Confused with signals with unexpected frequencies in LNA design based on BGU7003

I designed a LNA circuit based on BGU7003 as you can see in the attachment. The simulation results at 1279MHz and 1576MHz are also shown in the attached image. I then injected single tone of 1279MHz and 1576MHz respectively and tried to see the results on the spectrum analyzer. Then the weird thing happened. I saw some signals with unexpected frequencies as you can see in the attached image. It looks something like resonance. Did anyone see this kind of phenomenon before? What are these unexpected signals and where are they from? Is it because the gain of LNA is too high or there are some noises inside the power or ground to the LNA? I have never seen this before so it really confuses me. Could anybody have a look and help? Thanks.

-You couldn't do a well matching circuit,S11 and S22 metrics are very poor.Re-design them..
Show how did you implement the LNA, ( PCB,Real-Schematic,Components' familiy etc)
If you have implemented the LNA as shown , it's wrong because you do DC short circuit the Input.Check the application schematic carefully.

It looks like it is oscillating at about 1MHz. Check your DC bias circuits.

Thank you for your reply. Please see the schematic and PCB in the attached images.

Yes the matching is not good, but it should be stable and there is no reason such kind of harmonics or oscillating should happen in my understanding. I'm thinking whether I should add a BLM or an inductor in the power rail so that a LC low pass filter can be built with the 0.1uF capacitor which is already there near the LNA power supply pin. Will this help?

Your LNA is oscillating due to the S21 being too high. There is NO WAY that you will be able to achieve a stable single device design with a gain greater than 90. This is a typical rookie mistake, thinking that if 10dB of gain is good, 20db must be better and a simulation will happily show you that the design is possible. However, reality often raises its ugly head and slaps the rookie upside the head. Any RF engineer can design an LNA with 10dB of gain. An expert RF engineer can design an LNA with 12.5 - 13dB of gain. No one can design a stable LNA with 20dB of gain.

Thank you for your reply @SLK001. Yes I am quite newbie in RF design. I wondered whether the bizarre harmonics or oscillating was because of the too high gain, but there is another RF chain on the same board which has exactly the same circuit as the chain shown here, and there is no such kind of harmonics or oscillating happening on that chain. I also designed a carrier board with the same LNA, same matching networks, and there is no harmonics or oscillating on the carrier board either. Therefore I am not sure whether the too high gain is the cause of the issue. In my understanding, if the gain is too high with this matching network, such kind of issue should also happen on another RF chain and on the carrier board as they all have the same matching.

That's a weak statement. The SAW might cause extreme mismatch at the LNA input.

Hi volker@muehlhaus, thank you for your reply. I checked again the SAW and it says in the data sheet that the SAW is matched to 50 Ohm at both input and output. There are DNP components for the the matching of SAW just for compatibility purpose. Therefore I put zero Ohm resistors in serial for the DNP components. Could the zero Ohm resistors on the RF signal path change the impedance dramatically so that caused the oscillating issue?

I meant the out-of-band impedance. I would check for stability instead of just assuming it. On the other hand, I don't agree with SLK001 that 20dB must be instable.

I used BGU7003 as a GPS 1.5GHz LNA. The amplifier have stable high gain, but as any other high frequency MMIC its stability is sensitive to the bias decoupling and also is sensitive to the layout design.
You wrongly put 0.1uF decoupling capacitors on the bias lines (C232, C18). A 0.1uF capacitor at 1.5GHz looks as anything as you want, but not like a capacitor. Change the capacitors to values close to those that are recommended in the datasheet.
Depending on your layout and to get a wide band stability, you may need to tune these decoupling capacitors values or place in parallel other capacitors with different reactance at various frequencies.
Also I think the BLM2 ferrite bead properties are inappropriate for that place. Sometimes wrong ferrite beads values (together with wrong value decoupling capacitors) make RF circuits to oscillate.

I am mostly talking about single device (single stage) LNAs, such as with a transistor as the main device. Although I have no knowledge of the device, I believe that the package contains a two stage amplifier, simply because of the gain. Because of the (assumed) two stage design and because everything is together on a single die, a 20dB gain is claimed possible by NXP and I have no data to dispute that claim. However, proper design has to occur outside the device to maintain the stability at their listed gains. The datasheet for the device used here states that Maximum Stable Gain at 1.575GHz to be over 20dB, but they don't state the conditions. Is the device still stable under loadpull conditions at all phase angles? Is it still stable over the temperature extremes? These and other real world conditions make a 20dB design extremely dicey, at best.

Looking thru another NXP document, it states that the BGU7003 is a single stage device.

I said your matching circuit is not correct therefore you should review this circuit then you may run forward step by step.
I have kicked a starting circuit for you and I have used "BGU7003_S_NF_2p5V_10mA" s-parameters file ( you can of course change with yours).Playing with this circuit, you will find more consistent responses.

there are some FETS which can provide more than 18 dB associate gain @ 2 GHz

Don't just concentrate on the in band performance, althouih that is important, you need to look over the complete frequency range for which the device has gain and check for stability. It will almost certainly only be conditionally stable so make sure that it does not see impedances input and output that make it unstable.
As vfone says look at the 0.1uF capacitors you have in the DC bias, they are not the right value to use here for RF decoupling, something around 100pF or so is more appropriate. Just a single 0.1uF and 33nH resonates at 2.8MHz, and you have a other capacitos and inductors there to consider. If you don't have much experience of high frquency design you are safer following the application note circuits as closely as you can. If you must put capacitors in parallel then try to damp out any potential problem resonances. Put the 0.1uF at the supply input followed by a low value resistor then the high frequency decoupling close to the 33nH choke. Look out for resonances between parallel capacitors too, almost every one gets bitten by that at some time.