Hey everyone,
I've been working on an open-source hardware platform called T-DSP built around the Teensy 4.1, and this community has been a huge resource along the way.
Building serious desktop or rack-mounted audio gear from scratch is hard. Not the creative part - deciding what you're building, shaping the sound, the workflow, the controls. That part is fun. It's everything underneath that kills momentum: power, grounding, codec selection, USB host/device, MIDI isolation, analog/digital separation, I/O connectivity. Most existing projects are constrained by audio quality, and designing a full platform that gets all of that right is a serious undertaking. T-DSP solves for that foundation so you can stay focused on building something great. Desktop soundcards, audio interfaces, digital mixers, synthesizers, vocal processors, effects units - without reinventing the wheel every time.
The architecture is a backplane hosting a Teensy 4.1 and an ESP32 with a TDM expansion bus for chaining audio I/O modules via ribbon cable. Teensy runs the DSP via the Audio Library. ESP32 handles the UI layer (display, encoders, buttons, networked control via companion apps or web-served pages) and talks to the Teensy over serial. It also runs Bluetooth Audio A2DP (and WIFI), so wireless audio input is built in. Keeping the two MCUs separate keeps DSP latency clean and the UI responsive.
Current modules:
*Older t-dsp prototype on the bench during codec debugging - the ribbon cable connecting the backplane to the module is visible.
A few shoutouts to community members who have been part of this journey: @palmerr for his Audio Toy project, probably the closest thing to what I'm building here; @DD4WH for questioning my use of an 8-layer board (which I found out the hard way he was right); @h4yn0nnym0u5e for helping me debug the TAC5212 module (among other things); and @Paul, @houtson, @Bob Larkin, @Blackaddr, @Pio, and @chipaudette for contributions and inspiration along the way. I'm more than likely forgetting others I've communicated with or stalked over the years - thank you to everyone in this community.
If you've worked on something similar or have thoughts on structuring modular, reconfigurable audio systems with the Teensy Audio Library, I'd love to hear how you'd approach it.
Everything is open source (CC BY-NC-SA 4.0), designed in KiCad. Website: t-dsp.com | GitHub: github.com/t-dsp
Happy to answer questions or dig into the hardware decisions in the thread.
Jay
I've been working on an open-source hardware platform called T-DSP built around the Teensy 4.1, and this community has been a huge resource along the way.
Building serious desktop or rack-mounted audio gear from scratch is hard. Not the creative part - deciding what you're building, shaping the sound, the workflow, the controls. That part is fun. It's everything underneath that kills momentum: power, grounding, codec selection, USB host/device, MIDI isolation, analog/digital separation, I/O connectivity. Most existing projects are constrained by audio quality, and designing a full platform that gets all of that right is a serious undertaking. T-DSP solves for that foundation so you can stay focused on building something great. Desktop soundcards, audio interfaces, digital mixers, synthesizers, vocal processors, effects units - without reinventing the wheel every time.
The architecture is a backplane hosting a Teensy 4.1 and an ESP32 with a TDM expansion bus for chaining audio I/O modules via ribbon cable. Teensy runs the DSP via the Audio Library. ESP32 handles the UI layer (display, encoders, buttons, networked control via companion apps or web-served pages) and talks to the Teensy over serial. It also runs Bluetooth Audio A2DP (and WIFI), so wireless audio input is built in. Keeping the two MCUs separate keeps DSP latency clean and the UI responsive.
Current modules:
- T-DSP Core - 4-layer backplane, the accessible entry point. Teensy, ESP32, MIDI, Ethernet, USB host/device, Optical, and 2 T-DSP Expansion Headers. The Core schematic is based on known working designs from previous revisions, so the foundation is solid. At the time of this post: schematic and placement are finalized, routing is still in progress.
- T-DSP Desktop Pro - 8-layer stackup because analog and digital are integrated onto one board, requiring more aggressive isolation. At the time of this post: design is complete and ready to build - I've held off due to 8-layer fabrication costs and the current tariff situation on imports into the USA, but the files are there if anyone wants to run it.
- T-DSP TAC5212 Pro Audio Module - stereo ADC/DAC built around the TI TAC5212 with open-ended IO header for integration into audio backplanes. At the time of this post: previously built and tested, small issues resolved on the current revision. Planning to build it alongside the Core.
- T-DSP IO 2x2 Combo - balanced XLR/TRS I/O on the TDM bus. Reference design showing what an I/O module looks like on this platform.
- Mic Array Module - reference design, in progress.
*Older t-dsp prototype on the bench during codec debugging - the ribbon cable connecting the backplane to the module is visible.
A few shoutouts to community members who have been part of this journey: @palmerr for his Audio Toy project, probably the closest thing to what I'm building here; @DD4WH for questioning my use of an 8-layer board (which I found out the hard way he was right); @h4yn0nnym0u5e for helping me debug the TAC5212 module (among other things); and @Paul, @houtson, @Bob Larkin, @Blackaddr, @Pio, and @chipaudette for contributions and inspiration along the way. I'm more than likely forgetting others I've communicated with or stalked over the years - thank you to everyone in this community.
If you've worked on something similar or have thoughts on structuring modular, reconfigurable audio systems with the Teensy Audio Library, I'd love to hear how you'd approach it.
Everything is open source (CC BY-NC-SA 4.0), designed in KiCad. Website: t-dsp.com | GitHub: github.com/t-dsp
Happy to answer questions or dig into the hardware decisions in the thread.
Jay
Last edited: