mmwave hazard query

karthik s60

Other Parts Discussed in Thread: AWR1443, AWR1443BOOST, AWR1243BOOST, AWR1243, IWR1443, IWR1642, AWR1642

Hi,

1. If we want to use mmWave sensor for medical purpose, will there be any hazard for human beings?

2. What is the safety limits in which we have to use?.. like the min distance between human body sensor? power we can use?

regards

Karthik S

over 7 years ago

0 Radio-Joe over 7 years ago

TI__Mastermind 37500 points

Hello,
The Tx output power is 12dbm per Tx antenna. The gain directivity of the Tx antenna, would be used to develop the mmwave output power.
In most cases, the Tx power envelope with the 76-81Ghz or future 57-64Ghz the normal antenna gain < 20db, results in < 32dbm output power.
You would have to look at the proximity to the patient for different frequency ranges and output power.
Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

We are thinking of using isotropic antenna which will have radiation only about 5-10 cm of its vicinity. So we will try to keep antenna gain less, so that output power is less.

The sensor will be at 5 cm from the patient, with min detectable power after reflection.

--'You would have to look at the proximity to the patient for different frequency ranges and output power'...Where can I find these details?

regards

Karthik S

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello
Each country or area can have a radiated RF emission standard.

The EVM User Guide lists.
The AWR1443 and AWR1243 evaluation modules (AWR1443BOOST and AWR1243BOOST) are in
compliance with Directive 2014/53/EU. The full text of TI's EU Declaration of Conformity is available here

The European RF exposure radiation limit is fulfilled if a minimum distance of 5 cm between the users
body and the radio transmitter is respected.

E: The EUT has been tested in the 76 – 77 GHz band (2 Tx at a time) at a maximum peak
power of 26 dBm EIRP, and in the 77 – 81 GHz band (1 Tx at a time) with maximum peak
power of 21 dBm EIRP across the temperature range of –20ºC to 60ºC.

Another resource - www.who.int/.../

Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Thanks Joe,

I will go through the link.

One query: Is it possible to control the energy of signal emitted by mmWave sensor?

regards

Karthik s

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello
The Tx outputs have a Tx power backoff control, to reduce the Transmit power for the entire chirp.
The range is from 12dbm to 0dbm in 1db steps.
You can see this in the mmwave SDK User Guide description of the profileCfg, and also swra553 Application note.
Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

Thanks for info.

Just need one more clarification- what is the min range resolution of AWR/IWR? some T2E discussion say 3.75 cm.Is this the min distance that can be detected by RADAR or it is the ability of sensor to detect 2 objects which are sep by 3.75 cm?

1.I want to detect a range of an object range, starting from 50 cm in steps of 1 cm till the distance becomes 2 cm. it is possible with AWR/IWR right?

2. If I want to build any system using mmWave AWR/IWR what are essential things I need like mmWave EVM board-299 USD will include all SW part required for configuration? do we need to buy it separately?

3. If I want to measure the distance of human face, Can this sensor able to give distance from nose and that of other parts of face. Ideally I want to get the 2 distance of sensor nose distance and sensor chin distance? If we have 3.75cm resolution with 4 GHz between nose and chin.?

Regards

Karthik s

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
1) The IWR1443, IWR1642 have a 4Ghz bandwidth. In October the 60Ghz device with special conditions can have > 4Ghz BW. If you want a lower range resolution value, you need a higher RF bandwidth.

In the TI, www.ti.com/.../tools-software.html, software tools you can look at the labs and TIDEP TI designs for examples.

The Out of Box demo, and High Accuracy lab, both have processing tools. The IWr1443 Best Range Resolution is a place to start.

I would start with the IWR1443 (no DSP, higher IF) or IWR1642 (with DSP, lower IF), depending on your custom processing needs. The IWR1443 with the Out of Box demo, has a best Range Resolution example, I would use that for your first concept test.

You can follow swra553, for understanding the chirp parameters. Saving the OOB configuration to a file, and doing modifications, and reloading the configuration file allows you to tailor the chirp parameters.

Another approach is to use the mmwave Studio, and a separate DCA1000 to capture an process the raw Radar Rx data from the 4 antennas. You would normally use Tx MIMO Tx1, Tx3, with 4Rx to get the best azimuth resolution. mmwave Studio has additional capability to provide signal processing, detection, and classification.

2. The EVM, and the Out of Box demonstration software are the initial items.
If you want to do further Radar Rx processing the DCA1000 is needed, with mmwave Studio.

3. the detected object is the portion of reflected radar energy. If you are trying a facial recognition, you can't resolve separate objects based on range resolution, there are other velocity and phase components. You might find that aspiration distorts the mouth and nose view, to allow them to be separately detected. The point cloud visualizer and experiments may help.

Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

Thanks for the elaborated information.

Could you pls explain last part, what is the extent that I can separate the chin and nose distance using phase signal, if human face is stationary they I feel we will not have any phase change in the signal.

with 60 GHz higher RF band is it possible to satisfy my requirement of separating nose and chin distance?

Basically tell me what I am suppose to do, if my radar want to detect nose and chin as separate object and should give distance separately?

Referring to fig above Is it possible to find distance R1 and R2 separately. take the width of projection is 2 cm. and flat portion extends to 6 cm comparable to nose and chin.

I was going through tuto doc by Mr.Sandeep rao from TI and it was mentioned that, if 2 objects sep by 3.75cm and are at different distance wrt sensor the phase of reflected chirp would be different, if same principle is used for nose and chin, nose is projected front than chin so the distance between sensor and nose will produce reflected chirp with f1 and phase1 where as chin gives f2 nearly equal to f1 and phase2. With phase info can we process the signal to get range information?

"'. The EVM, and the Out of Box demonstration software are the initial items.
If you want to do further Radar Rx processing the DCA1000 is needed, with mmwave Studio"-With only EVM and out of box demo SW can we process phase information?

what extra processing can be done with DCA100,?

regards

Karthik s

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
1) in some of the Industrial toolbox labs, there is a vital signs demo. This demo uses doppler processing to detect respiration and heart beat rate. Given the person is breathing, using the vital signs demonstration, and then customizing it, you may be able to detect aspiration as changes in the nose.

2) the higher chirp bandwidth can result in a finer ranger resolution.

3) you have distance, velocity, angle - if the objects are stationary you need better range resolution. You can experiment with Vital Signs demo, to see if there is a nose related micro doppler.

Related to range, velocity, and phase - there is a detection process where objects > detection conditions are active objects. You may need a method to detect the nose, lips, chin as separate objects, a point cloud can be tracked , the center of a point cloud can be subtracted from the center of another point cloud, and a distance can be computed. Object detection can be done on distance, velocity, and phase information.
The Vital Signs Lab uses microdoppler to detect movement .

You can use the point cloud for tracking, and to compute point cloud differences.

The Industrial Toolbox - "www.ti.com/.../tools-software.html" has access to TI DEP (designs) and labs that have examples for range, velocity, and phase,. The Out of Box demo discusses the TDM MIMO and BPM MIMO that provide for 15 deg angular resolution.

Once you can identify signals separately to get a Point Cloud, this detected object to another detected object can provide a distance.

The DCA1000 is used to capture or forward over Ethernet the LVDS data that is output from EVM. The initial configuration forwards the mmwave Rx DFE output or stores it.

Regards,
Joe QUintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

1."Related to range, velocity, and phase - there is a detection process where objects > detection conditions are active objects. You may need a method to detect the nose, lips, chin as separate objects, a point cloud can be tracked , the center of a point cloud can be subtracted from the center of another point cloud, and a distance can be computed. Object detection can be done on distance, velocity, and phase information"-

What I am thinking is to, process signal reflected by nose and chin if possible to detect, since there will be projection of nose the reflected wave will have one phase and chin will have one phase by keeping dmin as 4 cm between nose and chin. And try to map this phase change with extra distance travelled.

2. Could you pls explain the concept of point cloud

3. I need a kind of application, where I want to get range value(not on plot) in number like 4 cm. I want to compare and if its less than some threshold I want to generate a high signal form any I/O pin- Can it be done with only mmWave IC which has DSP and ARM? If yes How?

4."The DCA1000 is used to capture or forward over Ethernet the LVDS data that is output from EVM. The initial configuration forwards the mmwave Rx DFE output or stores it"---could you pls elaborate this?-

I want to get Range FFT as shown in the above figure.

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello

1) If the person is breathing deflections in the nose area or cheeks could be detected for doppler as well as position.

The IWr1642 having an internal DSP, and the ARM control processor, can perform more post processing than the Hardware Accelerator and the ARM control processor. is the Industrial Control toolbox with experiments and labs for testing out some of your ideas. "dev.ti.com/.../

2) The point cloud is data about the objects detected by the mmwave sensor processing, see page 5 of "www.ti.com/.../spry311.pdf"

- range, velocity, angle about a detected object

3) the IWr1642 with the DSP can be used to develop algorithms after detection. There are both digital communications Control UART, MSS Logger UART (used in Visualizer over USB), I2C, and SPI, and GPIO that can be used for output.

4) If you put the mmwave sensor EVM on a tripod, and adjust if related to the target face, using the Best Range Resolution, you should be able to adjust the parameters (following swra553 - different chirp items) to get short range range and velocity information.

I have used the 4fps, side and front view. You can try this with other experiments.

front view (mmwave selfie)

side view

example config file

sidefront_1443.cfg

Regards,

Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi,

Thanks Mr. Joe for explanation with experiment.

The last point is it like, face is considered as object in front of mmWave sensor?

what does side and front view means? in point 4 explanation?

if I need onboared processing with IWR1642, but I have to drive this with 4 antennas for customized application, how should I go forward? simultaneously I want to measure range from all 4 antennas, I mean using both TX s in IC.

How exactly the signals received in 4 receivers is differentiated when signal is processing?

regards

Karthik

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello

Front view, in this case the radar antenna is inline with the face, ie the nose is closer to the radar antenna.

Side view, in the case the nose is further away than the cheek on the face,

Regards,

Joe Quintal

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
There are examples in the swra544a app note for TDM and BPM MIMO. We get better azimuth angular resolution, this is the case on the Visualizer where we have 4Rx, 2Tx, and 15 deg azimuth resolution. In the mmwave SDK, there is a profile_2d, TDM and profile_2d_bpm that can be loaded into the visualizer for testing.
Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

Thank you so much for the experiment now I have some idea on my concept with mmWave. Just to clear some more

1. Is it possible to change the scale of plot, In above fig range plot is in meters, I want to use it in cm steps. So that I can see good differentiation between nose and cheeks.

2.Could u pls try the same exp, with step changing to cm, I don't see the range value of nose in current range power plot, only I can see 4 points which are after 0.5m.

3. Any examples available for using GPIO pin on mmwave?

4.In my case I only want to get the range information, so can I use only one Tx and One RX to get5 the first and second plot or at least the range vs power (db) plot?

5. Could u pls show what and can be done with single TX Rx combination with same setup as above?

Please don't mind, Just need clarity so that I will be sure for proposal of solution with your mmWave sensor.

Regards

Karthik s

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
1) Eventually the Radar chirp Frequency will establish the range resolution. In the mmwave Industrial Toolbox, there are Labs and Experiments that discuss zoom FFT concept for measuring < range resolution limit.

2) A point cloud - is a set of parameters defining a detected point - " en.wikipedia.org/.../Point_cloud "

3) your picture seems to be a front view, if you are trying to detect the image of a nose and chin for biometric purposes, you would store a normalized point cloud. You would normalize position, and have to have some method so the angle from the radar antenna to the sensed person is nearly the same. You could then compare a normalized repositioned average point cloud to a stored parameter set, and set a GPIO for red(not OK)/green(OK). In this concept this has an analogy to an optical scanner, looking for a matching pattern.

4) The mmwave capture can collect the Radar Rx, this is external processing in the TSW1400 + Devpack, or DCA1000. If your application fits within the IWr1642, there is an onboard DSP that can perform the detection, point cloud extraction. Again look at the mmwave Industrial Toolbox for Lab and Experiment examples.

Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

Sorry for late reply.

"3) your picture seems to be a front view, if you are trying to detect the image of a nose and chin for biometric purposes, you would store a normalized point cloud. You would normalize position, and have to have some method so the angle from the radar antenna to the sensed person is nearly the same. You could then compare a normalized repositioned average point cloud to a stored parameter set, and set a GPIO for red(not OK)/green(OK). In this concept this has an analogy to an optical scanner, looking for a matching pattern"

---------I want to find the distance from tip of nose to antenna and chin to antenna as 2 separate range values, and if it is less than some threshold I want to make one GPIO pin high else it is kept at low.This is my application.

Can point cloud help me for this? How?

2. In the mmWave selfee image shown above the marked points indicate nose and chin? if yes then I can see example to have range plot in terms of cm values so that I will get the zoomed value.

regards

Karthik S

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
I would setup the IWr1443 for Best Range Resolution, save the config file to the PC file system, and check by recording the output from the visualizer, and then post processing. OR there is a data capture example on the mmwave_SDK drivers, I would build that and capture the Radar Rx from CCS. You can then post process this data.

If instead you record from the visualizer, you could save ~100 frames to the PC, and do some post processing.

Under the TIREX Industrial mmwave toolbox. There are some other demos you can try. If you utilize the mmwave SDK demos, there is a radar rx capture, you would use Tx1, Tx3 in a MIMO configuration (either TDM or BPM) to get better angular resolution.

Because of the HPF1, HPF2 response there is a null in front of the radar within a few cms. You could collect data from the data capture example, and post process to zoom into the object, however the ZOOM FFT provides a single point, you would need to calculate the zoomed FFT over more samples.

Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

We have thought of using IWR 1642 due to on board DSP and ARM.

we will try to set only best range resolution which is max of 4 cm by adjusting chirp signal. And put it in front of face and get the reflected signal, taking the FFT and zooming it with more samples and if we could able to get the nose point and the part of chin which is 4 cm apart from nose and try to differentiate(Using the visualizer and DSP processing).

Is it possible to do in this way?

Can I map your first sentence of last post with my explanation?

Could you pls explain same thing wrt IWR 1642, the procedure remains same or different?

"Because of the HPF1, HPF2 response there is a null in front of the radar within a few cms. You could collect data from the data capture example, and post process to zoom into the object, however the ZOOM FFT provides a single point, you would need to calculate the zoomed FFT over more samples."- what does this mean?

Can these fine detection tried with AWR1642 as well? ... since we are having little difficulty in getting the IWR so

I got to know today you also have sepeicifc ODS version of AWR with wide FOV.

Can industrial applications like my application be implemented with ODS , since we want to onlty detect range. What is the difference between AWR1642 BOOST, IWR1642 BOOST, AWR1642 ODS interms of performance.

regards

Karthik

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello Karthik,

The HPF1 and HPF2 portion of the mmwave sensor are the same.

The main mmwave sensor difference is that you have 2 Tx, and 4 Rx (1642) and 3Tx and 4Rx (1443). The extra Tx is used for elevation.

There is a MIMO application note swra544, that discusses the TDM and BPM MIMO to get better azimuth angular resolution.

The DSP can be used for other functions in the 1642 applications.

The IWR1443 has a higher DFE output rate, 18.75e6 complex 1x, the IWR1642 has a 6.25e6 complex 1x rate.

The BOOST EVM 1443,1642 have a medium range ~50 meter usage, in the EVM User Guide, the antenna pattern is displayed.

In the ODS EVM the antenna is setup for both azimuth and elevation with equal coverage.

You could try the ODS EVM also, to see which antenna pattern works best for your application.

In your example a dlelectric lens may help to focus the mmwave FMCW radar, " https://www.thorlabs.com/NewGroupPage9.cfm?ObjectGroup_ID=1627 "

Regards,

Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

I wanted to understand on how to do the post processing, after recording the 100 frames from the, visualizer.

Please help me to do as follows: As discussed I will set 1642 for best azimuth resolution, and let node and chin tip will have reflected waves which will be captured from visualizer. Where do I have those vales? How do i take it to CCS for processing.

I want to get the range value of nearest object and want to compare with threshold and if it is less then it, GPIO pin should be made high. Please help me to achieve this...How do I go about it , I did not get any doc related to GPIO ports enabling/disabling.

Could u pls send link for zoom FFT lab? I searched in your site but could not get it.

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
The data from the Visualizer is dependent on what you select on the configuration displays at the bottom left (Plot Selection). The Visualizer values are the detected values, and the range FFT magnitude. Scatter Plot, and Range Plot selections, you should be able to see point cloud data for scatter plot. You should have the 1D FFT results for the range plot.

There is a different mmwave SDK where we can collect the L3 Radar Rx data - data capture

Using the mmwave Studio / Radar Studio, and capturing the LVDS data from the BOOST EVM (additional hardware needed Devpack+TSW1400 or DCA1000) you can use additional post processing software tools.

Related to industrial mmwave sensors there is a TIREX industrial mmwave toolbox - " www.ti.com/.../tools-software.html " bottom right button

Regards,
Joe Quintal

0 Radio-Joe over 7 years ago in reply to Radio-Joe

TI__Mastermind 37500 points

zoom FFT is the High Accuracy Lab in the industrial mmwave toolbox, you need the proper version for 14xx or 16xx device.
Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

What and all can be done with only IWR 1642 dsp processor ? Can I do those point cloud processing what you told earlier witout DCA1000?

If in case I am getting the plot of range, where do I have the value for comparison?

Let us say I have a signal reflected from face and I will get the range or angular plot as shown in your expt images. Where will I have the range value which will be displayed on plot?

I would like to process the data in, inbuilt DSP itself to get zoomed FFT and range value which are of 4cm resolution, I should be able to get 2 range FFT peaks which are 4 cm apart when I set the mmwave to max resolution of chirp config. What I am telling can it be acheived without DCA1000? I mean only with IWR 1642 DSP and ARM?

Plsase explain different points you have got in range and angluar plot of mmwave salfee? i see 2 very nearby objects is it 2 points on face or 2 different objects( Image is attached in last 2 post)

Regards

Karthik S

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
The Out of Box demo software source code, is provided for both ARM+HWA only 1443, and ARM+DSP 1642 the mmwave SDK. If you want point cloud, you have to detect and aggregate object features. If you want to scan a surface with better than min Range Resolution, you need phase processing. There are training sections under TIREX mmwave sensors, that discuss these topics.
"training.ti.com/mmwave-sensors-ti-resource-explorer

Zoom FFT you would most likely use the IWR1642 version.

In your picture , face front view, you would need to reduce the Field of View with a lens. The min Range Resolution is based on the Chirp Bandwidth. When you have smaller than the Range Resolution, this has to use phase processing.

I suggest you start with the Gesturing for Phase processing, and AWR Occupancy detection for point cloud.

Regards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

I agree with your point of Phase processing.

I wil just tell my req in a line: In mmWave selfie, I want to measure a distance from sensor to nose tip. what I have to do in terms of processing.

If I use point cloud and subtract the centres, I will get distance between nose and chin I need accurate distance between nose tip and sensor.

I went through vital signal lab but could not identify the part of phase processing.

Since FFT of signal reflected from phase would look like single peak as far as I understood and I am interested in the signal reflected from nose to sensor since I need shortest distance.If it is leg and from the tip of the finger. Please help me in this.

I went thorough the 2 labs you mentiond. But could you pls help me which .C or .H file I have to look into for phase information?

And which file should be need to be considered for modification in case of customization?

It woulf be good if you could show which part of code I need to see, for comparison of range value for threshold and activating GPIO..when I studied doxygen doc all are explained wrt memory. So how exactly will I compare the values?

regards

Karthik

Regards'Karthik

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello
Please repost your current question, It is not seen in the E2E link I have
REgards,
Joe Quintal

0 karthik s60 over 7 years ago in reply to Radio-Joe

Expert 1240 points

Hi Joe,

I agree with your point of Phase processing.

I wil just tell my req in a line: In mmWave selfie, I want to measure a distance from sensor to nose tip. what I have to do in terms of processing.

If I use point cloud and subtract the centres, I will get distance between nose and chin I need accurate distance between nose tip and sensor.

I went through vital signal lab but could not identify the part of phase processing.

Since FFT of signal reflected from phase would look like single peak as far as I understood and I am interested in the signal reflected from nose to sensor since I need shortest distance.If it is leg and from the tip of the finger. Please help me in this.

I went thorough the 2 labs you mentiond. But could you pls help me which .C or .H file I have to look into for phase information?

And which file should be need to be considered for modification in case of customization?

where do I have the range value for comparison?

It woulf be good if you could show which part of code I need to see, for comparison of range value for threshold and activating GPIO..when I studied doxygen doc all are explained wrt memory. So how exactly will I compare the values?

regards

Karthik

0 Radio-Joe over 7 years ago in reply to karthik s60

TI__Mastermind 37500 points

Hello,
There is no one mmwave software that includes phase processing , or object tracking on small features.

You can download the people counting demos, and study the tracking with the IWR1642. After you have studied and worked with this,
The profile command will need to be modified to mazimize to 5Mhz Ito get the highest range resolution, at a close distance.

Specific signal processing for your application following the concept of point cloud, detecting objects and constructing a surface profile, should be asked with a new thread, so the proper application engineer for detection answers your questions,,

REgards,
Joe QUintal

Sensors

Sensors forum

mmwave hazard query