Forma clock on forma

[hobnail]

Does the calibre Kobo driver keep the clock on the Forma updated? I have wifi turned off on my Forma and yet the clock seems to be very accurate. I'm used to devices' clock drifting if they don't use some time service (ntp, etc.).

[davidfor]

No the calibre drivers have no interaction with that. As far as calibre is concerned, the device is a USB MSD. Calibre uses the device id to identify which ereader it is, and hence which driver to used to read the book details, but, otherwise, it is just a drive.

It could be happening at the OS level, but, I doubt it. Honestly, there is no excuse for the clock in a device to be less accurate than your microwave or a non-smart watch.

[compurandom]

Originally Posted by davidfor

It could be happening at the OS level, but, I doubt it. Honestly, there is no excuse for the clock in a device to be less accurate than your microwave or a non-smart watch.

The technology to make small stand alone digital clocks that are accurate within a second per year has existed for at least 30 years. Why we have PCs that require network time synchronization to prevent their clock from drifting by minutes per day mystifies me.

[NiLuJe]

FWIW, Nickel does an ntp sync at the very least during its startup process. And I imagine as part of a sync, too.

[hobnail]

Originally Posted by compurandom

The technology to make small stand alone digital clocks that are accurate within a second per year has existed for at least 30 years. Why we have PCs that require network time synchronization to prevent their clock from drifting by minutes per day mystifies me.

I would have said that maybe it's interrupts from devices but on the Unix systems that I worked on the clock interrupts ran at the highest interrupt priority. Maybe the clock interrupt routine sometimes takes longer than an interrupt cycle; on the original Unix systems it was 60 times a second (which may seem like a lot of time but remember that cpus were much slower back then).

But then again, that was 30 and more years ago. I could bend your ears with stories of reel to reel tape drives and disk drives the size of a washing machine and whatnot.

[compurandom]

Originally Posted by hobnail

I would have said that maybe it's interrupts from devices

That's not raelly a good excuse. We have hardware real time clock chips, and and it should be trivial to sync your drifting system clock to that, but somehow that doesn't happen or the RTC is also broken.

I could bend your ears with stories of reel to reel tape drives and disk drives the size of a washing machine and whatnot.

I have a disk platter from one of those hanging on my wall. I think I destroyed all my 1/2" tapes but every now and then I find another one. Some times I'm sad I have no way to read them, but then, that's probably just as well considering the age of the data on them even if it is readable. I still have some 1/4" tapes I haven't gotten around to disposing of.

But now this is getting way off topic. I would hope the clock in the forma would be high enough quality to not need a sync more than once a month or so. I have never noticed the clock in any of my kobos being off, but then I sync fairly often.

[hobnail]

Originally Posted by compurandom

That's not raelly a good excuse. We have hardware real time clock chips, and and it should be trivial to sync your drifting system clock to that, but somehow that doesn't happen or the RTC is also broken.

On the original Bell Labs Unix systems I was on the way it worked (as I remember) is that the OS set a frequency for the clock to interrupt, and that's all it did was generate an interrupt 60 times a second. The OS then updated the time that was stored in the computer's ram. But that implies that you'd have to run the unix date command whenever the system had been down or rebooted and maybe I did but I don't remember that. What I mainly remember is that NTP was a godsend for keeping the clock correct. Although now that I think about it I do remember having the ntpdate command in the startup file to set the date/time.

[DNSB]

Originally Posted by compurandom

The technology to make small stand alone digital clocks that are accurate within a second per year has existed for at least 30 years. Why we have PCs that require network time synchronization to prevent their clock from drifting by minutes per day mystifies me.

Interesting. I've always found that unless you happen to have an atomic clock which requires a very high degree of temperature control, you are not going to see even close to that accuracy. Given that most atomic clocks are now kept close to absolute zero, small is not a word to describe them. The solutions commonly found are generally use either a quartz crystal or ceramic resonator. The accuracy is susceptible to temperature changes, aging and a shopping list of other factors. The accuracy is about 15 seconds per month. Yes, mechanical pendulum clock did manage 1 second per year long before the digital era. As does the Citizen Calibre 0100 watch though the $16,800 price and limited production make it a rarity. My current watch is able to do much better but then it's not really a watch but rather a receiver and displayer of radio time signals.

In my working environment, a drift of a minute is enough to cause IT's telephones to ring off the wall so I feel very grateful to the inventors of NTP and those folk who keep the servers up and running.

[compurandom]

Originally Posted by DNSB

The accuracy is about 15 seconds per month.

Even that is more reasonable than a lot of PC's I've seen.

I've hit more than one laptop and desktop that couldn't even keep drift under 3 seconds an hour.

[DNSB]

Originally Posted by compurandom

Even that is more reasonable than a lot of PC's I've seen.

I've hit more than one laptop and desktop that couldn't even keep drift under 3 seconds an hour.

I remember looking at the innards of one computer where the RTC was controlled by a ceramic resonator with, from the manufacturer's data sheet, an accuracy of ~0.5% or ~18 seconds per hour. Automachron to the rescue.

[hobnail]

The builtin windows ntp seems to work well enough for me. After installing a Windows system I change the server from time.windows.com to 0.pool.ntp.org. Not withstanding how long it takes me to find where that setting is buried.

[DNSB]

Originally Posted by hobnail

The builtin windows ntp seems to work well enough for me. After installing a Windows system I change the server from time.windows.com to 0.pool.ntp.org. Not withstanding how long it takes me to find where that setting is buried.

Automachron goes back to the NT days when Microsoft's time client was not very usable. I still had it installed on a Server 2003 machine when we decommissioned it in July. It only took 9 years for the beancounters to approve a replacement for the time scheduling/management program on it.

Personally, I generally use ca.pool.ntp.org when I'm entering a single NTP server.

[JSWolf]

My Dad has a digital alarm clock that keeps the time accurate using the OTA time signal. IT also means he doesn't have to change the clock's time twice a year.

[hobnail]

Originally Posted by JSWolf

My Dad has a digital alarm clock that keeps the time accurate using the OTA time signal. IT also means he doesn't have to change the clock's time twice a year.

I have two; one is a bedside alarm clock, the other is on the wall, a combination indoor/outdoor thermometer and digital clock. My understanding is that the signal is WWVB which is in Fort Collins, Colorado. I'm in the San Francisco area and it surprises me that the signal can be read all the way out here. And I guess it works for the East Coast as well. Wikipedia says, "Fort Collins is situated at the base of the Rocky Mountain foothills of the northern Front Range, approximately 60 miles (97 km) north of Denver, Colorado, and 45 miles (72 km) south of Cheyenne, Wyoming. Elevation is 4,982 ft (1,519 m) above sea level." How that signal makes it across the Sierras here in California beats me; ionosphere bouncing?

[DNSB]

Originally Posted by JSWolf

My Dad has a digital alarm clock that keeps the time accurate using the OTA time signal. IT also means he doesn't have to change the clock's time twice a year.

That's like my "watch" which is why I commented it was a receiver and displayer of radio time signals. The 60kHz signal is well below the frequencies most people are familiar with and from my watch documentation, it synchronizes once per day at ~10:00 UTC when the coverage from WWVB is near it's maximum.