soundcheck wrote: > I was briefly looking into different "NO-DAC" projects ongoing on the > net. > > Basically these use Amanero or JLSOUNDS USB-I2S/DSD interfaces with > direct DSD processing capabilities, feeding the DSD stream into a basic > LowPassFilter. On the first glance that looks fairly simplistic. > There are different projects with different complexity > (FlipFlops/transformer/tubes/...) though. > > Since the PCM-DSD conversion requires serious horsepowers, using > squeezelite+sox-DSD on e.g. a PI2 might not be the way to go. > > In this case a powerful server needs to do the conversion. Most of the > discussions I looked into were referring to HQPlayer > as the DSD engine of choice. HQPlayer seems to have quite well done > filters in place. And is also using/can use GPU horsepower for the > calculations. > > There are two big questionmarks. > > 1. I have no clue how the sox DSD-fork compares to HQP. I don't know if > it would make sense to try the sox approach at all, since the conversion > is a crucial part of the story. > > 2. I'm also not sure if a project > LMS&sox(dsd-fork)->squeezelite->JLSOUNDS-USB->LPF @ up2 DSD256 would > work. > > > Let me know if anybody went that path. > > Thx
Sure it works on LMS and, as Kimmo reported, You coud try exactly this configuration using Daphile. As far as i know, the italian forum where my projects are hosted is hosting one of the more advanced NO DAC procjects, using JLSOUND and HQPLAYER, but I'm quite sure authors are writing also in some international fora on the same matter. Someone here tried to compare HQP and this fork of SOX. sSound quality apart, SOX resulted much more resource hungry and using 'high' or 'very high' quality makes impossible to perform conversion in real time on a i7 server, you should then use 'fast' setting for that. Similar results exploited Archimago in a trial of HQP vs JRiver vs SOX, but in his findings, in terms of measured 'fidelity' to the original signal, SOX was the best, for what it matter. The author of the SOX fork explained that the process is like that: 1. first PCM Upsample to 88200. 2. second PCM Upsample to xxx, depending on final DSD format. 3. Digital filtering applied on the result, using a modified 'dithering' algorithm (I suppose is not a dithering at all). Maybe I'm wrong, but double cascode upsampling in 'fast' qualiy is likely 30-35% passband reduction plus digital filtering. No idea on what Jriver or HQP are doing, but I, as for now, I'm right with old plain PCM (by the way, I'm not a fan of PCM upsampling too). ____________________________________________________________ SB+, Klimo Merlino + Kent Gold, Monitor Audio Studio 20 Gold SE+, Klimo reference and DIS Interconnect. ------------------------------------------------------------------------ marcoc1712's Profile: http://forums.slimdevices.com/member.php?userid=34842 View this thread: http://forums.slimdevices.com/showthread.php?t=97046 _______________________________________________ unix mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/unix
