strange NF measurement result

Hello, friends.
I've measured NF of LNA recently and got very strange NF plot.

What are the reason of strange oscillations of NF coefficient ?
When i monitor output spectrum, LNA doesn"t seem to oscillate.

Thanks in advance.

Have you checked the oscillation until fmax of the transistors ?

The LNA match bandwidth is normally narrow, if your bandwidth is (2-1)/1.5GHz=67%, it seems too wide. What is your objective bandwidth?

Hello
Thanks for your replies.
I've monitored output spectrum up to 26.5 GHz (max frequency of spectrum analyzer) with span of 1GHz - so the output noise floor was near -55 dBm - i've noticed nothing strange.
Actually my working bandwidth 1575 - 1605 MHz. At these frequencies NF oscillations are highly attenuated (as you can see from the picture), but present a little. So i'd like to know the reason.

so that's OK, I think nth is wrong.

I guess NF calibration is wrong or there are something strange on the cables...
NF can not drop down to -1 dB..mathematically impossible.

Nf calibration has been done before measurement and after calibration i carry out one measurement with noise source connected directly to spectrum analyzer (as during calibration) - so i get 0 dB gain and 0 dB NF, as it should be. So i don't think this is due to wrong calibration. The similar phenomenon i get when i measure NF of microstrip line on PCB - although gain remains constant, for example -2 dB, NF hesitates, sometimes becoming less 0 dB which is physically impossible.
I guess this is connected with reflection waves in case of unmatched load.

It should be no problem except you use very bad performance cables or connector, because your freq is too low, less than 2GHz.

If the device has poor match and the cables to connect are are 1 meter with a dielectric of 2 you could get 100MHz ripple like you see. Put a 3dB pad at the end of each cable (device side) you should see NF+3 and less ripple.