Dear Audrius Merfeldas ,

Antennas have very limited noise especially when it is resonant antenna. The antenna bandwidth will be limited and consequently its noise power. Also as the antenna gets directional its noise pick up will be smaller. Accordingly antenna array could have lower noise power than single antenna.

The addition of low noise amplifiers causes an appreciable increase in the noise because the noise generated in the amplifiers.

So, the antennas with the LNA will have the highest noise power.

Best wishes

Your third option is always better. Practically, In RF devices like BTS, Antenna feeders are connected with CDU(combining and distribution Unit) ,Then amplifier, after that TRU( Transrate adaptive unit) then power unit (-48 volt DC).But, It may not be a general concept......

Dear Audrius Merfeldas ,

Antennas have very limited noise especially when it is resonant antenna. The antenna bandwidth will be limited and consequently its noise power. Also as the antenna gets directional its noise pick up will be smaller. Accordingly antenna array could have lower noise power than single antenna.

The addition of low noise amplifiers causes an appreciable increase in the noise because the noise generated in the amplifiers.

So, the antennas with the LNA will have the highest noise power.

Best wishes

Thank you for answers. In my case lna is required, but as I assume it is better to place lna to each individual antenna and later combine RF signals camparing with the case of using single lna after combining few antennas. I assume that lna noises will be reduced as I will sum up few lna signals due to random nature of noise. One more note: lna noise will be suppresed even if I will use single antenna connected to lna and then wilkinson with another lna connected to opposite shoulder loaded with 50ohm resistor at its input. Am I right?

I would like to remind you about phased array style satellite antennae (those are the flat antennae which you can buy everywhere for practically the same price as a parabolic dish e.g. for 10 Ghz sat receivers) They have a very similar performance both in directivity (mainlobe) bandwidth and noise temperature as the parabolic dish.As for narrowband and resonant antennae : We have to discriminate between noise temperature (losses make noise) noise power and noise power density. Often resonant (receiving) antennae have a noise temperature as ambient (300 K) .Sat receiver antennae on the roof of a house have noise temperatures around 30 to 70 K + noise temp of the LNB.Those 30 ..70 K depend on the direction in which the antenna beam goes and essentially reflect the microwave losses of the atmosphere. Electronic amplifiers can have noise temperatures (10 K) well below their physical temperature (often 300K) since they are not in thermal equilibrium.In general a private amplifier for each antenna element is better(we use this where ever possible this way for highly sensitive devices for beam diagnostics) but also more expensive and complex (e.g. in a cryo and vacuum environment); thus the final decision depends on many parameters..but with a good power combiner system the first amplifier after the combiner tree is often good enough.

From a design perspective, SNR in antenna/array has to considered if only noise generated by antenna/array itself is involved w/o any external noise. Mathematically, the Signal component in SNR is effective collecting area times incident field intensity. The Noise component in SNR is generated by losses in antenna/array. Arrays makes larger effective collecting areas compared to single antennas and, thus, their Signal component in SNR is much larger for large array sizes. However, arrays require special networks to combine signals from all element ports into a single output port. Such real-world power combining networks are spatially spread behind array apertures and unavoidably lossy and, thus, can increase the Noise component in SNR. Note that losses in array elements themselves could stay comparably low like in well-designed single antennas. Adding LNAs in array power combining network helps to mitigate losses and minimize Noise in SNR. Art of designing sensitive arrays like those in radio astronomy or SATCOM or deep space COM is balancing properly the Signal and Noise contributions to SNR budget by selecting proper/better design approaches for apertures and combiners (beamformers). For example, patch antennas are not good for sensitive applications because of high losses in microstrip printed conductors and so on. There are some second order effects in arrays caused by mutual coupling between array elements through EM fields that could cause additional issues for SNR to mitigate. Equivalent to SNR is effective noise temperature mostly used in link budget equations.

Option 3 likely. However, one has to be conscious of various factors depending on the frequency range, etc.

There is one important, common and often not understood consideration - let's illustrate it by extreme (rather imaginary-abstract) example:

Let's say we have very clear, good quality microwave parabolic antenna looking up the sky in dry climate high up in the mountains. The noise temperature it will see will be not far from Big Band microwave of 2.7K. At the same time one may use some cable with 10% power loss (not bad indeed?) to deliver it to some LNA ( let's say amplifier is nearly ideal Tnoise

Thank you for response. I have made experiment, please check the diagram attached. Using two RF amplifiers loaded with matched 50 Ohm at RF inputs and resistive power combiner connected to SA spectrum analyser. I was expected to get reduction of noise level as I connect both amplifiers comparing to only one turned on. Basically I leave the same setup, but turn on power for both or only single one. There was no improvement as I turn on both amplifiers, but due to random nature noises should cancel at least for 3 dB. Any suggestions?

In this layout the noise power should be 3 dB higher, but if the antennas are combined properly to get a narrower beam pattern the signal should be 6 dB higher, so SNR goes up by 3 dB. Both of these powers will also see some combiner loss.

If your antennas are pointing in different directions so they don't "help" each other, then the SNR for both will be 3dB worse.

But in this case I just want to reduce noise of amplifiers. The losses mainly matters in front of first amplifier, later they will have less influence. From my point of view adding two noise source should cancel each other due to random phase and amplitude. Am I not right? For this case do not take any signal in to consideration, only noises.

You are not right. Noise power always adds. Any noise power from both antenna channels will add together in the combiner. If you want the lowest noise level and don't care about the signal then don't connect to anything. The noise will be the lowest possible and there will be no signal.

That was a silly example to show that you must consider the signal as well. It is nearly always the signal to noise ratio that matters, or something similar. Combining signal sources together usually improves signal to noise if the signals add coherently (all instantaneous voltages add up, not subtract) and the noise doesn't. Putting in an amplifier always increases the noise immediately after it, and increases the signal by a little bit less, so it always makes the SNR worse. To get the best possible, use a low (added) noise amplifier. They probably have small gain. You can follow this with a high gain amplifier if you need more gain because the noise contribution of the second amplifier is effectively reduced by the gain of the first amplifier.

Thank you

There are at least 2 ways of adding signals together, and they have different results. You can add voltages, which can be either positive or negative, or you can add powers, which can only be positive. If you add voltages, and the signal parts are all the same, then the signal voltage goes up by N and the signal power goes up by N squared. The noise part adds together randomly, with the statistics of a random walk which was described by Einstein, so that the mean noise voltage increases as the square root of N, and the noise power by N. The signal power gains over the noise by N. If the display is adjusted so that the signal occupies most of the range of the screen then it looks like the noise goes down, but actually both have gone up. Dividing the voltage by the number of samples has a similar effect.

If powers are added, then the signal powers add, and the noise powers add, so the is no change in signal-to noise ratio, I think, but what does happen is that the spread of the noise is reduced the more samples are taken. This is similar to the reduction in the standard deviation of the mean with the number of samples taken. This means that if lots of samples are combined then small variations in the signal can still be seen even though the noise may be large because the spread in the noise is smaller than the change in the signal.

70 years ago incoherent (power) combination was more common. Nowadays coherent (voltage) addition is nearly always used. Coherent addition is what is used to generate a narrow antenna beam from lots of antennas with wide beams, as would happen in your example.