Re: Airspy noise figure


(She screws up her face thinking furiously before creebing that measuring RMS signal plus noise and noise within an FFT spectrum summing the power for every FFT bin within the specified bandwidth should potentially be VERY accurate when reduced to a noise figure when using an SDR. Thanks for provoking me into analyzing your statement when it "bothered me" a little. It will likely be most accurate when your bandwidth of interest includes 1000 or more FFT bins and you diddle the calculated values to account for the FFT weighting algorithm. And I bet Leif has already done this in the past.)

{^_-} Joanne

On 20180313 10:06, doug wrote:
There's another method of getting a noise figure number, if you have a
pulsed RF source: carefully adjust the source so that the detected rf
pulse is tangential. What this means is that
the average noise at the bottom of the detected pulse is equal to the
average noise of the baseband, where there is no pulse. The value is
equal to 8dB above signal equals noise. Now if you
know the bandwidth of the receiver, you can determine the noise figure,
since you know that at 1 Hz BW, a perfect receiver would have a
sensitivity of -174 dBm. For every increase in
bandwidth, the s/n ratio decreases by the value ratio. For example, at
1KHz bandwidth, a perfect receiver would have s=n at -144 dBm. This
method is not as precise as a calibrated noise source,
but it's close.
--doug. WA2SAY, retired RF engineer.
On 03/13/2018 05:46 AM, jdow wrote:
Given a choice I'd prefer to try liquid funnygas er Helium for the
cold source.

{^_-}  Just being imaginative and silly again.

On 20180312 09:15, David Eckhardt wrote:
Yes, there is nothing to compete with the hot/cold 'noise' source.
kTB is very specific.  However, knowing the information bandwidth
introduces errors for our amateur receivers unless we rigorously
measure that as well. I've always preferred the hot/cold noise
source, but don't keep LN2 around.  However, these days, most larger
grocery stores carry dry ice.

Dave - WØLEV

On Mon, Mar 12, 2018 at 4:06 AM, Marcus D. Leech <mleech@...
<mailto:mleech@...>> wrote:

    On 03/11/2018 09:00 PM, David Rowe wrote:

        Thanks Dave - sounds like you have a great experimental set
up there for
        NF measurements.  For anyone interested in NF measurements,
I've blogged
        on my humble experiments here:

        I mentioned your results in the post Dave, hope thats OK.



    I'll point out that single-point noise-figure methods are rather
    error-prone, which is why the industry has standardized on the
    method, in which
       you have TWO sources of known noise power distribution.

    For noise figures above 3dB, you can usually use a high-quality
    at room temperature -- whose PSD will be -174dBM/Hz, and then a
       noise source, usually in the 20-35dB ENR range.  You take
    measurements of both, and use that to determine the noise figure
of the DUT.

    Once you start going to noise-figures below 2-3dB, it's better to
have a
    "cold load" as one of the noise sources.  Labs that care about
noise figure
       measurements for "good" LNAs usually use an LN2-cooled load
for the
    "cold" load, and again a calibrated noise head for the "hot" load.

        On 12/03/18 10:18, David Eckhardt wrote:

            I have a calibrated noise source good to 18 GHz and several
            precision attenuators (HP, both single value and
switched).  Knowing
            the noise output of the noise source, I attenuate until
the noise
            just goes into the 'noise', tangential noise.  The
            between the noise source known output and the
attenuation  is my
            noise figure.

            Dave - WØLEV

            On Sun, Mar 11, 2018 at 10:27 PM, David Rowe
            <mailto:david@...> <mailto:david@...
            <mailto:david@...>>> wrote:

                 Thanks Dave,

                 Could you tell me a little more about how you
measured NF?



                 On 12/03/18 08:37, David Eckhardt wrote:

                     I've measured both the R1 and my present R2.
They both come in
                     around 4 dB.  I believe 7 to 8 dB is a bit much
based on my
                     experience with them.

                     Dave - WØLEV

                     On Sun, Mar 11, 2018 at 9:34 PM, <david@...
            <mailto:david@... <mailto:david@...>
<mailto:david@...>>>> wrote:

                          Hi - I've been setting up a system to
measure noise
                     of SDRs
                          in real time, an extension of some similat
tests we
                     performed a few
                          years ago:


                          I'm getting a NF of 7-8dB for a new Airspy
R2 and Airspy
                     mini.  The
                          spec quotes 3.6dB.  Has anyone else
measured the NF?

                          I note the 3.6dB figure is exactly the same
as the front
                     end tuner
                          R820 tuner.  However a system noise figure
will always be
                          than the front end device.



                     --         *Dave - WØLEV
                     */Just Let Darwin Work/*

            --             *Dave - WØLEV
            */Just Let Darwin Work/*

*Dave - WØLEV
*/Just Let Darwin Work/*

Join to automatically receive all group messages.