Time loss in DS12887 RTC...any solutions??

ds12887 rtc

Hi,
I am using the DS12887 RTC to make a small real time clock. My problem is that
its accuracy is not what is expected out of a clock. I have already made the clock
with an AT89S52 and am planning to put it to day-to-day usage.

The data sheet says the loss of accuracy is about (+/-)1 min per month.
The exact technical statement(from the datasheet) is something like:

The real time clock will keep time to an accuracy of ±1 minute per month during data retention time for the period of tDR.

Click to expand...

Now I have 2 questions:

1. What does this retention time or tDR mean? Can anyone please explain?
I looked at their datasheet timing diagram, but sadly could not make it out

2. Don't you think, this time loss of 1 minute per month is too much for day-to-day usage??
Is there any solution I can implement in software to counteract this loss?
Like bumping the time by half a second or so every night at 12:00 AM?
Or is there any hardware solution which can increase the accuracy?
Like placing the 12887 chip close to the micro?

Any insights or solutions/suggestions would be greatly appreciated.

regards,
Seemanta

nvram ds12887

Hi,
Data retention time is specified for the internal battery as 10 years, which means the static NVRAM will retain data written into it for 10 years without external power. I think, since the RTC also runs on this internal battery in the absence of power, the ± 1Min. per month applies only if the RTC is kept without activating with external power.
Your idea to correct on a regular basis may work if the Time drift is predictable.
Regards,
Laktronics

tdr rtc

The time loss or gain for these parts is due mainly to two factors.

1. Temperature changes. This varies the frequency of the crystal oscillator. The reason your wrist watch is much more accurate than that is that your body keeps it at a constant temperature. Reduce the temperature changes and your accuracy will improve.

2. Crystal stability and proper crystal loading. Several years ago we used a Dallas part and had a very similar problem. The issue was that the crystal's load capacitance did not match the capacitance expected by the chip. This caused the crystal to oscillate at slightly the wrong frequency and the clock accuracy was terrible. Dallas recommended an Epson crystal, but it was never available. We finally worked with the crystal manufacturer to add some additional capacitance to our circuit to make their crystal work.

To make your correction factor idea work, you will need to calibrate each device. This is because, you do not know which direction it is drifting. Once you know the drift direction and magnitude for a particular device, you need to store it in a non-volatile memory so it can be used as needed in the future.