Why do things break?

I can understand disc drives eventually losing alignment, because rotating spindles may wear out bearings unevenly, but of the three computers on my desk the only moving component is the fan add-on that I bought for the Rpi4 (Pimoroni’s fan-shim). So what makes things break in devices with no moving parts? Is it changes in temperature, invisible cracks that widen, manufacturing inexactitudes? Perhaps no moving parts is too naive a notion seeing that microscopically everything is in constant motion?

I remember that about 40 years ago one Ivor Catt started a company called Anamartic (error-free in Greek) which was to produce self-healing memory chips. Is anybody producing circuits that can be scanned for estimated lifetimes, which can flag up when some part is about to fail? Machines that are grown sound like a better bet.

So what makes things break in devices with no moving parts?

Off the top of my head:

1. Temperature changes, particularly rapid temperature changes. The cumulative effect could directly damage a component, or perhaps distort it such that it no longer had the same electrical/magnetic properties.
2. Other environmental factors. For example, early epoxy-resin chips suffered from gradual water ingress in humid conditions: while this happened at an acceptably slow rate for domestic use, they were considered unsuitable for military projects.
3. The chosen technology may rely on ion movements (the classic example being a nickel-cadmium battery which would eventually fail due to the formation of nickel “whiskers” on the nickel electrode that shorted-out to the other electrode).
4. High levels of ionising radiation. Again, not something that’s important in domestic settings, but would be a consideration for (e.g.) spacecraft.

Is it changes in temperature, invisible cracks that widen, manufacturing inexactitudes?

Yes. Temperature cycling can weaken solder joints. This tends to be an issue on PoP devices (like the original Pi and Beagle) where the memory is stacked on the processor which is soldered to the board. Imagine, there are something like 70 connections between the two under there, and maybe something like a hundred between the SoC and the main board. The entire thing is the size of your thumbnail.

What has failed, and how do you know it is that part?

The wifi on my (Pine/ARM)book ceased working. I had successfully updated Manjaro-ARM using the wifi connection. The next time I used it there was no wifi – no access points shown. To begin with I did not know whether this was a hardware fault or some change of setting. I tried various Linux commands but they all reported that wifi was enabled; it just did not work. There was no problem with the internet. I could connect using the ethernet2usb adapter. Nor was there any problem with my router, next to which the pinebook was sitting, because other wifi-connected devices were working. I logged onto the Manjaro website and one of their support people talked me through reloading the network manager. No change. They decided in the end that it was a hardware fault. I explained all this to Andrew Rawnsley who had sold me the machine. With it he had also supplied me with another Linux distribution, KDE-Neon, on separate emmc memory. He suggested that I swap that in and see if the wifi came back. It didn’t, so I guess it has to be a hardware fault.

Inside the Pinebook the wifi circuit board sits next to the somewhat larger cpu board, and each has a metal radiation shield over the top, with tiny feet that clip on. Both are held in place by black sticking plaster that looks as if it has just come from the bathroom cupboard. The only suspicious circumstance that my untrained eye can report is that when I opened up the case the wifi shield was just a smidgeon askew.

If the shield isn’t secured, it may have shorted out a component.

That has to be like that. Predetermined breaking point (Sollbruchstelle).
What doesn’t break a long time doesn’t bring us money.
(Was lange hält, bringt uns kein Geld.)

Was lange hält, bringt uns kein Geld.

Succinctly put. I hope to cure the problem with a Linux wifi usb adapter. Pinebooks are no longer available even second-hand. They sold for a ridiculously low price. Then Pine brought out the Pro version, with a slightly enhanced specification, for twice the price – still a bargain, though. When you look inside the case you see where the cost-cutting went. The size of a laptop is evidently determined by its battery and by its screen. The dimensions of the Pinebook are just right for me, and there still seems to be masses of empty space under the bonnet, just like a Morris Minor.

I hope to cure the problem with a Linux wifi usb adapter.

All Linux I use are working with an Edimax EW-7811Un mini dongle.

…on the long term range (decades) I remember from the Amiga and C64 forums that one big problem are “leaking” capacitors/condensers that have to be replaced, as they spill some chemical substance on the board even…I guess sometimes we forget those boards is a bunch of chemicals/metals soldered together and of course affected by temperature/humidity/aging…

there still seems to be masses of empty space under the bonnet, just like a Morris Minor.

This inspired me to go make you a little video: https://youtu.be/13dgRsPvKbk

If the shield isn’t secured, it may have shorted out a component.

Alternatively, the shield could have doubled as a heatsink, and if off centre, may not have made suitable contact to be effective.
WiFi contains transmitters. The chips run hot. I’ve had a Livebox fail (WiFi died once it warmed up) and my original Vonets cooked itself a few summers ago when it was hot. Even with a heatsink, if ventilation is poor and it gets hot, WiFi devices can overheat, sometimes terminally.

[I was okay this summer, I have an under-laptop cooling fan gizmo that sucks air from around the Vonets and blows it over the Pi/PSU to keep it all cool]

as they spill some chemical substance on the board even

Pretty much every RISC OS machine Acorn ever made, and the little 3V battery inside…

This inspired me to go make you a little video

Thank you Rick. I am charmed by the glimpse of Felicity’s boudoir and her engine compartment.

And thank you Raik for Edimax EW-7811Un. I am expecting Andrew to send me a wifi dongle. Meanwhile I have a miniscule problem of no great importance. The Pinebook’s case is secured by 10 very small screws, four of which are shorter than the rest. One of them has disappeared, and probably fallen between floorboards. I have a neighbour who has a wizard’s cave lined with small boxes of resistances, racks of screwdrivers, coils of wiring, blow-driers, and gadgets of every description. He built his own portable telescope, and programmed its motor in Forth. I am pretty certain he will have boxes of tiny screws of every shape and size. I once asked a French friend what the French for oddments was. He said epluchure de ragotons . I may be missing a circumflex somewhere (obvious for many years some would say).

I suppose there are no pictures of the inside of the Pinebook!

Is there any info of the inside of the Pinebook 64 – differences between it and the plain Pinebook?

pictures of the inside of the Pinebook

See 2.44 minutes into this review made in July 2017.
The black rectangle at the bottom left covers the cpu board’s shield and the smaller silver rectangle below it is the wifi shield. On mine that is also covered in black sticky tape, which appears to be used liberally here and there at the discretion of the assembler.

The magic phrase you need is “teardown”.

Pinebook insides are here (towards the bottom): https://hackaday.com/2017/04/28/hands-on-with-the-pinebook/

Pinebook 64 is here (but it’s a bunch of links to Google Drive, how nasty): https://forum.pine64.org/showthread.php?tid=4523

Just a matter of interest it looks as if the Pinebook 64 can run in 32bit mode true/false?

Of course it can. Except some server chips (and perhaps some Apple chips too), all current 64bit ARM processors can do 32bit.

But can some of them not do user mode?

Just for interest what is the pecking order of start-up code of machines like the Pinebook?

And can Progs like uboot
be disassembled?

Depends upon the SoC. Usually some internal boot ROM will look for resources on removable media. This can be a bit braindead like the OMAP, or it can be smarter like the Pi’s BCM devices.

Why disassemble UBoot? Ask the manufacturer – it’s GPL…

Why disassemble UBoot? Ask the manufacturer – it’s GPL…

As I recall¹ there are variants of U-Boot around and the sites making those variants available tend to have a reasonably obvious link to the source.
That said I think the ARM mbed code includes a similar set of start up code that is a different licence² so development for inclusion might be easier.

¹ Last time I looked was somewhere around 2012. I was curious about what facilities U-Boot had that might be useful.

But can some of them not do user mode?

No the other way around! Some of them can’t do 32 bit privileged modes.

i.e. They have to run a 64 bit OS, but can support 32 bit user mode programs with suitable 32 bit libraries.

there are variants of U-Boot around

Oh, certainly. My (now ancient!) PVR runs a hacked version that displays a splash screen as the system starts, blinks the LED for… actually I forget now what that means, and also to look for, and handle, firmware updates from a special file on one of the media devices. Plus, of course, get the OS running. ;-)

I thought if I ever tried porting RISC OS “just because” (it’s an ARM9 clocking about 200MHz with a 640×480 framebuffer for composite video output in either PAL or NTSC), it wouldn’t be too hard to get the thing going using U-Boot.
[but I really don’t have the time]

so development for inclusion might be easier.

I suppose it depends upon the preferences of the people using the device. I don’t think either the OMAP3 or my PVR’s TMS320DM320 specify using U-Boot. The PVR just starts loading what is at the beginning of the flash, and the OMAP looks for a file called “MLO” which is the first stage bootloader that starts up U-Boot. It can be changed… https://witekio.com/create-your-own-mlo-for-a-beagleboard-xm/

there are variants of U-Boot around

Oh, certainly. My (now ancient!) PVR runs a hacked version that displays a splash screen as the system starts, blinks the LED for… actually I forget now what that means,

Prompted by the conversation, I looked on the ARM mbed github and lo – there’s a version of U-Boot

the OMAP looks for a file called “MLO” which is the first stage bootloader that starts up U-Boot. It can be changed

Indeed, hence my reference to the mbed_OS which has the startup and basic HAL. Might be halfway useful for an alternate dev board startup.

So what makes things break in devices with no moving parts? Is it changes in temperature, invisible cracks that widen, manufacturing inexactitudes?

My rule of thumb is that any fault is far more likely to be software than hardware because of its far greater complexity and capacity for non-deterministic behaviour. I’ve met too many software engineers who erroneously conclude that a fault ‘must be hardware’ just because their software works on one instance of a device but not another.

This, however, is a genuine limitation upon the lifespan of the chip internals which I still find incredible even years after learning of it at Broadcom.