add 48k filter design, small tweak to optimised scaling

[Allan-xmos]

No description provided.

[ed-xmos]

Changes look OK although I don't really understand the filter design process well enough to comment on the performance so I trust you have this covered.

Only comments are:

• Add a new changelog entry 5.2.0 to say that a 48kHz decimator script has been added
• Bump the module build info https://github.com/xmos/lib_mic_array/blob/develop/lib_mic_array/module_build_info to 5.2.0 also. My bad also - I forgot this for v5.1.0!!
• Add a link in the comments in the design_filter.py script to https://xmosjira.atlassian.net/wiki/spaces/CONF/pages/3805052950/32k+and+48k+Hz+lib+mic+arrayhttps://xmosjira.atlassian.net/wiki/spaces/CONF/pages/3805052950/32k+and+48k+Hz+lib+mic+array
• Do you know roughly what the MIPS increase is for this over 16kHz?

[Allan-xmos]

Do you know roughly what the MIPS increase is for this over 16kHz?

As a guesstimate from the table here https://github.com/xmos/lib_mic_array/blob/develop/doc/programming_guide/src/resource_usage.rst :

• whole 16k 65 tap is 10.65 MIPS
• whole 32k 65 tap is 12.90 MIPS
• so just stage 2 16k 65 tap is 2.25 MIPS (12.90 - 10.65)
• stage 2 32k 65 tap is 5.50 MIPS
• stage 2 48k 65 tap is 7.75 MIPS -> whole 48k 65 tap is 15.65 MIPS
• stage 2 48k 96 tap is 11.63 MIPS -> whole 48k 96 tap is 20 MIPS

From that it looks like it might be worth trying a shorter stage 2 filter, so I'll make a "small" version as well