• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 31 subscribers
  • Views 362 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

IWR1443: Scaling and Maximum Value of Magnitudes

Michael Linden
Intellectual 375 points
Part Number: IWR1443

Hi Akash:

  Since the prior associated thread has been locked, I've opened a follow-up thread. Here's a link to the original thread:

e2e.ti.com/.../iwr1443-scaling-of-magnitudes

Hi Akash:

  Yes. My application also requires presentation of the radar scene (amplitude versus distance). This "Range Profile" data will be the basis of of that graph. However, I need to understand the maximum amplitude of those values in order to scale the amplitude portion of that graph. In addition to that, I need to understand the maximum amplitude presented by the results of the zoom FFT (like dataPathObj.rangeProcStats.maxValue), and how that amplitude relates to the amplitudes in the Range Profile. In reality, knowing any two of these three items should be adequate. Does this make sense?

  Thanks, Michael

  • 0
    TI__Mastermind 23725 points

    Hello

    Are you able to access or identify the relevant code blocks and/or data structures that hold the range FFT and Zoom FFT results?

    Thank you

    Vaibhav

  • 0 in reply to Vaibhav Mahimkar
    Intellectual 375 points

    Hi Vaibhav:

      Yes, I'm sure I can get to all of the relevant data structures/objects using the debugger. When I view the Range Profile (output of the 1D FFT) using the debugger I can see that the bin for a strong nearby reflection will indicate a value of about 10600 whereas bins that are not associated with a reflection (~noise) typically show smaller values of a few hundred units or so.

      So, as previously noted, how can I calculate the largest value I should expect to see for a given configuration?

      I found the below "Scaling" information in the “Wave (mmw) Demo for XWR14XX” Doxygen information. However, I'm not sure how the variables discussed are related to the the configuration information in the "High Accuracy Lab". My attempts to use this information to arrive at a maximum value for the 1D Range Profile data has resulted in values that are far too small to be correct.

    "1D processing: If the HWA's FFT scale is set to where is the number of stages for which the scaling is enabled, and input to the FFT were a pure tone at one of the bins, then the output magnitude of the FFT at that bin will be ( is the FFT order) times the input tone amplitude (because tone is complex, this implies that the individual real and imaginary components will also be amplified by a maximum of this scale). Because we do a Blackman window before the FFT, the overall scale is about 1/2.4 of the FFT scale. This means for example for 256 point FFT, the windowing + FFT scale will be . For k=2 which is currently how it is in the implementation (no matter the FFT order), this will be 26.7. Therefore, the ADC output when it is a pure tone should not exceed +/-2^15/26.7 = 1228 for the I and Q components (even though HWA is internally 24-bit, the FFT output is stored as 16-bit before 2D processing, hence 2^15). The XWR14xx EVM when presented with a strong single reflector reasonably close to it (with Rx dB gain of 30 dB in the chirp profile) shows ADC samples to be a max of about 2000 and while this exceeds the 1228 maximum, is not a pure tone, the energy of the FFT is seen in other bins also and the solution still works well and detects the strong object."

      Thanks, Michael

  • 0 in reply to Michael Linden
    Intellectual 375 points

      I believe I've empirically arrived at the answers I need. For other's who may have the same questions, I'm using code that is based on TI's "High Accuracy Lab" and the size of my 1D FFT is 1024. By cranking the gains up, I've been able to determine that the 1D FFT Range Bins are saturating at 32767. I find this interesting as the Range Bins are held in an array of uint16_t, but 32767 is the largest positive number that can be represented by an int16_t. Regardless, experimentation clearly shows that 32767 is the maximum value of the Range Profile bins for my configuration.

      Likewise, I've also been able to confirm (at least for my configuration) that the amplitude calculated by the Zoom FFT (maxValue member) is consistently 16 times larger than the associated Range Profile bin (the location of which is specified by the maxIndex member).

      -Michael