An accurate seconds counter in VHDL

[omerysmi]

Hi,
I want to make a counter which counts seconds in VHDL, so first of all i make 1 Hz clock by clock divider, and then I count the clocks of the 1Hz signal.
The problem is that the 1Hz clock is not so accurate and I need it to be very accurate..

Do you have some advices for making an accurate seconds counter?

**By the way, I use altera DE1 with 50Mhz main clock.

 

[shaiko]

Use 2 counters (not 2 clocks).
One fast and one slow.
The fast counter increments with every tick of the 50MHz clock.
Every time it reaches 50,000,000 - it becomes 0 and the slow counter is incremented +1.

The accuracy of your system will always depend on the accuracy of the system clock - there's no avoiding that.
But what accuracy are you looking for? A 300ppm (not too accurate) 50MHz Abracon oscillator will still outperform your average Rolex...

 

[ads-ee]

The accuracy of your system will always depend on the accuracy of the system clock - there's no avoiding that.
But what accuracy are you looking for? A 300ppm (not too accurate) 50MHz Abracon oscillator will still outperform your average Rolex...

Unless I'm off with my calculation +/-300 ppm would result in about 2.6 hours off target for an entire year. Seems to me your average Rolex does better than that.

50MHz +/- 15000Hz (50E6 * 300/1E6)

so 1 sec would result in a range of 0.999700089975 to 1.00030009002, as there are 31536000 seconds in a year (365 days) the error is -9463.6389s to 9457.9625s or approximately +/- 2.6 hours per year.

 

[shaiko]

The equivalent of edaboard.com in the watch world...https://www.rolexforums.com/showthread.php?t=77535

 

[ads-ee]

Even at 5 seconds per month the Rolex will only gain/lose ~30 min per year unlike the 300 ppm oscillator design. You would need to use a 50 ppm oscillator to exceed the capabilities of the typical Rolex.

 

[lucbra]

how much do you want to pay for your accuracy?