- instead of storing chunk duration in ms store the frame count
and do calculations based on that
- add updated snapserver.conf for reference (#17)
- some code clean up
- move Kconfig entries to correct place
- remove unnecessary Kconfig entries
- add necessary dependencies in Kconfig files
Signed-off-by: Karl Osterseher <karli_o@gmx.at>
- change some function return values from int8_t to int32_t
- improve syncing
o try to mimic original algorithm used by badaix
- latency_buffer_full() call in player sporadically returns negative value
which led to a unnecessary vTaskDelay --> resolved
- use esp_timer_get_time() instead of gettimeofday() for timestamping
- enable IRAM optimazations for WIFI
- decrease WIFI RX buffers
Signed-off-by: Karl Osterseher <karli_o@gmx.at>
- disable shortMedianFilter
- decrease control loop's calls to adjust_apll()
- increase size of latencyMedianFilterLong
Signed-off-by: Karl Osterseher <karli_o@gmx.at>
o configurable through CHNK_CTRL_CNT
- move time sync messaging to timer callback so it will keep going to send and receive them as long as we are connected to snapserver
- change task priorities and core IDs of tasks
Signed-off-by: Karl Osterseher <karli_o@gmx.at>
- increase DMA buffer size from 1 chunk to 3 chunks settable through source code Macro CHNK_CTRL_CNT
- activate OTA server
Signed-off-by: Karl Osterseher <karli_o@gmx.at>
o copy component audio_board from ADF and create custom component from it
o copy component audio_hal from ADF and create custom component from it
o copy component audio_sal from ADF and create custom component from it
o copy component esp_peripherals from ADF and create custom component from it
- add fLaC support through xiph's original repository as a git module
- run player_task on core 1 with max priority
- create/destroy chunk queue as required automatically
- use mutex to access snapcast config and communicate this change by queue with size 1
- increase buffers for WIFI and LWIP in sdkconfig for NO_SPIRAM config
- add some more debug output
- detect snapcast configuration and init everything accordingly, e.g sample rate, chunk duration, ...
o calculate apll predefines in dependence of sample rate
o communicate these settings to interested parties
- remove typos
- minimize RAM usage of all components
- use both IRAM and DRAM in player component so we can buffer up to 1s on modules without SPI RAM
- support fragemented pcm chunks so we can use all available RAM if there isn't a big enough block available but still enough HEAP
- reinclude all components from jorgen's master branch
- add custom i2s driver to get a precise timing of initial sync
- change wrong usage of esp_timer for latency measurement of snapcast protocol
- add player component
- directly write i2s from sync task
o use DMA events to ensure buffer is empty on hard sync
- add i2s playback task
o try synced playback using i2s DMA events and task notifications
- drop short latency buffers
- change syncing to be more predictable using I2S_EVENT_TX_DONE
o also increase DMA length so i2s won't eat up so much processing time
o ensure at least one chunk is in DMA buffer
- add WIFI provisioning using Soft AP
- remove bug in server_now() leading to occasional unneccessary hard resyncs
- use shortMedianFilter to decide about hard resync condition
- drop ADF pipeline stuff related to playback, this introduced too much non deterministic delay
- add fast median calculation component
- increase LWIP buffers
- do apll change only when current setting is different than the wanted one
- improve snapcast_sync_task
- add esp timer to issue time sync messages to server every second
o block on semaphore which is given in timer callback
- start dropping adf piplines for decoding wirechunk part of code
o use write callback for output
o TODO: use read callback for input
- move time stamping of pcm chunks to decoder's write callback
- try to improve synced playback using APLL adjustments
- reduce diffBuf size to 1, bigger values seem to be problematic if packet loss is happening, maybe we need to detect this and clear diffBuf
- do early time syncs on boot to fill diffBuf
- playing with pipline ringbuf sizes and I2S DMA buffer sizes
- do some cosmetics
- improve server now generation
- improve playback synchronization
o use two audio pipelines so we can tap into decoded audio and apply collected timestamps from encoded chunks to PCM chunks in decoder write callback
o do hard synchronization on PCM stream and then start playback
o TODO: sample stuffing/skipping to keep a good sync
o TODO: for some reason playback hangs after a while on timestampQueue write
- use espressif ADF, remove external opus rep
o uses audio pipelines now
- change code to use flac decoder
- remove mersus code
- add first try of audio synchronization
o needed to sync timeofday to server on reception of server settings to avoid overflows in timeval calculations (int32_t on esp32 SDK)
o still a lot of TODO's in the code, but it's almost in sync, although there is quite some chunk skipping which I am currently working on
