RF Freq Error vs. Crystal Freq Error

[tony_lth]

HI, Gurus,
I used SX1280 LORA IC with a 52MHz/10ppm Crystal to output 2.4GHz RF signal.
The crystal freq error is 10ppm, and the 2.4G RF freq error measured is 30+ppm.
The FAE of the IC said, due to 2.4G/52M=47, i.e. 2^5, so the RF freq error should be 10ppm*5=50ppm.
I really doubt that conclusion.
I found in GSM system, the RF freq error is 0.1ppm with a 10ppm crystal.
I guess the RF freq error should be the crystal error + PLL error. In my case, the RF freq error should be 10ppm+5ppm(PLL)=15ppm.
Any comment?
Best,
Tony Liu

 

[FvM]

When you say "The crystal freq error is 10ppm", to which crystal spec do you refer? There are at least three ppm specs of a crystal, frequency tolerance stability over temperature, and aging. Supposing you refer to initial frequency tolerance, it's only valid for nominal load capacitance. As discussed in your previous SX1280 thread, it looks like the frequency deviation is caused by inappropriate load capacitance.

The FAE of the IC said, due to 2.4G/52M=47, i.e. 2^5, so the RF freq error should be 10ppm*5=50ppm.
I really doubt that conclusion.

Funny nonsense.

I guess the RF freq error should be the crystal error + PLL error.

A PLL won't cause static frequency error, just set a rational frequency relation between crystal and output.

 

[vfone]

I found in GSM system, the RF freq error is 0.1ppm with a 10ppm crystal.

In GSM, when the mobile is powered-on and send a message Random Access Channel RACH (first message from mobile to BS) only at that time the mobile is synchronized to the internal clock.
After the mobile synchronization to the eNB (base station) then, using a synchronization channel, makes the frequency of the mobile to be very accurate (the same as of the eNB).

Synchronization Channel - Wikipedia

en.wikipedia.org

 

[BigBoss]

If we assume \[ F{vco}=\frac{F{ref}}{R}*N \] then we take the differential of this equation, we find the sensitivity against crystal frequency.
So \[ \Delta F{vco}=\frac{N}{R^2}*\Delta F{ref} \]

 

[tony_lth]

@BigBoss,
Thanks for the equation.
The SX1280 datasheet say, F(RF)=N*F(XTAL)/2^18.
For 2.4GHz, with 52M XTAL, N=2400/(52/2^18)=12,098,953.
So according to your equation, delta(RF)=0.0002*delta(Fref).
Too good to be real.
Anyway, the idea is very helpful.
Best,
Tony Liu

 

[BigBoss]

@BigBoss,
Thanks for the equation.
The SX1280 datasheet say, F(RF)=N*F(XTAL)/2^18.
For 2.4GHz, with 52M XTAL, N=2400/(52/2^18)=12,098,953.
So according to your equation, delta(RF)=0.0002*delta(Fref).
Too good to be real.
Anyway, the idea is very helpful.
Best,
Tony Liu

My equations were valid for simple Integer PLLs but your IC uses Fractional one so the assumption is not correct.

 

[tony_lth]

Hi, BigBoss,
Thanks.
So for integer PLL, with bigger R, the delta RF is smaller. That is helpful.
Best,
Tony Liu

 

[FvM]

To be serious, I don't understand the whole calculation. ppm is measure of relative frequency error.

Presuming rational frequency relations \[ F{vco}=\frac{F{ref}}{R}*N \]

you get \[ ppm_{vco}=\frac{\Delta F{vco}}{F{vco}}*10^6= ppm_{ref}=\frac{\Delta F{ref}}{F{ref}}*10^6\]

 

[albbg]

As pointed out by FvM the accuracy of the frequency generated by the PLL is the same of that of the reference. Did you measure the accuracy of your crystal or just estimted it ?

 

[tony_lth]

Hi, albbg,
I just measured the RF freq with spectrum analysis. I didn't directly measured crystal freq, just read its datasheet.
Pls check the attachments, EXS00A -CS07103 is the original crystal recommended in SX1280 datsheet, and with it, the RF freq error at room temperature is -30ppm.
Another crystal, FW520WFMT1 is the replacement for the original one, and I haven't gotten it yet.
I hope it should be better than the original. In a few days, I will try it.
Welcome comments.
Best,
Tony Liu

 

[albbg]

It could be little bit better since the tolerances are tighten with respect to the original crystal.
In any case, as already said in post #2, the initial tolerance is only valid for the given load capacitance and for the nominal supply as well.
Furtermore you have to account for the deviation due to the temperature and the aging. In choosing the crystal you have to be aware of all these inaccuracies that your system has to be able to manage. You didn't mention your specs.

 

[tony_lth]

Hi, Gurus,
After handling the grounding around the crystal and replace the crystal,
now the RF freq error is less than +10ppm. It seems that the error can be even smaller, but I think that's OK.
many thanks.
Best,
Tony Liu