[SOLVED] Bandwidth for noise floor explain in direct rf sampling receiver

[Andrew77]

Hi all,
I have just a question: in a direct rf sampling receiver with a input filter from 30mhz to 500 mhz+ LNA and + ADC, what sets the bandwidth for calculation the overall thermal noise floor? Does the ADC affects the computation with its sampling rate? (Considering only the filter we can have: NoiseFloor=-174 +10log(270e6) + NFlna=-89,7+NFlna)
If I use the ADC can I decrease this value?
Thanks

 

[mtwieg]

You don't mention any anti-aliasing filter (AAF). If the bandwidth of the noise into the ADC is greater than the nyquist bandwidth (fs/2) of the ADC, then noise outside of the nyquist bandwidth will alias into the nyquist bandwidth, effectively raising the noise floor on the resulting sampled data.

 

[vfone]

This may explain the ADC Noise Figure in terms of SNR, Sampling Rate, and Input Power:

https://www.analog.com/media/en/training-seminars/tutorials/mt-006.pdf

 

[Andrew77]

You don't mention any anti-aliasing filter (AAF). If the bandwidth of the noise into the ADC is greater than the nyquist bandwidth (fs/2) of the ADC, then noise outside of the nyquist bandwidth will alias into the nyquist bandwidth, effectively raising the noise floor on the resulting sampled data.

Ok, Thanks but the receiver is intended to be used in the whole spectrum 30-500 MHz, If I sample to 1MHz of BW can I have a sensitivity of -105 dBm (ca)? No matter how many bits I will use (up to 24 / 32!)

 

[albbg]

The noise floor you are speaking about is referred to the input, that is no gain of the chain is taken into account. If you want to know if the ADC can affect the noise figure, instead, you have to consider also the gain to compare the level at the ADC input as dBFS (you have to know the voltage reference) with the floor of the ADC itself (roughly could be 1LSB).
If the incoming noise level is higher than that of the ADC, the noise of the ADC can be neglected.
In that case the noise floor will be given by -174+10Log(470e6)+NFlna (consedering adequate Nyqist filtering as said by mtwieg).
If the whole receiver chain is perfectly linear (that means no compression due to the power integrated over the whole 470MHz range), you can filter the acquired signal by means of a digital filter of 1MHz BW to achive a sensitivity close to -174+10Log(1e6)+NFlna+SNRmod

where SNRmod is the minimum signal to noise ratio given by the modulation you are using.