Devialet D-Premier D/A integrated amplifier

Most reviews are straightforward. One preamplifier or power amplifier replaces another. DACs are swapped out. A new pair of speakers takes up residence in the listening room.

But some products demand a complete revision of a system's architecture. Such was the case with Devialet's D-Premier ($15,995). Not only is this French product an integrated amplifier, with phono and line analog inputs; it has digital inputs and an internal D/A section. And since v.5.5 of its operating system, the D-Premier can also act as a high-resolution WiFi audio streamer, working with Devialet's Asynchronous Intelligent Route (AIR) client for Macs and PCs. At a stroke, the D-Premier replaces streaming program, USB or other computer audio interface, D/A processor, preamplifier, power amplifier, and many cables. And all is contained in a beautifully finished aluminum case about the size of a small pizza box and just over an inch thick.

I first saw and heard the Devialet D-Premier at the 2010 Consumer Electronics Show, but it was not until a year later that Audio Plus Services announced that it would distribute the D-Premier in North America. I received a first review sample in 2011, then a second, to use with the first as a pair of monoblocks, in early summer 2012—but first I needed to clear my decks of more conventional products that I was reviewing. I am now kicking myself for having waited so long.

Technology
Devialet SAS is a French company, founded in 2007 by Pierre-Emmanuel Calmel and Mathias Moronvalle, colleagues at Nortel France's R&D Lab, to develop a new type of amplifier developed by Calmel. Called ADH®, for Analog Digital Hybrid, this patented topology connects a small, high-voltage, but low-power class-A amplifier directly to the speaker, with then a parallel class-D stage providing the necessary current. This is reminiscent of the innovative "current-dumping" circuit developed by Quad in the mid-1970s, though the Quad circuit used a class-AB current amplifier. However, the AHD circuit differs significantly in detail from Quad's, and is considerably more complex. Extraordinarily, there are only two resistors and two capacitors in the analog signal path!

I discussed the D-Premier's topology with M. Calmel at the 2011 CES. The analog input signals are converted to digital with an A/D converter, a Texas Instruments PCM4220, running at 96 or 192kHz—the former is the default—before being applied to the volume control, which operates in the digital domain and is implemented in a 32-bit floating-point DSP chip, along with a soft-clipping function and crossover filters when required. All signals are then converted back to analog by two Burr-Brown PCM1792 chips—a high-quality, 24-bit, two-channel, current-output device operating at up to 192kHz. Just half of the DAC is used for each channel, and the current output of the DAC is converted to voltage with a resistor and fed directly to the class-A amplifier—the analog signal path from the DAC output to the loudspeaker terminals is only 2" long. In effect, the DAC swings the high voltage required to drive the speaker output, and the class-A amplifier therefore works at unity gain, as a voltage follower, so that its performance can be maximally linear at high frequencies.

To provide the current to drive the loudspeaker, a four-phase, multilevel digital amplifier—four switching stages, staggered in time—is added in parallel to the class-A amplifier. It is slaved to the class-A amp much as in a car's power steering, where the driver turns the steering wheel to indicate how much he wants the wheels to turn, and a servo-controlled hydraulic system actually turns the wheel.

Conventional class-D amplifiers suffer from high levels of ultrasonic switching noise riding on their outputs, which mandate use of a hefty low-pass filter between the output stage and the speaker terminals. In the D-Premier, there is no LC filter on the class-D amplifier's output; instead, the analog amplifier provides a very wide-bandwidth correction signal that cancels the ultrasonic switching noise that would otherwise be present.

The power supply is a 600W switch-mode type offering 2100W peak and incorporating full power-factor correction. Because of the high switching frequency, the planar transformer can be tiny. There is much more to the D-Premier's innovative and elegant circuit that I don't have room to discuss here; I refer you to a white paper that can be downloaded here. But the entire package offered by the D-Premier appealed to my sense of purity—it is no bigger than it need be to do what is intended.

Setup
I had to slide off the section of the D-Premier's top plate that covers the rear panel in order to be able to use my preferred XLO Reference 3 AC cable, the plug of which would have been too big to reach the recessed IEC mains jack. The D-Premier offers extraordinary flexibility in how its inputs can be arranged—see the diagram of its rear panel. Using the Configurator app, downloadable from the Devialet website, the user sets up the amplifier as he needs and burns the configuration as a text file to an SD card. Inserting this card in the rear-panel slot and turning on the amplifier updates its internal state. I used the factory default configuration, which offers two pairs of analog inputs (one of them phono, to be tested in a Follow-Up review) and five digital inputs: two TosLink, two S/PDIF on RCAs, and one AES/EBU on an XLR jack. There is an HDMI port, currently unused, and the D-Premier is WiFi capable.

The only control on the amplifier itself is a discreet On/Off/Sleep button at the center of the front panel. When the amplifier is on, a gentle amber circle is projected onto the surface beneath the front of D-Premier. All other controls are carried on the remote control. A large rotary knob adjusts the volume. A single button above the volume control duplicates the On/Off/Sleep button; three other buttons control Input Select (consecutive pushes cycle through the inputs, each starting with the volume control set to "–40.0dB"), Bass high-pass filter On/Off (when configured for use with a subwoofer), and Polarity Inversion. These buttons can also be used to control channel balance and tone-control selection, when the amplifier is appropriately configured. The small, circular, color display on the amplifier's top panel indicates the input in use, the volume level in dB, and the sample rate for digital inputs. If no datastream is present, the digital input's name illuminates in red; it turns black if valid digital data are detected. The D-Premier goes to sleep if no music has been played for 20 minutes or so.

To use the D-Premier's WiFi connectivity, Audio Plus had supplied me with an Apple Airport Express programmed to set up a network called "DevialetAudio." When turned on, the D-Premier looks for a network with this name and logs on. To play music over this network, you install the Devialet AIR app on your PC or Mac. This then handles the output of music files selected in iTunes, selecting the correct sample rate and transmitting the data over the network to the D-Premier. This runs automatically, and while it doesn't have to be open during use, as the screenshot shows, when open it displays the name of the file playing, its format, sample rate, and bit depth, and the playback time, as well as cover art and various network diagnostics. AIR is currently limited to 24-bit/96kHz, but a free upgrade to handle 24/192 files is promised.

When the D-Premier is connected to the DevialetAudio WiFi network, an iPhone/iPad app duplicates the remote control's functionality and the amplifier display, adding the words "My D-Premier" above the volume setting. Touching the input name brings up an Input Select menu, and again, selecting a different input reduces the volume to "–40.0dB." Touching the volume setting mutes the D-Premier; touching the Mute symbol unmutes the amplifier.

COMPANY INFO
Devialet SAS
US distributor: Audio Plus Services
156 Lawrence Paquette Industrial Drive
Champlain, NY 12919
(800) 663-9352
ARTICLE CONTENTS

COMMENTS
mphatic1's picture

John,

 

In your opinion, how does the Devialet's pre-amp/DAC section stack up against the Classe CP-800?  I ask this because they both appear to be cut from the same cloth so to speak; beautifully styled, apparent 18 bit resolution, switching amp/power supply, highly configurable, and relatively future proof.  And, when you start to add power amps to the CP-800, you get up around Devialet money anyway.

 

Regards,

Brad

JadenKrosis's picture

"the most extraordinary product I have reviewed for Stereophile."

Considering the very long list of other very extraordinary products you`ve had your hands on, this says alot!

rumnyc's picture

but in the end went with a slightly more conventional setup.

for exactly the same price as a D-Premier, I went with a Linn Akurate DSM and Mcintosh MC452 amp. I couldn't compare them back to back but I like the below features from the Linn:

4 HDMI input instead of 1 (yes I could have used an outboard HDMI switcher).

Ethernet input and DLNA server compatible (AIR feature was not yet released)

And the Mcintosh had way more headroom than the D Premier's amp.

My speakers are B&W 802 Diamond.

stgomes's picture

 

After some very good 1st impressions on sound quality driving "conventional" speakers (impedance curve not bellow 2-3 Ohm) I was curious to audition the D-Premier with my Martin Logan Summit X speakers. Sound quality is very good, until I pumped up the volume. At about 85db, the protection LED on the front panel starts blinking and the amp starts to mute.

I have reported this problem to Devialet and they say it should not happen, but my dealer had the same experience with other D-Premier units driving other Martin Logan models (Montis, Spire, CLX).

My guess is that this amp technology is simply not designed to drive impedance loads like the ones present in Martin Logan speakers, which drop bellow 1 Ohm in the high frequencies.

Looking forward to your comments.

John Atkinson's picture

Quote:
My guess is that this amp technology is simply not designed to drive impedance loads like the ones present in Martin Logan speakers, which drop bellow 1 Ohm in the high frequencies.

That's what I expect, too.

John Atkinson

Editor, Stereophile

BradleyP's picture

Notice that with the new SAM technology, two Martin Logan speakers have been matched to the Devialet--the Theos and the Electromotion. Now, I am wondering about Maggies. If Devialet can both drive and tame those big bass panels, then that would be even more amazing.

jlesnick's picture

Hey John.

Thanks for this great review. How do you feel the Devialet (and its new iteration) would stack up to a Pass X250.5/Classe Cp-800 combo?

Cheers,

Jon

John Atkinson's picture
jlesnicj wrote:
Thanks for this great review.

You're welcome.

jlesnicj wrote:
How do you feel the Devialet (and its new iteration) would stack up to a Pass X250.5/Classe Cp-800 combo?

I haven't heard the XA20.5 but the Pass XA60.5s have been my go-to amplifiers for the past year. The beauty of the Devialet is that it is just one box, with a digital section up there with the Classe. A review of one of the less-expensive Devialets is in progress.

John Atkinson
Editor, Stereophile

bapcha's picture

JA/Other reviewers:

You guys are a prime example of promoters of peddlers of snake oil. If an Airport (Mac) or a Ayre CD player feeds data to a D/A you claim to hear a difference. How can you - when the data is exactly the same?

Let's take Boulder - I like the look of their products, but how would a volume knob - polished seven times and clear-coated - improve sound quality? It is a bloody waste of money.

Cheers,
Bapcha (yes, I am a marketer who can see through your ____)

John Atkinson's picture
bapcha wrote:
You guys are a prime example of promoters of peddlers of snake oil.

Good morning to you, too.

bapcha wrote:
If an Airport (Mac) or a Ayre CD player feeds data to a D/A you claim to hear a difference. How can you - when the data is exactly the same?

Yes, the data may be the same, but the timing of when those data are presented to the D/A chip is also important. Variations in that timing are called "jitter" and result in distortion in the reconstructed analog signal. The right data at the wrong time is equivalent to the wrong data. There are several articles on this subject reprinted in our free on-line archives.

bapcha wrote:
Let's take Boulder - I like the look of their products, but how would a volume knob - polished seven times and clear-coated - improve sound quality?

No-one has said that it does.

John Atkinson
Editor, Stereophile

bapcha's picture

I will keep the conversation very respectful. I am a chip designer with ten plus years designing analog and digital chips and, and ten more marketing them and SSDs. Four issued patents (primary author). Most chips now have PLLs. After it locks, it does not matter if the jitter is 5 pico seconds or 50 pico seconds, the data can be PERFECTLY reconstructed (or the designer will lose his job, and chip cannot be sold).

You claim to hear differences in jitter between 5ps and 50ps - when the data is PERFECTLY reconstructed post-jitter.

YES - you can hear a difference in ANALOG products. If a DAC has a DIFFERENT analog output, it will sound different from a cheaper one with a not so well designed output.

Also, FYI, it is easier to match currents on chips without trying too hard (it is called a common-centroid layout where transistors are laid our perfectly symmetrically (x and y axes). It is not too hard to make chip-based output stages that sound better than discrete (but it is cost-prohibitive). I own Bryston/Ayre stuff with properly designed circuits, and sincerely believe it is impossible to do better (note the word impossible).

I have collected serial data from a $10K Boulder DAC and a $1000 Bryston Dac. I gave up after 10 GigaBytes. The data was exactly the same. So, 24x oversampling = 4x oversampling = 2x oversampling (I know that you know the Nyquist-Shannon theorem). It is a fool's game after one gets past the Bryston/Ayre A* (not the K*) price point.

Word length matters. Sampling rate does not. If it does, then I have a TV that shows pictures in X-ray.

Even James Tanner of Bryston has admitted that the differences between their BP10/17/26/SP3/SP2 analog stages are identical or have minuscule (inaudible) differences.

Bottom-line: Looks like I am half way agreeing with you and half-way disagreeing with you.

Sincerely yours,
Bapcha

John Atkinson's picture
bapcha wrote:
I am a chip designer with ten plus years designing analog and digital chips and, and ten more marketing them and SSDs. Four issued patents (primary author). Most chips now have PLLs. After it locks, it does not matter if the jitter is 5 pico seconds or 50 pico seconds, the data can be PERFECTLY reconstructed (or the designer will lose his job, and chip cannot be sold).

An impressive resume. So with respect, I am puzzled that when you refer to phase-locked loops (PLLs), you don't seem aware that a PLL circuit has a corner frequency, below which it increasingly fails to eliminate jitter in the incoming datastream. You can see the effects of this problem in the various digital products that I test for for Stereophile. Some are very effective at eliminating the effects of datastream jitter; others are wide-open to timing variations in the data presented to the DAC.

bapcha wrote:
You claim to hear differences in jitter between 5ps and 50ps . . .

We have made no such claim, any more than we have claimed, as you accused us in your earlier comment, that the finish of the volume control on a Boulder preamplifier affects sound quality.

John Atkinson
Editor, Stereophile

bapcha's picture

Thanks for your kind and timely responses.

Bottom-line: JA - we all have our biases. Bottom-line is the you run the best publication for reviews of music reproduction electronics. Today, a Cisco router can transmit ALL of the data from the Library of Congress from point to point - in about four seconds. Yet - there is room for endless back and forth for a trivial range of frequencies. Let's call it 0 Hz to 100KHz. When we have mastered 10s of gigahertz, we should be able to do this trivial range in frequencies easily, cheaply, perfectly, every single time - and yet we don't. My take is that it is due to the EMOTIONAL component of this trivial data "band" if you may.

I wish you and your magazine the best. (I know Robert Greene rather well. He is a colleague of my uncle's at UCLA math)

John Atkinson's picture
bapcha wrote:
Bottom-line is the you run the best publication for reviews of music reproduction electronics.

Thank you.

bapcha wrote:
Today, a Cisco router can transmit ALL of the data from the Library of Congress from point to point - in about four seconds. Yet - there is room for endless back and forth for a trivial range of frequencies. Let's call it 0 Hz to 100KHz. When we have mastered 10s of gigahertz, we should be able to do this trivial range in frequencies easily, cheaply, perfectly, every single time - and yet we don't.

That the transmission of digital data without errors is a given. But it is when those data are used to reconstruct an analog signal that problems in implementation arise.

bapcha wrote:
I wish you and your magazine the best. (I know Robert Greene rather well. He is a colleague of my uncle's at UCLA math)

Thank you. But I am wondering if you are confusing Stereophile with The Absolute Sound. Robert Greene writes for TAS, and has never contributed to Stereophile.

John Atkinson
Editor, Stereophile

bapcha's picture

Nice guy anyway. Took him a long time to switch to digital, and solid state.....

Sincerely,
Bapcha

X