FMCW Radar for Autonomous Vehicles | Understanding Radar Principles

[@mab7727]

Hands down, the best video on youtube explaining radars!

[@trinityflow]

Really nice job! Thanks. We at dSpace have auto radar test HW that can allow simulation of up to 1000m distance in various frequency ranges especially E-Band 75 to 82GHz (up to 5GHz BW) but over a short distance test set up in the lab! This allows auto/radar vendors to easily test HW for scenarios of e.g. FMCW, Doppler, MIMO => location, speed, angle, and differentiation of 2 cars ahead for example.

[@christianknutson3733]

This is an amazing explaination. Thank you for putting this together!

[@seb_3366]

Love the explanation with the signal chart in frequency and time domain. Thank you for that video.

[@abdolvakilfazli2488]

Very well put. I'm halfway finishing it and I just liked your video. Thank you so much.

[@ChiLLPiLL-g1i]

Amazing man. Absolutely perfect concepts.

[@jbrookley]

This is a fantastic video! Thanks for this!

[@kumark353]

Thank you Brian🙏, Great explanation to implant Idea and reason behind design than simple analysis

[@thepepperboi]

Super helpful. Thank you for putting this video together.

[@nassional]

Thank you for your work. One of the cleanest explanations about radar. I have a question, how did you filter the doppler beat frequency in Matlab environment? And how can I do this in practice?

[@mab7727]

This was legendary! THANK YOU, GREAT ONE!

[@michaelbafaro1953]

I agree with Chris!

[@habibihooman]

@Brian Douglas Hi Brian. You made a fundamental mistake at 8:54. You don't transmit an IQ signal. Transmitting sin + cos is the same as transmitting a shifted sinusoid, which does not solve any problem. It is like shifting the transceiver by quarter wavelength. What you need is performing IQ downcoversion in receiver. Otherwise nice video and beautiful presentation.

[@prescottluah6404]

Hi habi, actually my thought is that Brian had computed the Real and Imaginary (IQ) components of the Mixed signal being thereafter processed of prior transmitted signal. Hence, we merely need the Magnitude and Phase of mixed signal to derive In-phase (Real) and Quadrature (Img.) components in Complex Sinusoid form as indicator in relative to determine whether it's Approaching or Moving Away target, if there're any of other thoughts or opinions, feel ease to share, Thanks habi :)

[@supune]

Can you do a demo for processing of radar vs lidar vs sonar? Is the main determination of the technology to use cost and resolution?

[@aryachandanreddy120]

Just dope explanation

[@robertfitzgerald5432]

All nicely explained until 11:54 when you show the Beat Frequency as the difference between Tx (yellow) and Rx(pink). However the difference is the same wherever you are on the ramp. Therefore how do you use the beat frequency to deduce the range if it’s the same value along the time axis ?

[@StealthySpace7]

Very well done and explained

[@MrNess2911]

Amazing! I see the light 😊

[@FilipC75]

Truly amazing video.

[@arifistiakabeg]

Thank you. It was very helpful

[@tarcisio_menezes]

Awesome! Thank you!

[@DeeAviator1]

Nice Video!! thank you so much. Quick question -If we precisely estimate the received frequency, a Computer could estimate the direction of the target by simply subtracting the frequency of the received signal from the frequency of the transmitted signal? If so, then why do we need an IQ and beat frequency concept?

[@BrianBDouglas]

Good question! It's really hard (expensive) to measure the precise frequency when it's in GHz. But it much easier to measure a signal that is in KHz (like the beat frequency). If you did have a super precise computer that could estimate the receive frequency in GHz then I believe you could determine the Doppler that way. You'd still need IQ signals to determine direction though.

[@HungTran-zd3dd]

Thanks for the video. But I get confused at 12:36 sec. Could you explain more about the range doppler chart? I don't understand why we have a rising line in the chart, then we have another falling line? :/

[@georgederleres8489]

This is the only part of the video that was not well explained, I agree with you on this. I think I understood how these lines are derived: You have to find the two points that intersect with the Range,Doppler axis. We can find the point on the doppler axis by shifting the receiving curve on the right until we have 0 range (2 beet frequencies equal), and measure how much is the doppler-caused beat frequency there. Then, for the point on the Range axis, move upwards/downwards the receiving curve until doppler-caused beat frequency is 0 (2 beat frequencies have reverse sign). From these 2 points you can get the lines. Depending on the sign of these points relative to each other, you can get both the falling and rising lines.

[@mohamedtalha9790]

Thanks for these nice explanations.can you publish the codes you used to illustrate the idea ?

[@ejg6660]

Hi; great video series - I have watched up until ep4 (pulse-doppler) and the explanations have been great. Some questions: where you say the i and q signals are orthogonal, do you mean physically orthogonal, i.e. two perpendicular antennas (e.g. cross dipole)? If I'm not mistaken, wouldn't orthogonal, 90deg phase shifted signals result in circularly polarised radiation? Presumably it's not possible to produce i, q signals with a single antenna? Thanks in advance.

[@prescottluah6404]

Hi ejq, base on my understanding it's rather mathematical than physical, we may read on Complex Sinusoid form, as Brian had presented such that of single resultant complex field form (single poln.)

[@emmanuelpradera1598]

What do you mean when you say the object could have some doppler and cero range at 12:40? If range is the distance from the radar to the object it must have some value. I don´t understand this part. Thanks

[@edulchan]

How determine the direction if transmitte an IQ signal? That point I didn’t get it.

[@habibihooman]

He made a mistake. What you actually do in the receiver you perform IQ downcoversion. What you transmit is always a sinusoids. If you transmit a sin + cos then it is just a shifted sinusoid, which makes no difference with transmitting only sinusoid.

[@kawrx2002]

So beat frequency is proportional to radial velocity and direction is determined by which signal is leading in phase......how do you know which one is leading, e.g. is signal A shifted positive by pi radians or is B shifted negatively by pi radians?

[@benjyskinner]

The shift isn't bit +/- pi radians, it is +/- pi/2 radians, which makes it possible to determine which signal is leading.

[@sagaradoshi]

Hello, It is a very good presentation..Thanks for the video. I have a doubt at 9.3 sec. Firstly how we mix the transmitted and received signals? Because we have two transmitted and two received signals. So how the sine components look like? Any reference? Secondly, you mentioned that looking into which signal leads the phase after mixing (i.e., the real and imaginary) we can determine whether object is coming towards or moving away..this I couldn't understand. In Earlier part of video while explaining Doppler effect I understood that if freq of received wave is more than that of transmitted then object is coming near otherwise moving away. But with IQ signal concept I couldn't understand. Is it like if Real leads the imaginary signal then object is moving near otherwise its moving away? Looking forward for your support.

[@eddievhfan1984]

1. Only one phase needs to be transmitted from the RADAR antenna; the local oscillator used to generate the signal can provide the in-phase and out-of-phase components to be mixed with the received signal for motion analysis. Showing the two phases transmitted in the app was mostly for demonstration, since when the two are added together, they form a half-power wave shifted 45° from either reference phase. 2. The output of such RADAR systems to their signal processors is the Doppler frequency produced by mixing, not the raw RADAR frequencies on their own. The result of this mixing is that the magnitude of the shift is what's returned, not the sign of the shift. The only way to determine the sign (opening/closing range) of the Doppler shift without IQ encoding would be to directly measure the frequencies of the returned signals, or to also use the upper-sideband beat frequency (the ones up in the high GHz band). Since most RADAR transceivers are in the GHz band and the Doppler shift we're looking for is on the order of kHz, you'd need an exceptionally high-bandwidth receiver and ADC to measure it, driving up expense and electronics size. IQ encoding is considerably less expensive of a method, since all you need is an extra I/O line to the signal processor, and a phase-shift network/delay line and/or PLL to give you the quadrature component for mixing.

[@tungtritrinh1381]

@@eddievhfan1984 So the radar antenna only transmit the In-phase component of the I/Q signal ?

[@eddievhfan1984]

@@tungtritrinh1381 Essentially.

[@tungtritrinh1381]

@@eddievhfan1984 Can you give me a more in-dept explanation to how the Radar calculates the Quadrature beat frequency since the local oscillator only provide IQ component for the receive signal ? Im still confused cuz i dont study in Signal Processing.

[@eddievhfan1984]

@@tungtritrinh1381 It's not too complicated. As I mentioned, the transmitter only sends out one phase in its pulse. Phase shift of the reflection is caused by moving targets, while stationary targets will remain in-phase. (Fun fact: this can be used by itself to provide moving target indication (MTI) RADAR without detailed Doppler processing). The received signal is mixed with both reference signals separately before entering the signal processing stage.

[@LatifaSharifi-j1g]

In 12:50 , you said Sawtooth-like modulation doesn't provide a unique information about range-doppler? But I have seen many other tutorials that they take 2D-FFT and draw the range-doppler plot. Are they wrong?

[@LatifaSharifi-j1g]

or this ambiguity is only when we have either of zero range or zero doppler?

[@terrybrown3199]

Wouldn't multiple signals from several cars create interference between the cars?

[@tvpi-z1v]

The world is going opensource, you need to update your model

[@anim8dideas849]

One day, they already have plenty of donors.

[@marklundeberg7006]

The section from 8:45 to 10:00 is wrong and (IMO) too misleading. The physical signal we send is always real, and the signals we receive are real. We cannot transmit an imaginary signal, and we cannot receive an imaginary signal. Not only are the 'imaginary' signals on the upper two plots actual nonsense, but also the 'in phase' and 'quadrature' shown on those plots are literally misleading concepts. Especially the receive plot, that is a really misleading mistake in what in-phase and out-of-phase mean. What actually happens in this radar is we transmit the same real CW signal and we receive the same real doppler shifted CW signal. No changes at all there. The only thing we have improved is the demodulator: we send the receive signal through an I,Q demodulator which produces the I,Q complex values, i.e., it decomposes the received signal into one part that is in-phase with the transmission, and one part that is out-of-phase with the transmission. The I,Q then circle clockwise or counterclockwise on the complex plane depending on frequency shift direction, i.e. the I,Q values go up and down as shown on the lower plot. Only the lower plot is correct. So ultimately what the complex I and Q demodulation does is it allow us to distinguish the positive and negative doppler shifts, which is information that was in the reflected signal all along but that information had been discarded by the primitive beat frequency demodulation described earlier.

[@painpenguin]

Bro i want to ask something. What wave form just radar actaully recieve? transmitted+recieved , recieved or transmitted*recieved And if it is (transmitted+recieved , recieved) these two. Why we dont just use fft and get it?

[@marklundeberg7006]

@@painpenguin In general we want to think about radio frequencies in the GHz, which would take many gigabits per second of processing power to do FFTs. So that's why we do some mixing in the electronics first which reduces the bandwidth of concern. Actually for low frequency you can literally just measure and FFT. For example I have a cheap software defined radio that can do this up to 100 kHz if I remember right.