When the first standalone digital-to-analogue converter was launched back in the late nineteen eighties, the hi-fi world was aghast. CD players were still in their infancy, and the idea that you could split the disc reading mechanism from the digital conversion section and analogue output stage seemed the stuff of science fiction dreams. Over a quarter of a century later, this is now completely commonplace of course, but strangely there’s still an air of mystery surrounding hi-fi DACs.
In truth, making a modern digital converter is nothing to write home about. They can be assembled cheaply from powerful yet affordable off-the-shelf silicon chips, and this is why the consumer audio market is now so full of such products. Indeed, it’s so straightforward that according the dCS Technical Director Andy McHarg, “making a DAC is quite a lot easier than making your own turntable, actually.”
They’re simple to make because all the hard work has been done by the chip manufacturer – so much so that no specialist digital audio knowledge is needed, just basic electronics. This explains why almost all commercially available digital converter manufacturers use someone else’s technology, lightly repackaging it with their own box and branding. Suppliers such as Burr Brown, Crystal Semiconductors, ESS and Wolfson Microelectronics provide a free circuit with every DAC chip they sell. “This evaluation board,” explains McHarg, “has all the schematics done for you, so you can take the chips, lay them out, put them in a box with a power supply and you’re away.”
Silicon suppliers provide evaluation boards simply to showcase their chips. It’s a full circuit to which you simply add a digital input (for example, an S/PDIF or USB interface) on the front end, and an analogue output stage downstream. Available on eBay for as little as £70, the only additional costs are casework, socketry and a power supply. Rather like building a home-made PC, no knowledge of the complex processes that go on inside the DAC chip itself is needed to make such a thing work, it’s just a case of hooking it all up, just as the DAC manufacturer intended. As the chips have evolved, life has got easier. For example, the latest ESS Sabre chip doesn’t even need an external S/DPIF receiver chip – it comes built-in along with the digital filtering too.
Simple as this may be, dCS Director of Product Development Chris Hales explains that this approach can only get you so far. “You can easily draw up a circuit that will work in broad terms, but it requires careful thought and a fair amount of expertise to understand how to get the most out of the components. It’s not just a case of joining the dots and everything working perfectly. For example, there are certain signals which need to be close to each other, others which need to have short paths, and some which need to be distanced from others, and so on…”
In other words, it’s easy to do a decent DAC – but not a great one. Most hi-fi manufacturers feel the need to improve on the chip maker’s basic schematic, but find they’re not able to do take things much further. “Putting the evaluation board in a fancy box,” explains McHarg, is just a bit too obvious, so people take the reference schematic and usually muck about with the output filtering a bit. Generally they don’t do much more than that…”
Do It Yourself
“If I was going for a quick solution,” says Andy McHarg, “I would probably get a brand x S/PDIF receiver and a brand x DAC*, because the two were designed to work together. Most DAC chips have a digital filter built in, and you could use the reference analogue output stage specified in the application notes. And then you need some sockets from RS Components and a power supply – or you could even power one of these things off a USB port.”
With more time and budget, he says, “the obvious thing to swap would be the analogue output stage, because – if things aren’t right elsewhere – you can make a dramatic change to the sound just by shoving some tubes in there, for instance. Of course at this stage it depends on what you believe; you might want to have a super low feedback path or no capacitors for example – whatever your particular religion is, and how you want to market it…”
If most manufacturers design DACs in such a similar way, then why don’t they all sound the same? Chris Hales points out that this is down to the tendency of many manufacturers to do the detail wrong. “It’s possible to get good results from off-the-shelf chips,” says Chris Hales, “but for every company that has a professional and meticulous approach there’s another with the opposite. These parts can potentially perform well, but you have to apply them intelligently to take care of all those sensitive signals.”
Another common tweak is the clock, explains McHarg. “The S/PDIF receiver generates the clock for the DAC, and that clock is derived from the S/PDIF input, but a better solution would be to use someone else’s S/PDIF receiver and use a phase-locked loop. You can then use that clock for the DAC, or have a local oscillator that’s fixed frequency and feed the USB board on the DAC.”
Most designers add asynchronous USB at this stage too, although it costs extra for the USB receiver chip. “The USB input is a good one because you can buy the XMOS kit that will give you an asynchronous USB input. You would have to tweak about a bit with the clocking; the ideal scenario would be that you feed the XMOS with the DAC master clock. So you might need a bit of trial and error there, but it’s all fairly trivial glue logic”, adds McHarg.
For most manufacturers, the digital filtering is best left well alone. The semiconductor makers provide very reasonable sounding off-the-shelf solutions and most designers are reluctant to risk doing this worse. Digital filters have to be absolutely linear, so as to avoid the need to change the analogue filters later to compensate – this is intellectually lazy and introduces the risk of the performance varying over time as components age and operating temperatures vary. McHarg adds, “to put it bluntly, this sort of tweaking should be designed out of all DACs.”
Some will also try to attempt their own digital volume control, again with varying success. Any DAC should always perform this as late as possible in the processing chain, to give the most headroom for the filters. “Attenuating the signal before it enters the filters is a bad idea, as you are potentially reducing the input in terms of dynamic range”, says McHarg. “A volume control is typically one multiplication of many internal to a DAC, so sensible design would have 0dB having the best performance in terms of dynamic range possible.”
The choice of DAC chip that manufacturers use is often overplayed by the hi-fi press, what’s really critical is how well it is implemented. “The particular chip you’re using will have certain characteristics,” says Hales, “and clocking is a case in point – if you have a noisy or jittery clock then that’s going to affect the performance, and equally if your board is laid out in such a way that noise can get injected onto the clock signal between where it’s generated and where it’s applied to the DAC chip then that’s the same thing. Whether you’d call it an art or a science, there’s certainly a degree of expertise that’s involving in applying an off-the-shelf solution. Experience counts for a lot.”
If you’re sufficiently smart to get your chip to give of its best, then you’ll finally be able to drill down into the design of the DAC chip itself. Andy McHarg explains that each manufacturer chooses different trade-offs. “The brand X approach has quite low out-of-band noise; they don’t use much of a noise shaper but actually the distortion performance isn’t very good, and the brand X S/PDIF receiver has quite a lot of jitter on it. The brand Y chip is impressive thanks to its low distortion and low noise, but if you look at a brand Y DAC at 20kHz you always get out some aliasing products within the audio band at -100dB or so due to the way it does its filtering. The other thing is that it’s very power supply-dependent, so if you have any problems with this then it will show up somewhere.”
Brand Z chips are another popular and affordable option. “These are somewhere in the middle; they work pretty well and give reasonable distortion and reasonable noise. Again, you can run them in dual-differential mode so you can shove them together; you take two DACs fed the same signal, and sum them and then you halve the output, then the signal is always the same between the DACs but the noise is different, giving you potentially a 3dB signal-to-noise ratio improvement”, explains McHarg.
Take the long way home
For manufacturers acutely aware of cost constraints, using a bought-in DAC chip, following the supplier’s circuit suggestions and doing some light tweaking is the easiest way to market. Commercially, this approach makes a lot of sense, which is why almost every manufacturer does it this way – except dCS.
Instead, at the heart of every dCS digital converter is the Ring DAC – a bespoke digital signal processing engine designed by and unique to dCS. Instead of buying in other manufacturers’ DAC chips, the Ring DAC is a clean-sheet solution running code written and regularly refined by dCS engineers. It uses a network of FPGAs (Field Programmable Gate Arrays) programmed to run dCS firmware that does all digital filtering and digital-to-analogue conversion.
This system is deliberately over-specified to ensure superior performance to all off-the-shelf chips, with flexibility for the future built-in. “Ring DAC technology still outperforms all competing semiconductor solutions,” says Hales. “It betters them significantly in noise and distortion, and because we have ultimate control of everything we can refine the design and make improvements as we go along. Because we’re in control of the digital signal processing we can improve it. For example, the mapping algorithm – which is pretty crucial to the way the Ring DAC works – is something we change if we see fit. The filtering is completely under our control, so we’re not at the mercy of what’s in the chip that you buy, we can change it arbitrarily to whatever we want. We can even do that retrospectively and put that into products already into the field.”
Furthermore, by doing everything its way, dCS doesn’t suffer the vagaries of other manufacturers’ quality control issues, explains Hales. “You can buy in off-the-shelf chips and test them very rigorously, and it could well be that some don’t meet your requirements so you have to discard them, chances are that if you try to reject them on the supplier they’ll tell you they’re performing within spec, so hard luck! So you can weed out the particularly good parts and use those and get rid of the others. But the dCS approach means that if we don’t like the way something is performing then effectively we can make a change at the most basic of levels.”
“We’re doing it the hard way because we think it’s better,” adds McHarg. “There are fundamental performance advantages to doing it our way. Coupled to that, we have a culture here of measurement, inspection and highly rigorous design so when we see things happening that we don’t like the look of, we investigate them. That’s another thing that sets us apart.”
Chris Hales adds that Ring DAC technology is highly upgradeable. “Although the basic architecture remains the same on the new flagship Vivaldi, there are quite a few enhancements; the core of the Ring DAC circuit itself has changed significantly and there are some things around the edges which have been improved and refined. We’re constantly looking at test results that come off production, searching for things that can be improved. Across the whole dCS range, regular firmware updates bring additional functionality and/or sound quality.“
Doing a digital converter the dCS way is far more expensive than buying in off-the-shelf DAC chips from semiconductor suppliers, and requires a wealth of additional knowledge accumulated over a quarter of a century. But the results speak for themselves; the reason dCS DACs sound unlike everything else on the market is because they’re built in a dramatically different way. They always have been and always will be.