Compiler/OPT9221: Noise in phase-map for second modulation frequency

Semen Semenov

Intellectual 647 points

Part Number: OPT9221

Tool/software: TI C/C++ Compiler

Hello TI!

We made our-self board based on your demo-board OPT8241.

My problem is the presence big noise in phase-map for second modulation frequency in de-aliasing mode.

For first frequency it is seem OK.

It is phase-map for first frequency: (ind_freq_data_en = true, ind_freq_data_sel = false)

It is phase-map for second frequency: (ind_freq_data_en = true, ind_freq_data_sel = true)

Plain IR image the same scene:

Phase-map for both frequencies: (ind_freq_data_en = false)

The main registers values are:

{'5C' 'w' '02' '900C42'};
{'5C' 'w' 'B2' '000AAA'};
{'5C' 'w' '03' '101800'};
{'58' 'w' '83' '000026'};

{'58' 'w' '82' '0208D5'};
{'5C' 'w' '4C' '800012'};
{'5C' 'w' '29' '304040'};

{'58' 'w' '0C' '140000'};
{'58' 'w' '0D' '0F0000'};
{'58' 'w' '0F' '00041A'};

The main parameters for depth engine and time generator

ind_freq_data_en = false
use_dealiasing_mode = true
ma = 2
mb = 3
ka = 2
kb = 1
dealiased_ph_mask = 3
phy_test_enable = false
ind_freq_data_sel = false
illum_en_early = false
illum_mod_early = false
intg_duty_cycle = 18
mod_m1 = 20
mod_m_frac1 = 0
mod_m2 = 15
mod_m_frac2 = 0
mod_n1 = 2
mod_n2 = 2
mod_ps1 = 1
mod_ps2 = 0
pix_cnt_max = 133333
sub_frame_cnt_max = 2

Is it normal such noise level for second frequency?

over 6 years ago

0 Suramya Gupta over 6 years ago

TI__Expert 3975 points

Semenov,

Could you share the raw data for both frequencies?

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for tour answer!
Are you need 1 raw phase frame for 1-st and 2-nd frequencies?

0 Suramya Gupta over 6 years ago in reply to Semen Semenov

TI__Expert 3975 points

Yes Semenov,

Please share raw phase frame for both frequencies as well as the raw phase for the combined frequency.

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for your answer!

in attachment phase.bin 1 frame raw phase data,

phase_f1.bin 1 frame raw phase data on 1-st frequency, phase_f2.bin 1 frame raw phase data on 2-nd frequency,

phase.txt

phase_f1.txt

phase_f2.txt

raw data are calibrated by nonlinarity and commonoffset (not pixelwise) calibrations.

I loaded file extension ".txt" but real extension ".bin" i.e. plain binary file data 320x240 unsigned short values.

Data are written from left to right, from top to bottom.

0 Suramya Gupta over 6 years ago in reply to Semen Semenov

TI__Expert 3975 points

Semenov,

It seems that the second frequency data has multiples of 16 as phases. This is strange. Also, I found some outlier points, which have phase higher than 40000. Maybe, I'm not parsing the data correctly.
There is something wrong with the second frequency data for sure.
Can you confirm that you're using the latest OPT9221 firmware (v0.31)?

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for your answer!

I read 5C 01 register and got 0x000117, it is seem version 0.23.
Is it possible to change firmware through I2C interface?
Unfortunately hardware engineer is absent and I do not know how to do it other way.

0 Suramya Gupta over 6 years ago in reply to Semen Semenov

TI__Expert 3975 points

I'm not sure of the connections that you have on the board, but if the OPT9221 EEPROM is connected to the host processor via I2C the update is possible. If not, I can't think of a way that would do this.

Though we recommend FW 0.30, FW 0.23 will not show this problem. Is the Amplitude map okay for the second frequency?

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for your answer!
It is seem I gave you not correct data in last time.
In attachments below correct raw data file

The last free files are named phase.txt, phase_f1.txt, phase_f2.txt respectively. (in post it is seem a bit with other names)

Format data is the same.

But when I got raw amplitude data I do this:

unsigned short ampl = 0;

//some code 

ampl = ampl << 4;

If I do as in datasheet i.e:

unsigned short ampl = 0;

//some code 

ampl = ampl << 4;
ampl = ampl >> 4;

Amplitude values are very low.

But anyway amplitude values at second frequency are bad.

P.S. dealiased_ph_mask = 2.

0 Suramya Gupta over 6 years ago in reply to Semen Semenov

TI__Expert 3975 points

Hi Semenov,

I still see some problems with the images. The right side of the images is completely noisy. Here are the images:

amp_f1:

amp_f2:

For phase, the images look similar. Maybe, I'm not parsing the files correctly. They're supposed to be unisgned short images, right? What is the endianness? I tried both, this was the format where I could see the image. I still see values higher than 4095, though.

In the first images that you shared (not the binary files), what is the distance of the person (who is sitting) from the camera? The near object (with the checkerboard) does not seem so noisy. Maybe, the distance is causing the phase to wrap around?

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for your answer!

I am sorry for my mistake.

I fixed data format.

In attachments below correct raw data file.

0 Suramya Gupta over 6 years ago in reply to Semen Semenov

TI__Expert 3975 points

Semenov,

The amplitude is very low - it looks as if amplitude_post_scale (or iq_scale) of 4 is used.
With an intg_duty_cycle of 18, this shouldn't be the case. Is the light source giving enough signal? Can you check the power from the light source?

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for your answer!

I read register data from
5C1F : 0x054321
5C52 : 0x000000

It is seem iq_scale = 0, amplitude_post_scale = 0

more all, I reset registers by command to 5C00 0x000001.

I read register data from
5401:
max power: 0xD01000
min power: 0xD010FF
And it is seem clear that light power increases/decreases.

0 Suramya Gupta over 6 years ago in reply to Semen Semenov

TI__Expert 3975 points

Semenov,

If both iq_scale and amplitude_post_scale are 0, then amplitude should not be a multiple of 16. In the images that you shared, almost all pixels have amplitudes as multiples of 16.

I meant checking the actual signal coming from the light source. You'd need a photodetector for that (something like www.eotech.com/.../et-2030---silicon-photodetector).

It does seem that the signal is very small.

Suramya

0 Semen Semenov over 6 years ago in reply to Suramya Gupta

Intellectual 647 points

Suramya, Thank you for your answer!

I perform so for better amplitude view:

unsigned short ampl = 0;

//some code

ampl = ampl << 4;

Without this conversion amplitude is difficult to see in viewer.

Ok, I will check light power.

0 Subhash Chandra Sadhu over 6 years ago in reply to Semen Semenov

TI__Expert 3800 points

Hi,

Please let us know if this problem is resolved.

Regards,
Subhash

0 Semen Semenov over 6 years ago in reply to Subhash Chandra Sadhu

Intellectual 647 points

Subhash,

we tested phase for frequencies up 10MHz to 80 MHz with step 10 MHz.
We found the noise level vs frequency increases with increasing of frequency.
After we checked our scheme and we found that the electrical part after chip OPT9221 and before illumination module distorts the signal, i.e. the signal has no form of meander.
I think it is our own problem in electrical scheme.

Thank to all for help!

Sensors

Sensors forum

Compiler/OPT9221: Noise in phase-map for second modulation frequency