• State Resolved
  • Replies 13 replies
  • Answers 5 answers
  • Subscribers 31 subscribers
  • Views 289 views
  • Users 0 members are here
Support feedback
Options
Options
Related

Original question:

IWRL6432BOOST: I'm using IWRL6432 Boost EVM with DCA1000A to capture RAW ADC Data. However i see NO LVDS on Output

IWRL6432: A suitable EVM for IWRL6432FSPEVM

Shubham Jena
Prodigy 20 points
Part Number: IWRL6432
Other Parts Discussed in Thread: IWR6843AOPEVM, IWR6843,

Tool/software:

I am trying to make a mmWave radar solution to track multiple mouse/rats within 1-5mtrs, and need a good mmWave radar sensor to detect these mouse/rats, something which can help also track the behavior or might also track some vitals, without using a camera.
Now these mmWave radar might not be the most optimal case, but we are just doing a small feasibility research.
But I am confused for which Devkit to buy, as a few of them say connect with the DCA1000 EVM board and it says it has a higher bandwidth, and is too costly, what is the decent priced EVM board which can help me use all the possible functionalities of IWRL6432 , without compromising any features.

Also is IWRL6432 the best choice for my application or I should use any other ti-mmWave SoC ?

Please guide me.

  • +1
    TI__Mastermind 23177 points

    Hello,

    The DCA1000 is for capturing unfiltered, unprocessed, dumped ADC data for DSP algorithm development and research from scratch. You do not need need this, as the example demos found in the Radar Toolbox are meant to be ran with just the radar sensor.

    I actually think mmWave Radar technology is a solid choice here for detecting small pest animals in dark/hard to see places. Radar's strengths is detecting the most subtle of movements due to it high sensitivity to movement (doppler), and the fact it works under any lighting/weather conditions. The target application of Robotic Lawnmowers actually is similar to this due to the concerns of seeing small objects including squirrels/hedgehogs and low visibility (dirt/grass). Since this is more of a feasibility research kind of study, I would recommend the IWR6843AOPEVM as it is the higher performance device with a 4x3 antenna rather than 3x2. If things works out, then there can be considerations to move to the low cost/power consumption IWRL6432

    Best Regards,

    Pedrhom

  • 0 in reply to Pedrhom Nafisi
    Prodigy 20 points

    thanks a lot sir, solved all my doubts, also had one more question, these radars detect heart rate or respiration rate , from the targets body slowly expanding and contracting, but not moving on the ground. 

  • 0 in reply to Shubham Jena
    TI__Mastermind 23177 points

    Hello,

    It is from the target's chest expanding and contracting. As well as the beating of the heart. Radar is very good in the speed/doppler domain, so much so that we are able to get approximate vital signs information from a distance.

    Best Regards,

    Pedrhom

  • 0 in reply to Pedrhom Nafisi
    Prodigy 20 points

    so the IWR6843AOPEVM , is it capable of providing me the raw data which I can use to further process or train a model or an algorithm to do so for my target small mouse or rats, with low rcs signature?

  • 0 in reply to Shubham Jena
    TI__Mastermind 23177 points

    Hello,

    If you want to create your own model, you can go with the "raw data" approach, but for some guidance if you want to do it the way we do, we use radar micro-doppler. This is a concept you should look into as it is it'll be the key feature. Below is a presentation from one of our algorithm developers

    https://cms.tinyml.org/wp-content/uploads/ew2023/Muhammet-Yanik_tinyML-Asia-2023.pdf

    What I want to point out is from slide 6 which is the uDoppler Spectrum. If you do not have memory constraints and want the model to run on the PC rather than on chip, you can train a model by feeding in the whole micro doppler spectrum. The micro doppler spectrum is an output type of the Human Non-Human example demo for the IWRL6432. Unfortunately this example demo has not yet been ported to IWR6843

    Best Regards,

    Pedrhom

  • 0 in reply to Pedrhom Nafisi
    Prodigy 20 points

    okay, thanks, so IWR6432AOPEVM, i can buy this right , to train custom models and get raw data, or I should i buy IWRL6432BOOST kit, and do i need DAC1000 ?, i have a little tight budget.

    Or the IWR6843AOPEVM or IWR6843BOOSTEVM

     

  • 0 in reply to Shubham Jena
    TI__Mastermind 23177 points

    Hello,

    Go with just the IWRL6432BOOST, use the uDoppler spectrum TLV that is transferred over UART in the Human Non-Human example project. Download the toolbox using the web browser method.

    Best Regards,

    Pedrhom

  • 0 in reply to Pedrhom Nafisi
    Prodigy 20 points

    thanks for helping me so far, will surely go with the IWRL6432BOOST EVM, just one last thing, this is capable of detecting (after proper filtering and dsp stuff) sub-Millimeter level changes [contractions from body, to measure a vital like resp or heart rate] of a mouse/rat standing still within 1-5mtrs, and if so then what should I expect and be prepared for, just a direction or some resources to have a practical understanding.

  • 0 in reply to Shubham Jena
    TI__Mastermind 21071 points

    Hi,

    Pedrhom is out of office today. He will be able to get back to you when he returns. 

    Best regards,

    Josh

  • 0 in reply to Josh Dye
    Prodigy 20 points

    thanks, sure no problem

  • 0 in reply to Shubham Jena
    TI__Mastermind 23177 points

    Hello,

    I would not say quite sub-millimeter but more in the 0.5cm to 1cm domain level of changes when looking at a XYZ point cloud. I think the easiest way to go about this if not running the model on chip is to just feed in that uDoppler spectrum as a heatmap. The pattern of that heatmap will be different for different objects. It is the way these sub-centimeter levels of changes move in a pattern that can differentiate humans from not humans and especially static objects.

    Best Regards,

    Pedrhom

  • 0 in reply to Pedrhom Nafisi
    Prodigy 20 points

    so to track respiration rate for a mouse or a small bird, what data should I collect and what pipeline should I follow to begin with, and which DevKit should I buy, just simple steps and concepts 

  • 0 in reply to Shubham Jena
    TI__Mastermind 23177 points

    Hello,

    Ideally just pointing the radar sensor at the targets you want data of and recording their uDoppler spectrum as a heatmap is what I would consider easiest. IWRL6432BOOST EVM

    Best Regards,

    Pedrhom