-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
While I'm waiting for parts to arrive I've been taking some time to design the case. Here's a sneak peek render of what it might look like.
The case will be 3D printed and I think it just looks cool to have the insides visible in a transparent case. Not a final design of course, I still need to add face buttons and add a lot of changes inside for the individual parts.
The case will be 3D printed and I think it just looks cool to have the insides visible in a transparent case. Not a final design of course, I still need to add face buttons and add a lot of changes inside for the individual parts.
Would you be able to showcase it on Youtube once you are done? I have been looking for a template that I can follow for so long. I was planning on using Asrock's 4x4 Ryzen 4000u box. I have been plagued by how I should attach a portable power supply.
It looks like you are managing heat well, but the hot end of that heat pipe in contact with the LCD will eventually result in a discolored LCD.
My recommendation would be have a second fan blowing air (or sucking air) between and across the processor side of the motherboard, while the main heatsink cheats air in to cool the processor.
My recommendation would be have a second fan blowing air (or sucking air) between and across the processor side of the motherboard, while the main heatsink cheats air in to cool the processor.
Are you trying to build your own Aya Neo?
I'm actually considering one of those Asrock boxes for an upgrade later on. The 4500U box looks good enough for this case use. Though the Tiger Lake and Iris XE also look interesting.I might be able to record a video of it in action. I still have lots to do with setting up the non-stock cooler and batteries which I still didn't get yet. The PSU will probably be plugged to a daisy chained power jack that's separate from the stock NUC's. And as such will use a different wall PSU to ensure proper and safe charging.
I'm trying to picture what you mean with the fan placement. The processor side will face the back of the screen, there's barely room there for a second fan when it will be assembled. Are you saying that end of the heatpipe (that's directly above the CPU) will get more hot?
Last edited:
This mad lad tried to put the insides of a PSP into a larger handheld so he can game on a bigger screen. Wasn't completely successful but it's a good attempt. It got me some hints as to how I can make my handheld better as I go on.
He posted a link to a 7" touch screen on AliExpress that has a decently small control board that you can tuck away easily. The HDMI cable doesn't need to be routed as far as my current screen. Any screen that would have a connector that can be routed inwards is a improvement for me.
My HDMI thin ribbon cable did arrive and it's just long enough so that it can be routed into place properly. However even these adapters stick out 1 cm from the port, possibly to be compatible with all possible enclosures. Those who use these super slim cables for vertical SFF cases probably know by now. In a handheld system, the tolerances for tight spaces are even smaller. The 1 cm is enough to make it look a bit wonky, like a dongle that's clearly not part of the original case.
Some ideas to hide this protrusion are to make the entire case shell wider, from 110mm to 120mm. It will just not look as sleek as before. Another idea is to rotate the NUC board 90 degrees and reroute the I/O. But this adds even more cables into the mix and also the laptop cooler mount needs to be positioned differently.
He posted a link to a 7" touch screen on AliExpress that has a decently small control board that you can tuck away easily. The HDMI cable doesn't need to be routed as far as my current screen. Any screen that would have a connector that can be routed inwards is a improvement for me.
My HDMI thin ribbon cable did arrive and it's just long enough so that it can be routed into place properly. However even these adapters stick out 1 cm from the port, possibly to be compatible with all possible enclosures. Those who use these super slim cables for vertical SFF cases probably know by now. In a handheld system, the tolerances for tight spaces are even smaller. The 1 cm is enough to make it look a bit wonky, like a dongle that's clearly not part of the original case.
Some ideas to hide this protrusion are to make the entire case shell wider, from 110mm to 120mm. It will just not look as sleek as before. Another idea is to rotate the NUC board 90 degrees and reroute the I/O. But this adds even more cables into the mix and also the laptop cooler mount needs to be positioned differently.
A quick update:
I have the 12V 3A battery pack, and (with the help of an adapter plug) it powers on the system with screen without a problem.
Windows does not detect any use of battery power such as displaying a battery indicator or remaining charge. I don't know if it's hardware-related or something to do with the Windows settings so that's something I'll have to look into.
I also have a new SSD which it's now using with a fresh install of Windows.
I have the 12V 3A battery pack, and (with the help of an adapter plug) it powers on the system with screen without a problem.
Windows does not detect any use of battery power such as displaying a battery indicator or remaining charge. I don't know if it's hardware-related or something to do with the Windows settings so that's something I'll have to look into.
I also have a new SSD which it's now using with a fresh install of Windows.
Last edited:
Windows doesn't even know a battery is connected, it just sees it like any other power source, you'll want a battery pack which allows you to remotely monitor its status (if such a thing even exists). Personally I'd just pick a battery pack with its own status LEDs (or other indicator) and not worry about the OS knowing the battery status.
Unfortunately it doesn't work like this. If you take a notebook for example, the battery has a small chip inside reporting its status to the OS e.g. via SMbus (therefore notebook batteries have a few more pins, not just + / -)
But maybe you could use HWinfo or a similar software to check the battery status. Some NUCs can monitor the DC input voltage. If your NUC has this function and the powerbank output voltage does drop when discharging you could use this as a rough indicator.
But maybe you could use HWinfo or a similar software to check the battery status. Some NUCs can monitor the DC input voltage. If your NUC has this function and the powerbank output voltage does drop when discharging you could use this as a rough indicator.
Last edited:
I'll give HWinfo a try. The battery pack does have its own LED indicator status with 5 LEDs. I don't know how accurate it is, but it was 4 out of 5 LEDs when I first plugged it in before charging and measured with the multimeter to be 11.5V. I let it charge for a while until all 5 LEDs came on and the voltage went up to 12.5V.
It's been working fine so far. I had the desktop on for 10 minutes, then played a Steam game (Mirror's Edge) for about 20 minutes and power status went from 5/5 to 3/5. Then I put the PC to sleep to let it charge more overnight. I didn't yet try seeing how far it can ran before it powers off. In previous runs (before using the battery) MSI Afterburner on screen display showed the CPU using 15-17W during gaming. Then I have to factor in the motherboard, screen, SSD, memory, etc. I've yet to monitor how much Vdroop it has under these loads.
One nice thing about the battery is that it fits along with the motherboard almost like a perfect fit with no room to spare beneath the screen. This does make the whole unit a bit wider than I had planned but I like the footprint that it takes up. It lends itself to a more modular build.
The photo shows a view of this setup from upside down. These cables stick out from the top side when in normal use.
If I'm sticking to this layout, it will mean that the custom heatsink and mount will no longer be used because all these parts are completely stock. And hopefully even the power cables can be replaced with something more low profile to match the video cable.
The photo shows a view of this setup from upside down. These cables stick out from the top side when in normal use.
If I'm sticking to this layout, it will mean that the custom heatsink and mount will no longer be used because all these parts are completely stock. And hopefully even the power cables can be replaced with something more low profile to match the video cable.
Based on the layout from the last photo, here is a concept design for a portable system that is designed to be more modular.
It comes in at 32mm thick, making it one chunky boi. But it also makes the main body smaller, great for 3D printers with a print area limited by 200 x 200 mm.
Without the need of custom coolers, many NUC boards can fit in here quite easily and still have the front side open for any front I/O the board may provide.
Not yet in the design are side I/O ports which will provide a pass-through to connect controller inputs. Instead of seeing it like an extra beefy version of a Switch tablet I also like to think of it as a small, highly portable All-In-One computer. The controller parts are not as "snap-fit" like Joy-Cons but a semi-permanent fit that require screwing on for a more sturdy attachment.
I will also open it up for front SD card slots by directing the SD slot on the board with a cable. A common complaint I've read about some ultra-mini PCs is the lack of SD card slot.
-------
I have bought two Switch joysticks, breakout boards, a bunch of little momentary pushbuttons, cables, and a Teensy LC microcontroller. With these I plan to first build a rudimentary game controller which can later function also as a mouse/keyboard combo depending on modes of input.
It comes in at 32mm thick, making it one chunky boi. But it also makes the main body smaller, great for 3D printers with a print area limited by 200 x 200 mm.
Without the need of custom coolers, many NUC boards can fit in here quite easily and still have the front side open for any front I/O the board may provide.
Not yet in the design are side I/O ports which will provide a pass-through to connect controller inputs. Instead of seeing it like an extra beefy version of a Switch tablet I also like to think of it as a small, highly portable All-In-One computer. The controller parts are not as "snap-fit" like Joy-Cons but a semi-permanent fit that require screwing on for a more sturdy attachment.
I will also open it up for front SD card slots by directing the SD slot on the board with a cable. A common complaint I've read about some ultra-mini PCs is the lack of SD card slot.
-------
I have bought two Switch joysticks, breakout boards, a bunch of little momentary pushbuttons, cables, and a Teensy LC microcontroller. With these I plan to first build a rudimentary game controller which can later function also as a mouse/keyboard combo depending on modes of input.
Last edited:
Things just got more interesting. I got a deal for a Coffee Lake NUC for under $200. It has a discrete GPU, a RX 540, which would be about on par in performance with Vega 8. In other words, it should breeze right through most games at 720p.
That NUC has similar I/O including internal headers and a better power jack placement. However it has soldered RAM and will require modifications to the case design. Still...
That NUC has similar I/O including internal headers and a better power jack placement. However it has soldered RAM and will require modifications to the case design. Still...
Just a heads up, I received the Crimson Canyon NUC a bit ahead of schedule, on Sunday actually. I was surprised to see that it came with a 340 GB SSD by Intel, in addition to the 1TB hard drive. Windows 10 is pre-installed on the SSD with the hard drive completely empty and ready to use. I had to install drivers to get it gaming-ready.
Something to note is that with the portable display at least, the colors were all tinted green and purple as if the blue channel wasn't there. I thought the screen had broke, but it's the lack of a proper video driver. The Intel onboard graphics are disabled on this computer because it relies on the RX 540 so we're looking at an odd side effect with using the HDMI out without those drivers.
I first set up older Radeon Adrenalin software, but replaced it with Radeon Settings which gave me all the power tuning options I needed. Afterburner and Riva Tuner was missing a few items to monitor the GPU power and clock speed, but Radeon Settings overlay can cover the rest. At full power the GPU uses 25W, while the CPU can take up to 15-20W. Add in the other devices and it can draw at least 50W from the wall at stock settings.
It's a good thing this has more power limiting options and ones that are easier to understand than the Skylake NUC because this computer uses more power at stock settings. Pretty much how it came bundled with a 90W adapter while the other one only had a 60W one.
I reduced the GPU power limit by 25%, and the GPU obeys the limit well and it drops from 25W of use to ~19W. Clock speed is around 900-930 Mhz. A 40% reduction drops the power use down to 15W and clock speeds are around 650 Mhz. Even here, benchmarks run more smoothly than they did with the Iris 540. You could drop power by as much as 75% but this will be too much sacrifice in performance. I believe these are also the same range of power settings available for the Aya Neo's iGPU because I've seen videos of tests of its iGPU in modes ranging from 5W to 25W.
All things considered, this NUC definitely has more potential to show, but at the same time you have to be more mindful of its power usage. Its I/O is also fairly taking more space vertically so the handheld will need to be taller in some parts.
I will actually follow through completing the case and system for the Skylake NUC since that one is easier to fit parts in a smaller space, and it's also not as power hungry. Then I will use that as a transition to modifying the case for the other one.
Something to note is that with the portable display at least, the colors were all tinted green and purple as if the blue channel wasn't there. I thought the screen had broke, but it's the lack of a proper video driver. The Intel onboard graphics are disabled on this computer because it relies on the RX 540 so we're looking at an odd side effect with using the HDMI out without those drivers.
I first set up older Radeon Adrenalin software, but replaced it with Radeon Settings which gave me all the power tuning options I needed. Afterburner and Riva Tuner was missing a few items to monitor the GPU power and clock speed, but Radeon Settings overlay can cover the rest. At full power the GPU uses 25W, while the CPU can take up to 15-20W. Add in the other devices and it can draw at least 50W from the wall at stock settings.
It's a good thing this has more power limiting options and ones that are easier to understand than the Skylake NUC because this computer uses more power at stock settings. Pretty much how it came bundled with a 90W adapter while the other one only had a 60W one.
I reduced the GPU power limit by 25%, and the GPU obeys the limit well and it drops from 25W of use to ~19W. Clock speed is around 900-930 Mhz. A 40% reduction drops the power use down to 15W and clock speeds are around 650 Mhz. Even here, benchmarks run more smoothly than they did with the Iris 540. You could drop power by as much as 75% but this will be too much sacrifice in performance. I believe these are also the same range of power settings available for the Aya Neo's iGPU because I've seen videos of tests of its iGPU in modes ranging from 5W to 25W.
All things considered, this NUC definitely has more potential to show, but at the same time you have to be more mindful of its power usage. Its I/O is also fairly taking more space vertically so the handheld will need to be taller in some parts.
I will actually follow through completing the case and system for the Skylake NUC since that one is easier to fit parts in a smaller space, and it's also not as power hungry. Then I will use that as a transition to modifying the case for the other one.
Oh, I didn't realize the new NUC was Crimson Canyon/Cannon Lake. Certainly hope its performance and power efficiency are okay, given how utterly broken the 10nm process was at the time. At least it was a bargain! Looking forward to seeing this build come together. Btw, I think your 32mm thickness is pretty good in terms of ergonomics. Thinner might be better for packing away, but not for holding in human hands
Btw, for the controllers, have you considered modifying an off-the-shelf Switch accessory like the Hori Split Pad? I can attest for its ergonomics, button placements etc. being quite good, though I would guess hooking it up to a PC would be quite the undertaking.
Yea, 32mm might feel thick for some as a handheld system, when you compare it to mass-produced units, but it's still thinner than a 2-slot GPU (38mm-40mm). Plus a lot of DIY handhelds I've seen come in even thicker. The li-ion cells would also contribute to that, because not only do you need room for the batteries but also for the circuit board for the buttons and spacing them apart so that no shorting is possible.Oh, I didn't realize the new NUC was Crimson Canyon/Cannon Lake. Certainly hope its performance and power efficiency are okay, given how utterly broken the 10nm process was at the time. At least it was a bargain! Looking forward to seeing this build come together. Btw, I think your 32mm thickness is pretty good in terms of ergonomics. Thinner might be better for packing away, but not for holding in human hands
Btw, for the controllers, have you considered modifying an off-the-shelf Switch accessory like the Hori Split Pad? I can attest for its ergonomics, button placements etc. being quite good, though I would guess hooking it up to a PC would be quite the undertaking.
The controversial Smach Z prototype has its battery pack in an electrical 4S configuration, but physically split into two pairs of two cells on opposite sides.
This makes the handles thicker and more rounded so to their credit it makes ergonomic sense. It also divides the current between both pairs and these are properly balanced packs.
This 2+2 setup is one of the two possible battery configs I will try to use, the other being just group all 4 batteries together underneath the screen.
As for using a third party controller, I'm going to skip that approach for now. I've already bought a bunch of tactile switches and spare Switch joysticks along with some breakout boards. These are the some of the boards I got and they are tiny, half the size of each joystick!
I also have these generic FPC boards, which are somewhat bigger but still possible to wire up.
So first I will attempt to completely integrate a mouse input using a microcontroller to connect the joystick.
The custom program will eventually support several operating modes. Borrowing an idea from another PC handheld, the GPD Win 2 has a switch in the middle that lets you go between "Mouse" and "Gamepad" modes so you can have a good range of inputs suitable for a PC. Even without a touch screen, you should be able to navigate the cursor comfortably with a joystick.
For making the controller hardware, I decided to use the Teensy LC microcontroller as it can be used as several hardware serial peripherals, and has more than enough I/O pins for the controls. I've only had it for a couple of days but programming for it is very straightforward for me, having used other Arduino based boards.
The main obstacles came with wiring the hardware itself with the buttons and joysticks. But I got most of it sorted out yesterday and I now have a mouse peripheral that can be controlled with a Joycon stick.
Aside from the construction of the hardware itself, the main remaining goals is to finish programming and wiring up all the controls, and making the UX as good as possible. For example, while I successfully made the joystick button into a left mouse button, it's really awkward to use in practice. Even accounting for a "dead zone", it is hard to keep the cursor still while pressing the button. It would be better to map the mouse buttons to shoulder buttons.
Pictures will soon be added to this post.
The main obstacles came with wiring the hardware itself with the buttons and joysticks. But I got most of it sorted out yesterday and I now have a mouse peripheral that can be controlled with a Joycon stick.
Aside from the construction of the hardware itself, the main remaining goals is to finish programming and wiring up all the controls, and making the UX as good as possible. For example, while I successfully made the joystick button into a left mouse button, it's really awkward to use in practice. Even accounting for a "dead zone", it is hard to keep the cursor still while pressing the button. It would be better to map the mouse buttons to shoulder buttons.
Pictures will soon be added to this post.
Project PortaNUC just passed a big milestone- a custom battery pack has been made and it works very well with the computer. Using four 18650 cells in a 4S configuration, I can now supply more power to it and can last much longer on a single charge.
The batteries are Samsung 32E, 3.2 Ah cells that can handle up to 6 amps of continuous discharge. This is well above the maximum power these kinds of NUCs tend to draw, even when gaming and benchmarking. In total I can get about 47 watt-hours, which is at least 40% more than what is available in the 12-volt pack.
When I first got the batteries in the mail, they arrived in small plastic cases for 2 batteries each. I then checked their voltage, which was 3.8 V for all of them which is high enough to do a power test, but they were not fully charged. After this test, I used a 5A power adapter designed for battery packs like these, and they were fully charged in 20 minutes.
Here are two of the batteries in one of the cases they shipped with, and the balance board and 2S battery holder. Each of them came with two bare wire leads.
The battery holders are chained together for a 4S configuration, plus additional wires are added to fully connect the balance board.
Here is the PC powering up with the 4S battery pack. (you can also see the USB microcontroller in the upper right corner)
The batteries are wired up to the balance board and has a single barrel jack which connects with a cable to the NUC. I am using the same Y-splitter cable from the previous pack which allows me to charge and use the computer at the same time.
The other part I made progress in is with the custom controller. Most of the challenge turned out to be in the hardware itself- these parts don't come with instructions and had to look up guides online on how to read the pinout from the joysticks. Also, these wires are very temperamental on the breadboard. Most dropped inputs or stuck inputs were a result in a sloppy positioning of the wires, more so than software bugs.
The console modding store BitBuilt sells custom joystick breakouts which are much smaller than the generic FPC boards. I used the generic ones for testing here because they used the standard wire pitch of 2.54 mm and the small boards use 2 mm.
The wiring setup, which eventually turned into a rat's nest of wires when I started adding pushbutton switches to add more functionality to the controls.
If I were to do a complete test controller, I will really need to consider a protoboard with soldered wires, or perhaps even custom PCBs since for this use case they're not all that expensive. Given that most of my problems with these controllers have been with loose or finicky wires, changing the software for a good setup and calibration is much easier to work out.
The batteries are Samsung 32E, 3.2 Ah cells that can handle up to 6 amps of continuous discharge. This is well above the maximum power these kinds of NUCs tend to draw, even when gaming and benchmarking. In total I can get about 47 watt-hours, which is at least 40% more than what is available in the 12-volt pack.
When I first got the batteries in the mail, they arrived in small plastic cases for 2 batteries each. I then checked their voltage, which was 3.8 V for all of them which is high enough to do a power test, but they were not fully charged. After this test, I used a 5A power adapter designed for battery packs like these, and they were fully charged in 20 minutes.
Here are two of the batteries in one of the cases they shipped with, and the balance board and 2S battery holder. Each of them came with two bare wire leads.
The battery holders are chained together for a 4S configuration, plus additional wires are added to fully connect the balance board.
Here is the PC powering up with the 4S battery pack. (you can also see the USB microcontroller in the upper right corner)
The batteries are wired up to the balance board and has a single barrel jack which connects with a cable to the NUC. I am using the same Y-splitter cable from the previous pack which allows me to charge and use the computer at the same time.
The other part I made progress in is with the custom controller. Most of the challenge turned out to be in the hardware itself- these parts don't come with instructions and had to look up guides online on how to read the pinout from the joysticks. Also, these wires are very temperamental on the breadboard. Most dropped inputs or stuck inputs were a result in a sloppy positioning of the wires, more so than software bugs.
The console modding store BitBuilt sells custom joystick breakouts which are much smaller than the generic FPC boards. I used the generic ones for testing here because they used the standard wire pitch of 2.54 mm and the small boards use 2 mm.
The wiring setup, which eventually turned into a rat's nest of wires when I started adding pushbutton switches to add more functionality to the controls.
If I were to do a complete test controller, I will really need to consider a protoboard with soldered wires, or perhaps even custom PCBs since for this use case they're not all that expensive. Given that most of my problems with these controllers have been with loose or finicky wires, changing the software for a good setup and calibration is much easier to work out.
Last edited:
I haven't done any long-term tests yet. I will try to play some AAA games on it when possible and see how long it will run (Steam is currently being weird with wanting updates for games already installed in offline mode).
Yeah, I have to do the coding to get the inputs to read. The board is a Teensy LC which is a good low-cost solution when you want plug-and-play hardware peripherals. With the help of the TeensyDuino libraries I can get prototype code running quickly.
Similar threads
