[solved] A strange boot issue with H2

sdack

Hello,

I own an H2 board and have run into a strange boot problem.

I can boot the Ubuntu image for Xfce (Bionic 4.19.31 2019/03/25) and get it to run at HDTV resolution with connected ethernet and where it starts updating itself. I can open terminals, Firefox and I can see the unattended upgrades process running in the background, etc.
Similar can I boot Armbian (5.83 Debian Stretch 4.19.38). Here I can watch the boot messages and login as root after giving it the required new password and a new user account.

The obscure part is that it will only do all of this when I plug keyboard and mouse into specific USB ports and it cannot be any keyboard or mouse, but it has to be a specific setup. The keyboard and mouse it is working with is a Corsair K65 RGB keyboard and a Zalman mouse where the mouse is connected to the USB utility port of the keyboard.
Because it is a programmable LED keyboard does it draw more power from USB than a standard keyboard, which is why I'm trying to use other keyboards with it. In the end do I want to run the device headless and without a keyboard or mouse, but it won't boot when these are not connected.

I'm aware of power supplies sometimes being too weak, but since I can get it to boot do I assume it is not a problem with the power supply nor the SD card or heat. All tested mice and keyboards work correctly on a standard PC. A checksum verification of the SD card with the downloaded image checks out correctly. The amount of heat produced is slightly above "hand warm" and is being conducted away from the chips. I don't have any voltage meters around and so cannot measure the exact power draw, but seeing how it boots a graphical desktop with connected ethernet, hdmi monitor and plus a power-hungry keyboard repeatedly, as well as run it for more than half an hour, am I assuming the power supply is working correctly.

Does anyone know why the Tritum SBC is being picky in regards to the connected keyboard and mouse and why it won't boot when these are not connected?

loverpi

Do you "see" anything when you boot without USB keyboard/mouse attached?

sdack

No, I don't get any signal on HDMI. I do see the internal red LED as well as the two LEDs on the ethernet port being lit, but I don't get any life from it.

sdack

I think I've solved the issue.

I have noticed that the internal red LED radiates a solid, smooth light when it boots successfully. While during the failures it looked as if it had a bit of noise to the light. I'm assuming that either the H2 board or the power supply causes a bit of noise on the power line under low load. This may explain why under a high load it boots with everything working. It may also be that the Corsair keyboard, being an expensive gaming keyboard, comes with its own stabilisation of the 5V power line. It may explain why it requires the utility USB port of the keyboard to be connected as well (with the mouse on it).

I've opened an old, unused D-Link router and unsoldered the first nice looking capacitor, which happens to have 470uF, and now connected it between pin 2 and 6 of the GPIO header (between 5V and GND). This seems to smooth the noise out of the power line. Don't ask me for the math or the electronics. I haven't touched a soldering iron in perhaps 30 years, but I still seem to have the skills. *lol*

I'm going to run it like this for a while and see if it holds up. So far does it boot headless and with hdmi, and with and without keyboard or mouse.

By the way, according to the specs does the board have a 10 uF capacitor between 5V and GND. Could it be this is a little bit too low and it's what's causing people trouble with some power supplies?

loverpi

Depends a lot on the design of the power supply. Which power supply are you using? Quality varies drastically and cheap ones coming out of China do some scary hacks to cut costs.

sdack

Well, that's how it is. Nobody wants to spent money on a simple thing such as a power supply or cares for what name is on a power supply. They then all come out of China. The makers are forced to compete solely based on the price, which means they either keep pushing it or they get pushed out of the market. It's neither a pretty or a perfect world for them.

This is something Libre Computer should be looking into, because it may help them to gain an edge among their competitors. Companies such as ASUS for example did start by making a name for themselves by providing reliable hardware and going the extra mile. Providing hardware, which is able to forgive intolerances and comes with a high reliability lets a maker gain popularity and trust among customers. It simply causes less worries for a customer when they know a product just works and doesn't first require them to buy a special power supply. And let's be honest, blaming random power supply makers in China just doesn't look as good as providing reliable hardware.

Anyhow, I'm sorted here and all happy. I hope Libre Computer keeps providing these boards, and picks up the above mentioned quality traits. They've started strong and I hope they continue strong.

Increasing the size of the capacitors between the power lines (5V and 3V) and GND shouldn't be much of a cost factor and can help a long way, especially since people will want to add a variety of hardware from different makers to the GPIO port and do all sorts of things with it.


Here is a picture of it. A 1 mm copper plate is used as a heat spreader and is glued onto the chips with thermal tape. It conducts the heat onto the two "legs" of the aluminium case and the case serves as a heat sink. The capacitor, which is connected with two wires to pin 6 and 2, can be seen here, too.

loverpi

1) Gotta be careful when you slap a copper plate on like that. There's components with z height higher than the SoC and DRAM. You will definitely short it and cause board problems.
2) Capacitors won't help. Capacitors have really low capacitance that do not filter voltage drops > 0.00001 seconds. You need a battery or ultra cap if your power supply sucks and even then it most likely will cause low enough voltage to cause the running system to fail. Blaming the power supply is correct and only way to rectify the situation. You can also blame physics.
3) Asus Tinkerboard is horribly designed since it draws > 2A and uses MicroUSB. The voltage drop is so significant that you can run a program and crash the board no matter what power supply you use. Try running cpuburn-a7 on any Tinkerboard and let me know how it goes. With the best power supplies, USB devices will still disconnect due to low 5V rail. Let's not talk about the quality of their design.

sdack

To 1) and 2) can I tell you that it works and so proves you wrong. And yes, you shouldn't do it with a plate that is too large, but this has been true for all heat sinks in any computer. It doesn't matter if it's a 1 mm plate or an entire copper block. It's just common sense to pick the right size. If the boards had components in between the chips that were too high then it would cause a problem with any flat heatsink and so require a specialised heatsink. That's not the case here.
To 3) can I not comment in regards to ASUS's Tinker board, because I don't have one and I did buy the H2, because I didn't want a fast SBC, but there certainly are power supplies on the market with 3A output and more. All power supplies I have ever held in my hands do label voltage and max current for this very reason, including the no name power supplies. Again, some decisions one makes only require common sense.

I then wouldn't call ASUS's Tinker board horrible, but I do share the opinion that speed and some other factors shouldn't be primary factors for SBCs. There's already a lot of other hardware on the market that outperform SBCs, such as a standard PC to name an obvious one. I do however believe that SBCs need to be reliable and robust and able to work even under more rugged conditions than lab conditions and office conditions. This means it also shouldn't be picky in regards to power supplies, but this is just my opinion.

loverpi

I am a professional electrical engineer. Capacitors act as low pass filters. They prevent sudden changes on the voltage rails based on the capacitance which is very small. Attach a oscilloscope to the 5V line and you will see that when the power supply goes out, it will still cause enough of a transient voltage drop to cause system errors.
Just because the board still works or seems to work is not an indication that there's no data corruption and other errors so if reliability and stability are your core concerns. That solution is not it. Getting a proper power supply is the correct solution. Not adding capacitors to the board.

sdack

No, I'll just buy a proper SBC next time. Everyone knows power supplies can be tricky, so I'll better find an SBC that's more solid. I shouldn't have bought an early (or the first?) revision of your boards. Thank you.

I want to be able to plug an SBC into wall plugs such as these and not first need to buy an "Apple iPowerSupply" or whatever to run it.

loverpi

Just curious, which SBC do you deem "proper"? There are no small SBCs with the capacitors you mentioned.
You are entitled to your opinion but we cannot support your conclusion since it is flat out wrong from electrical engineering standpoint. Proper power supplies and stable voltage rails are necessary for any computer big or small.

sdack

That's what I'm trying to find out, which is why I've bought the H2 for evaluation and it obviously isn't.

Take the Corsair keyboard then as a counter example. It has it's own processor builtin and runs entirely off USB. It filtered out the noise and allowed it to run as well allowed your SBC to run. The people at Corsair are obviously smart enough to foresee real world conditions and built a keyboard, which runs even when there is noise on the power line.

The power supply may be at fault, or it may not. But both power supply and your SBC failed to filter the noise. So if you want to call my power supply improper then you have to call the SBC improper, too, and only the keyboard got it right, or else your just a being a twat who tries to feed their customers BS.

Or take the D-Link router from which I've taken the capacitor. Although the router is mainly builtin from smd components did D-Link include full size capacitors, one which sits directly at the power line and also smooths out noise I'm assuming. They didn't just leave it to the power supply to filter the noise and didn't built a device only from smd components like a cheap no name manufacturer. They're known for building good hardware.

D-Link and Corsair are popular brand names and they didn't get their popularity by accident. You on the other hand are new and are still trying to make a name for yourself. You won't be making one however if all you do is talk, blame and feed your customers BS. You are also neither grateful for the feedback you've been given nor have you been helpful to me and all you've done so far is to blame Chinese power supply makers and assume my power supply was at fault.

Good customer support is passionate about their customers. They're grateful for feedback, they collect it and pass it on to the engineers who will then consider it for their next revision. Good customer support is always kind and friendly and they don't take stabs at competitors like you did by calling i.e. ASUS Tinker board horrible. They provide proper customer service. You're just not that good my friend.

loverpi

Our comment about the ASUS Tinkerboard is factually correct and testable. It is using high powered out-of-order cores and a power connector that isn't able to handle the current necessary to run at full speed.
Your opinions are incorrect and are not accurate or relevant. Keyboards are simple devices and the microcontrollers are flex voltage anyway. They can disconnect and reconnect without issue since they don't hold state.
This is not customer service. We are professional engineers and not feeding you "BS". Banning for violating rules: https://forum.loverpi.com/discussion/593/general-guidelines#latest