Thermal noise addition after a combiner.

I am trying to understand how the thermal noise floor adds up after a combiner. I am first dividing the input signal into 0 degrees and 90 degrees by hybrid-90. Does the noise floor should also be divided as 0 degrees and 90 degrees phase difference at the two output of hybrid 90?

and then the two paths are added using a combiner. Should the noise at the output of combiner be a 3dB addition or should remain same? I am getting same noise floor at the output and I don't know why?

Any comments are welcome.

You should specify the setup more accurately. Which noise sources are you considering as "thermal noise floor"? How are you measuring the noise? With a hypothetical noiseless receiver?

A lossless divider/combiner without termination resistors should be expected not to add noise to the input signal and maintain the SNR, so that after recombining both paths you get same noise level.

I am using SystemVue (from keysight) software to simulate. I have a source which gives me a signal and source thermal noise which is -173dBm/Hz at room temperature. Now when I pass this through a hybrid 90. i am getting the same noise floor at two outputs, i.e., -173dBm/Hz and the signal is degrated by 3 dB at each output (which makes sense because it has divided the signal equally to two ports of the hybrid-90).
Then i am passing it to two-way 90 degree combiner so my signal level is increased by 3 dB at output of combiner but the noise floor remains same. Why it remains same? shouldn't the noise also added up by 3dB?

both the hybrid 90 and splitter are lossless.

You forget that you're using ideal elements and a simulator.
A simulator is a principally nothing but a special calculator.Noise Floor changes never since ideal element are used.The simulator computes ideally power loss for each branch that is 3dB for each and gives the result without any additional noise contribution.

No. Thermal noise from the different paths is uncorrelated and doesn't add up.

While calculating noise in a cascaded system, you can search for noise temparature calculations. Not in a simulator but with hand calculations, you can check how much noise temperature each element is adding onto the system.

But here, it is really simple fundamentals: at the same temperature, thermal noise of a 1-port with Z=50 ohm is always the same, no matter if the 1-port is built from one resistor, or from 1 million resistors combined with couplers.

Yes, you are correct.

I meant it as an alternative way to understand how the noise is adding up in a circuit. When looking from ideal components and simulators the idea of noise adding up may be confusing; hence using noise temperature concept and translating it into noise figure, could be helpful. Calculating a receiver's response while A-) antenna looking to sky and B-) looking to earth, and calculating the respective noise temperatures was the thing where I understood about noise.

Noise of two coherent sources adds COHERENTLY...i.e. 6 db higher after combining.
Noise of two in-coherent sources adds only 3 dB higher.
Not sure if that helps, but just saying...