[ntp:questions] 500ppm - is it too small?

David J Taylor david-taylor at blueyonder.not-this-part.nor-this.co.uk.invalid
Fri May 8 06:47:16 UTC 2009


David Woolley wrote:
[]
> A reasonable expectation of a cheap crystal is 50ppm static plus
> <10ppm temperature dependent.  When one gets >500ppm it suggests the
> problem is rather worse than poor crystal tolerancing.

I am inclined to agree.  One reference I found was this:

  http://www.maxim-ic.com/appnotes.cfm/an_pk/3566

including the statement about 32KHz crystals:

"There is nothing inherently wrong with using an RTC for timekeeping. 
However, the time will only be as accurate as the reference used. 
Unfortunately, the typical 32.768kHz tuning-fork crystal does not provide 
much accuracy over a wide temperature range. Due to its parabolic nature 
over temperature (Figure 1), this accuracy is typically ±20ppm at room 
temperature (+25°C). This is the equivalent of gaining or losing 1.7 
seconds of time each day, or 10.34 minutes per year. As Figure 1 shows, 
accuracy decreases at more extreme high and low temperatures. The typical 
accuracy at these temperatures is much worse than 150ppm, which is the 
equivalent to losing almost 13.0 seconds of time each day, or over 1.3 
hours per year."

Based on earlier experience, I would expect the higher-frequency crystals 
used for CPU-clock etc. to be rather better than this.

BTW: Windows can keep quite accurate time in some versions, Windows 2000 
and XP for example, however I haven't tamed Vista or Windows-7 as yet. 
It's not as good as FreeBSD, though.

Cheers,
David 




More information about the questions mailing list