RTOS/AWR1243: High DC component problem in IF Response (ES2.0 cascaded system)

Hello, Any help or suggestion from AWR team will be helpful.

+ additional comments

After tests, I confirmed that the system can measure the object's range, doppler, angle in ourdoor environment, even if this happens.

However, the sensitivity of receiver is not good so only corner reflectors can be detected with low SNR...

Hi,

Our cascade expert should be able to review your questions early next week

thank you

Cesar

Hi,

I checked if this is solved with increased ADC start time.

The magnitude of DC component is slightly decreased when the ADC start time is set to 10us, but no big difference.

Other thing I tried is to take average between ADC samples in one chirp.

This also shows slight change but no big difference.

I want some feedbacks from cascade system experts about this issue as form of

1) possible solution (what we can try)

2) is it only limited on ES2.0

3) or nobody experienced/reported this phenomenon

On the other hand, I found out two signifant issues related cascaded system during tests, and I will post a new thread about new issues.

(doubled frame period / captured data chip order is changed randomly with parameter change)

Best regards,

Hi Kim,

DC offset from ADC IQ imbalance is another possible cause of this problem. If you have the TX disabled, then the TX-RX coupling or a short-range "bumper" reflection is probably not the issue. Also, for short-range coupling like TX-RX or bumper, depending on the chirp resolution, you will usually see this peak somewhere in the first few range bins, but not usually the DC bin itself. 

If you look at the ADC IQ samples, is there a DC offset to the I and Q channels? At least for post-processing, I would recommend estimating the DC with an average of the data over each chirp and subtracting out the average. 

As for eliminating the offset to begin with, can you please check if the "Enable RX ADC DC offset calibration" option has been set in the AWR_RF_INIT_CALIBRATION_CONF_SB API call prior to RF initialization? 

For example, using the TI Cascade RF EVM (AWR1243P ES3.0) I have you can see the difference between enabling and disabling RX ADC DC offset cal.

Below is a single chirp IQ FFT response without RX ADC DC offset cal running. I have all of the TX disabled as well, so TX-RX coupling should not be an issue here. Log scale is dBFS. I see a large DC offset between the IQ samples and a large peak in the zero bin. 

Same scene, single chirp IQ samples and FFT response after running RX ADC offset cal. 

Thank you,

-Randy

Hi Kim, 

And just in case you have not seen the calibration description app note yet: 

http://www.ti.com/lit/an/spracf4/spracf4.pdf

Thanks,

-Randy

Dear Randy,

Thank you for your kind comment.

I will check whether ADC DC offset calibration is turned on/off, try same thing with you and share the result soon.

Best regards,

Hi Randy,

I run some tests based on your recommendation, and found out these result.

However, due to my company restriction in uploading things, I can upload the figures tomorrow.

1. I found out that the configuration didn't enabled RF INIT CALIBRATION.

2. After enabling calibration with calEnMask 0x17F0, the power of DC component is decreased significantly in FFT result.

3. The I/Q data plot shows similar result with your figures. (both before/after calibration)

4. However, I have still problem and can't get the same result with the figure that I mentioned. 

(https://e2e.ti.com/support/sensors/f/1023/p/663579/2446259)

I'm doing some tests and I think its related to HPF setting and IQMM calibration. I will also share this result with you tomorrow.

One more question. Does IQMM calibration still not recommended to ES2.0 or do we have any other ways or updates?

Hi Randy,

Below figures are the result after enabling boot-up calibration.

1) calEnMask 0x17F0

2) calEnMask = 0x7F0 *(IQMM calibration is set to 0)

HPFs are set to 175/350kHz, and the sampling rate is set to 18750 with complex1x mode.

I see no big difference between IQMM calibration, but the DC components are still remain.

After changing HPFs, as you can see below, still there is DC component even if the filter works in low IF region..

What do you think of this phenomenon?

Best regards,

Hi Kim, 

Glad to see the cal cleaned things up to some degree. Are you seeing any difference between the master, slave or single-device operation after calibration is applied?

Thank you,

-Randy 

Hi Randy,

Unfortunately, I can't see differences between the master/slave chips.

Since I'm using only radar sdk, I can't verify single-device operation.

Thank you for your support.

Best regards,

Hi Kim, 

With the radar SDK you should be able to test out single-device operation. This is setup with the AWR_CHAN_CONF_SET_SB_ICD API command by specifying the CASCADING_CFG field. 

CASCADING_CFG = 0x0000 -> single device mode

CASCADING_CFG = 0x0001 -> cascade master device mode

CASCADING_CFG = 0x0002 -> cascade slave device mode 

Can you set CASCADING_CFG = 0x0000 and then verify the DC seen?

Thanks,

-Randy

Dear Randy,

First of all, I think there are some misunderstanding between us regarding the radar SDK.

I'm so sorry for confusing expression. What I tried to say was I'm using "Processor RTOS RADAR SDK"

Until now, I built an appImage with some changed parameters in the massive codes, boot the system and get some raw data by "capture usecase"

I didn't try operating the cascade system by issuing API commands step by step.

Therefore, I think I need some time to check this because I need to setup some "API based environment" for our h/w setup.

Please wait a little.

Thanks,

Hello Randy,

I checked the IF response in single device mode.

As you mentioned, there are differences in the IF response between cascading configuration..

below figure is when I configured the master chip as single device mode, on the same cascade board.

and I think this IF response is very ideal result.

Do you think our customized cascaded board has problem?

Best regards,