Tidy support for 8/16 mics and parallel decimators#197
Tidy support for 8/16 mics and parallel decimators#197ed-xmos merged 16 commits intoxmos:developfrom
Conversation
…16-channel mic array
Resolves a buffer overrun (out of bounds array access) exception and ET_ILLEGAL_RESOURCE exception.
Works with 8 or 16 mics in DDR. Uses decimator that processes samples on multiple threads. Added API for enable/disable assertion (for debug purposes). Optionally support button input for selection of which two MICs are output to the I2S DAC. Port/Pin configuration is done via app_config.h settings. Set MIC_TILE in app_conf.h to match desired configuration along with other options. When using on Tile 0, USB may not be available and push buttons and LEDs are not to be used (LEDs will indicate MIC Data for 4 of the 8 data lines). For button mic selection functionality is provided via another pin (X0D12, J12 - Pin2) with an internal pull-up applied. When using Tile 1, another pin must be used to drive CODEC_RST_N high (using the test point on the board), in this example, X1D38 (Port 1O) is used for this purpose. A 3k3 series resistor is used on this pin which is sufficient to override the weak pull-down on the board while avoiding signal contention.
This change has not been tested, and there needs to be a decision on whether this decimator should be part of the mic_array namespace or left as an application specific one. Which option is chosen likely dictates whether the name MyMicArray is appropriate.
There appears to be issues with Prefap.hpp being included in the API header when defining a custom class template.
|
This adds the 16ch de-interleave which will close this #193. It has a passing unit test and I have tested it in hardware too. These changes do not change other core mic_array features so I think the risk of merging without a full CI run (which doesn't exist yet! #198) is low. There is also a sample app to show use of the parallel decimators which does not run without specific hardware and so it is not so useful. Perhaps in time, the 16 mic aggregator could replace it or perhaps the 16 mic sum/delay example. However I do not think it is urgent. |
|
This will also allow deletion of https://github.com/xmos/lib_mic_array/tree/jcarrier/feature/16-mic-support |
| static | ||
| i2s_restart_t app_i2s_restart(void* app_context) | ||
| { | ||
| static unsigned do_restart = 0; |
There was a problem hiding this comment.
How does the value of do_restart ever change?
| </Link> | ||
| </Links> | ||
| <ExternalDevices> | ||
| <Device NodeId="0" Tile="0" Class="SQIFlash" Name="bootFlash" Type="S25FL116K" PageSize="256" SectorSize="4096" NumPages="16384"> |
There was a problem hiding this comment.
I'm a bit surprised at the inclusion of the Type, PageSize, SectorSize and NumPages arguments. I thought that recent changes Tools 15.2 meant that we no longer needed these arguments and that their inclusion caused problems. However, my notoriously bad memory means that I am not sure of the facts, and I see that the other .xn files in this repository look similar to this one as far as this line goes.
There was a problem hiding this comment.
You are correct - this shows it https://github.com/xmos/sw_xvf3800/commit/4349ede0fd3b0527462de01da0402cdcc5322aa9
I will fix!
xhuw
left a comment
xhuw
left a comment
There was a problem hiding this comment.
Your changes since Michael's review appear to solve all of his comments. I didn't look at any previous changes.
You might need to get a super admin to work around Michael's "needs work" this approval might do the job though
|
I "dismissed" Michaels review so now it's mergeable |
4 participants
https://xmosjira.atlassian.net/browse/AP-12