On Board Computer Clock Failure v2

[CarinaPDX]

I can appreciate your reluctance: I delayed switching for years (from a Mac). But I finally got fed up with the planned obsolescence that is an increasing part of the Apple strategy; if I have to pay top dollar for my computer I expect it to last beyond a few years, and to not have to replace it because of Apple's decision to abandon it. I bought a high-end Windows laptop, which I knew I would not like (and it more than met that expectation). So I installed Linux Mint Cinnamon desktop in a partition on the internal SSD, and once started I never looked back. My Windows boot has been broken for at least 6 months now; I'll get around to fixing it at some point but clearly I do not rely on it.

The driver issue is interesting. In the past Windows had the broadest support for devices, and I think that is still the case. The Mac had problems because many developers simply didn't support their products on the Mac and so didn't write the drivers for it. Linux was in a similar position but through the years has either added open source drivers or in some cases is supported by proprietary drivers. Several of the devices I used with the Mac required using Linux/Unix drivers to work (thanks to the Mac being Unix). Today the situation has changed: orphaned products often don't have drivers available for the latest releases of Windows or MacOS, while Linux continues to offer the older drivers. It is often the case that Linux is the only platform for older devices. I use a wide variety of devices on my computers and I currently know of only two driver issues: the internal web cam has occasional horizontal lines (which I am hopeful is resolved in the newer releases - TBD) and my Canon scanner is not supported (but then again it is no longer supported by any OS AFAIK). When the Mac stopped 32-bit support the older drivers no longer worked and many were never re-written for 64-bit; I believe Windows 10 is in a similar situation. The main distros of Linux are also dropping 32-bit kernels, but updating open source drivers is at least possible without vendor support. Things are still in flux but so far I find that Linux is better than a Mac for driver support. Windows is best in that regard, but even with little usage Windows maintenance wasted more of my time than Linux.

Someday when you have some time make a list of the devices you need support for, and then check for Linux drivers before proceeding. (It is probably easier to make a bootable flash drive, plug in a device and see if Linux can find the driver.) While you are at it take note of how many devices you have that are no longer supported by Windows.

Until then I do understand the reluctance. And in the end I can't guarantee that you will find it an improvement, or that you will find the drivers you need. But I do know the pain that is Windows - sorry if you feel stuck there.

Greg

[Reefmagnet]

Motherboard clock batteries, be they button cells or super capacitors, will fail more often in applications where the computer spends most of it's time switched off. Some motherboard designs seem to not suffer this malady, but others will drain the battery quick smart if left off for long periods.

I'd never use a 12 volt PC directly on a 12v boat's DC supply. I'd be inclined to suggest it will likely be unreliable in the long term. The only PC I've found so far that will run off the boat's 12V DC supply happily is an Intel NUC designed for 19VDC input. Ironically the 12VDC Atom based NUC's won't run off the boat supply - they complain and refuse to boot. Go figure! New NUC's are as dear as poison, but used ones are cheap and a simple SSD hard disk upgrade will see them run happily for years. On my boat I have a NUC I3 which, IMO, beats the pants off Celeron/Atom based mini PC's and uses minimal power (about 10w on for average use). For low power applications I use the Raspberry PI which is perfect for non demanding stuff.


As to Linux vs Windows, it just happens that I don't have one I3 NUC, but 2. One is running Windows 7 and the other Linux Mint (although I've recently become a fan of Linux Lite for low end PC installs). For general use on a boat, both OS's are fine - with caveats.



Windows has the advantage that it's familiar to most people with the notable advantage in a real world application that there is a lot of software that will run on Windows but not Linux. For example, ProAIS software used by many AIS units for testing and configuration does not have a Linux equivalent. Windows main disadvantage is that it is heading in a direction that assumes more and more with each update that it has an unlimited internet connection - something difficult to provide on a bona fide cruising boat!


Linux, OTOH, has a more hands on approach to setup and use that may not appeal to everyone. On the plus side, it is less demanding of hardware, is extremely customizable and doesn't need to constantly "phone home" if one chooses for it not to. It's also more robust, and you're not likely to have your Linux powered PC based chartplotter reboot at 2 AM for a Candy Crush update just as you are midway through that tricky channel in inclement weather.

[a64pilot]

Originally Posted by Dockhead

Sounds reasonable. But if that's all true, then why would I have nuked TWO different system clocks?

Maybe you didn’t, but I am certain that all computers run off of low voltage, much lower than 12V and the power supply should be what gets cooked and nothing past it.
I would smear a bit of silicone grease on the coin battery, likely it’s a 2025 or 2032.

[Reefmagnet]

Originally Posted by a64pilot

Maybe you didn’t, but I am certain that all computers run off of low voltage, much lower than 12V and the power supply should be what gets cooked and nothing past it.
I would smear a bit of silicone grease on the coin battery, likely it’s a 2025 or 2032.

Computer subsystems run a variety of different voltages. The issue is that most of these mini computers are designed to run off an external regulated power brick and the designers of the motherboard base the input voltage design and protection on the specifications of the brick they choose to use for the PC. Remove the power brick from the system and all bets are off.

[a64pilot]

Sure but even without the brick there is still a power supply, and it should reduce voltage to whatever is required, but feed the computer voltage out of spec, the power supply should be what goes down, not the CMOS clock.
CMOS clock inop, if the battery hasn’t been changed, my bet is bad battery. I wouldn’t say the clock is bad not until after a known good battery was installed.

I know what your saying about a power brick, my first TV was 120VAC, but had a 12V power brick, I cut it off and ran the TV off of my battery, which was always, always above 12V.

If that is what he did, as in bought a computer and tossed the brick, then I’d say he ought to get a power supply that outputs 12V. I thought he bought “12V” computer meaning one that was meant to run off of 12VDC, and 12VDC items are really most often 14VDC. 12V boats most often have 12V batteries that are fully charged when at 13v, and the electrical system is 14V, so 12V is really a misnomer, because nothing is 12V.
24 is even worse, cause that’s actually a 28V system.
That’s something that I never got most pilots to understand.

[Reefmagnet]

Those chunky power bricks tend to be fairly precise with their output for the most part. And, as you say, this puts devices designed specifically to run on them at odds with a battery supply which could be anywhere between 11 and 15 volts and the possibility of including an assortment of random electrical noise. Cheap consumer computers are - no doubt - designed and built to a price point so it's not unreasonable to assume they may be lacking in additional componentry to handle connection direct to the boat's house bank.



The reason why I think the 19VDC NUC's work so well on a direct battery connection is because they weren't designed to be cheap and compete with media centres and part of the design criteria is suitability for industrial applications. It makes some sense that they would have been designed in to handle, among other things, more variable input supply quality.

[muttnik]

Originally Posted by chris95040

But I’m not sure they are actually getting time of day from it or just using some evenly spaced pulse train from the GPS to improve accuracy.

Both. What they're describing is creating a stratum 1 clock using gpsd with a GPS receiver with a PPS output (and ntpd/chrony). GPS sentences give coarse time which not least due to limitations on serial transmission of an ascii string will be offset and contain a huge amount of jitter. PPS gives accuracy. Not only will this allow the machine the GPS is hooked up to to get accurate timing, it can be used as a network time server. So have this running on a raspberry pi and you can feed accurate time to your other onboard devices (incl. windows pcs) as well as GPS

You don't have to use gpsd: a few years back I created a stratum 1 clock using GPS data fed to ntpd from kplex for coarse time and kernel pps. However the gpsd method is well documented and straight forward.

[CarinaPDX]

Yes, it is easy enough to create a network time server with a Pi - my brother has multiple Pi's and GPS modules collecting data. Most GPS devices that consumers own don't have a separate PPS signal (pulse per second) leaving the unit. Typically these used serial 0183, often over USB, for the GPS data and a second serial line for the pulses if available. Today devices often implement PPS through the same USB as the data stream, and GPSd at least can support that. In order to get the best results the Pi needs to be in a temperature-controlled box (box with a thermostat and small bulb) and should be using a minimalist Linux install. Using more "capable" computers with larger OSes and more complex hardware yields worse results, as interrupts interfere with timely processing of PPS. Of course I am talking about lab-grade timing here; I can't think of anything on a boat that would require this level of accuracy. It is impressive that a $20 GPS attached to a $35 Pi can replace laboratory time sources that cost in the many thousand dollar range.

Greg