Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Ligo]

I'm testing MC 22 for linux on my old laptop and i'm very happy both with Linux and with JRiver. I never tried Linux neither JRiver before and now i've become curious to make some experiment.
I've seen a version of Jiver for Raspberry and would like to try it on a PI 3. Before i go on and buy the Raspberry, i would like to know if the version of JRiver for Raspberry allow the use of DSP, particulary Resampling and Convolution, and in this case if the CPU and RAM of the PI 3 model are supposed to be enough.
The second option, in case JRiver won't allow DSP on Raspberry or it won't have enough power to make them run, is to use Raspberry with mpd-upnp as a renderer and JRiver as the server. Do you think that this configuration will work?

P.S. I apologize for my english, it's not my language.

[mwillems]

I run MC on several raspberry Pi's in my house.  I can confirm that ordinary builtin DSP in JRiver (i.e. parametric EQ) works great on the pi.  Ordinary resampling is mostly fine, but DSD resampling on the fly will not work (well). Convolution may work fine or not depending very much on the filters you're using.  Very long or involved filters will probably not work very will, but short filters doing mild phase linearization would probably be fine.  All convolution filters are unique, so the only way to know for sure is to test for yourself.

[Ligo]

Thank you.
I use filters for mixed phase room correction generated with DRC and they are 65536 taps.
On my laptop, an old core duo with Lubuntu and low-latency kernel, using SOX Resampler, Convolution and parametric EQ, i have 25% of CPU usage.

[mwillems]

Thank you.
I use filters for mixed phase room correction generated with DRC and they are 65536 taps.
On my laptop, an old core duo with Lubuntu and low-latency kernel, using SOX Resampler, Convolution and parametric EQ, i have 25% of CPU usage.

That will probably not work well on the Pi.  The Pi2 had about 1/8 the compute power of a late Core 2 Duo, and the Pi 3 is only about 150% faster than the Pi 2.  So, even assuming that JRiver can efficiently use all cores, 25% utilization on a Core 2 (depending on the Core 2) will likely translate to abou 130% CPU usage on a Pi 3, which is obviously not viable.  You might be able to reduce the load to a manageable level by reducing the filter length, but as you know that has other consequences.

[Ligo]

There is some way to reduce the filter length but it's quite difficult to do and the result isn't assured at all. May be that the second option to use the rasp as a renderer would be viable. My goal is to take away the usb from the signal path and the rasp with the spdif board seems a good way to do this with few space requirement and in a cheap way.