Noise floor and Rx Gain

All,

I’m trying to set all USRP and RFX2400 parameters comparable with Micaz
in
the context of receiver sensitivity. But I’ve observed a non-linear
behavior
of ADC.

By using usrp_fft.py, we can see noise floor level. When I change a
rx-gain
in a range of 0~45dB, noise floor level does not change. However, in a
range
of 45~90dB, noise floor linearly increases according to rx-gain. Does
this
affect receiver sensitivity? My thought is that high-gain of ADC
introduces
additional noise figure. Please give me a clue. Thanks in advance.


View this message in context:
http://old.nabble.com/Noise-floor-and-Rx-Gain-tp27288261p27288261.html
Sent from the GnuRadio mailing list archive at Nabble.com.

On 01/23/2010 09:48 AM, Yong J. Chang wrote:

affect receiver sensitivity? My thought is that high-gain of ADC introduces
additional noise figure. Please give me a clue. Thanks in advance.

There is nothing nonlinear about this. This is how all receivers will
behave, USRPs, other SDRs, and all radio receivers in general.

The first thing to understand is that noise figure is a function of the
gain setting.

The second is the difference between the displayed noise floor and the
noise figure. Noise figure is a function of the difference between
signals and the displayed noise floor, not the absolute displayed noise
floor.

It is more instructive if you put a weak signal in to the receiver when
you look at the fft display.

As you increase gain from zero, the displayed noise floor does not rise,
but your desired signal does. This indicates that the noise figure is
improving by roughly 1 dB for every additional 1 dB of gain.

Then there will be a range of gain settings for which 1 dB of additional
gain causes your desired signal to rise by 1 dB, but the displayed noise
floor will rise some amount less than 1 dB. In this range you are still
improving the noise figure, but by less than 1 dB per 1 dB of gain
improvement.

Next there will be a range of gain settings for which a 1 dB increase in
gain will result in your signal going up 1 dB but the displayed noise
floor will also rise by 1 dB. This indicates that you have already
reached the minimum noise figure, and that increasing gain will no
longer improve the noise figure.

Finally, if your signal is strong enough, there will eventually be a
gain range at the top for which increasing the gain by 1 dB will no
longer cause your signal to increase in amplitude by 1 dB. This is
known as gain compression, and it indicates that you have too much gain.
You will start to see strong intermodulation products here because of
the nonlinearity.

Matt

Thanks for your great answer Matt. I have another quick question.
We know that the controllable range of rx-gain is 0 to 90dB in RFX-2400
d’board.
But I only can find a programable gain amplifier in ADC chip whose range
is
0 to 20dB.
Where is the received signal actually amplified according to the
‘rx_gain’
value?

Thanks in advance again.

Matt E. wrote:

improving the noise figure, but by less than 1 dB per 1 dB of gain
longer cause your signal to increase in amplitude by 1 dB. This is
Discuss-gnuradio Info Page


View this message in context:
http://old.nabble.com/Noise-floor-and-Rx-Gain-tp27288261p27290059.html
Sent from the GnuRadio mailing list archive at Nabble.com.

On 01/23/2010 01:28 PM, Yong J. Chang wrote:

Thanks for your great answer Matt. I have another quick question.
We know that the controllable range of rx-gain is 0 to 90dB in RFX-2400
d’board.
But I only can find a programable gain amplifier in ADC chip whose range is
0 to 20dB.
Where is the received signal actually amplified according to the ‘rx_gain’
value?

Yes. There is a 70dB analog programmable gain amplifier in the
Quadrature Demodulator chip, the AD8347.

Matt