Brilliant Corners #9: a DAC and a Streamer from France's Totaldac

To misquote Morrissey, some knobs are better than others. The Manley Neo-Classic 300B amplifiers that I've been listening to, for example, have a knob marked "feedback" that goes from 0 to 10. I've learned so much from using it that I've come to believe that if your amp doesn't have such a knob, it should. You see, the higher you set this control, the better the amp will measure. Applying more global negative feedback to these amps lowers their nonlinear distortion and noisefloor, increases their bandwidth, renders them less sensitive to the speaker's impedance variations and otherwise makes them more stable and efficient. In fact, by applying lots of feedback to an amplifier, it's possible to reduce distortion to barely measurable levels.

So what's the problem? Well, a few turns of the knob suggest that negative feedback isn't as useful as it appears on paper. The Manley website urges the listener to dial in a "tasteful amount of feedback." For me, that's about 3dB, which tightens the bass without affecting the listening experience adversely. But turning the knob past 3dB progressively reduces my ability to enjoy the music: It robs it of color, texture, presence, and drama until these ravishing tube monoblocks begin to remind me of a receiver from the early years of solid state. Well, that might be an exaggeration, but you get the idea.

As it reduces the total amount of distortion, negative feedback adds higher order harmonics that sound nothing like the simple harmonics we associate with musical instruments (footnote 1), and after a while the brain begins to call bullshit on the idea that dialing in more of it is bringing us closer to the recording. The knob on the Manley amps makes these relationships—which usually reach audiophiles mainly in the form of theory and opinions—audible to anyone with ears.

Feedback is a large and complex topic, but I'm bringing it up to poke at the question of purpose: What is an amplifier's job? Swiss industrial designer Max Bill, who is best known for the Bauhaus-inspired clocks and watches he designed in the 1950s, once said, "The basis of any aesthetics should, above all, be function. An exemplary object should serve its purpose under all circumstances." I own one of the minimalist mechanical watches Bill designed for German watchmaker Junghans, and I enjoy wearing it because it is beautiful. I would argue that being beautiful is in fact its purpose—a $10 quartz watch and my phone both tell the time more accurately and reliably.

A home audio component works the same way. Its purpose isn't to play back music accurately—however you might define that—but to provide enjoyment. And by enjoyment, I mean the feelings of pleasure, surprise, and inspiration that can arise as you listen to your hi-fi. About this basic truth, a friend who works in the audio business recently remarked that what he sells are "basically expensive sex toys." He didn't mean this as a knock on our hobby—quite the opposite, in fact. How many things do you own that reliably give you pleasure? While submerging an amplifier circuit in gobs of negative feedback increases some types of measurable accuracy, it also demonstrably reduces its capacity to provide enjoyment, at least for me. And unless John Atkinson has been keeping it to himself, I'm not aware of a suite of measurements that measure a component's intrinsic listening satisfaction.

The disconnect between measured performance and listening has been on my mind for the past several months as I auditioned digital components from Totaldac (footnote 2). DACs in particular are overachievers when it comes to measurable accuracy; even the DAC chip in my long-since-retired iPhone 4S sounded surprisingly, consistently competent, especially when you consider its minuscule size and cost. But digital sources continue to struggle with providing pleasure—the kind of juicy, watermelon-in-an-ice-bucket-on-a-hot-August-day pleasure that you can experience by listening to, say, a 60-year-old 45 of the Fendermen playing "Mule Skinner Blues."

Many DACs nail resolution, transparency, frequency extension, smoothness, imaging, and dynamic muscle, but nearly all—even the very expensive ones—fall short of achieving tonal density, of pressurizing the air in the manner of a good record player or an actual musical instrument, which is sensed by our bodies as physical presence. Most of them tend to turn half-and-half into skim milk. And some particularly unsuccessful ones make the music so insubstantial that it feels like it's playing from behind a sheet of glass.

For me, the sense of weight, texture, and presence is a big part of what makes reproduced music sound real—and real fun. I don't know how one might go about measuring presence, but you sure sense it when it's gone. And I have never been entirely convinced by DACs that use tubes in the output section to add back a measure of this missing goodness.

In Brittany, near the tidal island of Mont Saint-Michel, Totaldac's Vincent Brient has been trying to redress this situation through a rather extreme approach to designing digital components. "Listening to digital should remind the listener of real concerts," he wrote to me in a recent email, "where [listeners] are constantly surprised by timbres, dynamics, presence, contrast, and frequency bandwidth. It should not be just an analytical and visual experience, with nothing more than a soundstage. After all, when a musician is playing in the next room, you don't have a soundstage, but the sound is still magic."

Brient's solution is to create DACs using R-2R networks instead of integrated chipsets. These networks are made of Vishay metal-foil resistors with a variation tolerance of 0.01%. As you might imagine, these high-precision devices aren't cheap, and the Totaldac d1-unity DAC I've been listening to contains 100 of them (whereas the four-box d1-sublime DAC, which I heard and enjoyed in Munich, uses 600!) (footnote 3). Given that the d1-unity boasts no exotic digital filters, features, or technologies—it tops out at a resolution of 24/192 and offers DSD as an option at extra cost—and goes for an impressive €11,500, my hopes for its sound quality were rather high.

Before I get to that, I should mention that the d1-unity's compact, elegant black box, which comes with a small outboard power supply and weighs an unassuming 15lb, also arrives shorn of frills. Besides a rather plain plastic remote, there's not much to talk about. Until 2012, Totaldac didn't offer a USB input. It does now, using technology from XMOS, though I found the AES3 input to sound pleasantly meatier and more colorful.

Brient also sent me his d1-streamer-sublime (€9100), which offers a network input and the expected digital outputs and functions as a Roon endpoint, which is how I used it. It also tops out at 24/192, and it passes DSD via DoP. Its streamer and reclocker boards, as well as its software, were designed in house. Except where noted, I listened to the d1-unity DAC and d1-streamer-sublime together. And one note about the cables that were thoughtfully included with these components: Though the Totaldac AES3 and Ethernet cables look impressively constructed, I preferred the sound of my AudioQuest Diamond alternatives by a clear margin.

I started listening with a stream of the soundtrack to Todd Haynes's 2007 film I'm Not There (16/44.1 FLAC, Sony/Qobuz), in which six actors—including Cate Blanchett!—play Bob Dylan. The soundtrack is a Dylan tribute album full of unusually inspired casting decisions and surprising arrangements. A favorite is "Señor (Tales of Yankee Power)," performed by Willie Nelson and Calexico with a verse in Spanish sung by Mexican-American troubadour Salvador Duran. I first got a feel for the Totaldac's talents during Nelson's guitar solo. I've heard him play Trigger, his Martin N-20 acoustic guitar with a hole worn through the soundboard just above the bridge, on dozens of recordings and in person. Through the d1-unity, the instrument's unmistakable sound came through with all of its nylon-string pluck and woodiness intact. But the guitar body also sounded rich, dense, and distinctly solid, as it does through a good record player and on stage. Hearing it hanging between my speakers produced what my brain had assumed was a distinctly analog thrill. The French DAC was allowing me to revel in one of the most fun illusions of reproduced music—the realistic presence of voices and instruments—using a digital signal. This was cool!


Footnote 1: See, for example, solid state designer Nelson Pass's take at passlabs.com/technical_article/audio-distortion-and-feedback.

Footnote 2: Totaldac, Tel: +33 6 18 03 14 08 Email: totaldac@totaldac.com. Web: totaldac.com.

Footnote 3: An earlier version of the Totaldac d1 D/A processor was reviewed in January 2016Ed..

ARTICLE CONTENTS

COMMENTS
georgehifi's picture

JA Any measurements coming for this?

Cheers George

John Atkinson's picture
georgehifi wrote:
JA Any measurements coming for this?

I don't usually measure products that are covered in Stereophile's regular columns.

John Atkinson
Technical Editor, Stereophile

georgehifi's picture

Bummer, would have been great to see what was good/bad with the measurements.

Cheers George

JRT's picture

George, to your point about measurements, which can be interesting and useful in de-obfuscating problems, a couple of the somethings to look for in DACs utilizing switched R-2R resistor arrays would be presence of glitch impulse energy added to the output signal, and settling time degraded by deglitching circuits. I wouldn't know if the performance of this DAC suffers any such problems. Useful measurements and analysis could highlight or eliminate some concerns.

You might find interesting an Analog Devices engineering note on the subject of deglitching switched R-2R DACs at the following link.

https://www.analog.com/en/technical-articles/deglitching-techniques-for-highvoltage-r2r-dacs.html

Here is the introductory section of that excerpted to the block quote below.

In an R-2R DAC design with supply voltages exceeding ±5V, large voltage glitches (up to 1.5V) can occur during the DAC's major-carry transitions. These glitches can propagate through the output buffer amplifier and appear at output. The slewing of the level shifters that control the top (VREF+) and bottom (VREF-) single-pole double-throw switches (S0 to SN) causes the glitches (Figure 1). If each switch of the "inverted" R-2R ladder were turned on and/or off instantaneously, glitch amplitudes at the DAC output (or input of the output buffer amplifier) would be small. However, switches do not switch instantaneously; in fact, to avoid crowbar current between outputs of the two reference buffers, the switches employ a break-before-make connection. The associated time delay can produce very large glitches during DAC code transitions that degrade the dynamic performance specification 'glitch impulse energy.'

A lower voltage could scale down the magnitude of the problem, perhaps also at the cost of reducing dynamic range. The cause of the glitching is in the switching.

georgehifi's picture

"I wouldn't know if the performance of this DAC suffers any such problems."

That's why I like to see for myself, not speculate from others subjective opinions.
To do so is like purchasing a new car without looking at all it's measured specs, even though your only going to drive it.

Cheers George

music or sound's picture

maybe it is expected to not measure well and therefore not a regular column?

JohnnyThunder2.0's picture

to relegate certain components to "columns" and not "features" to keep their measurements a secret. I mean if the truth came out then the entire audiophile universe would collapse in on itself in a matter of hours ! How silly. I would think that it's a practical thing - as in JA1 only has enough bandwidth to do his exhaustive and detailed measurements and listening tests to a few components a month. Totaldac, Border Patrol, AudioNote UK, Lector, etc. are all brands that have legions of music loving fans and receive great subjective reviews but do not measure "well." But that's a different story and one that isn't going to be dealt with here. This is not new news and to expect differently just means...well it means, just go read ASR.

hollowman's picture

yeah, JA's Measurements would be welcome.

PeterPani's picture

you would have to consider the Brilliant Corners on 15ips R2R from Tapeproject, too.

JRT's picture

The product in this category that most interests me (currently) is the Eversolo DMP-A8 ($1980), which seems far more capable at much lower pricepoint.

https://www.eversolo.com/

This post is not SPAM. I have no affiliation, no financial interest, no money on any horse in the race.

Metaldetektor's picture

Great column!

I love Totaldac for their rich, dense tone. Hits me like my favorite old grappa. But I can only listen to it occasionally (just like I don’t drink 8 glasses of that grappa every day) - it has a distinct flavor that can distract. Soft, a little sweet, the opposite of a crisp white. Maybe it’s better balanced with the horns they use in their best system. This newer version should also be better than mine from a couple years back.

For that dense musical tone but a balance that might better suit non horn systems, Playback Designs is wonderful. Like Totaldac, I doubt it measures great but so be it.

P.s. - love that soundtrack, the Jim James / Calexico track is the one that takes me home.

Alex Halberstadt's picture

Thanks for the comment. The Jim James track is just amazing.

georgehifi's picture

Metaldetekto: "I love Totaldac for their rich, dense tone."
"But I can only listen to it occasionally"
"It has a distinct flavor that can distract. Soft, a little sweet"

Thanks Metaldetekto:
What you said in a couple of words tells me more than the 2 page review does, and about how and maybe why it was not maybe measured.

Cheers George

hollowman's picture

Product link:

https://www.totaldac.com/D1-unity-en.htm
-non-oversampling DAC compensation filter activated or disactivated by remote control.

totaldac's picture

A new amplifier and the matching driver have just been released by Totaldac.

The Amp-1-sublime is the new Totaldac reference amplifier, using solid state components exclusively.
The triode stage used in the previous generation of Amp-1 is now upgraded by an innovating transistor gain stage, using very low component count (all discrete) and no global feedback.
A very high voltage (240Vdc) is used to give the maximum dynamic to the signal and its power supply is based on a discrete regulator as well.
The circuit uses 0.01% VAR Bulk Metal® Foil resistors from Vishay in the critical voltage gain stage.
The gain stage is then followed by an equally innovating and powerful transistor buffer, using no global feedback as well.
This design combines no global feedback, low noise (0.2mVrms), high gain (33dB), high bandwidth (400KHz) and high current (driving 1ohm) thanks to its 2500VA/25kg live-power transformer.
These characteristics produce a transparent, fast, live and natural sound without harshness and at the same time with punch and weight in bass.
The Amp-1-sublime amplifier is available as stereo amplifier with RCA inputs or as fully balanced monobloc with XLR input.
amplifier page: https://www.totaldac.com/amp-1-en.htm

A new d1-driver is also released to perfectly match the quality of this new amplifier.
The d1-driver-sublime has no volume control because it relies on the DAC volume control but in case the complete system gain is too high (DAC + amp + speakers + room) an analog -15.5dB attenuation can be activated using a switch at the back panel.
This attenuator uses 12 pieces of 0.01% VAR Bulk Metal® Foil resistors from Vishay.
In case the complete system misses some gain with some tracks recorded very low, the latest Totaldac DACs offer a +12dB digital gain option.
This 2 selectable gain in a Totaldac system allow to optimize all configurations.
driver page: https://www.totaldac.com/D1-driver-en.htm

celef's picture

i do not get the first part, that some coloration sounds better then some other coloration? then how much coloration is preferable, a lot or little less or something in between? and will this coloration be suitable at all situations?

georgehifi's picture

"then how much coloration is preferable"

How long is a piece of string??

No coloration is ideal to me, as it's closer to the real thing and the way you were meant to hear it.

No??? then get yourself a graphic equalizer

Cheers George

barrows's picture

You really need to be a bit more educated before making absolute statements such as this:
"As it reduces the total amount of distortion, negative feedback adds higher order harmonics that sound nothing like the simple harmonics we associate with musical instruments (footnote 1), and after a while the brain begins to call bullshit on the idea that dialing in more of it is bringing us closer to the recording."
To become more educated I would suggest that you look into Bruno Putzeys thoughts on feedback. He has shown with his designs that there is point where one adds even more feedback, and ALL of the harmonic distortion products disappear.

X