Arduino internal clock inaccurate??

Hi Jon,

Interesting project!

Will it not damage the mechanism though? Maybe something like that is worth a lot of money.

You didn't say how accurate you wanted : )

Anyway, the Arduino clock is ok, but not spectacular. It may not be precise, and you'll notice time drifting slightly (perhaps a minute per week or a bit more - enough to be irritating).

A nice device is the DS3231, you could expect a minute or two of gain or loss per year; therefore in comparison the Arduino clock is an order of magnitude worse.

In practice the Arduino clock is more worse than that, because it is affected by change in temperature (the DS3231 is temperature compensated, so affected less).

Actually, after all that, you would have achieved a clock circuit that is still an order of magnitude poorer than the 250-year old Harrison clocks : ) which were only out by a second per month,

but the DS3231 circuit costs a tiny fraction of course and doesn't take a lifetime to build.

There are other ways to have precision clocks (e.g. GPS based or NTP based) depending on where you wanted to go with this.

Information on how to use the DS3231 is here. One important point is that it has 3.3V logic, not 5V like the Arduino Uno. So you may wish to use a different board than the Uno.

There are some 3.3V Arduinos, but I have not used them so I wouldn't be able to recommend any. There are Arduinos (and other single board computers such as Raspberry Pi - although this

could be considered over-the-top for a clock) with network capability or with Ethernet shields, so NTP could be used with them. A google search on 'arduino ntp clock' revealed some hits.

Thanks!

To answer the first question. "Will it damage the clock and maybe it's worth a lot of money?

I have plans to put the drive on the proper gear with the least amount of drag/pressure on the coupling pieces and gears in the clock.  Also on top of that, I will be making a rubber breakaway/absorption coupling in between. If something get's out of whack and jams the gears, it will separate. As far as the value. The clock was severely damaged in all other aspects other than the gear train and of course no parts available. LOL! I will make it far more valuable than it would have ever been as a collectible clock. To give you an example. the last project will be for sale for at least 6 figures. ;-)

As far as accuracy, a minute per a week wouldn't be bad considering the original would have been about a minute a day. LOL! But seriously, I like the sound of the 1 to 2 minutes per year but I will definetly need some resources to learn how to set the clock module up paired with an Arduino etc., which I'm sure I can find. Is the DS1307 basically the same thing but with the soldering points?

I do plan on purchasing 2 Arduino's as a permanent part of the project that is soldered in and smaller like the Pro mini 328. One will be the "Chime Arduino" and have a music program for the chimes. It will be signal'd by the main Arduino when the chime needs to be sounded off since there aren't enough outputs on the one Arduino for all the electromechanical devices in the project. Also I want to be able to leave the clock running and upload musical programs to the chime Arduino as I write them. On top of that I want to have some signal buttons to make tunes when I push them for show and tell. :-) 

I will also be making a voltage divider circuit to create the power needed for the different devices so making a 3.3v line won't be an issue if needed.

Cool stuff on the arduino ntp clock search!

Thanks again for all the time and info. It is very much appreciated!

Hi Jon,

Great, sounds like an excellent idea with the rubber coupling. DS1307 is more popular than DS3231 for hobbyists, so more chance of finding info on arduino/DS1307 projects. Less accurate than the DS3231, but more than the Arduino on its own, so a good compromise. The DS1307 integrated circuit is available on hobbyist-friendly soldered boards, so no need to solder the IC yourself, you could just get a board.

As Robert Peter Oakes says, regardless of the method you use to obtain your clock, you'll need a strategy to either directly drive your stepper motor from the clock (hard to do) or perform a compensation (e.g. daily) to correct for any time (probably seconds) gained or lost.

The way I'd possibly approach that (never tried it) is to keep a local count in a variable (i.e. still use your arduino clock and arduino events/interrupts or whatever constructs arduino's have to pulse your stepper motor), and compare with the RTC (DS3231, 1307, or whatever method you choose) daily (or some other interval e.g. hourly).

I'd also pre-determine an amount of time (e.g. 1 minute) that would make 1 second invisible to the average observer.

So, daily or hourly, after the comparison of local vs RTC time, if the delta is (say) one second or more, then for the next minute I would run my local clock at a speed that I can correct for 1 second. After the minute had completed, I would then compare again, and repeat, until the delta is less than one second.

Anyway, that's just an idea - maybe there are other more refined ways of doing it.

Hi Jon,

Great, sounds like an excellent idea with the rubber coupling. DS1307 is more popular than DS3231 for hobbyists, so more chance of finding info on arduino/DS1307 projects. Less accurate than the DS3231, but more than the Arduino on its own, so a good compromise. The DS1307 integrated circuit is available on hobbyist-friendly soldered boards, so no need to solder the IC yourself, you could just get a board.

As Robert Peter Oakes says, regardless of the method you use to obtain your clock, you'll need a strategy to either directly drive your stepper motor from the clock (hard to do) or perform a compensation (e.g. daily) to correct for any time (probably seconds) gained or lost.

The way I'd possibly approach that (never tried it) is to keep a local count in a variable (i.e. still use your arduino clock and arduino events/interrupts or whatever constructs arduino's have to pulse your stepper motor), and compare with the RTC (DS3231, 1307, or whatever method you choose) daily (or some other interval e.g. hourly).

I'd also pre-determine an amount of time (e.g. 1 minute) that would make 1 second invisible to the average observer.

So, daily or hourly, after the comparison of local vs RTC time, if the delta is (say) one second or more, then for the next minute I would run my local clock at a speed that I can correct for 1 second. After the minute had completed, I would then compare again, and repeat, until the delta is less than one second.

Anyway, that's just an idea - maybe there are other more refined ways of doing it.