Rev2 Results

In addition to the power switching circuit issues I had with the Rev1 circuitboard, there were a few other issues I had that motivated a respin of the board relating to diagnostic capability:

• LEDs for showing the status of the different voltage domains
• I/O-controlled LEDs from the ESP32 
• A way to probe the master clock signal
• Ground references for the Saleae logic analyzer channels

These would have been useful for the sake of determining the board state without having to attach scope probes or a USB-to-UART connection. 

Although I was able to order bare boards and hand-solder them myself for a low cost, I found that the assembly time for each PCBA was a bit prohibitive. This led to me only soldering 2 assemblies out of the 5 bare boards I had received, and with each assembly I made, I had some concern about the reliability of my solder joints.

For Rev2, I decided to get the boards fabricated at PCBWay! A few years ago, I had used PCBWay for the fabrication of bare boards alone for a different project, and had a good experience with their customer service. Bare PCBs take around 4-5 days to process, and the PCBAs take around 24-25 days. A huge thanks to PCBWay for sponsoring this project and covering the board and assembly costs. 

One thing I really appreciated with PCBWay's assembly service was the level of support and communication I had with the fabrication and assembly teams. I had some issues with my PCB Gerber files, including design rules that did not properly enforce the minimum trace width that I intended to set. Once the board had passed engineering review and the PCBA was in review, the assembly team asked for clarification about the orientation of some polarized caps and diodes on the board (which my silkscreen didn't show) and to check an inconsistency in my BOM regarding which ESP32 version to use. Lacking an engineering team, I - as a solo designer - appreciated the feedback and communication, which was always polite, to keep the process running.

In Rev2, in addition to implementing the diagnostic changes from REV1, I swapped some choices for some passive components and headers for ease of manufacturing. I also added a nifty logo and the load switch ICs mentioned earlier.

When I received the circuitboards, one of the first pieces of behavior I had to debug was the power sequencing. When PVDD and DVDD of the TAS5711 aren’t sequenced with the correct timing, configuration of the TAS5711 fails, and I receive NACKs when I start writing to its registers over I2C. I found that the PVDD net was capped at 5.6V rather than matching the 18V input voltage, which was a little concerning until I read the datasheet for the TCKE905ANA,RF more thoroughly. It was also super useful to have up to 5 professionally-fabricated assemblies that I could use for this debugging, thanks to PCBWay’s sponsorship, rather than having only one or two hand soldered assemblies to work with.

As it turns out, the TCKE905ANA,RF is designed for 5V nominal input voltage and clamps its output at 5.6-5.7V. A different available component that could work for an 18V input/output voltage is TCKE920NL,RF which has an overvoltage clamping voltage of 22.2V. This has a slightly different fault response from the TCKE905ANA,RF, but only the NL variant of the 20V eFuse IC was available on Digi-Key. To keep consistent fault behavior I may switch to the TCKE920NL,RF for both switching circuits, or use TCKE903NL,RF for the 3.3V switched circuit.

More to come once I fix the power sequencing issues!