BGA291 integrated solid particle to CF Express

       This open source project follows the GPL3.0 open source agreement. If it is used commercially, please donate 20% of the net profits to charity projects. The time and amount of the donation must be clear and verifiable. And the modified hardware must also be open source. Violation of the GPL agreement will result in legal liability, so please be informed! ! ! At the same time, any modifications made based on this must be clearly marked as being modified by me and @揕you大哥.

       The adapter board was modified from the project of @凯有大哥. Thank you very much dalao for your selfless sharing. The specifications and information of CF Express are much harder to find than the general M.2. You need the shell purchase link and the original point definition. If you are waiting for information, you can click on the blue words to go to the big brother’s open source page to request it. It will not be provided repeatedly here.

The changes are as follows:

1. All traces were redrawn and widened from 5.9mil to 7mil. The differential lines of the PCI-e part have adjusted spacing and length in accordance with high-speed signal specifications to shorten the line length as much as possible. The double-sided panels have to span the power plane divided area. At least ensure that the projection of the traces does not turn when crossing, so as to minimize interference, and Ground holes are provided at equal intervals on both sides of the differential line to ensure that the high-speed signal return path is as short as possible.
2. The packaging of all peripheral components has been replaced. The original packaging did not have silk screen, and the markings were all on the document layer. After the replacement, there is silk screen for easy identification.
3. The harder-to-buy MP2143 has been replaced with the easier-to-buy RY8120 and FR8205 (RY8120 is recommended because it is cheaper and the feedback resistor value will not be weird). External filter capacitors should be added as much as possible. After all, the NVME protocol disk is very important for power supply. Stability requirements are higher.
4. A number of deleted empty points have been filled in, including several ground points. The routing adjustment of the differential line has also made some positions so that some empty points can be put back in. It is not easy to have false soldering when the points are complete. Particle displacement, seesaw problem.

SSDs that can theoretically be used to transplant particles:

Intel particles: Star Dance P200, Overclocking 3 D200

SK Hynix granules: BC501, BC511, BC711

Samsung particles (seems to require an external crystal oscillator, which is currently not defined and may not be used): PM971, PM971a, PM991, PM991a

KIOXIA granules: BG3, BG4, RC100

I haven’t built a board for testing yet because there is no device that can insert a CFe card. Feedback is welcome; the Buck part should be no problem. I have used the same chip in other boards.

Please choose 0.6mm for board thickness! Otherwise it may be too thick!