Noise Figure and Gain of HMC996

Dear support,

I measured the Noise Figure (NF) and the Gain (G) of the HMC996LP4E on the EvalBoard using a noise-gain analyser. The HMC996 was biased at Vg1,2 = -0.75V, Idd = 120mA, Vctrl = -4.5V, Vdd1,2 = 5V, operating bandwidth 9.2 - 9.5 GHz.

The measured results are reported in the picture below, as you can see the Gain is the typical, but the NF is more than 1.5dB higher than the specification in the band 9.2GHz to 9.5GHz.

Could you please comment on that? In particular, are the specification for the EvalBoard that different in terms of NF respect to the component specification (NF=2dB typ.)?Why this discrepancy?

Do you think the measurement setup is not correct?

Could the component not working properly?

  • 0
    •  Analog Employees 
    on Nov 17, 2020 9:54 PM 3 months ago

    Hi luca.banchi,

    During characterization the component is measured on an evaluation board so the plots on the datasheet were taken on that product's eval board. My initial thought would be that it is a set-up issue as the discrepancy is quite extreme compared to what we got during characterization. Could you give more more details on your set-up? Also are you de-embedding this measurement? De-embedding is critical for noise measurements so if you could give me more details on how you are doing that it would also be helpful.



  • Dear Dan02, thank you for your ans.
    I double-checked the measurement setup that is simply composed by a noise generator directly connected on the evalboard input and a noise gain analyser (R&S FSV40) connected on the output by means of a series of two coax adapter and a DC-block, see picture attached


    In this configuration I got the following results for the NF of the HMC996LP4E on the EvalBoard with respect to the Idd:

    NF (dB)
    Idd / freq 9.2GHz 9.3GHz 9.4GHz 9.5GHz
    90mA 3.2 3.3 3.3 3.4
    100mA 3.2 3.4 3.4 3.3
    110mA 3.3 3.5 3.3 3.5
    120mA 3.3 3.4 3.5 3.4
    140mA 3.4 3.5 3.4 3.4

    After the calibration I tested also a commercial 10GHz LNA in order to check the validity of the calibration process.

    Moreover I tried to insert a 3dB attenuator on the input of the EvalBoard HMC996LP4E measuring a 3dB increment on the NF and 3dB reduction of the G, as expected.

    Do you think I missed something in the de-embedding process?

    With the EvalBoard (s/n: 2018052912518) there was no test data report, is it possible that the component was not working properly?

    Thank you


  • 0
    •  Analog Employees 
    on Nov 20, 2020 7:11 PM 3 months ago in reply to luca.banchi

    Hi Luca, 

    Do you have a VNA available where you can check S-parameters to verify the part is good? Also I would suggest measuring a bullet through to verify the NF system and also installing connectors on the EVB so you can de-embed and compensate the board loss.

  • Dear Dan02,

    in the picture I reported the measured S parameter by means of our VNA

    I double check the gain (S21) with the Noise-Gain Analyser and it seems to be reduced compared to the previous measurements.

    Maybe something went wrong with the chip, or with the manual bias procedure...

    Could you comment on that?

    At this point, in my opinion, going for a replacement part could be a good choice, what do you think about that?

  • 0
    •  Analog Employees 
    on Nov 24, 2020 6:39 PM 3 months ago in reply to luca.banchi

    Hi Luca,

    I agree with your assessment, something is certainly wrong, the gain is significantly lower than expected. Going for a replacement part is probably the best way to go but I just want to make sure your manual biasing procedure/sequencing is okay. Could you give me more details on that?

    - DanO

  • Hi Dan,

    Sorry for late reply, I missed your message!

    Regarding the bias-up and down procedure, I went through the following steps:

    - set Vctrl to -4.5V, Vgg1,2 to -2V and Vdd1,2 to 5V

    - increase Vgg1,2 together untill Idd1,2 reach 120mA

    I got Vgg1,2 of about -0.75V

    This is more or less what the DS reports.

  • +1
    •  Analog Employees 
    on Dec 2, 2020 10:03 PM 2 months ago in reply to luca.banchi

    Hi Luca,

    Okay, your sequencing/biasing is correct and your Vgg12 value of -.75V also seems as expected. At this point I would suggest going with a replacement part as it does seem like we're dealing with a problem relating to the DUT. The only other variable would be a bad EVB but I think that is less likely.



Reply Children
No Data