Facing Prblem with Noise figure of Cascaded System

[kamranraja222]

Hello everyone,

I am working on a system which has following configuration:

Attenuator(7dB),Amplifier(17dB),Attenuator(9dB),frequency multiplier(18dB loss),Bandpass filter(3dB loss),Amplifier(26dB)

This system works as X3 frequency multiplier(100 MHz to 300 MHz). I am bound to use the initial attenuators and amplifier(but I can alter their values). The problem I am facing is with Noise figure of the system. If I changed the Noise figure of the system by changing the values of the attenuators(increased or decreased), the measured noise level remained at same level, which is the also the same noise level of actual configuration. The input power level and output power level remained at same level. Please help me out. Thanks and best regards

 

[biff44]

the output level is not changing since you are almost fully compressed...most likely in the X3 frequency multiplier.

The term "noise figure" may not be a good one, since you clearly have a non-linear system. It you look at the output noise on a spectrum analyzer with sufficient dynamic range, you will see a noise floor. That noise floor is a combination of:
1) the 100 MHz signal source amplitude and phase noise. Usually, the amplitude noise is 10 dB or more BETTER than the phase noise.
2) The noise generated by your input amplifier, as it is modified by the nonlinear X3 multiplier. The multiplier may clip any AM noise, and may turn it into PM noise. AND the multiplier will take any phase noise at its input and make it look 20 LOG (3) = 9.5 dB worse at the output.
3) the noise generated by your output amplifier, which is likely to be ~ equal parts phase and amplitude noise.

So if I were to guess, the main noise source in your system is the oscillator phase noise, which is multiplied by 9.5 dB. So adding a linear attenuator pad, which has NO effect on phase noise, will have no change in the measured output phase noise (so long as the multiplier has enough RF input power to keep multiplying)

The most likely

 

[BigBoss]

Cascading NF specification of a system is valid only and only if the system is linear.( or the since the system is in linear region)Otherwise, non-linearity will bring more or less additional noise and this relation will not be valid anymore as in your system that includes a high nonlinear element frequency multiplier.

 

[kamranraja222]

what type of alterations should i adopt to improve the noise floor of my system? Conversion loss of multiplier is equivalent to noise figure of multiplier, so if I improve the conversion loss,it will improve noise floor of the system? However in this case the phase noise will remain at same level. Basically I need to reduce the spur levels of the system.

 

[kamranraja222]

Another observation: When I directly connected my oscillator(100 MHz) to spectrum analyzer, it gave me the noise level of -95dbm and signal power of 8dbm. Then I connected my diode multiplier,bandpass filter and amplifier, it again gave me the same noise level of -95dbm and signal power of 8dbm. In my opinion it should raised the noise level. I am confused at this point.

 

[biff44]

Another observation: When I directly connected my oscillator(100 MHz) to spectrum analyzer, it gave me the noise level of -95dbm and signal power of 8dbm. Then I connected my diode multiplier,bandpass filter and amplifier, it again gave me the same noise level of -95dbm and signal power of 8dbm. In my opinion it should raised the noise level. I am confused at this point.

you are very likely measuring with the wrong settings, OR have a broken spectrum analyzer. Try reducing the RESOLUTION BANDWIDTH to 100 Hz and make the same measurement, and see if there is a difference.

 

[kamranraja222]

RBW is 300Hz used for this measurement. May be the noise level is limited by spectrum own noise level. This will confirm, If I somehow raise the noise floor of spectrum