More > JRiver Media Center 22 for Windows
NEW: SoX Resampling
michael123:
No option for apodizing filter?
No option to specify the filter per sampling frequency? If we're talking about 'established best mode', linear phase is today preferred for high-resolution streams
SOX btw requires quite a lot of hardware processing, interesting how did you achieve responsiveness especially with the memory play
sorepinky:
I do not know whether upsampling is what SoX is supposed to be used for. Downsampling is what I see discussed on the various forums. Nonetheless, while upsampling with SoX from standard CD (ripped to standard compressed FLAC) to 176.4, Windows 10 Task Manager shows MC22 using between 0.2% and 1.5% CPU on a 7 year-old first generation i5 HP desktop [JRMark (version 22.0.8 ): 2785] - and that's with Play from Memory selected (not for sonic reasons) and a couple of PEQ LF filters. With the SoX option unchecked, the numbers are similar. When playing a 192 FLAC (i.e. without any resampling) the numbers are also similar.
I can't tell if I hear a difference in sound quality. Too subjective, but it sounds fine. Some user-configurable options would be nice.
kr4:
--- Quote from: JimH on July 15, 2016, 12:48:08 am ---No. The problem you're having could be CPU or bandwidth related.
--- End quote ---
Aha. I spoke too soon. I relocated the NAS to the same switch shared by the MC server and it worked immediately.
However:
1. It seems to depend on the DSP settings. With a DSD256 5 channel file converted to PCM and DiracLive, the success depends on the target resolution.
a. Converted to 192/32 or 96/32, there are interruptions and bufferings every 30 seconds or so. Dirac on/off makes no difference.
b. Converted to 176/32 or 88/32, everything plays. Dirac on/off makes no difference.
2. In all cases, the CPU/memory usage seems similar (5-22%, mostly in the teens, and memory usage at 2-4 GB). However, when I look at the individual cores, only 4 are running and 0 and 7 have most of the load. Cores 1,3,5,7 are “Parked.”
3. On the other hand, network behavior varies strangely. In case 1b (good playback), network usage averages 2% but peaks in the 85-100mb/s range. In case 1b (interrupted playback), the average is the same but the peaks are in the 50-60mb/s range. Changing from direct to play from memory changes the network activity pattern but not the peak values.
I do not know what all this means or what may be significant but all these observations are with the same NAS connection described above.
kr4:
OK. In response to my original post here, SoX does make a difference and all playback (tried so far) seems to be fixed.
pschelbert:
Hi
okay, so its about signal-processing, may be not everybodies topic.
The DAC industry is now offering it in chips and also as a feature several seletable filters in some DACS on the market (reconstruction filters). To design such a filter yourself via JRiver and activate it during playback, what I described would be needed.
clearly I tried a lot of things. Offline to use different filters is no problem. However if you want it fixed in a playback system, since I use JRiver, I would like to just integrate user designed filters (myself).
Otherwise than that, I do not see any need for a different resampler then the one which is built in in MC21.
Actually, if you just want to hear music as you say, there is no need at all for a resampler. The DAC does it all...
Peter
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version