Why the @10K PN is bad?

[tony_lth]

Hi, Gurus,
I desgined a LO at about 15G.
Its phase noise is bad @10KHz, but @1K & @100Hz is very good.
I wonder the reason for this case.
Attachement "15G" is the PN of the 15GHz, and there is a peak @10K. In theory, it's alomst about 3dB variation from theory value. My PLL bandwidth is about 400KHz.
Attachment "100M" is the ref of VCO, from the VNA. It should be better from calculation of 15G PN. Maybe it be better if I open the "coherent" option of the VNA, right?
"PN" is the several points of the PN curve. All are better than ADISimPLL values except @10Khz.
So I wonder maybe there are some other root causes of this case?
Best,
Tony Liu

 

[BigBoss]

I think the VCO itself has worse PN beyond 10kHz and therefore PN of free running VCO beyond this point is dominant and PLL filter cannot correct this.
Also, using a Signal Analyzer is not a good idea instead of using a Phase Noise Analyzer. There are some constraints in my humble knowledge
For instance -74dBm/Hz @100Hz Off-set does not seem convincing to me. I have doubts..

 

[dick_freebird]

A reference part would be handy here. Maybe Mini-Circuits has something
in the same vein with datasheet and apps support that would validate or
help you improve your setup. Two unknowns and one equation no bueno.

 

[tony_lth]

thanks, maybe it's VCO problem.
Since I simulated with a 20dB improvement on ref, but final 15G PN only improved about 5dB.