How to qualify a cystal unit?

[konloun]

HI, guys:
Plz help. I want to know how a good/bad crystal affect a GSM/WCDMA rf system?
to my knowledge, there is Phase Noise, so the modulated quality. and any other thing?
plz forgive my poor english. Thanks.

 

[DeboraHarry]

Frequency stability is often important, which changes both due to aging and temperature. I'm no expert on this. I don't think phase noise is set too much by the crystal, but rather by the oscillator that it is built into.

However, I'm not sure how critical frequency error is on things like mobile phones, as I believe data about the frequency error is sent to correct the phone. That data comes from a service provider, which is obtain by them using GPS, which comes from cesium clocks in the sky.

Other things that can be important are things like the time for an oven to warm up, but again, that is probably not applicable.

Ask on the time-nuts mailing list if you want a more exhaustive answer. There are people there who are seriously into precision time/frequency. Many have cesium clocks. Some have more than one.

One guy Tom, has a hydrogen maser, and has done things like climb mountains with a cesium clock strapped to his back, so he could see the clock speed up at altitude as predicted by Einstein.

http://leapsecond.com/pages/atomic-tom/

 

[MK28]

Well, crystal is used as a master clock of your system..if you are familiar with basic PLL system then you know that reference clock/master clock is fed into the system:

system clock>Phase detector>Low pass filter>VCO------>output
[Note: it has a feedback loop meaning sample of output will be feedback to PD and PD will take the difference--pls read abou it]

There are quiet a few kind of crystal such as VCTCXO (internal temperature citcuit, thick in size, additional components), VCXO (thin in size, no internal temp. compensation, temp comp. is done in SW algorithm) and DCXO (digitally controlled and no external component, temp comp is done via SW algorithm). Now, when you pick a crystal you walk through its S-curve whcih shows the crystal performance (corner samples+ordinary ones) @ different tempratures. To have a robust system, all these crystal should be varified on all set of conditions making sure frequecy error (LB: +/-90Hz, and HB: +/-180Hz or so , or 0.01ppm I beleive w.r.t carrier freuency).

Bottom line is your frequency offset with the network should be within the limit as we do not want to see your DUT causing interference with other uses assigned in different frequency.

 

[drkirkby]

or 0.01ppm I beleive w.r.t carrier freuency).

Bottom line is your frequency offset with the network should be within the limit as we do not want to see your DUT causing interference with other uses assigned in different frequency.

You will not get 0.01 ppm with a standard crystal. Here's the specs taken from the data sheet of a 20 GHz microwave network analyzer.

Range
8719D 0.05 to 13.51 GHz S-1
8720D 0.05 to 20.05 GHz S-1
8722D 0.05 to 40 GHz S-1
Accuracy (at 23 °C ± 3 °C) ±10 ppm S-1
Stability
0 °C to 55 °C ±7.5 ppm C
Option 1D5 ±0.05 ppm C
Per year (aging) ±3 ppm C
Option 1D5 ±0.5 ppm C

Those in bold are the high stability timebase, with the oscillator in an oven. If you need 0.01 ppm then you need at least a rubidium. That's clearly impractical with things like mobile phones. As such, I believe there is software frequency correction, so the crystal does not need to be at such precision.

That correction information comes from a cesium.

 

[MK28]

@drkirby: I mean 0.01 pm w.r.t GSM900 or 1800MHz band leading to +/-90Hz or +/-180Hz limit/requirements by 3GPP.

 

[FvM]

GSM 05.05 requires mobile stations to track the base station frequency with 0.1 ppm accuracy (the said 90/180 Hz numbers). This is setting requirements for short time stability, not absolute crystal accuracy.