SPYY005 article on The fundamentals of millimeter wave sensors

Hi,

My first basic question is how delta phi (eq. 16) is measured.  If the scatterer is a point scatterer, then a sinusoidal wave form (as described by eq 7) is generated by each received signal at the receiving antennas.  The leading or lagging phase difference for these two signals can be determined by monitoring the zero crossings.  If the scatterer is complex in nature comprised by many point scatterers of various strengths, zero crossing is not an option.  How does ti measured delta phi?

My next question is from figure 12.  It indicates that the most accurate aoa is determined from a near boresite condition.  In this condition, delta d is very small making delta phi small too.  Considering a single point scatterer (as in figure 11, moving from the left to the right), the zero crossings between the waveform become very separated and reducing ones ability to determine the delta phi or aoa.  The attached plot is a measurement of such a condition.  Each waveform is from an antenna close to one another as depicted in figure 11.  As the scatterer is positioned over the two antennas (delta d=0), the determination of delta phi becomes very difficult.

Please advise.

Thanks.

Al

HI Al,

  There is more information on angle estimation at  https://training.ti.com/mmwave-training-series  . This might answer some of your questions.  [ Look at module 5 of  the training video "ntroduction to mmwave sensing : FMCW radars"].  

Sandeep

Hi Sandeep,

Thank you for your response.  I have looked at the mmsensor material, performed simulations and measurements.

Although my true application can be accomplished by determining aoa, basic range determination may suffice.  However, after my investigation,

I am slightly confused by the presentation in spyy005 for aoa determination.  One can readily relate aoa to the antenna spacing and delta d using

sin(theta) =delta d/l

where l is the antenna spacing and delta d is the difference of the path difference for the two antennas.  These paths can be determined

by ftting the beat signal from each receiver.  However in spyy005 (attached spyy005_p7.gif) states "in a phase change in the peak of the range-FFT". 

I think this is misleading if I have not missed something.  It is the difference between the transmitted and received signals which when mixed generates a beat

signal.  It is this beat  signal when transformed indicates what the beat frequency is and hence the distance traveled.  Delta d is obtained when both beat frequencies

(distances) are obtained and subtracted.

In spyy005, the traditional phase difference concept is presented to determine aoa.  I see no place for this material here.  This material applies to a fixed frequency

signal which is not the case here concerning frequency chirping.   I hope that it is not implied that the difference in phase for the two beat frequencies (determined from the fft)

is used in equation 15 to determine delta d.  If so, this is wrong.

In another document, spyy006 (attached at spy006_p4.gif), reference is made to using high resolution techniques such as music and espirit for aoa estimation.  I am familiar with

these techniques.  Can you tell me how they have been applied for aoa estimation?

What are the differences between the automotive and industrial chips AWR1443 : IWR1443 and the AWR1642 : IWR1642?

Can I get pricing and availability for these chips?  Is there any guidance on the design of the micropatch antennas used on you development modules?

Thank you.

Al

Hi Sandeep,

I forgot to attach the images I referred to.

Al

Hi Al,

Changes in distance cause changes in both the FREQUENCY AND PHASE of the beat signal. However for VERY SMALL changes in distances (such as the delta d in our context), the corresponding change in the beat frequency can be very hard to measure. In these scenarios a more accurate estimate can be got by comparing the phase of the beat signal [i.e. comparing the phase of the FFT peak corresponding to the signals from the two antennas]. So the delta_phi in eq(15) does refer to the phase difference between  the FFT peaks (range-FFT if the objects are stationary, but more generally Doppler-FFT peaks).

For a discussion on phase :

a) See for e.g. slide 26 of  https://training.ti.com/sites/default/files/docs/mmwaveSensing-FMCW-offlineviewing_1.pdf . 

b) You can also look at module 2 of the mmwave training material which focuses on phase

Because of the way small changes in distance are reflected in the peak of the FFT's, it turns out that the math of angle estimation is similar to the traditional phase difference concept that is discussed in fixed frequency contexts. [You can think of the mixer in the radar as undoing the 'frequency chirping' and after that the math is analogous]

Sandeep

Hi Sandeep,

Thank you for your reply.  I derived the IF/doppler expression and recovered close to what you have.  I have attached the derivation as doppler_derivation.pdf.  The difference is in the constant phase term.  You have a delta R variable and I have an R variable.  The doppler frequency term is the same.  Please see equation 7.  I believe the delta R variable is in error.  Please correct me if I am wrong, if you wish.  I do not know how one would have a "delta R" component in this case.

I then did a comparison between a measurement and equation 7.  I used a downhill simplex method to recover the leading amplitude coefficient, the bandwidth factor, the R factor and the heuristically added constant phase term for any mixer impact.  I have attached the comparisons for your review.  The comparison is excellent.  The red curve is the measured data and the green curve is the calculated result using the curve fitted results.  The fft recovered factors for the beat frequency and R were similar to the curve fitted result.  If I did not add the constant phase term for the mixer impact, I would not get a very good comparison.

I have ordered a development kit from d3 engineering.

Aldoppler_derivation.pdfcomp.pdfcomp_begin.pdfcomp_end.pdf