We have been working nonstop this month and the last to get DVT2 into production and assembled before Chinese New Year. All factories take a holiday from February 5th to February 18th. Chinese New Year is on February 11th this year.
The process of verifying everything again was really worth it. We discovered an issue with the LCD pinout. Luckily, we had not yet ordered the PCBs and could change the pinout on the connector in less than half a day.
There were over a hundred changes and refinements made to the schematic, layout and BOM. Many were minor cosmetic changes to the schematic and some were major including moving and re-routing some components to accommodate the new antenna. We also swapped some data lines in the LPDDR3 Memory in DVT2 to make it function properly.
The latest schematic and layout has been open sourced and released on GitHub. You can click here to check out the latest files. We will release the design files after we start shipping Pocket P.C.
Getting a PCB Fabrication House to produce 10-Layer impedance controlled PCBs before Chinese New Year was a difficult challenge. We tried numerous PCB services and companies we know or have worked with in the past and none could start production until after the holiday. Many companies are trying to get existing orders out of the door and are not taking new orders.
Fortunately, one of our team members checked with one of his long term partner factory and they agreed to accommodate us.
As part of the normal process of auditing a new factory, we took a number of pictures. Here are a few.
There were a number of Engineering Questions (EQ) that were raised that required us to carefully check the potential issues before providing a response on how the fabrication house should proceed. This is part of the process of getting PCBs manufactured. Since our board uses some complex technologies such as a few Wafer Level Chip Scale Packages (WLCSP) that have very fine ball spacing, we had to get the manufacturing tolerance correct to avoid issues during assembly. One issue we had is that their processes can’t accommodate white soldermask for the WLCSP component footprints. Their solution is to use a green soldermask around these components while the rest of the PCB will be white. This will not be the case for Mass Production as the fabrication house we use for M.P. has more advanced machinery capability of higher tolerance.The color white was chosen for the PCB so that it can reflect the keyboard backlight LEDs thereby making them appear brighter. If we had chosen a different soldermask color, the true color of the LEDs would not come through.
Our perseverance is paying off. We are getting a quick-turn around on the PCBs and will be doing assembly of five pieces this week.We will be sending another update shortly after we have had some time to test the PCBs and let you know the results.
After a period of intense effort, we are back with an update with the latest developments with Pocket P.C. From onboarding a new factory to sourcing components, manufacturing a D.V.T. followed by board bring-up and formal verification all took much longer than originally anticipated. Now that those are all completed, we have a lot to share with you.
Our first production run with our new factory went extremely well. We manufactured 25pcs for a D.V.T. or Design Verification Test. We are extremely impressed with their production facility, processes, competence and skill that they demonstrated in getting the PCBs assembled. We were pleasantly surprised how thoroughly the factory performed during every step of the process.
After assembly, we did a formal verification which took a few weeks. We carefully examined every aspect of the project down to every single component.
As a result of exhaustive testing, we determined that we will have to do a second D.V.T. This will add a few weeks to our shipping timeline but we feel it is worth it to resolve a few critical issues that we encountered. We want to resolve these issues and test that the solutions are sufficient before starting a P.V.T.
We accidentally swapped a few data lines in the PCB. This meant the RAM, SD, and one of the USBs did not operate as expected.
A newly added power protection component interferes with the enclosure preventing it from closing properly.
The LoRa antenna selected was not optimal for the location on the PCB. We had a meeting with the antenna manufacturer and they recommended a bigger antenna with a different layout requirement. The advantage of this antenna is that it is the same part for both low and high frequency bands. This antenna should also allow for better range than the previously selected component.
Following our formal verification, this week we had a productive meeting with representatives from all the factories involved in the manufacturing Pocket P.C.’s components. We wanted to ensure that everyone is on the same page. Each factory made their recommendations on how to improve the product and manufacturing process.
As you can see above, the keyboard is almost there.
Here are a few things we noticed and will be correcting in the next version:
The status LED is a little too bright which causes the blue to bleed into the main keyboard when it’s on. This is solved easily by increasing the resistance of the LED resistor to make the LED not as bright.
A backlight cover film is missing over the LED where the newly added “+ and -” buttons are which is why the keyboard is slightly brighter there. We will correct this in the M.P. keyboard cover film.
The keyboard was silkscreen by hand which means it is not as sharp as the M.P. version and the black is too thin. We will ask them to do two passes of black in the M.P.
There was a software file glitch at the factory which caused the rightmost button on the bottom to show question marks instead of the actual characters. We will ask them to correct and carefully review the silkscreen stencil against the proof before doing the silkscreen in the future.
Some characters are hard to read so we will be making them bigger in the final version.
Something which we believe we haven’t mentioned and would like to make you aware of is that the back cover screws into the front cover with 10 screws. This means that Pocket P.C. can be easily taken apart without great risk of damaging it during disassembly.
The assembly factory made a suggestion that we use hexagonal or triangle screws but we informed them we did not want to make it difficult for users to open by requiring a special screwdriver. We will most likely end up using phillips screws if there are no assembly issues during mass production that would require otherwise.
Besides the main issue of swapping data lines, we carefully reviewed the schematic and layout in its entirety and made a list of changes for D.V.T. 2.
If you are interested in what changes we will be making in the next revision, you can follow along with our changelog here:
Further, we are commissioning an outside engineering firm to double-check everything. They will review the schematic and layout to make sure we don’t miss anything and they will give us an outside opinion on the overall design.
We decided to scrap our initial Arduino-based firmware for the Keyboard/System controller in favor of QMK (Quantum Mechanical Keyboard) Firmware. You learn more about this firmware here: https://qmk.fm/
According to their website, “[t]he goal of the QMK software project is to develop a completely customizable, powerful, and enjoyable firmware experience for any project – keyboard or otherwise – and to provide helpful, encouraging, and kind support and feedback for people with any software development experience.”
Most of QMK itself is under GPL v2, but it includes software that are under various other licenses. We intend on upstreaming our changes to the original repository when they are ready.
As for the main software running on Pocket P.C., we will be working with Bootlin, an embedded development firm, to upstream everything from U-boot specific changes including Device Tree, Linux Kernel changes including Device Tree, Buildroot configuration and an Armbian build configuration.
This means that not only will we be supporting Armbian in addition to Debian but this should make it much easier for other distributions to be supported once the changes are accepted upstream.
If anyone has emailed us in the past few weeks, we will respond to all your questions soon. We apologize for falling behind in responding to customer inquiries but please know that we have been extremely focused on doing what needs to be done to get an excellent product shipped to you soon as possible.
We’ve taken a lot of pictures and videos which we have to edit together and are planning on sharing those soon.
Thank you everyone for being so patient while waiting for us to ship Pocket P.C. We want to ship a well-tested well-built product to you. This means that we have iterated through multiple prototypes and revisions before deciding to enter mass production.
The latest 3D Prints of Pocket P.C. to proof the mechanical CAD design have arrived. This will help us check to see if we have the correct dimensions for the PCB, battery and display. Please take a look and let us know what you think in our Popcorn Computer Community.
This week the Pocket P.C. crowdfunding campaign hit the exciting milestone of three-quarters funding. We are gracious for the support of over 180 backers who are excited about defining the future of Linux handhelds!