How does the phase noise of the LO effect IF accuracy?

[csopi]

Hi,
I have a roughly 1.1 GHz signal to be downconverted to 100 MHz by mixing it with a 1 GHz local oscillator. I am not sure how to choose the performance of the LO.

In particular: let's assume the LO has a jitter of 100 fs rms. At 1 GHz this corresponds to a frequency error of 100 kHz. Does this mean that after mixing I will have an error of 100 kHz? If yes, how to improve the performance?

 

[trurle]

Hi,
I have a roughly 1.1 GHz signal to be downconverted to 100 MHz by mixing it with a 1 GHz local oscillator. I am not sure how to choose the performance of the LO.

In particular: let's assume the LO has a jitter of 100 fs rms. At 1 GHz this corresponds to a frequency error of 100 kHz. Does this mean that after mixing I will have an error of 100 kHz? If yes, how to improve the performance?

Jitter is the specification usable for very high-bandwidth (comparable to LO frequency) signals which is likely not you case.
In case of imperfect oscillator you are going to observe your output frequency randomly drifting as you described above, but the drift amount is poorly constrained by "jitter" specification. You should use "phase noise" specification instead. Usually, IF frequency drift is reduced by PLL circuit which reduce low-frequency components of noise by stabilizing local oscillator with the help of crystal or atomic oscillator operating at lower frequency.

Very simplistically, point of phase noise curve crossing the 0dB line indicates your oscillator expected frequency deviation.

 

[csopi]

Jitter is the specification usable for very high-bandwidth (comparable to LO frequency) signals which is likely not you case.
In case of imperfect oscillator you are going to observe your output frequency randomly drifting as you described above, but the drift amount is poorly constrained by "jitter" specification. You should use "phase noise" specification instead. Usually, IF frequency drift is reduced by PLL circuit which reduce low-frequency components of noise by stabilizing local oscillator with the help of crystal or atomic oscillator operating at lower frequency.

Very simplistically, point of phase noise curve crossing the 0dB line indicates your oscillator expected frequency deviation.

Many thanks! I have looked into phase noise, and now I understand that I should calculate the total rms phase error, which describes the "average" deviation of the system. And here comes the problem: how to calculate this? I have found various tutorials on manufacturers' website, most of them is either incomplete or upright erroneous. Could you please help me performing this calculation for the following oscillator? Let's say we have a noise of -120, -150, -165 dBc/Hz at 100, 1k, and 10k Hz away from carrier. Manufacturers do not seem to further elaborate on the issue, but I am not sure how to derive a meaningful value out of this...

 

[trurle]

In your particular case, you likely will see linewidth about 0.1 Hz (assuming worst case random-walk slope 40 dB/decade below 100 Hz). Your drift for 1 day will be at least 10 Hz even in case of perfectly stable temperature.

 

[the_emi_guy]

... let's assume the LO has a jitter of 100 fs rms. At 1 GHz this corresponds to a frequency error of 100 kHz.

If we assume the LO has jitter of 100fs rms, this means that periodically your LO will be 100fs rms ahead or behind, in time, an ideal reference oscillator. But you have not specified how often this will occur. Once a day? Once per millisecond? Jitter has both an amplitude (fs, ps, UI etc), and a frequency, and you are not specifying the freq (and I'm not sure where you are getting the 100KHz from).

Could you please help me performing this calculation for the following oscillator? Let's say we have a noise of -120, -150, -165 dBc/Hz at 100, 1k, and 10k Hz away from carrier. Manufacturers do not seem to further elaborate on the issue, but I am not sure how to derive a meaningful value out of this...

The phase noise at the output of the mixer (dBc/Hz) will be the same as the LO phase noise (dBc/Hz). Is this what you were looking for?