Re: [Discuss-gnuradio] Number sink unit to dBm

[Marcus D. Leech]

On 28/04/2011 1:55 PM, Matt Ettus wrote:
> There is a full WBX characterization with USRP1 here:
>
> http://code.ettus.com/redmine/ettus/projects/public/documents
>
> Noise figure with N200/N210 is 1-2 dB better.
>
> Matt
[Just because I'm in an expansive and tutorial-esque frame of mind this 
afternoon].

Equivalent noise figures of ADCs are (relative to RF amplifiers) *just 
awful*.  Equivalent noise figures of ADCs can be anywhere between
  20 to 30dB (or more).  For ADCs of equivalent "quality", the 
equivalent noise figure decreases with increasing number of bits, so all 
else
  being equal, you'd naturally expect that the equivalent noise figure 
of the 12-bit USRP1 ADC to be noticeably worse than the 14-bit
  N2xx ADC, and in fact, that is the case.  This is a natural 
consequence of ADC "physics".

Now the total noise figure of a system tends to be dominated by the 
noise-figure of the first gain stage in a sytem, that is because the noise
  figure at any given stage N, is always conceptually *divided* by the 
gain of the previous stage (N-1), so:

    Ntot(N) = Ntot(N-1) + (Noise(N) / GAIN(N-1))

From this, you can see that the first-stage noise-figure and gain is 
quite important in determining the overall noise-figure of the 
system--as you
  move further away from the antenna, a given stage noise figure 
contributes less and less to the overall noise figure, provided that it is
  preceded by a low-noise gain stage.

With boards like the BASIC_RX and LF_RX, there are no gain elements, so 
observed noise-figure is entirely dominated by the less-than-wonderful
  equivalent noise figure of the ADC.   Which means that for those 
boards, you need some low-noise gain ahead of them to achieve decent
  weak-signal performance.  I've used 40dB of low-noise gain (ERA-3 
GaAs MMIC), with 25-45MHz bandpass filters to make repeatable
  observations of galactic background noise at 38MHz with a BASIC_RX 
and USRP2.

For other cards, the RX chain typically has an LNA at the head of the 
chain, followed by some amount of variable-gain amplification.  Something to
  keep in mind about variable-gain amplifiers in the RF domain is that 
their equivalent noise figure gets *worse* as you reduce their gain.
  That's typically because the gain-control is implemented as an 
electrically-controlled step attenuator, and the noise-figure of an 
attenuator
  is identical to its attenuation--so a 10dB attenuator "contributes" 
10dB of noise figure at whatever stage in the gain chain it is placed.
  This made obvious in the plots for the WBX--the noise-figure curve 
approaches a limit of 5dB as the gain is increased, with a peak noise
  figure of roughly 22dB at the lowest (0dB) gain setting.