• State Not Answered
  • Replies 4 replies
  • Subscribers 74 subscribers
  • Views 127 views
  • Users 0 members are here
Attachments (0)
Nordic Case Info
  • Case ID: 345285
Options

Improving Channel Sounding Accuracy with nRF54L15 Boards

NickDeLeenheer

Hi,

I am currently conducting measurements using channel sounding with my nRF54L15 boards. I am at the end of my master thesis, so there is not so much time left to really come with IFFT distance estimate algorithms etc., and my supervisors don't expect me to do this asswell. The outdoor performance is quite good, but I am still encountering a random error that results in overestimation (estimating a larger distance than the actual distance).

However, when measuring indoors—particularly in complex environments like halls with many reflections—the range estimates are significantly less accurate.

I noticed that adjusting parameters like the number of antenna paths (n_ap) is not straightforward, and I feel like it is difficult to modify on my own due to the limited available information. Also, since I am not required to develop specific algorithms for the data, I am unsure how much I can contribute to improving accuracy.

Initially, I hoped that using multiple frequencies would mitigate the effects of obstacles. However, it appears that reflections still have a major impact on accuracy.

Could you provide some insights into how the Channel Sounding algorithm selects the shortest path? Specifically, does it prioritize the path that reaches the device the quickest (which would ideally be the direct Line-of-Sight path), or does it select the signal with the highest RSS (Received Signal Strength)?

Any suggestions on how to enhance indoor accuracy would be greatly appreciated.

Kind regards,

Nick De Leenheer

Parents
  • 0

    Hi Nick,

    so there is not so much time left to really come with IFFT distance estimate algorithms etc., and my supervisors don't expect me to do this asswell.

    Understandable. 

    The outdoor performance is quite good, but I am still encountering a random error that results in overestimation (estimating a larger distance than the actual distance).

    So you are not getting errors in the logs or anything, just inaccurate results?

    Could you provide some insights into how the Channel Sounding algorithm selects the shortest path?

    It doesn't use RSSI. It uses either the I/Q data from the signal, or the RTT. I believe our solution currently uses the phase slope, the tech behind is rather similar to our propriatery and older "nordic distance toolbox" which you can see a webinar here that describes how it works. Channel sounding is expected to work a bit better than the NDT in theory (ie. with a good algorithm etc).

    We currently haven't focused much on optimizing the algorithm on our side for the channel sounding samples, and it is easy to see - like you have - that they are lacking. In NCS 2.9 we also said so in the documentation for the samples here. 

    The accuracy is not representative for Channel Sounding and should be replaced if accuracy is important.

    Though in NCS 3 we've apparently improved it a little bit.

    So to summarize:

    • Make sure that you are using NCS v3.0.1
    • Maybe try a median filter?
    • Consider purchasing a third party algo for channel sounding
    • Consider using the older and propriatery Nordic Distance Toolbox instead, which does get you better results with what is freely available, though it does require you to get NRF52 DKs - it doesn't run on nRF54.

    If these suggestions doesn't work for you, I can try asking the relevant R&D team if they have any suggestions for how to get better results. Though I think this would involve more work for you, and if your thesis doesn't focus on how to optimize channel sounding I think it sounds like a waste of time.

    Good luck no matter what you end up with! Let us know if there is anything else Slight smile

    Regards,

    Elfving

  • 0 in reply to Elfving

    Hi,

    I am currently installing NCS 3.0.1 and would really appreciate some guidance on what is available in this toolchain that could help me.

    Is there a new sample related to channel sounding that I should be aware of? Also, I noticed there are mentions of implementing improved algorithms, but I am finding it difficult to find real-world examples or case studies of people who have succeeded in achieving better accuracy. Do you know where I could get in touch with someone who has hands-on experience or expertise in this area (possibly a third party)?

    I have also experimented with a kind of median filter where I take the smallest distance estimate over N measurements. Unfortunately, this did not lead to improved accuracy outdoors. Perhaps it could manage better accuracy for indoors.

    One of the biggest challenges I am facing is dealing with reflections, which significantly reduce accuracy. Are there known methods to mitigate this issue?

    I would really appreciate being put in contact with someone who has successfully achieved better accuracy, or at least someone who could explain the inherent limitations. I am particularly curious about how the Phase-Based Ranging (PBR) samples are selected, as I suspect that in complex indoor environments, multipath signals might be incorrectly chosen, leading to inaccurate distance estimates.

    Looking forward to your insights and suggestions.

    Kind regards,
    Nick De Leenheer

  • 0 in reply to NickDeLeenheer
    NickDeLeenheer said:
    Is there a new sample related to channel sounding that I should be aware of?

    These (here and here) are the ones we've made, and there is also one directly from Zephyr. 

    NickDeLeenheer said:
    Also, I noticed there are mentions of implementing improved algorithms, but I am finding it difficult to find real-world examples or case studies of people who have succeeded in achieving better accuracy.

    I wouldn't know about any case studies etc. What I was referring to regarding this is that you can perhaps purchase something, though I wouldn't know from whom, and I shouldn't recommend a certain 3rd party here either way.

    NickDeLeenheer said:

    I have also experimented with a kind of median filter where I take the smallest distance estimate over N measurements. Unfortunately, this did not lead to improved accuracy outdoors. Perhaps it could manage better accuracy for indoors.

    For a median filter you would typically take the middle value, though if you are using the phase slope then this idea could make sense, as you can think of the results you get as a maximum.

    NickDeLeenheer said:

    One of the biggest challenges I am facing is dealing with reflections, which significantly reduce accuracy. Are there known methods to mitigate this issue?

    There is probably a lot to consider here and multiple ways one could do this. Kalman filtering for instance. Though doing any of this is a larger undertaking.

    Regards,

    Elfving

Reply
  • 0 in reply to NickDeLeenheer
    NickDeLeenheer said:
    Is there a new sample related to channel sounding that I should be aware of?

    These (here and here) are the ones we've made, and there is also one directly from Zephyr. 

    NickDeLeenheer said:
    Also, I noticed there are mentions of implementing improved algorithms, but I am finding it difficult to find real-world examples or case studies of people who have succeeded in achieving better accuracy.

    I wouldn't know about any case studies etc. What I was referring to regarding this is that you can perhaps purchase something, though I wouldn't know from whom, and I shouldn't recommend a certain 3rd party here either way.

    NickDeLeenheer said:

    I have also experimented with a kind of median filter where I take the smallest distance estimate over N measurements. Unfortunately, this did not lead to improved accuracy outdoors. Perhaps it could manage better accuracy for indoors.

    For a median filter you would typically take the middle value, though if you are using the phase slope then this idea could make sense, as you can think of the results you get as a maximum.

    NickDeLeenheer said:

    One of the biggest challenges I am facing is dealing with reflections, which significantly reduce accuracy. Are there known methods to mitigate this issue?

    There is probably a lot to consider here and multiple ways one could do this. Kalman filtering for instance. Though doing any of this is a larger undertaking.

    Regards,

    Elfving

Children
Related
Terms of service