Network Audio Players/Servers

To the list of brands that have the wherewithal to code a custom DAC and combine it with a novel hardware solution – Chord Electronics, dCS, Meitner, Mola Mola, MSB and PS Audio among them – joins Grimm Audio. Its boldly titled ‘Major DAC’ is no conceit, for its performance earns its place on the front row of the digital parade ground. Like many of its officer comrades, Grimm Audio has opted for a high-speed processing solution, combining upsampling, digital filtering, truncation and noise-shaping to, in this instance, derive a PWM datastream that addresses a matrix of ‘bit converters’. The precise implementation is, of course, unique to the Major DAC. Calculators on standby...

All digital inputs are subject to 8x integer upsampling/digital filtering at an elevated wordlength – for example 352.8kHz and 384kHz with 44.1kHz and 48kHz sources, respectively, at well above 32-bit resolution. Next comes a series of upsampling stages, to 128fs, in concert with a truncation of the wordlength to 1.5 bits supplemented by a single-stage 11th-order noise-shaping. The latter is vital if an adequate S/N is to be recovered across the audioband from the ~9dB remaining after the wordlength is reduced to 1.5 bits.

Grimm Audio’s literature says 1.5 bits, but the truncation is mathematically closer to 1.6 bits (or 1.59 bits if we are being especially picky). Why? Because a 1.6 bit wordlength is required to define three discrete pulse widths per (up)sample, each width describing one amplitude level. This 1.6-bit/128fs (5.6448MHz and 6.144MHz) bitstream then addresses not one but 16 parallel current sources (1-bit ‘DAC cells’) – each handling the three widths – at 512fs. Intriguingly, Grimm applies a minuscule time offset between each DAC cell to provide an FIR filter-like advantage when the output signal is summed in the analogue domain, further suppressing a deal of this noise-shaped ultrasonic hash before the audio signal reaches the final (4th-order) balanced output filter. PM