Log My handheld PC projects

[CC Ricers]

Current project log: Intel Compute Stick handheld

Core m3 6y30, 5.5" FHD OLED screen

Original post:

This build log is spun off my previous PortaNUC thread, which is also going to be a SBC computer inside a custom case made for portability and gaming in mind. While the PortaNUC uses a more powerful PC, it came with more complications with regards to portability and power management. This thread is for "portabilizing" (as they say in the console modding community) a Chuwi LarkBox, a PC with an even smaller footprint than the NUC, and lighter power consumption.

Current dimensions for the handheld console are set at 190 x 90 x 29mm. The LarkBox mainboards will be positioned to the left and batteries to the right. I will show in a later post how this layout would be beneficial. Compared to other SBC handhelds, it will look more or less like a larger, thicker Anbernic RG351P or PowKiddy RGB10 Max. But since the hardware is x86, I can both play some PC games natively and stream other games from another computer.
Here's a render of the design in progress (still need to add many cutouts including extra face buttons):

I plan to use it in conjunction with an Elecrow 5" IPS touch display with a resolution of 800x480. No need to hassle around with sound with its built-in speaker.

This resolution is very close to ideal for emulating older consoles as well as playing older PC games. It's no Steam Deck but at least it's more portable!

For power, I plan to use the TalentCell 12V 3000mAh power pack. I have done successful tests of it running solely on its power and can run for up to 2-3 hours depending on programs used.

I will take apart the power pack enclosure and place the batteries and power board inside my own case. These use three 18650 Li-ion batteries and determine the thickness of the console. If I were able to run the PC off a single LiPo battery perhaps the console could be slimmer. But I haven't yet managed to get it to power on successfully with one so I will continue to investigate this. In the meantime the Talentcell will have to do.

Some inspirations include the custom cases built by the Wii modding community, as well as a few one-offs that put PC components in handheld cases. This has actually already been done with the GMK NucBox which has very similar hardware. YouTuber Sfdx Show crammed the NucBox hardware into a modded Game Boy case and also have it run off battery power shown here:



The vertical Game Boy form factor works well with the hardware, but I plan to make the console in a horizontal format, as it is easier to integrate shoulder buttons.

 

[Valantar]

Subbed! Looking forward to seeing this come together

 

[CC Ricers]

Subbed! Looking forward to seeing this come together

Thanks, I think I have a much better handle on how to make this portable PC compared to the PortaNUC. The parts being smaller helps.

I've added more face buttons and tweaking the dimensions some more for the case design. Here's the case as it looks in three different colors, all which are locally available for me to buy as PETG filament.

I went for the boring gray filament but the green also looks nice to me, as it also has a hint of teal-turquoise in it. As I'm finalizing the overall dimensions I started to notice it resembles the Odroid Go Super in size and proportions, only mine's going to be extra thicc because of the batteries used.

Speaking of portable consoles for retro gaming, I already have an image of Batocera loaded into an SD card and the screen I'm using is great for older systems. 480 vertical pixels makes integer scaling easy. Also, it only draws 6-8 watts on the wall while gaming, which with 30 watt hours gives me up to 5 hours of game play.

 

[CC Ricers]

Here's a preview of the layout of the main components of the LarkBoy. The two boards are lined side by side with the ribbon cables connecting them. The board on the left is the I/O expansion board and the one on the right is for the CPU. This is so that the headphone jack and SD card slot are available from the left hand side of the console. The two USB ports will be on the inside connecting the microcontroller and LCD screen. They're offset a short distance from the top because I need room for the shoulder buttons.

I'm taking advantage of the height of the console shell by placing the L and R shoulder buttons in the usual configuration found in game controllers. The power outlet will also likely be placed on this side.

 

[CC Ricers]

A couple updates to the design and now I have a top piece that I think is ready to print and try out. There are clips and notches to keep the screen in position, but to keep it from falling inwards, as well as the controls, I need to model the middle piece further which will connect using screws and tabs on the sides.

 

[CC Ricers]

The first attempt to print the PETG top shell was a failure. I still have to hone in on the best settings for PETG printing, even after I printed a small test cube and that turned out fine. Printing on a rough steel bed, the top right corner warped and curled up, but the sides that stuck to the bed stuck too well. It was very hard to remove the part from the bed after giving it time to cool off, and snapped apart when I pulled. I tried wedging a knife underneath but at the same time did not want to scratch the bed.

I was hoping to get a nice textured finish for the bottom most layer but turns out it's not gonna be that easy. But at least the bezel mostly fit the screen.

Next time I will either attempt to print it upside down, at the expense of more print time spent on supports since this is a hollow shell piece. I'm hoping that printing it that way will make it easier to remove the supports. Another option would be using the glass bed heated at a higher temperature.

 

[CC Ricers]

My next attempt at printing the shell turned out much better. The extruder temp for PETG was correct, but the adjustments for the print bed needed tweaks in order to make it work. I actually printed two shells because the screen bezel was a little too tight in the first try, which causes visible bending. I loosened the tolerances for the screen dimensions and after that, it looks better.

The first of these revisions came out with almost no issues, but a slight imbalance in the bed made the left side's finish look rougher.

The second print (with the screen pictured above) was after another bed balancing and both sides are now more uniform in look and feel.

I went back to using a glass bed and despite what I've read from others printing on it, the PETG did not have problems sticking to the bed. I used the bare glass, no adhesives. The key was just warming the bed long enough. In case anyone is curious I printed on a stock Ender 3 with 238 degrees C set for the extruder and 90 C for the bed.

Now with these prints coming out well I can continue modeling the other shell pieces and make them fit to the top. (or maybe try out some other colors besides the boring gray ) The middle shell piece is where the custom electronics will go. I had originally decided to use protoboard, but its grid spacing limits how I will have to position the buttons, which is rather important for a gaming device this small.

For more flexibility I will need to either use a custom-made PCB, or the more crude but still effective method of attaching the tactile switches directly to the button caps. then securing them with backing boards screwed into the face plate. The backing boards would also be 3D printed and can be tightened or loosened to achieve the desired button push feel.

When I can accomplish this I can wire the buttons to the Teensy microcontroller and program the joystick functions.

 

[BaK]

I went back to using a glass bed and despite what I've read from others printing on it, the PETG did not have problems sticking to the bed. I used the bare glass, no adhesives. The key was just warming the bed long enough. In case anyone is curious I printed on a stock Ender 3 with 238 degrees C set for the extruder and 90 C for the bed.

Congrats on the tweaking and the print! The shell looks much sturdier now compared to the failed print, but still seems very thin!

 

[CC Ricers]

Congrats on the tweaking and the print! The shell looks much sturdier now compared to the failed print, but still seems very thin!

Yeah, the quality makes me more eager to try PETG prints in other colors. I think transparent filaments would look especially nice for this.
I was trying to find a balance with making the screen bezel at the top and bottom as thin as possible but without making the part too weak.

Most handhelds have a two-piece shell design which gives each piece enough wall material to be sturdy. But I went with a three-piece design because it's easier to iterate and make adjustments for a first attempt at something like this. I could have easily gone with making the shell two pieces but I wanted to reduce print times for each piece, and a simple box shell with posts to put the screws in can lead to more re-prints if you don't get the positions exactly right.

The middle piece will have the most thin walls but also attach with additional printed boards to hold the buttons in place. Put together, this middle area should be strong enough to not bend.

Right now I'll be using standoffs as much as possible to fasten all the pieces together. Once I have the finished product I could go ahead and work on a two piece design with more confidence that all the screw locations will be correct and use threaded inserts.

 

[BaK]

Thanx for the explanation!
I understand that once all will be attached together the whole assembly will be sturdy enough. Can't wait to see that!

 

[CC Ricers]

I got the middle piece printed last night, and after trimming the screen tabs with a Dremel, fitted it back into the top piece and attached the middle piece. Altogether there are six screws holding both parts together. It's now looking more like a finished console (even though there are still more things left to work on).

The screen fit is very good with no adhesives. It doesn't feel loose, even when shaking around the whole thing.

A front side view shows the speaker holes for the mono speaker in the middle. I think this will need re-designing too, because the walls between the holes are very thin.

Remember that the bottom piece is still missing, so this handheld will be ~5mm thicker than it looks here when it's done. But these two parts already provide a good mockup for how it will feel holding it in your hands.

The area for the shoulder buttons did feel weak, but it's less of a problem after attaching the main top shell piece. One area that is still flimsy is the top center of the screen. There is some flex in here because there are no supporting tabs like there are in the other corners or opposite side. A reprint of the top shell would have the additional tabs for a more secure, sturdy feel.

There's also a side port for the TF card on the left. but ports are still missing for the power button and battery charging, and auxiliary buttons for different input modes, so this is not the final version. For now the focus will be on lining the Lark Box motherboards and designing the parts to support the buttons from the inside.

I'll be getting a different color for PETG filament and try different colors to print out.

 

[CC Ricers]

The extra piece I printed for the screen also works as a mock bottom piece to get the full size of the console.

I realized there that it needs to be made at least 1mm thicker because the 18650 Li-Ion battery cells won't fit beneath the screen otherwise. The batteries are what is really determining the max thickness of the unit and it would be made more slim and sleek if the LarkBox used 5V instead of 12V for power. 5V makes it possible to use a single cell instead of three, so I could use a LiPo there instead of the Li-Ion.

The three cells are from the 12V power bank which has an additional board to distribute both 12V and 5V power. The 5V can be useful to power the CPU fan independently. I could not find the exact battery model but it has to be approx. 2600 to 3000mAh for each cell, if total capacity is 30 watt hours.

I also have a massive 10000mAh LiPo that I can use, and does work well with my Atom-powered stick PC. I also have tried using a boost converter for 12V input to the LarkBox. However, the LarkBox didn't like it. The PC would power on, and get to the boot screen logo, but then restart, and go through that boot loop over and over again.

The power might be spiking during the boot sequence which the LiPo doesn't like. The spec of the LiPo did say that it is limited to 3A current, which multiplied by 3.7V means you can only hope to power up to 11.1 watts. In an article comparing the power draw of a Raspberry Pi 4 to a Celeron J4105, the Celeron uses twice as much power when all 4 cores are in use. And the Larkbox uses a J4115 a higher clocked version of the Gemini Lake family.

Somewhat regrettably, had I bought the very similar GMK NucBOX the battery space issue could be solved more easily because since that uses 5V DC instead of 12V. All you would need there is a single 3.7V cell, negating the need for a balancer circuit, and with a more modest boost to 5V power everything in the console quite easily.

 

[CC Ricers]

It's been a month since my last update and there's been some bad news. I have made more progress in designing the case for the Larkbox after a few minor revisions and getting the thing to fit almost perfectly within the size constraints. But with installing the motherboards to test fit in the case, I probably screwed on one of the parts too tight and broke the board. The power LED doesn't turn on and there is no video output and I'm going to assume it's dead for now.

Here's how the case looks from my last prints:

As you can see, it's a pretty good fit. But with the computer no longer working I will have to go back to the drawing board, somewhat. I have had a few PC sticks I've tried out and will be using one of them as a replacement. It will require a redesign of some parts, but the top piece should be mostly the same and it will be easier to arrange the parts along the bottom side.

The computer I will now be using is the Intel Compute Stick with Core m3 CPU. The CPU is somewhat weaker than the J4115 of the LarkBox but its GPU is better. It does have more limited expansion with only 1 USB port and SD card slot, but it consumes less power and can be powered by 5V instead of 12V which will make managing battery power easier.

 

[Valantar]

It's been a month since my last update and there's been some bad news. I have made more progress in designing the case for the Larkbox after a few minor revisions and getting the thing to fit almost perfectly within the size constraints. But with installing the motherboards to test fit in the case, I probably screwed on one of the parts too tight and broke the board. The power LED doesn't turn on and there is no video output and I'm going to assume it's dead for now.

Here's how the case looks from my last prints:

As you can see, it's a pretty good fit. But with the computer no longer working I will have to go back to the drawing board, somewhat. I have had a few PC sticks I've tried out and will be using one of them as a replacement. It will require a redesign of some parts, but the top piece should be mostly the same and it will be easier to arrange the parts along the bottom side.

The computer I will now be using is the Intel Compute Stick with Core m3 CPU. The CPU is somewhat weaker than the J4115 of the LarkBox but its GPU is better. It does have more limited expansion with only 1 USB port and SD card slot, but it consumes less power and can be powered by 5V instead of 12V which will make managing battery power easier.

Oh, that sucks :/ Are the PCBs really that thin and fragile? That's really something to watch out for then. Hope the compute stick redesign doesn't require too much work!

 

[CC Ricers]

Oh, that sucks :/ Are the PCBs really that thin and fragile? That's really something to watch out for then. Hope the compute stick redesign doesn't require too much work!

I'm not sure exactly what broke because there are no cracks on the board. But I think one or more traces were damaged and that's what's stopping the PC from powering up.
Unlike that PC, I don't plan to remove the casing with the compute stick. It will go completely inside the handheld and I'm figuring out the best positioning for it in order to get all the other parts to fit.

 

[CC Ricers]

I finally have some good news now. The Larkbox is working again! I cleaned up the old thermal paste on the chips and wiped it with some alcohol. It powers up again and can get back to using it for the custom case. The only thing that stopped working was the power LED. This isn't a really a problem as there are other ways I can use to indicate power status.

 

[CC Ricers]

After some consideration, a few days ago I finally bought one of those 5.5' 1440p replacement LCD panels that @thewizzard1 had recommended me a while ago. I could only buy these from China so it'll probably take longer in getting this part. Nothing wrong with the 5" 480p panel I already have, but I am interested in a panel that has much higher resolution and slimmer profile.

The board, speaker and other components add to the bulk of the 480p panel, and will just compare the two to figure out which I'll end up going with. Modern desktop OSes also aren't easily navigable in 480p. I'll probably run the 1440p display at half the width/height or use some desktop scaling since the screen is small.

 

[CC Ricers]

The screen arrived much sooner than expected. I guess the arrival date of December 8 is a worst case scenario, but ordering parts from China it's harder to take a good guess. It comes with a DSI to HDMI controller board which is about as big as the screen is on its shorter sides, and with the included cables it's good to go. The default orientation is portrait (must be primarily used for phones) and it's no AMOLED but the colors still pop and the pixel density is really good.

You don't want to run a desktop at full resolution with 100% scaling because the text and UI will be too tiny. Booting to a Pop! live OS it automatically scales to 200% and you need to set the landscape orientation manually. Some other distros may not do the scaling automatically.

Here it is displaying Batocera. Again, manual adjustments are required and I need to go into the config files to rotate the display.

This screen is video only, no speakers and no touch panel but I think that is a worthy tradeoff to maintain a slim a profile as possible for the handheld. I have some tiny speakers I can wire up to the headphone jack and for touch, that's TBD. Maybe I will not go with a touch overlay and use a different form of input for controlling a cursor.

Now that I have this screen working I am curious about buying additional ones or better replacements. There are spare screens being sold for Voxelab 3D printers, and one of them looks exactly like the one I have in the display, circuit board, and ribbon cable. But it also has a tempered glass cover to protect it.

So I also bought one of these replacement displays for the glass cover. If it works, I will have a better display that won't scratch easily. If not, no big deal as it was just a purchase of $20.

 

[CC Ricers]

Found a third candidate for a screen, the 5.5" Sharp LS055T3SX05. This one is 1920x1080, still plenty of pixels for a screen of this size. Reason I'm considering this one is that it can be found with a LCD controller that has 3.5mm audio jack to wire it independently. But also, I have found a quirk with my current 1440p LCD, that's impeding the potential for this computer.

As I've previously said, the default orientation is portrait. While it's possible to switch to landscape mode via software, the only available resolution for landscape mode for is the native 2560x1440. For desktop use this isn't a problem, but the iGPU is too weak to run 3D games decently on that resolution, and if the game doesn't support internal resolution scaling, you're pretty much out of luck. I was hoping to scale down from 1440p to 720p full screen resolution, but this panel does not support hardware scaling. On Batocera I especially noticed this with the emulators for more powerful consoles- Dreamcast, GameCube, and Wii games felt more sluggish and no software config can override the hardware setup, it always sticks to 1440p. The games would mostly run fine and at normal speed on a 480p panel.

It is why I will consider the 1080p panel as well. Far fewer pixels to draw and comes with its own audio out in the controller board, while still being less bulky than the Raspberry Pi screens. In retrospect I should've gotten this panel instead since 1440p is overkill for the size. I just jumped on that one because it was more readily available.

 

[CC Ricers]

Well, the other LCD panel did not work as expected. It is probably functional, as, it could display an image, but it would be too dark to tell because the backlight does not work. Behind the screen, the three wires that power up the backlight were cut from the panel end, and this wasn't visible in the seller's photos because the dark glass overlay hides it.

So, I have an unlit replacement panel for a resin 3D printer that I do not own Perhaps I can send it to someone that needs one.

For other options I am also receiving a OLED touch panel from DFrobot. This one is definitely designed for maker projects, such as this sweet handheld Hackintosh. I know I've been spending more, but It's all part of the trial and error process!