This project runs on the STM32F4Discovery (STM32F407 variant) in combination with the STM32F4DIS-EXT expansion board. The on-board MP45DT02 MEMS microphone is sampled and filtered to produce a 16 kHz PCM audio signal. This is streamed over UDP via the LAN8720 Ethernet Phy on the expansion board. A simple Python script is provided to receive and play the audio stream. This project was a learning exercise and should be considered as little more than a proof of concept. If it is of interest to you, may also want to check out some relevant blog posts written along the way. 1. Update submodules git submodule update --init --recursive 2. Extract LWIP The LWIP source code should be extracted from the ChibiOS repository. Assuming the current working directory the top level of the repo, issue the following commands to extract LWIP. This assumes 7z is available on your system. cd lib/ChibiOS/ext 7z x lwip-1. 4. 1_patched. 7z cd - 3. Configure as Required Modify stm32_streaming/config. h with the desired static IP configuration.
There is no actual RTP implementation. This wasn't always the case as I had hoped to let VLC take care of playback for me. I made some headway implementing what I considered the minimum required functionality from the follow specifications: Real-Time Streaming Protocol (RTSP) ( RFC2326) Real-Time Transport Protocol (RTP) RFC3550 Real-Time Transport Protocol RTP Real-Time Control Protocol RTCP Session Description Protocol (SDP) RFC4566 However I eventually gave up on this approach due to a mixture of getting fed up, and VLC appearing to expect non-standard RTSP headers. Python PyAudio Playback I hoped for a low latency audio stream but ran into some issues with PyAudio/PortAudio when trying to play 20 ms audio frames. The callback from PortAudio was trying to consume data a lot faster than it was being provided. Buffering a few packets alone didn't appear to alleviate PortAudio's hunger, and neither did bumping the frame/chunk size. So both approaches were taken. Note that this may be something to do with my specific hardware or the VM I was attempting to play it back through.
Debugging audio playback wasn't something I particularly wanted to do - which is why I originally planned on using VLC to avoid it altogether. Finally, there is no Python code in place to handle lost / out of order frames. The RTP sequence number in the packets is completely ignored.
These are both provided in hexidecimal format, with leading zeros present as necessary. stop stop a running audio stream stm32_streaming/audio/mp45dt02_processing. c This file handles (over) sampling the MP45DT02 MEMS microphone as well as running a CMSIS FIR filter over the data. The FIR coefficients can be found in stm32_streaming/audio/autogen_fir_coeffs. c which are generated using utils/ stm32_streaming/audio/audio_tx. c This module tracks when 20 ms of audio has been collected from the MP45DT02. Once a full payload has been stored, it requests rtp/rtp. c add a RTP header to the payload and then transmits the data over UDP to the address previously specified by the user. Dependencies The STM32 binary requires the following libraries: ChibiOS for RTOS + HAL LWIP for IP/UDP/TCP CMSIS for FIR DSP These should be met by following the steps listed above to build the project. In python_playback there are a handful of Python scripts which will request, recieve and play the audio stream. There are also some additional debug modules.
Sitemap | asgardiatv.space, 2024