AWR1642BOOST: People Counting Demo Don't Run

Hello Amanda,

Thanks for your reply.

1. Yes, I made sure the COM ports are correct. In fact, when I opened the visualizer, it defaults to the correct ports, and there is no way for me to change it from the pull down menu.

2. I don't have CCS installed, so I tried to run the lab0005-high-accuracy-16xx from SDK version 2.3.0, and it does not work also. The status says that it is waiting for data.

I then tried to reflash the EVM with the same demo with SDK version 2.4, and the status changed to Hardware Connected, but no object can be detected either.

Does this indicate that my board is defective? I just received my board a few days ago, and if it is the case, how can I get a replacement board?

Thanks for your help,

Yohannes

Hello again,

Just wanted to give an update on what I have tried this morning, and still no luck. In order for me to run the OOB demo, I downloaded and installed mmwave_sdk_01_02_00_05, and flashed the EVM with the xwr16xx_mmw_demo.bin found under the .../packages/ti/demo/xwr16xx/mmw directory. Flash procedure was successful as seen in the screen shot.

I then removed POS2 jumper, recycled the power, and hit the reset button until the yellow LED turn off and back on.

I launched the visualizer, and the status indicates "waiting for data". Note that the COM ports match with what shown on the device driver. It seems that I have received a defective EVM with my purchase. How do I process the RMA (I purchased this board from Mouser Electronics).

Thanks,

Yohannes

Amanda,

Thanks for your suggestion. I believe you're right that my board is an ES2.0 based on the marking on the chip. I followed your suggestion to run the OOB demo with the 2_0_0_4 SDK version, and at least I am getting some data back to the PC. I did get an LED error lit up consistently when I run the demo (see picture). I am not sure what this error means though, you have any clue?

I will try to run the people counting demo again, assuming that my board is ES2.0. I will use industrial toolbox version 2.4 instead and will let you know. I thought I have tried this before without success, but will try again to make sure.

Thanks,

Yohannes

Hello again,

The out of box demo works with SDK2.0.0.4, with the exception that the NERR_OUT LED was turned on. The visualizer was also indicating that it detects objects, although I did not spend too much time testing it.

I downloaded mmwave_industrial_toolbox_2_4_0, flashed the ..../lab0011_pplcount_quickstart/xwr16xx_pcount_lab.bin, removed POS2 jumper, and recycled power, but ran into the same problem. The fact that I can get data back with the OOB demo, but not with people count demo points to there are certain settings that is not correct with this demo. Any suggestion I can try?

Thanks,

Yohannes

Thanks Amanda. Looking forward for you response.

-Yohannes

Hi everyone! I also have problems with running the example of Peple Counting Example on AWR1642. I recorded a videos to describe my problems.

1)The port gets locked at the moment the firmware starts downloading

2) Method to unlock the Port.

3) Prebuilt binary xwr16xx_pcount_lab.bin download from mmwave_industrial_toolbox 2.4.0. The program doesn’t respond.

4) People Counting Demo from mmwave_industrial_toolbox 2.4.0. works in debug mode only, but the system is not able to detect objects

www.youtube.com/watch

5) Error in People Counting Example Evaluation. Analysis of received raw data

6) Error in People Counting Example Evaluation. ADC raw data analysis

Please help me solve these problems and run an example.

Hi Yohannes.

You could not describe in more detail what you do to run an example of counting people. Thank you in advance.

Sure, I try. First thing is that you need to find out the version of AWR1642 board you have. In the user manual, it asks you to check for a sticker that says ES2.0, but in my case there is no such sticker, and my board is actually an ES2.0, The best way, as suggested by Amanda, is to look at the chip marking: Can you check the device markings on the package of the chip does the bottom left corner read "502A" (ES2.0) or just "502" (ES1.0)? 

I can only explain what I do as my board is the ES2.0. If yours is not, try to follow the direction in the manual on the SDK to install. In my case, I installed CCS and made sure all of the components are the latest. I don't even bother to match the version to what was in the document (such as compiler version, etc). Make sure you install the mmWave SDK the latest version, and Industrial Toolbox the latest version. You can then follow the direction in the People Count Lab 11 user manual for the details. I outline what the process below:

1. Import the People Count project to CCS from the .../lab0011-pplcount directory of the latest Industrial Toolbox (version 2.4 in my case)

2. Rebuild the _dds and _mss project

3. Use Uniflash to flash the generated xwr16xx_pcount_lab.bin into the EVM. Remember to flash you need to put the jumper on both the SOP2 and SOP0, and power cycle.

4. After successful flash, power the board off, remove SOP2 (only SOP0 is installed) and power cycle the board.

5. Run Matlab, go to the .../lab0011_pplcount_gui and run the script main_pplcount_viz. You will need to modify the controlSerialPort and dataSerialPort according to what is shown in the device driver

In my case, the trouble I had was that I tried to run the demo using the prebuild binary. This doesn't work. Once I rebuilt the binary, using all the latest SW, it works just fine.

-Yohannes 

Hi Yohannes.
Thank you for your detailed answer.My board is an ES2.0 since of the chip does the bottom left corner read "502AC".Therefore I decided to repeat your actions, although I must admit that I managed to launch a radar on version 2.3.1 but this is a little lower. After repeating all you described the radar starts working and the application gui considers the received packets but do not see people. I heve a question tu you , you use default settings or write your own.
Now I'll tell you how to run the example on version 2.3.1 I changed one setting parameter in profileCfg (I changed <idleTime>" from 30 us to 100 us).

Hi Konstantis,

Glad that you get the radar is working. The default setting works (i.e. it can detect people), however it wasn't optimized for my experiment. It detects too many 'ghosts', and it took too long for the 'ghosts' to disappear from the counting. I have seen also that the counting is one off, it thinks there is 1 person even though nothing is detected. I modified the SNR threshold and the state machine counters, and for the most part it works well for my experiment.

One of the most important thing is the HW setup for proper operation. The tilt angle, height, and room configuration play a major role in correctly counting the people. All other SW parameters pretty much just make the experiment smoother.

-Yohannes 

Konstantis,

When you say that the "application gui considers the received packets but do not see people." are there:

A) no detection points

or

B) there are points but the tracking isn't working correctly so you don't see the bubbles?

If A, I believe it it is likely that the incorrect compiler version is being used. Please see this e2e thread for similar issue:

If B) Please refer to  Yohannes' reply above as well as the tuning guide

Best,

Amanda

Hi Yohannes.

Thanks for your reply. Would you send to me a compiled project.Here is my mail stremitelnu@mail.ru.

Yes, that's correct. The code was optimized for 7.4.16 compiler and there is a cycle penalty with the new compiler that makes it incompatible for the People counting and traffic monitoring demo.

Best,

Amanda