• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 52 subscribers
  • Views 134 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.

CC2745R10-Q1: Adaptative ranging algorithm

Paul Chen
Prodigy 40 points

Part Number: CC2745R10-Q1

Dear TI Support Team,

We're using the Channel Sounding (CS) feature on the SimpleLink CC2745R10-Q1 and need clarification on the various ranging algorithms.

image.png

We require detailed documentation on the NN (Neural Network) Algorithm's internal operation, including the specific channel features (inputs) it utilizes. We also need the exact memory footprint (RAM/Flash) of the algorithm. 

In addition , can you confirm that the Adaptive Algorithm intelligently blends the results of IFFT, NN, and MUSIC?

If so, we did a series of testing and the Adaptive Algorithm yields an accurate distance (close to ground truth), while , its individual components NN and IFFT return zero (0), and MUSIC returns a poor/incorrect value. Could you provide an explanation ?

Thank you for your technical support.

Sincerely,

  • 0
    TI__Expert 8925 points

    Hello Paul, 

    1. A detailed documentation of the NN internal operation cannot be provided. The NN is developed by our team, and we are not providing more details about the internals of the algorithm. I can work on getting memory numbers for you though. 

    2. The Adaptive Algorithm uses the NN and the Music algorithm. From a high level, MuSIC will be used for LOS, and NN for NLOS. 

    3. MuSIC is effective for LOS situations. Were you testing MuSIC in NLOS environments? Additionally, all values are 0 for NN and IFFT? 

    Additionally, what SDK are you using? What antenna configuration are you using? 

    Please let me know. 

    Thanks,
    Isaac

  • 0 in reply to Isaac Larson
    Prodigy 40 points

    Hello Isaac, thank you for your reply

    1. If possible, yes, could you provide us some information about the memory usage for the NN algorithm?

    2. Is the switch between LOS and NLOS determined by the quality of the signal? If possible is there a way to know when it use the Music or the NN ?

    3. We were doing test in LOS and NLOS and yes for the NN and IFFT we had 0 

    We are currently using the 9.14.0.39 SDK, and we are in a 2x2 configuration: two antennas for the sender and two antennas for the receiver.

  • 0 in reply to Paul Chen
    TI__Expert 8925 points

    Hello Paul, 

    1. This data can be seen within the .map file, when looking for BleCsRangingNet. 

    2. I am mistaken. My last response is not accurate for the algorithm operation. Currently, we are treating the algorithm as a black box, and TI will not be disclosing the operation of the algorithm. 

    3. I have checked with my team to see if NN, MuSIC and IFFT work independently. I will get back to you next week (12/1) due to US holiday. 

    Thanks,
    Isaac