Maker Pro
Maker Pro

Crystal accuracy

kong

Sep 26, 2010
122
Joined
Sep 26, 2010
Messages
122
Is a higher frequency crystal going to provide a more consistent clock signal over time than a lower frequency one?

Is there a relationship between frequency and consistency over time?

Because I know quartz watches use 32.768khz crystals but I thought I read somewhere that the higher the frequency you use the more accurate the timebase will be
 

Harald Kapp

Moderator
Moderator
Nov 17, 2011
13,700
Joined
Nov 17, 2011
Messages
13,700
A higher frequency does not by itself guarantee a better accuracy. You'll have to compare the specs of the crystals/oscillators used.

I thought I read somewhere that the higher the frequency you use the more accurate the timebase will be
This concept is valid if the accuracy of the crystals is comparable and the timebase in question is generated from teh crystal oscillator's frequency by digitally dividing it.
A higher frequency needs more dividing. Since the error of the oscillator is divided, too, the resulting timebase has better accuracy.
But a good 32.768kHz crystal may be able to deliver better accuracy than a bad 10 MHz crystal.

btw: 32.768=2^15 and therefore ideally suited for generating a 1Hz timebase for clocks while on the other hand requiring only a modest amount of power (as compared to higher frequency crystal oscillators and subsequent dividers).
 

kong

Sep 26, 2010
122
Joined
Sep 26, 2010
Messages
122
Oh, that makes perfect sense. So one would ideally look up the datasheets for each crystal and compare specs/correct capacitance values before swapping a smaller crystal for a larger one.

But there is credence to the idea if the dividing is done digitally (which is what I was assuming) and if the crystal quality is suitable.

What exactly is one of the parameters that one would look for in crystal quality?

I know temperature can be the pin in the neck for crystals, although wristwatch manufactures have managed to take advantage of this by using the body's regulated temperature to reduce instability of the crystal and hence maintain the time.
 

BobK

Jan 5, 2010
7,682
Joined
Jan 5, 2010
Messages
7,682
I am not sure what Harald is thinking, because his posts are usually quite accurate. But the way I read it, he is saying that dividing down a crystal will improve it's accuracy. This is not so. If the crystal is + or - 20 ppm (which is typical), the divided down frequency will also be + or - 20 ppm.

So what your are looking for is the frequency tolerance of the crystal, and perhaps the temperature coefficient. The frequency tolerance is usually stated in parts per million or ppm 10000 ppm would be 1% accuracy. A typical 20ppm crystal is accurate to 0.002%. Which sounds pretty good, until you see how many minutes that is per year. It works out to 10 minutes per year.

The temperature coefficient tells you how much the frequency will vary with temperature. This is stated as ppm / degree Centigrade. The original tolerance is typically measured at room temperature (20 degrees). If the actual temperature was 30 degrees you would multiply the temperature coefficient by 10 to see how much the frequency would shift.

Bob
 

KrisBlueNZ

Sadly passed away in 2015
Nov 28, 2011
8,393
Joined
Nov 28, 2011
Messages
8,393
I agree with BobK. The important parameter is the crystal error expressed as a percentage, or usually, in parts per million (ppm). The actual frequency is irrelevant, though it's true that a 32 kHz oscillator uses a lot less supply current than a 10 MHz one, for example.

If you want detailed information on the subtleties of crystals, I suggest you look for application notes from the manufacturers. Go to Digikey or Mouser to get a list of manufacturers, then search their web sites for application notes or other documents that go into detail about factors that affect accuracy such as aging and temperature.
 

duke37

Jan 9, 2011
5,364
Joined
Jan 9, 2011
Messages
5,364
There are many orientations of the crystal. A capacitor is used to trim the oscillator to the correct frequency and then an orientation is chosen to give minimal frequeny change with temperature.

If high accuracy is required, then an orientation is chosen with minimal frequency change at an elevated temperature and the crystal and oscillator are mounted in a temperature controlled oven.

My LCD watch (30 years old?) gains about half a minute at each time change. I rarly wear
it, it just sits in a drawer. I dont like to be reminded of time passing.:)
 

kong

Sep 26, 2010
122
Joined
Sep 26, 2010
Messages
122
Great info, thanks guys. So really it comes down to temperature and tolerance, not initial frequency.

I look forward to building more circuits utilizing crystals.
 

(*steve*)

¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd
Moderator
Jan 21, 2010
25,510
Joined
Jan 21, 2010
Messages
25,510
One of the most important issues with variation in accuracy (as opposed to absolute accuracy) is temperature.

Watches often achieve good stability because they're attached to a device which maintains a reasonably close control on temperature (This device is called a human).

In other devices, the crystal is placed in an enclosure that is maintained as close as possible to a constant temperature.

After you've done that, you then need to consider power supply variations, crystal aging, and the tolerance/stability/aging of other components.
 

Harald Kapp

Moderator
Moderator
Nov 17, 2011
13,700
Joined
Nov 17, 2011
Messages
13,700
Yup,
pms are ppms, regardless of the divisor.



I'm not sure what I was thinking right tehn. Sorry.
 
Top