The motherboard for the active line array is where the amp modules, the DSP module, the MCU module, ADC and DAC modules all “plug in” and get interconnected. There is also some active circuitry with power supply chips, isolator chips and selector ICs to allow using multiple SPDIF sources. And of course, it has a lot of filtering and terminations for the serial data from the ADAU1466/7. Because of its size, it is more expensive than the other boards in this project, so it hasn’t undergone as many revisions. The motherboards described here are still evolving somewhat, so the designs are not “final”. However, they work well enough for building prototypes, and it shouldn’t take much more testing to finalize the designs.
And yes, there is more than one motherboard. There was an initial design using the ADAU1467, followed by a board designed for the ADAU1466, followed by another ADAU1467 board that is still being debugged as of late January 2026. This write-up will get updated once the new ADAU1467 board is fully tested, but in the meantime these two most recent boards will be documented.
The ADAU1466 populated motherboard, mounted on an aluminum plate next to the power supply, is shown above. The picture below is the ADAU1467 board that is assembled but does not have the modules installed.
The schematic for the ADAU1466 motherboard is shown below, but you can get a clearer view by downloading the KiCad files. The ADAU1467 board is similar, with the exception of isolator chips that were added for the analog modules. The isolators show up on page 2 of the schematic, and that page is shown below. The ADAU1467 motherboard also includes a small OLED panel to display the status of the amplifiers.
There are two design challenges for these motherboards. The first is signal integrity for the serial audio. Since the design requires either TDM4 (ADAU1467) or TDM8 (ADAU1466), the serial audio is changing at a very fast rate–either 6MHz or 12MHz. Routing this high-speed data to all the amplifiers requires keeping the lines as short as possible and using resistive terminations to avoid reflections. For both motherboards, the signals are terminated with pull-up and pull-down resistors. The values selected are 220ohms for the pull-down and 390ohms for the pull-up. These values result in a Thevenin equivalent of 120ohms into a 1.2V source. Since the SSM3582 uses 1.8V logic, this seemed like a good value for the motherboard. Is this an “optimal” or at least a very good value? I don’t know–that is why we build prototypes, and it is something that I still need to verify. The Analog Devices evaluation board uses series 33ohm “damper” resistors rather than termination resistors, so maybe that would work even better.
The other design challenge is noise on shared ground and supply lines. Connecting the digital and analog modules to the shared power planes rarely works without some digital noise bleeding into the analog signals, so the motherboards break the power planes into different sections that are connected at a single common point (“star ground”). However, this scheme doesn’t always work, and the only sure-fire solution has been to use an isolated DC-DC converter for certain circuits. To ensure the analog out and input are absolutely quiet, I added isolator chips (ISO6740), but this board hasn’t been debugged yet, so this design is still untested.
Interestingly, I have not been able to eliminate noise from the ESP32 without using an isolated DC-DC supply dedicated to just that module. The Bluetooth radio creates pulsating noise that you can faintly hear in the amplifier output unless the supplies are completely isolated and connected at just one point. I didn’t expect that much noise from the ESP32, but sharing circuitry with a radio is apparently a challenge.
The ADAU1466 motherboard has some resistors that are too close to the connectors and the SPDIF transmitter chip used and incorrect pinout. Those problems were fixed on the ADAU1467 board, but both boards missed the extra isolated DC-DC converter needed to eliminate some low-level noise from the radio in the ESP32. The schematics will eventually get updated for these two motherboards, but in the meantime, here are the design files as they currently exist. There are quite a few SMD parts on these boards, but the passives are 0805 and they are spread apart enough that they are easy to solder with a hot air station. There are no JLCPCB files for these boards, but if you want to get them assembled, it isn’t difficult to format the BOM and create the layout files from KiCad. I used OSHPark to manufacture these boards, as they offer the best value for 3 boards given the current tariffs on the Chinese suppliers.