HifiZine
The enthusiast's audio webzine

Mid-priced DACs – how to choose

Back in November of last year, I wrote a piece for Blogcritics called “Happy Holidacs,” in which I surveyed what I consider to be mid–priced Digital–to–Analog Converters or DACs. That original survey has continued to grow, and when I mentioned the current version with its eighty ninety–some entries to editor John Reekie, he opined that it was “mind boggling.” That got me thinking…

If John, a quite technical guy and dyed-in-the-wool audio geek, found it a bit overwhelming, what would you, gentle reader, think? This month’s ramblings are dedicated to those of you who are either considering the addition of digital componentry to your rig, or are considering an upgrade.

These days, DACs come in a wide variety of packages and price points, with a broad range of features and sometimes questionable benefits. Since, with any transducer, you truly get what you pay for, let’s start with price. Yes, there are US$200 DACs that do an OK job, and there are $400 DACs that are quite good. That said, the sweet spot for me is from $500 to $2000, where you’re purchasing really good performance and some very desirable features, all without emptying your bank account.

In general, throwing more money at the problem buys you better build quality and/or features. More than any other factor, the cost of a product’s bill of materials, all the stuff that goes into its manufacture, determines the selling price. Building with better stuff means higher cost of finished goods.

Better build quality can mean higher quality components, such as:

  • metal film capacitors and low noise resistors and potentiometers a.k.a. “pots”
  • a stepped attenuator instead of a pot
  • a NOS DAC chip or multiple DAC chips
  • audiophile op–amps or discrete Class A circuitry
  • advanced grounding and noise isolation techniques

All of these usually contribute to better overall sound.

Let’s look at “features,” which might include:

  • fully balanced I/O and/or headphone output
  • the addition of a better quality headphone amp, either a discreet design or one based on “audiophile” monolithic op–amps
  • DSD or extremely high sample rate (8 or 16x) capabilities
  • analog input(s) for a pre–amp function
  • a low noise, linear power supply
  • a spiffy display and/or meters

If you, as I do, make frequent use of high-quality headphones, then a better than average headphone amplifier is much appreciated. Due to the wildly varying loads that different headphones present to an amplifier, you really have to try the amp with your own cans. Speaking of stylin’ headphones, if yours accept a balanced input then that feature may be important to you.

Very high sample rate support is certainly interesting but, at present anyway, I find it a hollow promise. Yes, there are labels recording in DXD (see below), but that’s because DSD is such a pain in the tokhes to edit and EQ. Native DSD content for sale is, and will continue to be, a tiny minority of available content. In this day and age, 192 kHz is a perfectly fine maximum sample rate. In fact, I can really only recommend purchasing a new DAC if at least one of its inputs is 192k–capable.

Build quality is quite subjective. I get all excited about well routed circuit boards and nicely fabricated switches, because the former contributes to quieter performance and the latter to longevity. Unfortunately, attention to build quality can also focus on bling: thick, machined faceplates and stupidly oversized heatsinks; aesthetic industrial design; and hollow state for the sake of glowing tubes protruding from the chassis, not for better fidelity. You may scoff, but a study by colleague and audio geek Jon Boley and friends (http://lsbaudio.com/publications/AES128_AV_SDT.pdf) correlates visual quality to subjective audio quality. I would take his findings further by stating that visual appeal unconsciously influences our perception of audio quality as well. In short: more bling = more tendency to like the sound without recognizing the bias.

Other aspects to consider while weighing your purchase choices are support for common sample rates, number and type of inputs, USB Audio (Device) Class 2 (UAC2) compliance, defeatable upsampling, remote control and, for The Old Cheese, the form factor. These ancillary aspects can be quite significant. If you don’t have a computer set up for “computer audio,” never mind a FireWire–equipped computer, then you can ignore USB and IEEE 1394 inputs. If you don’t have a digital source that outputs I2S, a very high quality digital interconnect with separate data and clock also known as “I2S,” then you can ignore that too. Another consideration is sample rate support… I’m not talking about 352.8 and 384 kHz DXD, I’m referring to the AES standard sample rates; 44.1 and 48 kHz, and the 2x and 4x higher rate versions of those. Any modern DAC these days should support 192 kHz on one of its inputs. Beware of the 2 x 44.1 and the 4 x 44.1 sample rates, 88.2 kHz and 176.4 kHz respectively. Though only very quality conscious engineers track at these rates†, a handful of DACs don’t support them. My otherwise groovy little NuForce Icon HDP is one such critter, as it doesn’t do 88.2, and Arcam’s irDAC is another, not handling 176.4 kHz as an input sample rate.

I mentioned USB Audio Class 2–compliance, which allows plug and play operation at up to 192 kHz sample rates. Mac OS 10.6 and newer supports that standard, as does Linux, while Windows does not. As such, a Windows host requires that drivers be installed for proper operation of a Class 2 DAC.

Before we move on, a brief word about DXD or Digital eXtreme Definition, which is simply 24-bit linear PCM sampled at 8x. So, 176.4 kHz x 2 or 352.8 kHz and 192 x 2 or 384 kHz. Merging Technologies is the chief, er, only proponent of DXD technology in the pro world. As with the other superspeed tech, DSD, there are a handful of DACs that can run at those speeds; five in my mid–priced DAC table and, for crazy expensive, MSB and Playback Design come to mind. In the pro world, Digital Audio Denmark and Merging make the only DXD converter sets that I know of. For mere mortals, TEAC’s $1200 UD-501 supports DXD. Despite the similarity of their names, DSD and DXD are very different; DSD is one bit sampled at 64x 44.1 kHz or 2.8224 MHz, that’s MegaHertz as in a million times a second, while DXD is a 24-bit word sampled at 8 x 44.1 kiloHertz. So, one bit very fast versus 24 bits not nearly as much.

Sorry for the diversion… Some DAC designers pride themselves in using “NOS” chip sets. I’m not talking “New Old–Stock,” I referring to “non–oversampling.” Oversampling is a data conversion technique that employs a sample rate higher than that required by strict sampling theory. Unless you build a discreet digital–to–analog converter circuit out of individual parts, “discrete” as opposed to a chip–based “monolithic” design, you’ll be purchasing an oversampling DAC chip since, to my knowledge, they are the only type currently manufactured for audio applications. That said, there are those who feel the old skool, non–oversampling method of conversion is superior, and so usually employ new old–stock, non–oversampling (NOS NOS(?!)) converter chips. An interesting exception is Metrum Acoustic’s Octave, whose modern NOS design tops out at 176.4 kHz.

After all this raving about speeds and feeds, what’s a poor audio geek to do when it comes to deciding on a new DAC? You weigh all factors, with the majority consideration often going to cost. To that I’d add sample rate and input types. That will narrow down your choices significantly. Do I have any recommendations? Not yet as I have not heard most of the DACs in my table and, of the ones I have, it’s been at shows where I only have a few minutes to listen, with unfamiliar gear in a less than flattering room. Take heart though, I plan on cherry picking some of the standouts for in–depth listening and review. So, please stand by while I marshal the forces… I’ll be back with more DAC madness!

† Sidebar: A Sampling of Sampling Theory

Sample rate conversion, or SRC, is the process of changing audio sampled at one frequency into another, say 96 kHz into 44.1 kHz. A high rate to a lower rate is called “downsampling,” while going from a lower to a higher is “upsampling.”

Rant enabled: One moment please while I don my Nomex hat. Ahhh, I feel flameproof now…

On–the–fly SRC at a non–integer ratio is quite deleterious to quality and should be avoided at all costs. Out–of–realtime SRC processing is less funky since, unlike a realtime process, as much time as needed can be used to perform the computations.

Now that we have the background in place, let’s talk about those 2x and 4 x 44.1 rates… savvy engineers record and produce music for distribution on a variety of platforms, the most common being Compact Disc. So, it makes sense to start a project with the aim of producing a great sounding CD. If you’re recording at high data rates, it also makes sense to sample at an integer multiple of 44,100 Hz because that’s the CD sampling rate. 48 kHz is the standard for video distribution; DVD, BD, DTV and the like. Audiophiles and engineers know that sound is the red-headed stepchild of picture so, though audio is very important in a film and video world, it is frankly secondary. So, better to track a pure music project at 88.2, 176.4 or 352.8 k if you’re doing high rez, and convert to all your target distribution formats, than to record at some multiple of 48 k, and have to perform the hairy SRC ratio that is the difference between 44.1 and 48. Trouble is, recording “engineers” usually pick 96k because it’s more than 88.2, so “It must be better!” Usually, they’ve also never bothered to do their own SRC listening tests!

Before I disable rant mode, one last thing… As I mentioned, upsampling is the process of converting from a lower to a higher sample rate. Rather than tossing out samples and smoothing over the discontinuities, as is the case with typical downsampling, upsampling sticks invented samples into the data to “pad it out” to more samples per unit time. Upsampling is one of those features of player software and DACs that calls into question the quality of the anti–image filter(s) in a DAC. If your DAC “sounds better” with upsampling operating somewhere in the chain, your DAC is probably either a bit long in the tooth or, you purchased a DAC with a less than optimal filter design.


Readers' comments

    Just taken delivery of an LITE DAC-AH NOS and I have to say it’s been a total revelation.
    I have tried a few dacs, all with oversampling delta sigma architecture, at one brief stage I had a 384 khz oversampling dac which was very brittle sounding, to my ears none compare to this, this is just musical, the bass is incredible and there seems to be a realism, 3D solidity and a lack of “edge” to the sound that I have not heard from digital before. I am using my new NOS dac with foobar/sox x2 upsampling>hiface and an Indeed tube buffer for hi-fi PC sound, its working so well in my set up that I don’t think I will modify it, although I may fit a filtered mains IEC inlet.

    I will put my neck on the block and say if you do not have native 96/176.4/192-24bit type files, stick to NOS, the less oversampling/filtering of the digital signal the better if what I’ve heard is even only partially due to NOS. If I were being cynical I would have to say oversampling seems to have been a marketing ploy because if what I’ve heard is typical of NOS dacs in general then I can then think of no other reason to use it.

    • Hi 405line,

      Thanks for dropping by…Like any technology, technique or topology, blame the designer. Oversampling in and of itself is no “better” than NOS. In the hands of a talented engineer or team, each can really shine. The converse is also true; epic fail is just as likely. I like my DACs to play it straight; not messing with the sample rate of the source. That’s partially dogma, part philosophy, and part experience.
      That said, our hobby comes down to the fundamental, “What do _you_ like?” After all, it’s about enjoying your music.

      All the best,
      OMas

Leave a Comment