*improvement!*

Rev 2.1

Months of research and co-op helped improve this schematic quite a bit! It is certainly not perfect, but pretty good for a revision.

All of these files are on Github with tags for each version we deemed final enough to procure. We did our best to be good git stewards but still managed to break the repo once or twice.

Analog -> digital

Some problems from R1.x were hammered out
MEP

All of the lessons learned from a board and system level were pulled into the redesign of the entire system, but the Motherboard had significant changes.

MEP


There are so many ADCs on the SPI bus that a decoder is used to save Beaglebone pins for the chip selects. The LEDs are used for bring-up and software development, not for "flight" units. Initially the clock is as slow as possible to ensure all of the addresses map correctly. These LEDs were probably pretty detrimental to operation and it probably should not have worked. I guess the ADC CS pin is sinking the extra current?

MEP
One pain point for the ADC is some interesting behavior during testing. It locks up when supplies are not brough up in the right order. AVCC must always be greater than DVCC, but the VREF also has a play in the timing. The latter isn't in the datasheet but it is fixed by turning on analog-> digital -> VREF. A power sequencing chip ensures the Beaglebone has time to boot, then the beaglebone enables the rails. the issue. The 3V3 rails are not tied together this time. Each output powers half of the ADC bank, but they should have been designated A or B.

MEP