Hi,

Summary of the current audio situation:

LM4550 codec: there is always "white" noise at the output, no matter
what. Adam will try changing decoupling capacitor models to see if it
makes an improvement. Right now, there is no known solution to reduce
the amount of noise present with the LM4550.

WM9707 codec: after replacement with this new chip, weird behaviour and
noises were observed. I have tracked these problems down to the content
of the padding AC97 slots: the WM9707 requires them zeroed out, but the
LM4550 does not. The AC97 controller in the FPGA used to fill those
slots with whatever was present on the CPU bus at the time of emission -
which resulted in the noises being dependent on software activity. This
is now fixed in:
https://github.com/lekernel/milkymist/commit/1ebbb92f837c30534ecc003ace33d0302861dc20
After this fix is applied, the audio output with the WM9707 codec is
noise free, so we at least have a fix for the audio noise for the RC3
boards.

Using the otherwise footprint-compatible WM9707 would need a few PCB
changes:
* DNP 1M resistor on crystal
* add capacitor on pin 33
* add capacitors on pin 32
* route SPDIF to the internal connector (optional)
* DNP the parts for the nonexistent headphones amp
See the datasheet at:
https://github.com/lekernel/milkymist-datasheets/blob/master/Audio/WM9707.pdf?raw=true

S.

_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode
Twitter: www.twitter.com/milkymistvj
Ideas? http://milkymist.uservoice.com

Reply via email to