• State TI Thinks Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 2 answers
  • Subscribers 33 subscribers
  • Views 9874 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.

AWR1443: FMCW radar angle estimation

Zhe Wang
Prodigy 30 points
Part Number: AWR1443

Hi,

I have been watching TI training video about FMCW radars, and noticed angle resolution = λ/Ndcos(θ), where N is the number of receiving antenna.

So if d=λ/2 and θ=0, suggests angle resolution = 2/N, which means: 4 receiving antenna => 28deg, almost no much use in practice. if want 1deg of resolution, that needs tens even over 100 antenna!

Sounds simple math, but the results surprised me, so post here want double confirm.

Many thanks,

Zhe

  • 0
    TI__Guru 58035 points
    Hello Zhe,
    Your understanding is right, if there is only 1 TX and 4RX and a viewing angle of 180 deg the inherent angular resolution would be limited to 28deg. But TI device supports 3 TX , which allows you to realize 4*3 = 12 virtual receivers hence improving the resolution 3x. You can refer to the app note : www.ti.com/.../swra553 for some basic understanding of the virtual receiver array configuration.

    For extremely high resolution we also enable cascading of multiple chips to allow higher number of receivers.

    Regards,
    Vivek
  • 0
    Genius 4165 points

    The resolution you calculated is the "spatial resolution", which is defined by the physical world. However, we can increase the length of FFT in the angular domain to have smaller "frequency domain resolution" than "spatial resolution". For example in 1642 chip, there are 2Tx 4RX, so the "spatial resolution" you can get is 15 degree. If we use 64 points FFT to calculate direction of arrival, this will lead to the 3 degree (180/64) frequency domain resolution. I am not saying that the frequency resolution is accurate, but it is much more finer than 15 degree "spatial resolution". This is the same idea that people always use longer size of FFT than the number of samples.

    -Peter

  • 0
    TI__Prodigy 565 points

    Hi Zhe,

    With regard to your question there are three aspects to consider:

    (1)  One can increase the angle resolution of a radar by operating it in the  MIMO mode. For e.g. with M TX antennas and  N RX antennas, you effectively get MN 'virtual' antennas. In the AWR1443 EVM, there are 2 TX's and 4 RX's for azimuth resolution (=>8 virtual antennas) , giving you a native resolution of 2/8 radians = 14 degrees. If you  had an antenna configuration that deployed all the 3 TX antennas for azimuth resolution, you would increase the native resolution to 2/12= 10 degrees. The appnote http://www.ti.com.cn/cn/lit/an/swra554/swra554.pdf 

    (2) There is a difference between resolution and accuracy. "resolution" refers to the ability to resolve or separate close by objects. [In this case it is the ability to separate two objects which have the same range and velocity with respect to the radar but different angles]. Accuracy, usually refers to the accuracy of the  angle estimate when there is only ONE dominant object at a specific range and velocity.  The accuracy depends on the signal SNR and the antenna non-idealities. Typically, with a good SNR you can get very accurate angular measurements (several times better than the  angular resolution). Processing the angle-FFT using zero padding (and/or using interpolation across bins) can also  help improve the angle accuracy.

    (3) To resolve multiple objects , the radar needs to resolve them in only ONE of the 3 dimensions of range, velocity and angle. So two objects which are resolved in range OR in velocity  do not need to be resolved in angle. Radar's strength is its  great range and velocity resolution (AWR1443, can get a range resolution of a few cm's and velocity resolution of a fraction of a m/s). So in typical radar applications the range and velocity resolution will alleviate the need for  very high angle resolution.

  • 0 in reply to Sandeep Rao30
    Prodigy 30 points

    Thank you so much Sandeep, your reply made things much clearer.

    Now just one more thing I shall appreciate you could double confirm with me:

    if there were only ONE dominant object, it's angle can be estimated, some say, accuracy can be up to 1°

    but if the resolution were poor, say 10°, and there were 2 objects within this angle. Even though these 2 objects can be resolved by range or velocity, the angle of both objects can no longer be estimated.

    am I right?

    Zhe

  • 0 in reply to Zhe Wang
    TI__Prodigy 565 points


    Hi Zhe,
    If two objects are resolved in range or velocity, then their angles can be independently estimated using the  angle-FFT. [Does not matter how close in angle they are].

    The 2D-FFT resolves objects in range and velocity. So after the 2D-FFT these two objects will show up in different cells in the 2D-FFT grid (i.e the signals from these two objects no longer interfere with one another). Note that there will be one such 2D-FFT grid for each antenna. For each object, the angle-FFT is then performed on the 2D-FFT cell corresponding to that object, across antennas . Slide #64-70 of training.ti.com/.../mmwaveSensing-FMCW-offlineviewing_0.pdf also discuss something similar

    Let me know if this helps.

    Sandeep

  • 0 in reply to Sandeep Rao30
    Prodigy 30 points

    Thank you Sandeep, very helpful.

    Maybe you could give me more help if I tell you what my application scenario is. We have multi moving objects in a field of view of approximately ±50° and range within 50m. we want to draw a map to show each object where it is.

    Each object is unlikely have the same range or velocity, so as said can be resolved by range or velocity. Then we can use angle-FFT to measure the angle of each object - this sounds we don't need a lot receiving channels, that will make the system difficult - so what is the minimum receiving channel that your can suggest, and another question is, why many TI documents mentioned cascade use of the chip so that receiving antennas can be tens or even over 100?

    Many thanks,

    Zhe

  • 0 in reply to Zhe Wang
    TI__Prodigy 565 points
    Hi Zhe,
    For your stated application a single radar chip should be sufficient. You should not need any cascading.
    Sandeep