Post Go back to editing

Noise Figure

Category: Datasheet/Specs
Product Number: ADRV9003


I want to measure and verify the Noise Figure mentioned in the data sheet of ADRV9003.
What is the correct way to do so?

Thank you,

  • Hi Yoel,

    The NF measurement steps can be described as the following: 
    •  In TES, configure the profile. Set the Rx gain index at 255 (max gain, 0dB attenuation).
    •  Feed an in-channel tone at Rx1A. choose an input level such that to avoid clipping due to 18dB digital gain. Then observe the Rx output level in unit of dBFS from the TES. This value is displayed as “Fund Amplitude” under “Receive” tab. Denote the Rx input and output level as Rin (dBm) and Rout (dBFS), respectively.
    •  Put an 50 ohms terminator at the Rx1A input, then measure the integrated in-band noise power within a certain BW (eg. ±4kHz) in unit of dBFS from the Rx dumped data, denoted as N1(dBFS). This could be done through the “save” button under “Receive” tab and then use “VisualAnalog” tool to calculate the integrated noise power
    •  The tool can be found at: 
    •  NF is given by the following equation: NF(dB) = N1(dBFS) – Rout(dBFS) + Rin(dBm) – 10*log10(BW) – (-174 (dBm/Hz))
    •  This measurement includes the cable loss and SMA connector loss, which should be compensated. By taking those into consideration, the final NF should be calculated as: NF(dB) = N1(dBFS) – Rout(dBFS) + Rin(dBm) – 10*log10(BW) – (-174 (dBm/Hz)) – (cable loss + SMA connector loss).



  • Hi Rahul,
    Thank you very much for your quick reply.

    I don't have the Develop Board yet, so I can't use the TES.
    I have a custom board with controller and ADRV9003 which I can read the IQ data at the Rx port.
    I can also set the gain index, digital gain and ,more parameters.
    Is there any way I can measure it this way?


  • Hi Yoel,

    To calculate the noise figure using your script can be done by utilize the provided formula. This requires measuring the noise power within the desired bandwidth in dBFS, along with the power levels of the transmitted  in dBm and received signals in dBFs. Additionally, cable and SMA connector losses should be factored in during the measurements in dB. If your script can retrieve these parameters, you should be able to determine the noise figure.

    Noise factor(dB) = N1(dBFS) – Rout(dBFS) + Rin(dBm) – 10*log10(BW) – (-174 (dBm/Hz)) – (cable loss + SMA connector loss).



  • Hello Rahul,

    I'm using a matlab script that I wrote for this measurement.
    The script input is the IQ data and the strobe given by ADRV9003.
    I manage to read it correctly for sure (I've done some tests on it).
    I use the spectrum Analyzer Toolbox in matlab in order to plot the DATA sent from ADRV9003.
    Gain index is set to 255, and I measured with 18dB digital gain and 0dB digital gain.

    I transmitted a -80dBm (with 3dB cable and connector loos) tone into ADRV9003 and measured the power of the tone via Spectrum Analyzer toolbox.

    0dB digital gain: In order to check whether the Spectrum Analyzer works fine or not, I measured with digital gain equal to 0dB. The RSSI measurement (from ADRV9003) was exactly the same as the power measurement from the Spectrum Analyzer (-74dBFS).

    18dB digital gain: Assuming the Spectrum Analyzer works just fine, I measured once again with digital gain equal to 18dB. The power of the tone was -55dBFS, while the noise was -117dBFs (at RBW = 10Hz). RSSI measurement from ADRV9003 still was -74dBFS.

    By using your equation I get:
    NF(dB) = -117(dBFS) - (-55dBFS) + -80(dBm) - 10*log(10Hz) -(-174dBm/Hz) - 3dB.
    NF(dB) = 19dB.

    I made another measurement. This time I observed the S/N in two different setups:
    1. Signal generator transmit directly to ADRV9003.
    2. Signal generator transmit to well known LNA (G=40dB, NF = 1dB). The LNA output is the ADRV9003 input.

    This way I can measure the ADRV9003 NF. By subtracting the S/N gives the difference between the NF of the LNA and the ADRV9003.

    1. S = -55(dBFS). N = -117(dBFS). - > S/N = 62dB
    2. S = -10(dBFS). N = -94  (dBFS)  - > S/N = 84dB
    The difference between the S/N is 22dB which means the ADRV9003 NF is 23dB (22dB + 1dB[LNA NF]).

    I'm not sure if the second method is right, but either way the first methods gives higher NF (19dB over 15dB that expected at NB @1.5GHz)/
    Is there any way to measure it more precisely?