[ Home | Staff & Contacts | HiFi Playground | Listening tests | DIY & Tweakings | Music & Books ]

Audio Note DAC Kit 1.1

Music bits and music notes

or rather

Audio kits and audio notes

[DAC 1.1 kit]

[Italian version]

Product: Audio Note DAC Kit 1.1
Manufacturer: Audio Note UK
Approx. price: EU899 (built) / EU699 (kit)
Test sample: manufacturer loan
Reviewer: Werner Ogiers

Audio Note (UK) have traditionally been an outlet not only for normal (well ...) hifi components, but also for quality vacuum-tube based kits, ranging from affordable to most certainly not-so-affordable.

One of their latest kits, the digital-to-analogue convertor reviewed here, created quite a stirr as it followed in the wake of the new top-of-the-line DAC 5, a component which got very high praise from the UK cellulose-based press, and also a component that directly spawned the little DAC Kit 1.1.

DAC 1.1 backsideThe 1.1 is a so-called non-oversampling DAC (the term "1 x oversampling" proudly emblasoned on its brushed aluminium front actually is an oxymoron), which flies in the face of accepted DAC-lore, going back straight to the early days of CD, to that dark period around 1981. (For more on this read the technical insert.)

The device under test comes in a large, heavy, and rather rigid metal housing that screams "I'm a nephew of Ongaku". It also screams that its lineage goes back to HMS Dreadnaught, too. This is a serious piece of kit, if you know what I mean.

Contrary to the MP-DAC, however, it is not that much of a DIY job: the power supply and digital circuit boards come completed, and only a small analogue output board with a handful of components is left over to the handywork of the owner. Unless of course said owner got the DAC in possession by opting for the more expensive assembled version. Not that there is much need for the latter route, as the documentation delivered with the project is of exemplary quality, covering the circuit at length, offering step-by-step guidelines, and even a page with hints on the fine art of transforming lead blobs into quality electrical joints.


On over sampling

When a sampled signal, any sampled signal, is replayed over a DAC, the resulting analogue-domain signal spectrally contains the original information, say the band from 20Hz up to 20kHz, as well as an infinite series of copies (images) of that information, shifted up in the spectrum around half the sampling frequency fs and its integer multiples. With CD, and its 44.1kHz sampling rate, this means that there is music up to 20kHz, then a gap, then garbage from 24.1kHz to 64.1kHz, a gap, garbage from 68.2kHz to 108.2kHz, and so on. This is visualised in the the graph below:

Assuming that we can't hear above 20kHz, it is just fine to listen to such a raw output of a DAC. Our ear constitutes a perfect and sharp low-pass filter, or reconstruction filter, and all we would perceive would be a fine replay of the original sampled signal.

But our amplifiers don't like the spectral images above 20kHz. These can cause nasty intermodulation distortion which folds down into the audible band. And even if they didn't, it still would be an absolute waste of power trying to amplify the images.

Neither do our speaker's tweeters like the images. Tweeters are made to withstand power levels in the order of 5 to 10 Watts. This is fine, even with high-power amps, as the spectral contents of actual music are mostly lumped into the lower frequencies only. But if you confront a tweeter with a digital image, you hit it with all of the bass power of the original signal, magically up-transformed to a higher frequency. And it is not only the first image that hits, but also the second, the third, and so on. R.I.P. Tweeter.

So we really really have to filter the raw output of an audio digital-to-analogue convertor, not so much for sound quality's sake, as for the safety of our gear.

The first Japanese CD-players back in 1980 used sharp high-order analogue filter sections right after the DAC chips. Now it is a very tough job to make a high-order filter with analogue components, keeping an eye on overall cost, and hoping for a decent sonic transparency.

Philips, at the time not able or not willing to make a true 16 bit DAC chip, did things differently. Using oversampling, digital filtering, and noise-shaping they traded DAC resolution against time and hence avoided the need for a 16 bit part and for a steep analogue filter. The Philips machines - and the ones they spawned, notably the first Meridians and Missions - sounded better and the rest is history: quickly everyone adopted upsampling and digital filtering, inserting 1, or 3, or 7 interpolated values inbetween each pair of original samples.

But how does this all work? Say that you use 4-times oversampling. This raises the original sampling frequency from 44.1kHz (CD) to 176.4kHz. If you combine this with a sharp digital filter cutting above 20kHz, you can clean the whole band from 20kHz to 176.4kHz-20kHz or 156kHz from all nasties. All that is required after DA conversion would be a gentle analogue filter reducing the remaining components above 156kHz.

And everybody was happy for a while. The engineers because they had a seemingly simple solution to a real problem, and the salespeople because now they had a distinguishing specification in what so far had been a market of rather homogeneous products: the race of oversampling figures could begin. Philips with 4x and a true 16 bit DAC, Japan Inc. with 2x, UK-based Cambridge with 16x, Japan Inc. with 8x, then Belgian outfit Audio Discovery with a shocking 32x (8 TDA-1541 convertors in a massive 4-box CD-player looking not unlike Walker's Proscenium turntable), Wadia and Krell following later, and so on.

But wait a moment: all that oversampling and obscenely-high-order filtering is signal processing. And whenever you process a signal, you change it. Not necessarily for the better ...

Back to basics, thought some people a couple of years ago, and now we have a growing list of non-oversampling digital sources, combining the raw output of the DAC chips with a simple analogue filter only. Examples of such machines are the 47 Labs DA convertor, our own TNT Convertus, and the latest generation of Audio Notes. With Audio Note the return to non-oversampling ties in with their earlier efforts on transformer-based DAC current-to-voltage conversion: a transformer forms an interesting low-pass filter to be used with a non-filtered DAC.

La machina

Technologically, this DAC is about as simple as it gets. That is, the signal path is simple. The power supplies for the digital parts are endowed with some sophistication: three discrete regulators, built with opamps and pass transistors, and referenced from red LEDs, drive the Crystal CS8412 receiver and the Analog Devices AD1865 18-bit ladder DAC. There is one coax SPDIF input on BNC, and one AES/EBU input on an XLR connector, selected by a switch on the unit's rear panel. Outputs are unbalanced.

The current output of the DAC chip is passively converted into the desired voltage by just a resistor. This resistor's value ultimately determines the DAC 1.1's output voltage, and Audio Note provide information on how to modify the standard 2.5V by changing the conversion resistors. After this follows a CLC filter, its inductor ferrite-cored, and finally a common-cathode triode stage, delivering the necessary gain.

Since this is a no-oversampling DAC, that low-pass CLC filter must cut steeply, starting at a low frequency, lest the end-user fries his/her tweeters. I verified the response of the filter found here with Pspice, and the result is, relative to a 1kHz signal:

DAC 1.1 overview

frequency

relative level (dB)

1kHz

0

10kHz

+0.25

20kHz

-2

30kHz

-10

40kHz

-17

50kHz

-22

100kHz

-39

200kHz

-57

500kHz

-80

1MHz

-99

Observe the slight treble hump at 10kHz and the early roll-off at 20kHz, both probably of little consequence. As for the rejection of the potentially harmful images, anything attenuated more than 20dB can safely be ignored, as signal power is then cut with a factor of 100. The only source frequency band that is potentially troublesome here is the range 20Hz-10kHz, which is roughly reflected to the range 30-40kHz, where it is only filtered by as little as 10dB, meaning a reduction to 10%. In layman's terms this means that if your woofers are getting 100W, the imaging accounts for about 10W sent straight into the tweeters. In real life, though, these are often preceded with resistors, and their self-inductance precludes them from absorbing much power at higher frequencies, so in the end there is not that much of a problem at all.

With such a radical filter, though, it is worthwhile looking at its phase response too. The Audio Note people did a fine job, as the phase remains essentially linear up to 20kHz, after which it decidedly curves. This means that, rather than smearing fundamentals and their overtones by confronting each with a different 'time to get through', the filter tends to act as a constant time delay.

(Now just as an aside, the Hawk MP-DAC, with its highly-oversampled architecture, but also using passive CLC filtering, reads -3dB at 68kHz, and -73dB at 1MHz, with the overall phase response also remarkably linear. MP-DAC's bass is -3dB at 4Hz.)

Tube output circuit - click to enlarge

Yet, the DAC 1.1's design isn't entirely without vices. For starters, the output impedance from that tiny triode is a staggering 6.3kOhms, according to the manual (3.2kOhms according to Peter Qvortrup, which still is high). Further, the output coupling capacitor is a mere 0.47uF. Together these characteristics demand that the DAC always see an input impedance of over 100kOhms. Go too low with impedance, and curtailed bass, excess phase shift in the lower registers, squashed dynamics, and distortion are your share. So careful system matching will be your part, if you want to make the most of this convertor. Tube preamps tend to be fine, but many transistor inputs can go as low as 10kOhms, which would mean murder in this case!

Mods versus rockers

AN encourage the owner to replace the Crystal data receiver with the pin-compatible CS8414, which allows for reception of data up to 96kHz and at 24 bit (sigh) word lengths. Of course, while the DAC will happily convert these data, words will be truncated (rather than dithered) to 18 bit, and the passive filter will still cut above 20kHz unless you are prepared to re-design it.

Other modifications are possible too, even if the components specified for this entry-level DAC are not entry-level at all. Beyschlag metal-film resistors are used throughout, combined with Wima MKP capacitors in the filter, Elna Stargets for cathode bypassing, and finally Jensen paper-in-oils as output coupling caps. Audio Note themselves suggest an upgrade path from the basic 1.1: notably using tantalum resistors in a few critical positions, as well as going Black Gate or Elna Cerafine for the bypasses. This seems good advice. Moreover, I understand now that a Signature version is or will be made available, this one more in line with the DAC 5, employing an output transformer for AC coupling and reduction of the output impedance to a negligible factor.

But even more can be done: methinks that the Wimas are good candidates for replacement by something better altogether, think along the lines of Philips or LCR polystyrenes, or perhaps even Teflon caps, should you be able to procure them. And substituting for instance 2.2uF M-Caps for the Jensens will bring more load-tolerance in the bass department. Don't want to stop here? Given the vast amounts of empty space in the DAC 1.1's box, I would then suggest an all-new inductor-smoothed power supply with vast capacitor reserves for the advanced 'consumer' with a smoldering iron and a habit of, well, melting lead.

Now, if you're starting to be worried by these allusions to homedone alchemy and boiling cauldrons, please don't: I never said that the 1.1 isn't a complete DAC all on its own ... Just read on.

Audio hell?

Initially I teamed up the DAC 1.1 with my normal preamp, the Michell Argo/Hera. While renditions of vocals could be stunningly realistic, especially background vocals hitherto lost in the mix, overall performance was rather variable, and more, I really could not come to grips with the sound of this DAC: the words to describe it kept eluding me, except for the basic fact that I was NOT wowed by what I heard.

So out with the schematics and lo!, the Argo's CD input impedance was a neat 43kOhms, the other inputs going even lower. OK, 43k is only half of the 100k specified by Audio Note as a decent minimum, and as such still too high to cause obvious problems, I would think. But still, this was borderline, and some corrective action was needed as I am, in the end, very serious about reviewing. (And, no, the nice 47 Labs Gaincard resting chez nous was of no avail: that one is even 22kOhms and so it remained firmly in its box!)

How to proceed?

Try to lend another preamp? Possible. Come to terms with in a short while, enough to do a valid evaluation of the AN DAC, and then return it? Forget it. But hey, I had a bunch of selected JFETs, a nice 100k Panasonic for Audio potentiometer, big Elna Cerafine caps, a vacant breadboard, and some Erno Borbely-originated thoughts in my head. The ingredients of what could become one hell of a unity-gain linestage. Out with the soldering iron. Again ... Aah, the lengths we TNTers go just to please you, noble unknown audience. (I know I shouldn't complain. After all this AN DAC is supposed to be a kit, and so I simply had to solder something together.)

Listening to the As and Bs of DACs

With the new linestage operational and finally some time at hand, what better than conduct a nice AB comparison between the DAC Kit 1.1 and the Hawk MP-DAC reviewed last month?

Actually such a comparison makes utter sense, as both D-to-A convertors share some aspects of design, while differing widely in other aspects: the AN has a multibit DAC chip without any processing in the digital domain, the Hawk has a massively-oversampled delta-sigma heart; the AN has a simple tube output stage, the Hawk uses transistors, these too in a rather simple configuration; both of the analogue filters are passive CLC types, and both DACs devote more than average attention to the power supplies; to end, the two work entirely without loop feedback, and they source their digital data from the same Crystal receiver chip.

Some of the direct comparison sessions were precisely level-matched, of course, as the AN delivered an output of 2200mV into a 90kOhms load (315Hz, ref. 0dB), while the Hawk put down a slightly lower 1900mV. That's a 1.3dB difference.

Starting with Garbage's first CD, the Hawk MP-DAC delivered the music in an open, three-dimensional and positively cavernous soundscape, the electronic nature of the production very obvious, yet without a trace of hardness. Married to this was a nice, open and airy bass, although a slight lack in impact could be detected. The Audio Note DAC was smoother, less 3D although still nicely open, but overall somewhat too soft and too sweet, robbing the music from its industrial menace.

The next disc, something else altogether with Monteverdi's Vespra Della Beata Vergine set in the San Marco in Venise, gave rise to the same pattern: the Audio Note was fluid, quite large, and with nicely differentiated vocal lines. But the Hawk sounded huge, bold, with brassy brass, a more massive bottom end, and ultimately more alive. Then again, the Hawk also sounded coarser, a bit roughed-up.

On with Dead Can Dance's live album. Again the MP-DAC set down a large and powerful picture with a full bass, some roughness, and good vocals. But what happened next??? The DAC Kit 1.1 suddenly matched the MP-DAC in soundscape largesse, and although overall impact was slightly lessened, the Audio Note component better brought out the rhythms due to a more agile handling of the complex percussion in the first track, and to lightning-quick responses in the higher frequencies.

A first and careful summary would be that the Audio Note DAC excels with a very clean and 'real', though sometimes a bit thin, rendition of acoustical instruments, while its bass can lack some definition. Actually it all depends on the software. Take music driven forth by a deep and rumbling bass guitar, and the Audio Note has to let go of it. But if the track's fundaments are pitched a bit higher up the scale, then suddenly the AN becomes a rhythm meister, leaving the MP-DAC in its tracks.

As such Natacha Atlas' Halim betrayed the Audio Note, with a one-note-ish electronic bass and overall a slightly grey presentation. But percussion was very natural, and overall clarity and detail were superior to the MP-DAC, giving more insight in the humour of the track 'Ya Weledi': starting with a few bars of tango, moving over to sounds expected to be heard on Montmartre, then a genuine Western guitar, only to eventually burst out in arabica. All of this in perhaps 30 seconds. The MP-DAC was less precise, though more organic, while the analytical DAC 1.1 showed more contrast in those cross-bred musical styles.

With Introducing Ruben Gonzalez both DACs reached equilibrium again, each in its own fashion: the Hawk rounded, glowing, now grainless and with subtle brushwork and above all a powerful rendition of the piano as a large and wooden instrument. The AN didn't go that far, but excelled with more detail, better neutrality, and what the other did to piano, this be-tubed DAC did to trumpets.

On Gidon Kremer's Hommage a Piazolla the Audio Note took the lead. The solo violin was sweet, full bodied, with the strings as well as the wood audible. The overall impression was one of lushness, and this not at the cost of dynamics or clarity, which both were very fine indeed. Likewise Clannad's Nil Sen La, a live recording, got handled with agility by the Audio Note DAC, with fine detail, and just as much space as the warmer MP-DAC.

It took the lead, I said? Only to loose it again. Think of an endless Schumacher-Senna duel. The tapestries of synthesizers in the background of Loreena McKennitt's The Visit were more like, well, tapestries with the Hawk, while the AN lost out on drama and emotion on Greensleeves. Conversely, the Mary Black CD proved for once and for all that the Audio Note did have superior treble.

Almost by accident I fired up my Dynaco ST-35. It has recently gotten new tubes, but no further modifications ... yet. Now it allowed the DAC 1.1 to play 1) in an almost all-valve system, and 2) a system with dramatically reduced complexity (what with one triode, one JFET, two triodes and finally two pentodes as active elements? only five stages from DAC chip to loudspeaker terminals!). This brought definite gains in presence and immediacy, width, and breadth, even when the ST-35 - as is - is not the most neutral amp around. All in all this makes me think that the Audio Note DAC Kit 1.1 has been tailored with low-powered tube systems in mind, and much less for hyper-analytical transistor ones.

Conclusions

As you remember I called the Hawk MP-DAC one of the most musical digital devices I had encountered so far. The AN matches it, but with a wholly different character. The Audio Note 1.1 sounds smaller, is slightly limited in width and in depth, and sometimes a bit cooler. But on the other hand, against the MP-DAC's texture and slightly dark colourations, it is tonally purer and smoother, more neutral, with a treble quality unheard of from digital sources, with a more delicate and precise character, and with lively dynamics and agility. (Now I hear them Hawkpeople muttering I could not possibly judge the MP-DAC with my low standard of transport, but then I would reply the same Planet was in use for both DACs!)

Audible differences left aside, the DAC 1.1 is more finicky with respect to system matching. It really needs a high input impedance to deliver in the bass department. It also prefers simple tube amplification, eliminating as much devices as possible between the source and your ears; I can imagine that people wanting to use this DAC to bring some tube-warmth to a dried-out and thin solid state system may be left wanting. I came to this conclusion myself, but reading the few reviews of this DAC on Audio Asylum as well as later discussions I had with AN's upper note Peter Qvortrup point in the same direction.

The differences between both DA convertors are not huge, they are more in style of presentation than in absolute level of quality, and this is definitely a situation where things like personal taste and system matching come into play. Yet, as I heard it, I would give the edge to the Audio Note DAC Kit 1.1: it sounds lively, is well-built, and offers the jaded DIYer enough room for further experiments!

Now, does all of this mean that any of these DACs performs at the level of a Truly Great Turntable?

Of course not.

But do they make listening to CD worthwhile?

Yep, they do.

And thus endeth one of the toughest review periods I have ever had ...

(Shakespearean slant to honour and remember the late great Sir John Gielgud, and, curses, I forgot to play my Prospero's Books CD!)

System used

  • CD-transport: Rega Planet

  • linestages: Michell Argo/Hera, homemade unity-gain singled-ended zero-feedback class A JFET (...) preamp

  • power amps: Quad 306, Dynaco ST-35 (briefly)

  • loudspeakers: Quad ESL-57 on stands, 1m from back wall

  • Cables: Sonic Link digital, Audio Note AN-V interconnect

© Copyright 2000 Werner Ogiers for TNT Audio Magazine (http://www.tnt-audio.com)

[ Home | Staff & Contacts | HiFi Playground | Listening tests | DIY & Tweakings | Music & Books ]