing a movie. We want a sub-woofer explosion with brief VLF peaks of 148dB SPL, and we want our noise floor to be -8dB SPL equivalent. This requires a soundtrack dynamic range of 156dB. A 24-bit file cannot achieve 156dB dynamic range. A 32-bit file is required for 156dB dynamic range – something only HDR-A architecture can deliver.

 Sure. After Edison commercialized acoustic capture and delivery in 1887, his “wax and tin foil” dynamic range remained relatively stable at around 15-20dB (if that), within a very narrow frequency range. It wasn’t until 1925 — the advent of electric recording — that dynamic range and linearity improved by at least 10X (20dB). I think we can say electricity or electrical recording was the first paradigm shift of audio technology. Condenser mics, mic preamps, power amplifiers, electric record cutter heads, vacuum tubes, etc.

After the initial wave of electrical audio invention (1925-1935), dynamic range and linearity growth slowed. Then, around 1950, magnetic tape began its commercial ascent. Magnetic tape was the second paradigm shift of dynamic range and linearity, improving on hard-surface recording by another 20-25dB (>10X).

The 1960s and 70s saw some modest step improvements with Dolby, etc., but it wasn’t until the 1980’s that another paradigm shift occurred — digital recording. A-to-D and D-to-A. Digital recording bumped dynamic range again by 20-25dB. Since then, dynamic range and linearity have improved slowly — about 0.8dB per year on long moving average.

Multi-path HDR-A is the next chapter in the historic dynamic paradigm. It represents the largest ever dynamic performance improvement, eclipsing today’s best DAC dynamic range and linearity by 40dB (100X). Currently, this improvement is limited to D-to-A conversion. Over time, our goal is to apply HDR-A architecture to every link in the audio path, from microphone to power amp. We envision every element in the recording and delivery path to exhibit greater than 170dB of dynamic range and linearity, which will result in a systemic (end-to-end) dynamic range of over 160dB.

As we say: Once in a generation, a new audio architecture changes everything.

The only way to confidently determine low-path self-noise is by using the Johnson-Nyquist equation. It’s both simple and accurate. For instance, the broadband, unweighted self-noise of the D-1 DAC processing path (DAC IC => I-V conversion => Filter => Summer => Buffer) is measured at -116dBu. That’s measured, not calculated. We then run that -116dBu of quiescent noise into a -34dB passive attenuator. This calculates to -150dBu of broadband self-noise. However, our attenuator itself also exhibits resistive thermal noise. We use the J-N equation to calculate the “bulk” resistance thermal noise and add that to -150dBu, giving us our resultant low-path self-noise of ‑146dBu. There are no other noise generation sources in the circuit. Those familiar with the J-N equation can readily calculate the bulk attenuator resistance at 20C and 20-20,000Hz MBW.

What we’ve achieved happens only once in a generation, if that. The D-1 DAC represents a 100X (double-exponential) improvement to key audio parameters: dynamic range, linearity, noise-floor, etc.. Never in the history of audio has one invention improved dynamic range and linearity performance metrics to this depth. For instance, the 1980 paradigm shift of analog tape to digital recording was a roughly 20X improvement (25dB).

Somewhere around 2018, we sought independent testing of our early prototype, to confirm the outrageous performance improvements we were seeing. We hired the Vice President of Advanced Technology at Dolby as an independent testing consultant. His testing confirmed our results. His insight and suggestions were so good that we later asked him to remain as an advisor to the company, and he agreed. This was the beginning of the imersiv Advisory Board.

More recently, we recognized that having a diverse board of independent, impartial advisors would help us navigate our revolutionary new technology to market. For instance, multi-path architecture can (and should) be miniaturized into an integrated circuit, so we asked an IC expert to join the board — the President Emeritus of Ford Aerospace Microelectronics, with other executive roles at Intel and Texas Instruments.

We also wanted someone with deep experience in taking an audio company through a period of rapid growth, so we asked the former CEO of Blue Microphones to join the board (sold to Logitech for $100M), currently CEO of Bose Professional.

As a pro-audio company (Millennia), we have limited experience in the home audiophile market. We asked a highly respected luxury consumer audio veteran to join the board.

And we wanted a panoramic, big-picture perspective from a polymath with deep experience in pro-audio product engineering, musical performance, and business development + management.

The imersiv Advisory Board represents a hand-picked council of seasoned visionaries that will help us maximize our impact on this new journey.

The D-1 headphone output is an exceptionally low-noise, high-speed amplifier with 136dB open-loop gain. It’s designed to drive virtually any dynamic ear speaker. The headphone amplifier is not an imersiv circuit. While the headphone amplifier is exemplary in every way, it does not pretend to approach the paradigm-shifting performance of the D-1 XLR line outputs.

D-1 was designed to offer two varieties of headphone outputs, balanced and unbalanced. The standard HPA is an unbalanced ¼” stereo phone jack. Upon special order, the D-1 can be fitted with a 4.4mm TRRRS Pentaconn diff-balanced stereo headphone amplifier. This balanced HP amplifier uses the brilliant new OPA891 driver and sounds amazing. This optional TRRRS balanced output is actually the ideal HPA for the D-1, as the HPA driving source is inherently diff-balanced.

As of this writing, we are working on a dedicated 170dB imersiv headphone amplifier console. We have published an elegant patent on multi-path power amplification. This patent is the foundation of our upcoming multi-path headphone amps and mains power amps.

Glad you asked. This 10-gauge aluminum plate runs the entire inside length of the D-1 chassis (14in/35cm). It serves two purposes. Its main purpose provides RF shielding between the digital and analog PCBs. Its secondary purpose is a reminder to you (the end-user) that the analog output section sits at self-noise levels never before achieved in audio electronics. Not even close. This plate is your “note to self” to use the shortest XLR cables with the most effective shielding and ground techniques.

The Pro Menu (selectable by pressing the Setup and Select front-panel buttons) was developed for functions not normally used in home entertainment or audiophile listening. This “secret menu” will probably be interesting only to audio professionals.

As of this first FAQ revision, there are three functions in the Pro Menu:  (1) L/R Balance, (2) Harmonic Distortion, and (3) Mid Balance or Mid-Side (M/S) adjust. L/R Balance is handy for mastering labs that need to assure +/-0.01dB DAC channel matching. While home audio generally does not require such resolution, the home user is welcome to experiment with the Pro Menu functions. M/S and Distortion applications are addressed in separate FAQ entries.

Note that Pro Menu functions are not notified on the Main Screen. Use them at your own risk. Pro Menu settings will be lost when reverting to Default Settings.

During beta testing, one mastering engineer (who does mostly rock and metal) commented that he wanted a slightly more forward “center image”. That is, he wanted the “mono” or “middle” level to come up but the “stereo” or “side” information to remain unchanged. This is a job for M/S, or mid/side processing. Professional mastering engineers sometimes use M/S to enhance music, such as bringing up a vocalist without changing the overall stereo mix levels.

Home-entertainment applications would generally not use this feature.

During D-1 beta testing, a seasoned mastering engineer in Nashville compared the D-1 DAC to his 25-year-old Pacific Microsonics Model 2 DAC. The D-1 was vastly better in all respects. But his old DAC had a low-mid “thickness” or “richness” coloration, which made certain kinds of music sound “fuller” or more “finished”. I immediately recognized this as second-harmonic distortion giving music more apparent “body.” That’s what 2H distortion does, in small amounts.

I really liked the sound of his old DAC.

So the mastering engineer says, “my clients keep coming back because they LIKE the sound of my signal path. They LIKE the sound of the old PM2 DACs, and I don’t want to change that.” He then went on, “if you can add some coloration to your DAC that emulates my old DAC, then I would have the best of all worlds and would replace all the DACs in all my studios.”

So that’s what we did. We spent the better part of a year learning how to add different amounts and flavors of 2H distortion as a selectable feature of the D-1 DAC. We put a curated selection of these 2H distortion parameters in the D-1 Pro Menu, selectable as “A”, “B”, “C”, etc.. Go to the Pro Menu, select a 2H preset letter, then use the front-panel encoder knob to dial-in a desired amount of harmonic coloration.

Use care — a little goes a long way. At full blend, distortion levels can exceed 20% THD (!) Don’t overdo it. Sometimes just the slightest hint of “warmth” and “richness” can be the difference between “too thin” and “just right.” We’ve created the 2H presets such that numerical levels between 4 and 8 will likely be most common for pop music masters.

Yes, a digital signal processor (DSP) splits the incoming digital-audio signal into two pathways. One pathway manages the quiet audio, the low-path. The other pathway manages the loudest audio, the high-path. These two paths are separately D-A converted and then DSP-summed (combined) at the D-1 analog final output.

When a rising music program reaches a certain level (roughly -45dBFS / -23dBu) it must transition from the low-path to the high-path. Similarly, when a loud program gets quieter, it must transition from the high-path to the low-path. These transitions are accomplished by “cross-fading” — that is, we fade out one signal while fading in the other signal. The cross-fade algorithm gives an objectively linear result.

We’ve created a lab test unique to multi-path HDR-A processing. We call it CLE, or Cross-fade Linearity Error. The CLE test measures non-linearity between the low-path and high-path during cross-fade transition. Early in our multi-path research, we did a number of blinded listening trials to determine the ear’s sensitivity to CLE, i.e., the point at which cross-fade non-linearity becomes perceptible. It turns out that the ear is surprisingly insensitive to CLE with music program (music editors know this). The earliest perceptible onset of CLE was found to be a just noticeable difference (JND) of around 2.5dB for music and 0.9dB for a 1kHz continuous tone.

Based on these listening tests, we set our goal to create a maximum multi-path cross-fade error two orders of magnitude below JND. We achieved 0.003dB, nearly three orders of magnitude below the threshold of audibility. But to achieve this extreme matching requires calibration in your set-up, with your cables and gear.

There are two calibrations, one for line output and one for headphone output. After your initial D-1 installation, run calibrate on both outputs. That’s it, you’re done! Unless you change a component or cable that’s connected to the D-1’s inputs or outputs, you should never have to repeat the calibration cycle. While not essential, I think it’s a good idea to calibrate your D-1 after it has “warmed up” a bit, perhaps 10 minutes. In the rare case where a calibration does not complete, simply reboot the D-1 (using the rear panel switch) and do another calibration.

Much more could be written about D-1 calibration, but it is mostly proprietary, so we won’t…

Intuitively, yes. But the answer is no. An explanation is beyond the scope of this FAQ, and somewhat proprietary. Hint: if you read the patents deep enough, you can find the answer.

This is a long, complex conversation — better reserved for an engineering paper.

The short answer is this. In HDR-A design, there are two paths, a low-level path and a high-level path. When the music/program level rises, the DSP needs to know — in advance — when the level must transition from the low range to the high range processing circuits. If the DSP has no advance awareness, it cannot cross-fade the program in time, causing a potential overload to the low-path circuits.

To remedy this, the DSP delays (buffers) the incoming audio to determine if (and when) a low-path peak is about to occur. When the DSP “sees” a peak coming, it can then start an upward cross-fade from the low-path to the high-path, always at the precise moment required. This look-ahead processing requires a latency period.

In some professional applications, such as Foley, extreme low latency may be required (think gunshots). For such applications, the D-1 has provision for low-latency (<1mS) processing. When required, select the D‑1’s “low-latency” setting. In most all other applications (studio monitoring, home audio), use the D-1’s “normal-latency” setting.

Unless there is a compelling reason to use “low-latency,” always use normal-latency. You should not hear any sonic quality difference between latency settings, though certain objective performance measures may be slightly affected.

Something we learned in developing HDR-A topology is that multi-path digital signal processing and D-to-A conversion at higher sample rates may provide a slight perceptual improvement. But not always. And these improvements are very much program-dependent. To this end, we created two user-selectable internal processing paths. Select the path which, to you ears, sounds most spacious and transparent and realistic.

(1) In “Native” mode, the D-1 processes an incoming signal in its native sample rate. For example, if the native sample rate is 44.1kHz (CD, etc), then all digital signal processing and D-to-A conversion is done at 44.1. This is conventional DAC processing.

(2) In “Resample” mode, incoming signals are immediately up-sampled to 192kHz. All digital signal processing and D-to-A conversion is then processed at 192kHz. Native 192kHz (or higher) signals are not resampled, regardless of Resample setting.

Experiment with Resampling on lower sample rate digital audio and let us know what you think. There is no “right” setting, and in most cases you probably won’t hear a difference. It should be noted that Native processing tends to exhibit modestly lower objective processing artifacts, as measured on FFT plots. Note also that such artifacts, in either Native or Resample mode, are far below the threshold of hearing: in the -140 to -160 FS region.

Sure! Go ahead and start with all the normal DAC tests:  jitter plots, frequency responses, phase plots, filter responses, that kind of thing. Your results should rival or exceed any DAC made today. In some cases (e.g., linearity plot, noise), the D-1 will greatly exceed your test equipment’s performance limits. This is something you’ve never seen before in any DAC, and is quite breathtaking!

For broadband noise figure testing, you’ll want to build a really short XLR cable, perhaps no longer than 10-15cm. Use high-shielding cable, like Gotham 10561 GAC-2. To start the test, plug the test equipment output directly into its input and measure its own baseline noise. You might also build a zero-ohm XLR slug and plug that directly into the test gear input. Measure broadband (20Hz-22kHz), without a pilot tone, and without “weighting” filters. Make note of the test gear self-noise reading. Our AP2722 broadband residual noise reading is -119.5dBu (0.8uVrms).

Now, using your shorty cable, plug the D-1 DAC output directly into the test gear input and remeasure the noise. For reference, the output resistance of the D-1 is about 1.8 ohms. The D-1 noise reading should be identical to the test equipment’s self-noise reading (i.e., no change). This tells you that D-1 self-noise is significantly lower than the test equipment’s residual noise. If the D-1 was anywhere near test gear residual, you would see an additive effect, say 0.5dB or such, in your comparative numbers. For example, a reading of –120dBu test equipment residual combined with –125dBu DUT would read – 118.8dBu.

To measure even deeper into the noise floor, you will need to scale up using a high-gain, ultra-low-noise preamp, such as the Millennia HV-3C (–133dBuEIN). Compare the broadband self-noise reading of the preamp itself to the reading of the D-1 through the same preamp. This is somewhat of an art and not an exact science (see Art Kay’s seminal papers on measuring ultra-low-noise). Assuming you’re not picking up bench or cable noise (EMF fields, poor shielding, inductive effects, etc.), you should not see any additive noise. Caveat, it is extremely difficult to avoid environmental / bench / equipment / lighting / cable noise at these dramatically low nanovolt levels. Just moving one’s hand or body a couple inches can significantly alter a reading.

If you want to run baseline FFT’s, keep in mind that you’re working with a multi-path topology, i.e., a low-path and a high-path. When seeking low-path performance graphs, use a pilot tone below the cross-fade point. A -70dBFS pilot tone should be fine. When seeking high-path performance graphs, use a full-scale pilot tone. For the D-1, this is roughly +24dBu (actually, D-1 output can upwards of +29dBu, but we do not specify performance beyond +22dBu).

A visual waveform plot is another important multi-path test. The D-1 will deliver a visually perfect sine wave below 1uVrms. Unfortunately, today’s best test equipment cannot properly plot waveforms at this level. Our AP2722 “self-noise jaggies” become visible below 50uVrms plotted waveforms. To partially overcome this limitation, we use a Millennia HV-3C @ 60dB gain as an ultra-low-noise up-scaler. This allows us to cleanly visualize a waveform down to around 10uV (i.e., displayed at 10mV).

It’s instructive to compare DAC visual waveform performance at low-but-perceptual levels, such as 30uV. Run a 30uV equivalent digital sine wave into the D-1. Run the output of the D-1 into a high-gain, ultra-low-noise preamplifier, used for scaling. Then run the preamp output into your test equipment and observe the visual waveform plot. Of course, if you’re using a 30uV-equivalent original signal, your test equipment will now be displaying a 30mV waveform (+60dB or 1000X higher) due to the scaler. That’s OK. You’re looking for relative visual waveform purity, not absolute numbers.

Those stacks do look impressive. Hey, if a design team thinks they can improve D-to-A specifications by separating the main functions (conversion, clock, power, etc.) into multiple stacked-up chassis’s and gorgeous carved out aluminum billets, then I trust they’ve achieved performance specifications unachievable in their single-chassis approaches.  It would be instructive to ask these companies — specifically — which performance specifications were improved, and by what magnitude, and why multiple boxes were required to achieve this.

I hate to bust anyone’s illusions, but multi-path D-to-A conversion makes these head-spinningly expensive DACs technically obsolete. They could divide up a DAC into 5 or 10 boxes, but, compared to HDR-A architecture, these oligarch playthings would still exhibit profoundly inferior dynamic range, self-noise, linearity, and quiet-side THD performance. Multi-path HDR-A is a triple-patented, once-in-a-generation leap-frog on these most critical DAC performance metrics.

That said, we do use multiple boxes in the D-1. There’s a main aluminum chassis and an isolated steel sub-chassis inside the aluminum chassis. The main aluminum cavity houses the digital/DSP PCB. The fully-enclosed steel sub-chassis houses the analog/DAC PCB. This wasn’t a marketing decision. The analog topology achieves 40 nanovolts of output noise. But many environments (homes, studios, etc.) have at least 40nV of “airborne” energy (or “E-field”) generated by lighting, computers, TVs and audio gear, HVAC, appliances, routers, local RF, NAS, etc. Homes and studios can exhibit volts-per-meter of free-field noise. We put the sensitive 40nV circuits in a steel sub-chassis because steel is a better material to shield these common EM energies, while aluminum is better for shielding RF noise to/from digital circuits.

A long time ago, switching power supplies had issues that could cause problems in audio circuits. In the 1990’s, we tried many switching architectures in our Millennia high-performance professional analog designs. We decided to stay with linear power. Around 2010, on the advice of my dear friend Bruce Jackson, we revisited switching technology, which had greatly improved. We spent two years researching switch-mode power for ultra-low-noise preamplifiers, doing endless blind A/B comparison trials against our proven linear solutions.

We moved to our first switch-mode solution around 2012. We wouldn’t have done it if it didn’t sound (and test) perfect. Today’s SMPS technology can be ideal for audio, though our vacuum tube gear still employs giant toroid transformers and linear / shunt regulation. It’s no longer an ideological question of “linear” vs. “switching”. The question becomes:  which topology makes better audio sense for a specific functional design?

For the D-1, we opted for a superior audio-grade external SMPS – not just for ideal sonic and noise performance, but also because it provides 100% galvanic/earth isolation to the sensitive USB and Dante interfaces. Moreover, this power source is double downstream regulated by ultra-low-noise linear circuits, along with amplifier stages that provide further extraordinary levels of PSRR. The D-1 is a master class in state-of-art power and ground design.

Bottom line: we use the power technologies (XFMR, FWB, SMPS, LDO, Shunt) that achieve the highest sonic reality and electrical performance for a specific product and application. The idea that one power technology is better than another power technology, for all audio products, in all cases, is misguided.

It’s science, with a bit of art and tribal knowledge. But no magic. Bill Whitlock described the practical science of grounding decades ago. Before you do anything, or form an opinion, read his papers.

Where I would diverge from Bill is in the practical application of theory. In theory, standard practice always works. In practice, not always. The problem is that audio manufacturers are not uniform — many have clearly not read Bill’s papers. Different manufacturers can (and do) have different theories of audio ground design. And often we simply don’t know what ground topologies are lurking in our audio gear.

This means that following standard audio ground techniques may, or may not, give optimal results. Experimenting with audio (signal) grounding can sometimes improve performance. We never know until doing some tests, listening, measuring, and trying a few alternatives.

That said, there is one rule that must always be followed: never defeat the AC power earth safety ground. Never.

As we say, the magic happens in the low-path.

Any well-recorded music with high dynamic swings is an ideal candidate for HDR-A conversion. High dynamic range music spends much of its time at low levels, in the low-path. As we know, the imersiv low-path can exhibit orders of magnitude lower ambient-range THD+N than any legacy single-path DAC architecture (R2R Ladder, Sigma-Delta, MDAC, etc.). The improvement is plainly audible with high dynamic content program, especially in high-resolution monitoring rooms. The difference is often breathtaking.

Classical, jazz, and acoustic music would be primary examples of high dynamic program. Here are some specific examples of recordings we’ve found to be especially stunning on the imersiv D-1.

Firebird, Telarc CD-80059, 1979, Recording Engineer: Jack Renner. Note the spatial intimacy and expansion in the first movement.

Birds (track), Dominique Fils-Aime, 2018, Qobuz 0623339210904, Recording Engineer: Jacques Roy

Babylon Sisters (track), Steely Dan, MCAD-37220, 1980, Recording Engineers: Roger Nichols, Elliot Scheiner, Bill Schnee. Yes, I know. Cliché and hackneyed. We’ve all heard this track hundreds of times at audio shows and demos. But multi-path does something that I have never experienced, even on the world’s most lauded single-path audio systems. It was a bit shocking to hear, but the instruments, especially Bernard Purdie’s drums and Patti Austin’s vocals, take on a three-dimensional depth and clarity that I have never experienced – try it!

Soledad (track), Sera una Noche, MA Recordings, 1998, Recording Engineer: Todd Garfinkle. Stunning natural room depth.

St. James Infirmary (track), Louis Armstrong Plays King Oliver, Audio Fidelity, 1959. Producer: Sid Frey.  Vocal and trumpet microphone: Neumann SM2. Recorder: Ampex 350. (note: YouTube session video is not the Ampex recording – find an original master copy on CD)

Messiah – Unto to Us (track), American Bach Soloists, Delos B000BQ7JUC, 2005, Recording Engineer: John La Grou

Polarity, Hoff Ensemble, 2L 145, 2017, Recording Engineer: Morten Lindberg

Besides following good acoustic practices (phase coherence, time-alignment, equi-triangle listening position, proper room design and treatment), there are two fundamental keys to optimizing imersiv D-1 sonic performance.

The first key to optimal sonics is using a power amplifier with at least 20dB of self-gain, ideally 30dB or more. If your amplifier has level controls, turn them all the way up. Connect your D-1 directly to the power amp, and use the D-1 level control just like a preamp.

As we say, the magic is in the low-path. The goal here is to maximize the amount of time your program spends in the D-1 low-path, where the distortion and noise can be orders of magnitude lower than legacy single-path DACs.

The second key is simple: choose well-recorded program with high dynamic range. Music that “breathes” with life and energy, subtle nuance, timbre detail, and reverberant atmospherics, will be presented with a breathtaking realism that no legacy single-path DAC can deliver. Not even close.