How to do an accurate PLL calibration?

[stsiligg]

I want to callibrate a pll(74HC4046AN) as part of an mfsk demodulator for my diploma thesis.
17 different frequencies which correspond to 17 different analog voltages get in the phase
comparator. The pll have to lock each time at different frequency. I have tryed any
combination for RC filter but the best result is as shown at this picture(obrazki.elektroda.pl/14_1159529006.JPG).
Is there any way to have a better signal? Any help?

 

[timo67]

pll calibration

I think your modulating signal (with it's 17 values) has a too high frequency compared to the mean frequency of your FSK signal
Try to lower the frequency of the modulating signal (let us say FMOD) and in the same time trimm the RC constant of your LPF filter (first order RC LPF no more nor less) to verify :
FMOD = 1/2*PI*R*C

good luck

 

[stsiligg]

Re: pll calibration

Thanks for your reply. My frequencies at the baseband are placed from about ~ 20Khz to 90kHz. This square signal is transmited from an rf module at 433.92 MHz. After transited i receive the original signal (20kHz to 90kHz) and these fraquencies i want to convert into voltages. At this point i use the pll. What do you mean saying that the modulating signal has a too high frequency compared to the mean frequency? Do you know another way to make different voltages from different frequencies?

 

[timo67]

pll calibration

Do you mean you have 17 frequencies (all of them between 20KHz and 90KHz) let'us call them f1 to f17 ?
If it is so, my question remain: when your signal goes from one frequency to another frequency, this change in the frequency happens with a certain speed or cadence. That is what I mean by: " the frequency of the modulating signal".
If you know the exact valu of these 17 frequencies, you can opt for another solution:
You can make 17 different band pass filters each one tuned to one of these frequencies and followwed by an amplitude detector and a schmidt trigger. At the end of the circuit you have to draw a sort of DAC each input bit correspond to the outout of a schmitt trigger and has a weight different to the others.

good luck
And excuse me for my poor englich.

 

[stsiligg]

Re: pll calibration

The frequency of the clock, in other words the symbol rate (rate at which signal goes from one frequency to another frequency) is at 20kHz or (1/20000)sec, something which means low signal rate. I thought to use filters but these analog filters is extremely difficult to be calibrated, especially now where i would had 17different filters. Fore these reasons i prefer pll. Do you know if it would be good idea to use active filter?