Network Audio Players/Servers

Grimm Audio's innovative MU1 is not only a network-attached music storage device but, like the recent Innuos Statement [HFN Jan '20], Melco N10 [HFN Jun '19] and Roon Nucleus+ [HFN May '18] it also includes proprietary signal conditioning. In this case the 'conditioning' – upsampling and/or downsampling to 2x or 4x the base 44.1kHz/48kHz sample rates – is specific to the balanced AES3 digital outputs. The USB 2.0 output runs directly from the onboard Intel NUC 8i3BE 'mini computer' that hosts the SSDs and provides all the peripheral digital housekeeping. The AES3's proprietary resampling/re-clocking routine is executed separately on a Xilinx Artix-7 FPGA.

As the music data remains in the digital domain in and out of the MU1, and we are interested in any differences between the USB and 'conditioned' AES3 outputs, then this can only be inferred via a third-party DAC. This is complicated by any further jitter suppression/galvanic isolation/re-clocking within the test DAC, so the 'best' may not express a significant difference.

That was certainly the case here as our three AES/USB DACs – dCS Vivaldi One [HFN Feb '18], Mola Mola Tambaqui [HFN Nov '19] and Mytek Brooklyn [HFN Aug '17] – showed no significant difference in either distortion or A-wtd S/N via either USB or AES3 streams – right down to 0.00005-0.00007% (0dBFs to –30dBFs) and 118.5dB, respectively, for the Tambaqui. There were minuscule changes in jitter revealed by each DAC [Tambaqui, Graph 1] amounting to mere psec. The real explanation for the difference in sound between the MU1's USB and AES3 outputs lies in the first-stage 4x upsampling/filtering applied to the latter – a very high tap linear-phase filter revealed by its extended pre/post ripples [black impulse and response, Graph 2]. By contrast the unprocessed USB output invokes the slower roll-off linear phase filter inside the Tambaqui [red impulse and response]. PM