-
Notifications
You must be signed in to change notification settings - Fork 0
X13 Conversion
IMAGE OF PCB, potentially including its place on the relevant ROV
| PCB Summary | |
|---|---|
| Vehicle | X13, ROV Triton |
| Contributors | Bryce Sasser |
| Predecessors | X12-Power-Conversion |
| Success? | Yes, after several revisions |
Architecture Link
SID Link
REPO Link
The board converts power from the +48V +12V input to +5V and +3.3V.
The board input comes from the distribution board and the output is sent back to the distribution board.
What priorities did you have in your design? What design considerations did you have? What methodologies did you follow? (routing a differential pair, keeping something separate for isolation, etc)
The X13 board has two buck controllers to step down voltage compared to the old setup which used two buck converters. The old setup used its two DC-to-DC converters to separately drop to 5V twice, and one of the lines was again dropped to 3.3V. The new setup uses two steps to drop to 5V, which is dropped to 3.3V by a linear regulator. The buck converter keeps more circuitry, specifically the MOSFET, on board. The buck controller does not have the MOSFET as part of the IC, so it is more customizable. Every other design change was made around this switch. The new output allows much higher current, and the output voltage is about 5.3V in order to avoid errors with the Pi. The linear regulator was also changed from the previous year’s design. The new one was easier to solder and smaller. The conversion board no longer handles the 12V line. This allows the 5.3V and 3.3V lines to stay functional if the 12V line experiences issues. Also, power conversion no longer connects to Backplane at all.
The board was shrunk as much as possible from the previous year. In Eagle’s y, the board size was determined by the connection to Distribution. The horizontal dimension (Eagle x) is limited by the subconn. There is a notch in the bottom of the board that allows signal wires (between backplane and distribution) to get past.
The main circuit reference was given as an example on the buck controller data sheet on page 17 for v1 and v2 and page XX for v3.
Input: 48V at 1A or 12V at 2.5A Output: 5V at up 6A and 3.3V at 1A
Why did you pick certain components for your board? (If you don’t know the answer/were told, now is a great time to ask)
Buck Controller (LM5085MM/NOPB): The buck controller can handle up to 75V input. The upper limit on current is around 10A. The switching is handled by a PFET and inductor. link V3
PFET (DMP6023LSS-13DICT-ND): The PFET controls the switching of the buck controller. Using the PFET simplifies gate drive requirements and allows for 100% duty cycle operation. link
Linear Regulator (AZ1117EH-3.3TRG1): The linear regulator steps down from the 5V output of the Buck Controllers to 3.3V. It is the same linear regulator used for the Pi Shield. As mentioned, it is common and easy to solder. link
Working around the large inductors was a pain. Adhering to the circuit layout from the buck controller datasheet was necessary to keep the buck controller functioning with the right frequency. This kept the design locked in place for a large portion of the board.
The size of the board was restricted by the subconn. That is about the only restriction made by the mechanical side of things.
Many mistakes were made. The board had to be recreated because version 1 outputted over 6V instead of 5V, a problem not found in the test board. The problem could not be resolved easily, so the board was made again. In this new version, voltage was stepped down twice, to 18V first and then 5V, and a larger output capacitor was added. 18V was used because 18 is about one-third of 48 and 5 is about one-third of 18. These changes eliminated a heating concern for the inductor, cleaned up the output significantly, and, most importantly, provided the correct output.
The LM5085s ended up being rather flaky. When going 48V -> 5V, there was noticeable ripple, lots of noise, and rampant heating, especially at low loads. A second version was sent out with two buck controllers to go 48V -> 18V -> 5V. This fixed some heating and voltage ripple concerns, however the first controller (the 48->18 one) died several times and needed to be replaced. This is likely due to having to buck converters in series with a completely shared current load and the same switching frequency. To quickly get something working, a third version was ordered with the buck converter from previous years, the LM2678, present and connections to wire 12V in from distribution. The original brown out concern of the 12V line dying was resolved with finding the 48V line would dip and trigger the on-off circuit to turn the bricks off. After the on-off circuit was bypassed, the bricks no longer shut off and the 12V line was stable. So converting 12 to 5 was a viable solution again. The third version of just the LM2678 worked fine.
Each pin can handle 6.2A.
Connector test report
Calculators
- Any fun side details
Search keywords.
The purpose of the conversion board is to take the +48V from the distribution board (which gets it directly from the tether) and convert it to +5.2V and +3.3V. The new voltages are then sent back to the distribution board. The board is able to do this through a buck controller. In the previous year, a buck converter was used, which had more components on board the IC and made it less customizable. The more customizable buck controller allows more current to pass through than before to fix some brownout issues in the entire system.
Tools
Enclosures
Frames
Other
ROS