Robert Harley Explains the dCS Ring DAC

The Ring DAC, invented by dCS in 1987, is a brilliant solution to the challenge of converting digital data to an analog output signal. It is particularly well suited to high-resolution digital audio.

To understand the Ring DAC, let’s first consider how a conventional multibit DAC works. You can think of a multibit DAC as a ladder, with as many rungs on that ladder as there are bits in a sample. A 24-bit DAC will have 24 “rungs,” each one a resistor that corresponds to each bit in the digital sample. The resistors are connected to a voltage source through a series of switches; the digital data representing the audio signal open or close the switches to allow current to flow to the output or not. The currents of each rung are summed, with that summed value representing the audio signal’s amplitude.

The arrangement of the resistors and the voltage source result in a “binary weighting.” This means that each resistor lower down on the rung must effectively have double the resistance of the rung above it, and so forth, corresponding to the binary progression 1, 2, 4, 8, 16, and so on. In practice, only two resistor values are used; the resistor ladder forms a voltage divider that reduces the output voltage by a factor of two for each successive rung.

One problem with these so-called “R-2R ladder” DACs is that it’s impossible to make resistors with the precision required for perfect binary weighting. The result is that the tolerances in resistor values introduce amplitude errors in the analog output. Moreover, those amplitude errors will occur in the same places on the audio waveform. Compounding the problem, the errors are a greater proportion of the signal at low levels.

This problem becomes more acute the greater the number of rungs on the ladder. In a 16-bit resistor-ladder DAC, the output voltage of the least-significant bit (LSB) should be exactly 0.0000152 the value of the most significant bit (MSB). In a 24-bit converter the LSB value should be precisely 0.000000119209289550781 the value of the MSB. It is obviously not possible to achieve anywhere near this level of precision in resistor manufacturing. Any deviation from the precise resistor values, in any resistor in the ladder, translates to amplitude errors in the analog output.

The now-defunct UltraAnalog company addressed this challenge by driving its 20-bit DACs (which were composed of two off-the-shelf 16-bit DACs ganged together) with 100,000 different digital codes, measuring the DAC output at each code value, calculating the degree of error in each specific resistor, and then having technicians hand-solder tiny precision metal-film resistors on the ladder rungs to bring them closer to the correct value. I visited the factory and saw this heroic (and expensive) approach in action.

A DAC technology that doesn’t rely on binary-weighted resistor ladders is the one-bit DAC. This device converts a multibit code into a single-bit datastream that has two values, one and zero. Unlike a multibit DAC, the one-bit DAC’s amplitude precision is very high, but the one-bit DAC suffers from very high noise that must be “shaped” (shifted away from the audio band). One-bit DACs are also very susceptible to jitter.

dCS’s solution is the Ring DAC, which can be considered a hybrid of the two approaches. It is based on a five-bit code that drives resistors of identical value. Because the resistors in dCS’ Ring DAC are all the same nominal value, their actual values are very close to one another. The five-bit code has a much higher signal-to-noise ratio than a one-bit datastream and requires an order of magnitude less noise shaping.

Digital signal processing first “maps” whatever datastream is coming in (192kHz/24-bit or the 2.8224MHz 1-bit code of DSD, for examples) into a unique five-bit code. This five-bit code opens and closes one of 48 latches connected to a current source that drives one of five resistors of identical value. Because these resistors can never have exactly the same resistance, the Ring DAC employs an array of resistors and randomly shifts the audio signal between resistors in the array. The Ring DAC gets its name from this “passing around” of the signal from one resistor in the array to another, as in a ring. The effect is to convert what would be amplitude errors in the analog output into a very small amount of random white noise that is uncorrelated with the audio signal.

dCS’s latest version of this evolving technology, the Apex, is based on the same principles, but with a more advanced implementation. The current source, the summing stage, the filter, and the output buffer have all been redesigned in the Apex. In addition, single transistors in the Ring DAC have been replaced by compound pairs in the Apex. The circuit board layout has been optimized. The result is a DAC that is quieter than previous generations with more than 12dB greater linearity.

The Ring DAC is brilliant in concept and is executed at its highest realization in the new Apex. The commonality in sonic character between all dCS products—the density of information, the resolution of fine detail, the unique spatial qualities—are probably attributable in large part to the Ring DAC.