• State TI Thinks Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 4 answers
  • Subscribers 11 subscribers
  • Views 1026 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TRF370417: Noise Floor discrepancy from DS

Al B
Intellectual 2900 points
Part Number: TRF370417

From my customer...

Now we are facing noise floor issue of TRF370417.  We tested the EVB of TRF370417 and found the test result worse than the value specified in datasheet. could you please check if our test condition and ways are right? In the meantime, could you please show us the detailed test condition and how to get the performance of noise floor specified in the datasheet?  As this is a little urgent for our a NPI solution decision, we highly appreciate your quick response.

 

Test result:

Test results(PXA Avg type: RMS. Preamp on and NFE on) for your information:

2140MHz, offset 13MHz, ~ -158dBm/Hz.  Round 4dB gap with datasheet

2400MHz, offset 13MHz, ~ -157dBm/Hz.   Round 3dB gap with datasheet

3400MHz, offset 13MHz, ~ -154.5dBm/Hz.  Round 4.5dB gap with datasheet

  • 0
    TI__Mastermind 18065 points
    The Noise Floor measurement is indeed difficult to make. We typically use a phase noise analyzer to complete the measurement since it has the lowest noise floor, better than the spectrum analyzer with pre-amp. You can still use the PXA with pre-amp, but you need to mathematically subtract out the contribution from the spectrum analyzer. With a -169 dBm/Hz floor of the spectrum analyzer there is still a little bit of contribution.

    The largest impact is likely the phase noise of the LO source. This noise source can be eliminated by using a low noise signal generator and applying a stout BPF on the LO that provides significant attenuation at 13 MHz offset.

    LO drive level is also a key contributor. Ensure LO drive level is at the desired set point at the DUT, accounting for the losses of the cable and EVM trace loss. You can play around with the amplitude to see the effect of drive level on output noise.

    The second largest contributor likely comes from the BB source if it is connected to the BB ports. The BB ports need to be biased at 1.7V. Even if the BB generator is "off" there is likely still a lot of noise on those lines. We keep the BB ports connected to the BB generator through stout 200 kHz LPFs. This way the DC common mode is maintained but noise at higher offsets is significantly attenuated. Alternatively, you can supply the Vcm with a simple pull-up / pull-down voltage divider on each port.

    The last consideration is the power supply. Even "good" bench power supplies have high frequency noise. We use a special linear power supply that has very low noise and then add a stout power filter to lower the noise even further.

    All that said, I am not a proponent of duplicating these types of measurements. A real world ACPR measurement, for example, encompasses the effects of linearity, noise, and LO phase noise and is a better indication of device performance and usability.

    --RJH
  • 0 in reply to RJ Hopper
    TI__Intellectual 2900 points

    Customer response...

    Thanks for your coordination and your team’s quick response.

     From your email, we know the factors impacting the test result. However, from our test conditions in my email on Mar,8th, you can see we strictly follow the requirement in the datasheet and use the best instruments which we can found.

    Could you please let your team actual test this specification and share us the test conditions and result? Then we can try to duplicate your methods and check if we can achieve the typical value in datasheet?

    See attachedTRF3740417.docx

  • 0 in reply to Al B
    TI__Mastermind 18065 points

    If you need to independently test/verify the noise performance then I will summarize the requirements based on your provided block diagram which is consistent with my previous post.

    • Place 200 kHz LPF on I/Q lines (4 total) between BB gen and DUT
    • Place BPF at LO frequency between S/G and LO port of DUT
    • Place a power LPF between power supply and DUT; ensure 5V at DUT
    • Calibrate LO power to achieve desired LO drive level at DUT
    • Measure PXA noise floor with pre-amp on and input terminated; use this value to mathematically subtract out from DUT measurement

    This test requires some specialized filtering and is difficult to replicate.  That is why I generally recommend to test/verify a more real-world parameter like ACPR that will be more important for your system.

    --RJH

  • 0 in reply to RJ Hopper
    Prodigy 10 points
    We used a very good 2.4GHz LO which has -170dBc/Hz phase noise at 10MHz to 100MHz offset, even without narrow BW BPF filter, it is good enough and should not impact the modulator’s noise floor.
    The +5V power supply, we also measured the noise floor which is also close to -169dBm/Hz noise floor at 100MHz offset.
    And we provide a very good 1.7V DC common voltage to I+/I-, Q+/Q-, this common voltage noise floor at 100MHz offset is also close to -168dBm/Hz.
    At above conditions(no input signal), the modulator’s output noise floor is -156dBm/Hz measured 100MHz offset of 2.4GHz(2.5GHz).
    Which is worse than datasheet about 5 to 6dB.

    Could you explain any other reasons, or help to measure it in your lab.
    We really suspect how did 99TI get -161dBm/Hz output noise floor of this part.
  • 0 in reply to Runtao Liu
    TI__Mastermind 18065 points

    Let's assume the following:

    • LO performance is:   -170 dBm/Hz
    • Power Supply is:       -169 dBm/Hz
    • BB Source is:            -168 dBm/Hz
    • S/A noise floor          -169 dBm/Hz
    • TRF370x                   -161 dBm/Hz

    The combination will yield a measured performance of about -159 dBm/Hz.  That is still a little off from your measurement.  You are reporting -156 dBm/Hz.  Assume all the equipment values are correct that would imply the TRF370x is -157 dBm/Hz.  Let's say the device is really performing at that level.  There is some statistical variation on this parameter and I have seen values dip down to -159 or -158 dBm/Hz.  I have not seen them go as low as -157 dBm/Hz but this is not a parameter that we measure on the ATE so there is not a large statistical sample to draw from.  Let's assume that -157 dBm/Hz is within statistical variation of the device; is that a showstopper?  What system spec does this parameter impact that results in non-compliance or is this an exercise to validate all the parameters in the datasheet?

    If your parameters are correct then I do not see any other reason for the performance; it is valid to blame it on a low performance of the TRF370x.

    --RJH

  • 0 in reply to RJ Hopper
    Prodigy 10 points
    In fact, the IQ modulator is limiting our 800MHz BW ACPR measurement result. And I tried to increase the LO power from +8dBm to +12dBm, the noise floor and ACPR both improved around 0.4dB, this means the IQ modulator's noise floor is the system bottleneck, so we really need the datasheet noise floor level. We tried 3 IQ modulator part on same board, the noise floor is almost same, so we didn't find a statistical variation of the device. Is there any other reason, or the IQ modulator we bought is worse performance batch than your factory ATE test. Could you send us some better part which is measured in your lab first, then if it works, we can consider screen parts with higher cost.
  • 0 in reply to renqing song
    TI__Mastermind 18065 points
    This thread will be taken off-forum for further discussion and evaluation. --RJH