As chief engineer, Søren Kristensen is the technical guru at Soekris Audio. While his main business focus for many years has been designing and manufacturing embedded communication computers, he has always had a strong interest in high-end audio. Two things jump out immediately about the Soekris line of DACs. First, they’re all discrete R-2R sign-magnitude designs. In particular the dac1421 features a 27-bit precision ladder built with over two hundred 0.02%-tolerance thin-film resistors. Second, all Soekris products are designed and built in Denmark—no “designed in the West and built in the East” paradigm here—and yet the product line is remarkably affordable.
The R-2R sign-magnitude DAC technology was developed by Burr-Brown in the early 1990s to address the zero-crossing problem of a conventional R-2R DAC when one-half of the bit-switch circuits turns off and the other half turns on. This degrades signal-to-noise ratio of small signals due to the large noise contribution of switching transients. A sign-magnitude architecture takes a rather clever approach by actually starting at zero crossing and either adding or subtracting current from the resistive ladder to obtain the analog signal. On average, the smaller step sizes result in less noise and greater precision at low levels. To do this requires two internal DAC sections and the splitting of the binary bitstream into positive and negative polarities. The first bit of each binary word is designated as a sign bit with “1” denoting positive polarity and “0” a negative polarity; the rest of the bits specify an amplitude. One DAC section converts positive amplitudes while the other converts negative amplitudes. In the end, the output currents are summed to give the final analog signal. [See the sidebar for more technical details.—RH]
Søren tells me that he always liked the Burr-Brown series of sign-magnitude DACs, and that during 2014 he was thinking about designing a DAC for himself as a side project. It turned out that the PCM1704 chip, his first choice and widely considered to be the best, was designated as end-of-life by Texas Instruments, the parent company of Burr-Brown. With future production in mind, he decided against the PCM1704, but the problem was that weren’t any other similar chips available. Necessity being the mother of invention, and inspired by MSB Technology (at the time the only manufacturer of discrete sign-magnitude R-2R DACs), he set his mind on doing a discrete design. That ended up being the dam1021, the first low-cost R-2R DAC module for the DIY market, which is still being manufactured and sold today.
You may wonder, as I did, how Soekris was able to pull off a discrete R-2R DAC at an entry-level price of $899.One reason had to do with the availability of reasonably priced, precision (0.01%), surface-mount resistors from suppliers such as Digikey, which meant that no matching was necessary. Søren believes that one reason the PCM1704 was designated end-of-life was due to the expense of being manufactured on an old process line requiring laser-trimming. It’s also a question of smart engineering and manufacturing, says Søren, and personally taking charge of every operational aspect. The main PCB has been optimized for automatic manufacturing, using as much surface-mount technology as possible, and, hence, no wires. Parts were selected on a cost/performance basis rather than on audiophile reputation. All high-precision resistors in the R-2R network are highest-quality Vishay, but not better than needed. All important connectors are high quality and gold plated. Søren is not afraid to use parts some might object to for anecdotal reasons, like NP0 ceramic capacitors, MOSFET relays, and switch-mode power supplies. Costs are also being held down by using a direct online sales model, with the maximum of a single reseller between the manufacturer and the end user.
There are two SPDIF inputs (RCA and BNC), as well as a USB input, selectable from the front panel. The USB input is based on a standard XMOS interface chip, with the firmware modified to allow switching between Class 1 and Class 2 USB audio, the latter requiring USB 2.0, which is pretty standard these days. All of the digital inputs are routed to a field-programmable gate array (FPGA), which buffers and reclocks the bitstream and performs anti-aliasing filtering. It is followed by a digital volume control, whose output is clocked by an ultra-low-jitter oscillator driving the sign-magnitude R-2R resistor networks. The DAC can also accommodate a DSD input signal, which is internally converted to PCM by the FPGA.
The resistor network output voltage is amplified and buffered by a high-speed, all-discrete, zero-negative- feedback amplifier, operating in high-bias Class AB. The output stage can be switched to either line or headphone output, using MOSFET relays. When driving headphones, the amplifier is set for higher Class A bias current and gain. A crossfeed circuit may also be activated during headphone listening to control soundstage size.
There are four anti-aliasing filters that may be selected via the “Filter” button on the front panel. This gives the user enormous flexibility in tailoring the DAC’s sonic signature to suit particular tastes and system needs. The filter chosen is indicated by the LED color. Red is a sharp linear-phase filter, equivalent to a classic “brickwall” filter. Orange is a mix between linear- and minimum-phase filters. Green is a reasonably sharp, short-delay, linear-phase filter. And finally, the LED off indicates a soft Butterworth minimum-phase filter, which comes close to no oversampling (NOS).
By Dick Olsher
Although educated as a nuclear engineer at the University of Florida, I spent most of my career, 30 years to be exact, employed as a radiation physicist at the Los Alamos National Laboratory, from which I retired in 2008.More articles from this editor