I started this because I was using the upsampling built into
Squeezelite(libsoxr), not using SoX in LMS so I can't offer any clues as
to how to get that to work properly. The Squeezelite resampling option
is not the same as in SoX the program arguments, although the underlying
code is the same.

The one I'm using right now is mI:::28, which  is medium quality,
intermediate phase and 28 bit depth. I'm not quite sure exactly how this
translates to SoX arguments. 

The upsampling is done in Squeezelite on the Wanboard and then sent to
my own USB DAC, which uses a standard XMOS UAC2 interface, I2S is run
through isolators, on isolated side are low jitter clocks, reclocker and
the DAC chip is PCM5142 with very low noise regulators on all of  this.
The low jitter clock is sent back through an isolator to the XMOS
interface. 

The PCM5142 has several builtin filters, some  of which are better than
others. If you feed it 352.8/384 it turns off the  internal filters. So
by using resampling  in Squeezelite to 352/384 I can bypass the
implementation in the chip and just the software.

Even if upsampling to a lower rate it can still be advantageous,
upsampling  to 88.2 with the above paramters is a significant
improvement. Going  to 176.4 is even better but  the  best canbe 
achieved  by going high enough that the built  in filters are completely
bypassed. 

In all these cases the network traffic stays the same, but the USB data
rate goes up since the upsampling is being done in squeezelite. 

I have not had time to go into depth trying all kinds of different
parameters, I spent a few nights  trying different things and came up
with the  above. I've  been listening to it for some time now and am
still enthralled with what it  is doing. 

BTW the load on the Wandboard processor is about 8% when using this
setting. When using the default 20 bit setting it is about 4% and when
not doing any upsampling its about 2%.

Klaus, to your statement that upsampling  should  not  be necessary, the
answer  is of course YES. The issue is that as far as I can tell all DAC
chips with builtin filters are compromised sonically, the upsampling is
an attempt  to bypass these filters with an external filter that is more
sonically "transparent". So yep it IS a band-aid, but one that is
currently necessary for most DACs. 

If the internal filters are not disabled, exactly how the internal
filters interact with the external filter are going  to be very DAC
specific. In the case of the chip I'm using the filters get simpler as
the sample rate goes up so even though they are still there if you
upsample to an intermediate rate, the total result sounds  better than
the builtin filter going from 44.1. 

The best sounding parameters for going all the wayto 352.8 and 88.2 will
almost certainly be different, but I have not  explored it yet.

John S.


------------------------------------------------------------------------
JohnSwenson's Profile: http://forums.slimdevices.com/member.php?userid=5974
View this thread: http://forums.slimdevices.com/showthread.php?t=99088

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/audiophiles

Reply via email to