Home
JAQForum Ver 24.01
Log In or Join  
Active Topics
Local Time 21:33 24 Nov 2024 Privacy Policy
Jump to

Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.

Forum Index : Microcontroller and PC projects : PicoMite Thermal Camera

     Page 1 of 4    
Author Message
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 09:57pm 21 May 2024
Copy link to clipboard 
Print this post

This thread is about a Thermal Camera based on PicoMite and a MLX90641 camera module that can be purchased around 20 dollar at Digikey or the likes.

The proto is running on VGA, but will have to be tested for a while before I release the code. The final product will run on LCD (handheld). The idea is to create a plug in for the Game*Mite expansion connector.

So this is just a heads up of what is coming, using the limited color palette of the VGA Picomite. The camera is 16x12 pixels, and this has been upscaled to 31x23 pixels.
The screen refresh rate is 0.5 seconds.



Volhout
PicomiteVGA PETSCII ROBOTS
 
Gizmo

Admin Group

Joined: 05/06/2004
Location: Australia
Posts: 5078
Posted: 11:17pm 21 May 2024
Copy link to clipboard 
Print this post

Watching with interest.

Glenn
The best time to plant a tree was twenty years ago, the second best time is right now.
JAQ
 
OA47

Guru

Joined: 11/04/2012
Location: Australia
Posts: 926
Posted: 02:03am 22 May 2024
Copy link to clipboard 
Print this post

COOL  

0A47
 
Amnesie
Guru

Joined: 30/06/2020
Location: Germany
Posts: 396
Posted: 09:51am 22 May 2024
Copy link to clipboard 
Print this post

Cool, I like the VGA version too! For stationary use this is also good. I am sure there is a lot of math involved (interpolation?).

Greetings
Daniel
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 11:17am 22 May 2024
Copy link to clipboard 
Print this post

  Amnesie said   I am sure there is a lot of math involved (interpolation?).

Daniel


The interpolation is simple code.

The complex math is in the conversion from ADC values per pixel, to temperature. Where there is dependency on emission coeficient, 3.3V voltage, ambient temperature, compensation pixel behaviour, linearization... and this all per pixel. It took me days to implement it from the Melexis MLX90641 datasheet (look at it, 54 pages mainly filled with math and examples).

Currently the camera itself is set to an update rate of 2Hz. And the current code uses 190ms to read the pixel data through I2C (100kHz), 70ms to do all the math (mainly useing MATH array commands in MMBasic), 190ms to do the up-scaling and pixel display (BOX command).

The camera can work faster, the I2C can go to 400kHz, so I might be able to get to 4Hz by squeezing every last bit out of the performance. But for now, let's first make it work correctly. My ambient temperature calculation is few degrees off. So need to search into the math if somewhere I repaced a "+" with a "-"....??

Volhout
PicomiteVGA PETSCII ROBOTS
 
karlelch

Senior Member

Joined: 30/10/2014
Location: Germany
Posts: 172
Posted: 06:22pm 22 May 2024
Copy link to clipboard 
Print this post

Cool project!
Some time ago I played with an 8x8 thermal camera and Micropython. I remember optimizing the code was tricky. I was interested in blob detection (Video) and ended up writing C code for some routines (Description, German only, sorry).

Thomas
 
Martin H.

Guru

Joined: 04/06/2022
Location: Germany
Posts: 1113
Posted: 06:52pm 22 May 2024
Copy link to clipboard 
Print this post

  karlelch said  Cool project!
Some time ago I played with an 8x8 thermal camera and Micropython. I remember optimizing the code was tricky. I was interested in blob detection (Video) and ended up writing C code for some routines (Description, German only, sorry).

Thomas

nice Project, the discription
Google translated to dutch
Edited 2024-05-23 04:57 by Martin H.
'no comment
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 11:45am 31 May 2024
Copy link to clipboard 
Print this post

This is the status quo of the PicoMite Thermal Camera.

- 0.5 sec update rate (@252MHz CPU speed)
- "s" will make a screenshot
- 2 markers (X=higest temp, O=centre pixel)
- "c" zero's the screen (*)

Code

MLX90641_dev_09.zip

The code is compatible with VGA and ILI9341 320x240 LCD. Control is through keyboard ATM.
Support for Game*Mite controls and filesystem is next on the list.

Unit in operation (VGA), my finger is in front of the camera module.



I use a Seedstudio Grove IR camera module with MLX90641 with 110 degrees opening angle. Final design will use the 60 degrees opening angle part (software compatible).

This message is short, because my long message took too long, and I was kicked out of the forum, and lost all the message.

Volhout
Edited 2024-05-31 21:56 by Volhout
PicomiteVGA PETSCII ROBOTS
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 12:19pm 31 May 2024
Copy link to clipboard 
Print this post

Zero-ing.

Zeroing the screen is not really needed when temperature differences are large.
But the small differences between individual pixels can be equalized.

This is done by holding a piece of metal (at room temperature) in front of the camer and pressing "c". I use an old heatsink, so I know it is same temperature over the whole surface.

Volhout
PicomiteVGA PETSCII ROBOTS
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 02:01pm 31 May 2024
Copy link to clipboard 
Print this post

The current code upscales a 16x12 array to 31x23 by simply calculating the pixels in between in MMBasic by linear interpolation. Since there are only 16 colors in the VGA and in the framebuffer (needed to get speed to the LCD updates mainly) this gives a blocky screen. But it is the maximum that can be done in MMBasic in a refresh rate of 2Hz. Painting 700+ BOX'es and drawing the text etc..

If you see this (using probably the MLX90640 32x24 pixel array) and a piece of C code, you can start painting individual pixels.

demo code

The video is realy nice. Using all the colors, and calculating pixel RGB values using a fast algorithm, and immediately pushing the pixel out to the screen.

Still a lot can be improved in  MMBasic code, but I may need to work on a CSUB to do get closer to this demo.

Volhout
PicomiteVGA PETSCII ROBOTS
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 06:21pm 31 May 2024
Copy link to clipboard 
Print this post

Experiment with 4x upscaling...

The calculation and plotting takes more time
2x upscaling : 190ms
4x upscaling : 730ms

So the total update rate goes to 1 second.





Volhout
PicomiteVGA PETSCII ROBOTS
 
stanleyella

Guru

Joined: 25/06/2022
Location: United Kingdom
Posts: 2120
Posted: 07:47pm 31 May 2024
Copy link to clipboard 
Print this post

I went and bought a camera that looks like a kit for £26. no instructions. can't read as usb drive. could it be hacked for data ?





 
stanleyella

Guru

Joined: 25/06/2022
Location: United Kingdom
Posts: 2120
Posted: 09:20pm 31 May 2024
Copy link to clipboard 
Print this post

it could be a premade adafruit board with lcd and battery. figured usb and scroll through saved images.


half full mug of coffee
Edited 2024-06-01 07:34 by stanleyella
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 07:39am 04 Jun 2024
Copy link to clipboard 
Print this post

All,

Yesterday played with matrix multiplication for up-scaling. It is 2-3x faster than doing the math cell by cell, but it is consuming too much memory. For 2x up-scaling it is roughly 7kbyte, but for 4x up-scaling it is 20kbyte extra in arrays that contain the up-scaling constants.

Since up-scaling was not the dominant factor(*), it is back to standard math.

Volhout

(*) dominant factor upscaling 4x is writing 2790 BOX'es each frame, and re-drawing some text (=480ms). And I currently see no alternative way. I knwo many have the same color, but finding that out, and blit fewer larger boxes, costs also time. Maybe someting can be gained by determining the median color in the array, plotting the whole screen in median color, then plotting only the boxes that are different.
Edited 2024-06-04 17:42 by Volhout
PicomiteVGA PETSCII ROBOTS
 
Amnesie
Guru

Joined: 30/06/2020
Location: Germany
Posts: 396
Posted: 08:06am 04 Jun 2024
Copy link to clipboard 
Print this post

Hello Volhout,

If the memory problem could be solved,
why not leave the upscaling as an option, maybe there are applications in which the refresh time isn't too crictical?

Greetings
Daniel
Edited 2024-06-04 18:08 by Amnesie
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 08:56am 04 Jun 2024
Copy link to clipboard 
Print this post

Hi Daniel,

That is the idea. Upscaling 2x is working fine (also with fast matrix math).
It is the upscaling 4x that causes the problem. I have improved the screen update from 480ms to 350ms for 4x. Now the cell based up-scaling math becomes a player with 260ms.
I am trying to squeeze the code, and remove everything that is not needed.

I expect that the RAM is also used as a scratch-pad during matrix multiplications. As if Peter is doing the math MATH M_MULT in1(),in2(),out() in a copy of the out() array, and when done copies that into actual destination array. That means RAM memory should hold 2 copies of the (large) out() array (22kb size).

Since the scaling using matrix math is a 2 pass action (first up-scale horizontal, then vertical) I calculated I can save another 2 kbyte RAM by doing vertical first (the conversion matrices get smaller).

Last resort: up-scaling 3x (not 4x)

A lot of tech bla-bla....

Volhout
Edited 2024-06-04 18:57 by Volhout
PicomiteVGA PETSCII ROBOTS
 
Amnesie
Guru

Joined: 30/06/2020
Location: Germany
Posts: 396
Posted: 11:03am 05 Jun 2024
Copy link to clipboard 
Print this post

Volhout,

I read in the "PicoMite V5.09.00 release candidates" topic, that you've made huge progress regarding the speed thanks to Peters new implementations / adjustmens. This is great. I am happy to built this project too and would like to ask what thermal camera module I should buy for it. I will only use it with a VGA monitor.

There are two devices:

MLX90641 with an 16x12 array  

and

MLX90641 with an 32×24 array

other than that the Operating Temperature is different.

Also there are different view angles, I think this isn't cirical from your software point of view?

What thermal camera would you suggest, or is the code with the 32×24 array different?

Greetings
Daniel
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 12:26pm 05 Jun 2024
Copy link to clipboard 
Print this post

Hi Daniel,

The 32x24 array is the MLX90640, not the MLX90641.
I am currently using the MLX90641 (16x12 array).
It can be purchased with 2 opening angles, 55(-BCB) degrees and 110(-BCA) degrees.

I have (up to this moment) used the BCA version (110 degrees), simply because it is mounted on the SeedStudio Grove module. But I plan to build a Game*Mite adapter PCB that has the BCB unit on it.

But it depends on the application you plan to use it for. 110 degrees is wide angle.
The software is independent of the opening angle.

I saved another 1.8kbyte RAM (1 array shared between functions), and have to save another 2.2 kbyte to benefit from the BOX draw. But it is getting more and more difficult.

Volhout

P.S. Do not purchase the MLX90640 if you want to use my code. The MLX90640 has 4 times the pixels, so you do not need to scale 4x, as 2x is sufficient for the same result. But the 8 arrays that hold pixel calibration values are 4x as large also. It will present its own set of memory restrictions, and will definitely not run on my code.
Edited 2024-06-05 22:32 by Volhout
PicomiteVGA PETSCII ROBOTS
 
Amnesie
Guru

Joined: 30/06/2020
Location: Germany
Posts: 396
Posted: 01:43pm 05 Jun 2024
Copy link to clipboard 
Print this post

Thank you very much for this information! I just odered the  MLX90641 (16x12 array), like suggested.

Can't wait to try it  

Impressive work!

Geetings
Daniel
 
Volhout
Guru

Joined: 05/03/2018
Location: Netherlands
Posts: 4223
Posted: 02:03pm 05 Jun 2024
Copy link to clipboard 
Print this post

Hi Daniel,

Just for your convenience, this is the status quo. It has an update rate of 750ms in 4x up-scaling. When you plan to use it, make sure you have the I2C bus set correct. I wired the camera chip to GP0/GP1 (I2C), but you can configure for GP14/GP15 (I2C2) in line 7. IN my archive this is archived as MLX90641_dev_14.bas

 'MLX90641 IR camera support
 
 
 '-------------------- generic program defines ------------------------
 
 const SA=&h33    'I2C 7 bit slave address is &h5A
 I2Cport=1                 'I2C2=2 anything other is I2C
 scalefactor=4             '2=16x12 to 31x23
 num_colors=16              '6 or 8  'define video interface
 dim i,x,y as integer      'in stead of replacing all i with i%
 
 
 '-------------------------system initialisation ----------------------
 
 vga%=instr(mm.device$,"VGA")          '0 or not 0
 if vga% then
   mode 2
 else
   font 7
 endif
 framebuffer create     'use framebuffer when LCD faster screen
 framebuffer write F    'use framebuffer when LCD faster screen
 
 
 'I2C bus open at 400kHz
 if I2Cport=2 then
   'assume system I2C on GP14/GP15
   'setpin gp14,gp15,I2C2
   'I2C2 open 400,100
 else
   setpin gp0,gp1,I2C
   I2C open 400,100
 end if
 
 
 'arrays with values per pixel
 dim Po1(191)  'offset subbank 1
 dim Po0(191)  'offset subbank 0
 dim Pa(191)   'alpha
 dim Pkt(191)  'temperature impact
 dim Pkv(191)  'voltage impact
 dim Vir(191)  'voltage from sensor element and after conversion temperature
 dim Toffs(191):math set 2,Toffs()'Temperature offsets for manual calibration
 
 'debugging how much memory left
 'dim vzz(250) '2kb
 
 'define the colors
 if num_colors=6 then 'colors
   dim col%(5) = (rgb(blue),rgb(cerulean),rgb(red),rgb(orange),rgb(yellow),rgb(white)) 'tile colors
 else  if num_colors=8  then '8 colors
   dim col%(7) = (rgb(magenta),rgb(blue),rgb(cyan),rgb(green),rgb(yellow),rgb(orange),rgb(red),rgb(white))
 else '16 colors
   dim col%(15)=(0,&h4000,&h8000,&hff,&h40ff,&h80ff,&hff00ff,&hff40ff,&hff80ff,&hff0000,&hff4000,&hff8000,&hFFFF00,&hFF00,&hFFFF,&hFFFFFF)
 end if
 
 
 'scaling buffers and math matrices
 if scalefactor=2 then
   Dim scrn(31,22)    '16*2-1,12*2-1 = 31x23 screen pixels + 1 dummy
 else 'scalefactor = 4 (uses matrix math)
   dim Toa(15,11)                            'this is Vir() in 2D form
   dim b(61,15),c(15,44),d(11,44)            'arrays for matrix math
   fill_bd_4                                 'conversion matrices
   dim scrn(61,44)    '16*4-3 x 12*4-3 = 61x45 screen pixels + 1 dummy column
 end if
 
 memory 'debug
 
 '--------------------------- read camera constants -------------------
 
 get_res_corr
 dummy=get_vdd()
 get_cal
 get_Kta_Kv
 get_Po0_Po1
 get_Pa
 get_Vir
 get_conf
 get_stat
 
 
 '---------------------------- screen setup -----------------------------
 
 'place the legend on screen
 box 310,47,10,179,1,0,0               'bar graph frame
 v_size = 179 \ num_colors
 
 for i=0 to num_colors-1
   box 311,48+i*v_size,8,v_size,1,col%(num_colors-1-i),col%(num_colors-1-i)
 next
 
 text 0,232,"PicoMite Thermal Camera, C = cal, S = save, Q = quit",,8
 
 
 '-------------------------- MAIN LOOP -------------------------------
 'this is based on MLX90641 datasheet chapter 10.5 without STEP mode
 'the MLX90641 runs autonomous at 500ms (2Hz) default speed
 
 do
   
   'wait for new data by checking data available flag
   do
     pause 10
     status% = get_reg(&h8000)
   loop until (status% and 8) = 8
   
   
   'get actual subpage where data is available
   page% = (status% and 1)
   
   'get data (takes roughly 190ms)
   get_vir(page%)
   
   'clear data ready bit (just write the page bit back)
   set_reg(&h8000,page%)
   
   'convert IR readings to temperature according datasheet
   temp_conv
   
   
   'check if we need to re-calibrate
   if a$="c" then cal_array
   
   'calibrate the measured values to the reference
   math c_sub Vir(),Toffs(),Vir()

   'calculate extremes
   tmax=math(max Vir(),pix%):tmin=math(min Vir())
   Tlow=min(tmin,Ta-2):Thigh=max(tmax,Tlow+10)


   'show the colorfull picture on screen
   if scalefactor=2 then
     scale_up
   else
     scale_4
   end if
   
   
   'show the legend
   
   box 260,0,60,16,1,rgb(white),rgb(black)         'peak value in view
   text 270,4,"X "+str$(tmax,3,1),,,,rgb(white),rgb(black)
   
   box 260,17,60,16,1,rgb(white),rgb(black)         'centre screen value
   text 282,21,str$(Vir(88),3,1),,,,rgb(white),rgb(black)
   circle 272,24,5,,,rgb(pink)
   
   box 280,34,40,13,1,rgb(black),rgb(white)        'high value
   text 284,37,str$(Thigh,3,1),,,,rgb(black),rgb(white)
   
   box 280,226,40,14,1,rgb(black),rgb(blue)        'low value
   text 284,229,str$(Tlow,3,1),,,,rgb(white),rgb(blue)
   
   
   'the marker's
   text 20*(pix% mod 16)-2,20*(pix%\16)-3,"X",,1   'cross
   circle 162,102,10,,,rgb(pink)                   'ring
   
   'debug
   text 0,0,str$(timer,3,0):timer=0
   
   'fast copy of framebuffer to screen
   framebuffer copy f,n

   
   'do we want to save the current image ?
   if a$="s" then save image "screenshot.bmp"
   a$=inkey$
   
 loop while a$<>"q"
 
end
 
sub temp_conv
 
 local f_t(191)  'scratchpad temp
 local f_v(191)  'scratchpad voltage
 
 'process data according 11.2.2 from datasheet
 Vdd=get_vdd()                     'get Vdd 11.2.2.2
 Ta=get_ta()                       'get Ta 11.2.2.3
 
 math scale vir(),Kgain,vir()      'scale gain 11.2.2.5.1
 
 'IRdatcomp 11.2.2.5.3
 math scale pkt(),Ta-25,f_t()      'calc 1+Kta*(Ta-T25)
 math add f_t(),1,f_t()
 math scale pkv(),vdd-3.3,f_v()    'calc 1+Kv*(Vdd-3.3)
 math add f_v(),1,f_v()
 math c_mul f_v(),f_t(),f_v()      'multiply and result in f_v()
 if page% then                     'compensate for page offset
   math c_mul f_v(),Po1(),f_v()
 else
   math c_mul f_v(),Po0(),f_v()
 end if
 math c_sub Vir(),f_v(),Vir()      'and add to pixel voltage
 
 'emissivity comp 11.2.2.5.4 and IR gradient comp 11.2.2.7
 math add Vir(),-tgc*cp_pix_os,Vir()
 math scale Vir(),1/em,Vir()             'this is Vir_compensated()
 
 'Sensitivity normalize 11.2.2.8 and put values in f_t() for temp conversion
 math add Pa(),-tgc*alpha_cp,f_t()
 math scale f_t(),1+ksta*(Ta-25),f_t()   'this is alpha_comp()
 
 
 'Temp conversion (0-80C) 11.2.2.9
 Tark4=calc_tark()
 'since there are no math commands for square root and raise to the power
 'this is still in a loop, needs 55ms @ 252MHz (all above math is only 10ms)
 for i=0 to 191
   Sx=KsTo_3*( ( Vir(i)*(f_t(i)^3) + (f_t(i)^4)*Tark4 )^(1/4) )
   Vir(i)=((tark4 + Vir(i)/(f_t(i)*(1-KsTo_3*273.15)+Sx))^(1/4))-273.15
 next
 
end sub
 
 '--------------------------- display routines -------------------------------
 'these routines perform the visual representation of the data in Tob()
 
 'show 192 temperatures on screen in a grid of 31x23 interleaved
 
sub scale_up '2x
 
 local bc%,xm%
 
 'timer=0
 
 'seed the measurement data Tob()on the screen window scrn()
 'scaling math takes roughly 65ms this way
 
 for y=0 to 11
   for x=0 to 15
     scrn(2*x,2*y)=Vir(x+16*y)
   next
 next
 
 'interleave the horizontal even lines
 for y=0 to 22 step 2
   for x=1 to 29 step 2
     scrn(x,y)=(scrn(x-1,y)+scrn(x+1,y))/2
   next
 next
 
 'interleave the horizontal odd lines
 for y=1 to 21 step 2
   for x=0 to 30
     scrn(x,y)=(scrn(x,y-1)+scrn(x,y+1))/2
   next
 next
 
 'convert to colors (only picomite 5.08.00 and newer)
 math window scrn(),0,num_colors-1,scrn()   '6 or 8 colors
 
 'block size for 31x23 pixels on 320x240 screen
 size%=10
 
 '? timer,
 
 'show screen on mode 2 on picomite (takes roughly 130ms)
 'box 0,0,310,220,1,col%(com_col%),col%(com_col%)
 for x=0 to 30
   xm%=x*size%
   for y=0 to 22
     bc%=col%(scrn(x,y))
     box x*size%,y*size%,size%,size%,1,bc%,bc%
   next
 next
 
 'put some default values in the array
 math set Thigh,scrn():scrn(31,0)=Tlow
 
 '? timer
 
end sub
 
 'show 192 temperatures on screen in a grid of 61x45 interleaved
sub scale_4
 
 local bc%,xm%
 
 'timer=0
 
 'copy linear array into 16x12 array, and perform math to scale up (38ms)
 for y=0 to 11
   for x=0 to 15
     Toa(x,y)=Vir(x+16*y)
   next
 next
 math m_mult d(),Toa(),c()           'c() is correct vertical resolution
 math m_mult c(),b(),scrn()
 
 'set the pixels in the dummy column to smooth the screen colors when no
 'delta T is visible
 for i=1 to 44:scrn(61,i)=Thigh:next :scrn(61,0)=Tlow
 
 
 'convert to colors (only picomite 5.08.00 and newer)
 math window scrn(),0,num_colors-1,scrn()   '8 colors
 
 'block size for 61x45 pixels on 320x240 screen
 size%=5
 
 '? timer,
 
 'show screen on mode 2 on picomite (takes roughly 350ms)
 for x=0 to 60
   xm%=x*size%
   for y=0 to 44
     bc%=col%(scrn(x,y))
     box xm%,y*size%,size%,size%,1,bc%,bc%
   next
 next
 
 '? timer
 
end sub
 
 
sub cal_array
 math add Vir(),-Ta,Toffs()
end sub
 
 
 
 ' ------------------- get pixel cal data from the chip ----------------------
 'these functions read data from the MLX90641 and do some formatting
 
 
 'get status register
sub get_stat
 
 'get value
 Reg8000%=get_reg(&h8000)
 
end sub
 
 'get config register
sub get_conf
 
 'get value
 Reg800D%=get_reg(&h800D)
 
end sub
 
 
 'get voltages for the pixel array from subpage n (0 or 1)
sub get_vir n
 local i,j,yoff,xoff as integer
 
 xoff=n*&h20
 
 for j=0 to 5      'all 6 banks
   yoff=j*&h40
   for i=0 to 31   'of 32 pixels
     'get value
     Vir(i+j*32)=get_reg_16_2(&h400 + i + yoff + xoff)
   next
 next
 
end sub
 
 
 'get alpha(i) and adjust for row (0-5)
sub get_Pa
 local i as integer
 
 'adjust per row (32 pix) according 11.2.2.8
 for i=0 to 31
   Pa(i)=a_ref_row_1*get_reg_11(&h2500+i)/2047
 next
 for i=32 to 63
   Pa(i)=a_ref_row_2*get_reg_11(&h2500+i)/2047
 next
 for i=64 to 95
   Pa(i)=a_ref_row_3*get_reg_11(&h2500+i)/2047
 next
 for i=96 to 127
   Pa(i)=a_ref_row_4*get_reg_11(&h2500+i)/2047
 next
 for i=128 to 159
   Pa(i)=a_ref_row_5*get_reg_11(&h2500+i)/2047
 next
 for i=160 to 191
   Pa(i)=a_ref_row_6*get_reg_11(&h2500+i)/2047
 next
 
end sub
 
 'get Kta(i) and Kv(i) and scale them to be used for math
sub get_Kta_Kv
 local x,i as integer
 
 for i=0 to 191 'all 192 pixels
   x=&h25c0+i
   
   'get value
   Pkt(i)=(get_reg(x) and &h7e0)/32
   if Pkt(i)>31 then inc Pkt(i),-64
   
   'scale
   Pkt(i)=(Pkt(i)*(2^kta_scale_2) + kta_avg) / (2^kta_scale_1)
   
   'get value
   Pkv(i)=get_reg(x) and &h1f
   if Pkv(i)>15 then inc Pkv(i),-32
   
   'scale
   Pkv(i)=(Pkv(i)*(2^kv_scale_2) + kv_avg) / (2^kv_scale_1)    
 next
 
end sub
 
 'get Offset(i) for 0 and 1 compensated as in 11.2.2.5.2
sub get_Po0_Po1
 local x,i as integer
 
 for i=0 to 191 'all 192 pixels
   x=&h2440+i
   
   'get value
   Po0(i) = get_reg_11_2(x)
   
   'scale
   Po0(i) = (Po0(i) * 2^scale_os_r1) + pix_os_av  
 next
 
 for i=0 to 191 'all 192 pixels
   x=&h2680+i
   
   'get value
   Po1(i) = get_reg_11_2(x)
   
   'scale
   Po1(i) = (Po1(i) * 2^scale_os_r2) + pix_os_av    
 next
 
end sub
 
 'determine Tambient
function get_Ta()
 local dv,vbe,vptat,vptat_art as float
 
 dv = get_vdd()-3.3
 vbe = get_reg_16_2(&h580)
 vptat = get_reg_16_2(&h5a0)
 vptat_art = (vptat / (vptat*alpha_ptat + vbe)) * 2^18
 
 'formula from datasheet 11.1.2
 get_Ta = (((vptat_art/(1+kv_ptat*dv))-ptat)/kt_ptat)+25
 
end function
 
function calc_tark()
 Tr=Ta-5 'from datasheet in absence of better
 local tvk4=(Tr+273.15)^4
 calc_tark=tvk4-(tvk4-(Ta+273.15)^4)/em
end function
 
 
 
 ' ------------------- get specific data from the chip ------------------------
 'these functions read data from the MLX90641 and do some formatting
 
 
 'resolution correction from the ADC
sub get_res_corr
 RC = (get_reg(&h2433) and &h600)>>9
 RC = RC/((get_reg(&h800D) and &hC00)>>10)
 if debug then print "RC               = ";RC
end sub
 
 
 'actual Vdd measured by the chips ADC
function get_vdd() as float
 local v=get_reg_16_2(&h5aa), k=get_reg_11_2(&h2427), va=get_reg_11_2(&h2426)
 va=va*32 : k=k*32
 get_vdd = 3.3 + (RC*v - va)/k
 if debug then print "Vdd              = ";get_vdd
end function
 
 
 'this sub extracts calibration values from the chips EEPROM and formats these
 'for use in the math
 'formula's are derived from MLX90641 data sheets
sub get_cal
 local x
 local scale_row1,row1_max
 local scale_row2,row2_max
 local scale_row3,row3_max
 local scale_row4,row4_max
 local scale_row5,row5_max
 local scale_row6,row6_max
 local alpha_cp_sc, kta_cp_sc, kv_cp_sc
 
 
 x=get_reg_11(&h2410)
 scale_os_r1 = x>>5 : scale_os_r2 = x and 31
 
 x= get_reg_11_2(&h2411) * 32
 pix_os_av = get_reg_11_2(&h2412) : inc pix_os_av, x
 
 kta_avg = get_reg_11_2(&h2415)
 x=get_reg_11(&h2416)
 kta_scale_1 = x>>5 : kta_scale_2 = x and 31
 
 kv_avg = get_reg_11_2(&h2417)
 x=get_reg_11(&h2418)
 kv_scale_1 = x>>5 : kv_scale_2 = x and 31
 
 x=get_reg_11(&h2419)
 scale_row_1 = (x>>5) + 20 : scale_row_2 = (x and 31) + 20
 x=get_reg_11(&h241A)
 scale_row_3 = (x>>5) + 20 : scale_row_4 = (x and 31) + 20
 x=get_reg_11(&h241B)
 scale_row_5 = (x>>5) + 20 : scale_row_6 = (x and 31) + 20
 
 row1_max=get_reg_11(&h241C)
 row2_max=get_reg_11(&h241D)
 row3_max=get_reg_11(&h241E)
 row4_max=get_reg_11(&h241F)
 row5_max=get_reg_11(&h2420)
 row6_max=get_reg_11(&h2421)
 
 a_ref_row_1 = row1_max / (2^scale_row_1)
 a_ref_row_2 = row2_max / (2^scale_row_2)
 a_ref_row_3 = row3_max / (2^scale_row_3)
 a_ref_row_4 = row4_max / (2^scale_row_4)
 a_ref_row_5 = row5_max / (2^scale_row_5)
 a_ref_row_6 = row6_max / (2^scale_row_6)
 
 ksta = get_reg_11_2(&h2422) / 32768
 
 'emissivity
 em = get_reg_11_2(&h2423) / 512
 
 'general gain
 gain = get_reg_11(&h2424) * 32
 x = get_reg_11(&h2425) : inc gain,x
 Kgain = gain / get_reg_16_2(&h058A)
 
 'ptat
 ptat = get_reg_11(&h2428) * 32
 x = get_reg_11(&h2429) : inc ptat,x
 
 kt_ptat = get_reg_11(&h242A) / 8
 kv_ptat = get_reg_11(&h242B) / 4096
 alpha_ptat = get_reg_11(&h242C) / 128
 
 'control pixel
 alpha_cp_sc = get_reg_11(&h242E)
 alpha_cp = get_reg_11(&h242D) / (2 ^ alpha_cp_sc)
 
 x = get_reg_11(&h242F) * 32
 offset_cp = get_reg_11(&h2430) : inc offset_cp,x : map_16_2(offset_cp)
 gain_cp = get_reg_16_2(&h0588) * Kgain
 
 x = get_reg_11(&h2431)
 kta_cp_sc = x>>6
 kta_cp = x and 63 : if kta_cp > 31 then inc kta_cp,-64
 kta_cp = kta_cp / (2^kta_cp_sc)
 x = get_reg_11(&h2432)
 kv_cp_sc = x>>6
 kv_cp = x and 63 : if kta_cp > 31 then inc kta_cp,-64
 kv_cp = kv_cp / (2^kv_cp_sc)
 
 cp_pix_os = gain_cp - offset_cp*(1+kta_cp*(Ta-25))*(1+Kv_cp*(get_vdd()-3.3))
 
 'IR gradient compensation
 TGC = (get_reg(&h2433) and &h1ff) / 64
 RC_cal = (get_reg(&h2433) and &h600)>>9
 
 'Object temperature scaling
 KsTo_sc = get_reg_11(&h2434)
 KsTo_1 = get_reg_11_2(&h2435) / (2^KsTo_sc)
 KsTo_2 = get_reg_11_2(&h2436) / (2^KsTo_sc)
 KsTo_3 = get_reg_11_2(&h2437) / (2^KsTo_sc)
 KsTo_4 = get_reg_11_2(&h2438) / (2^KsTo_sc)
 KsTo_5 = get_reg_11_2(&h2439) / (2^KsTo_sc)
 KsTo_6 = get_reg_11_2(&h243B) / (2^KsTo_sc)
 KsTo_7 = get_reg_11_2(&h243D) / (2^KsTo_sc)
 KsTo_8 = get_reg_11_2(&h243F) / (2^KsTo_sc)
 CT6 = get_reg_11(&h243A)
 CT7 = get_reg_11(&h243C)
 CT8 = get_reg_11(&h243E)
 
end sub
 
 
 
 ' ------------------- conversion functions -----------------------------
 'this function makes use of the fact that changes in variable x reflect in
 'source variable
 
sub map_16_2(x)   '16 bit 2'th complement
 if x>32767 then inc x,-65536
end sub
 
 
 
 ' ------------------- read chip functions -----------------------------
 'these functions read data from the MLX90641 and do some formatting
 
 
 'return 16 bit value read from MLX90641 registers n, MSB first
function get_reg(n) as integer
 get_reg = read_reg(n)
end function
 
 
 'return an 11 bit value (remove hamming code) from MLX90641 register n
function get_reg_11(n) as integer
 get_reg_11 = read_reg(n) and &h7ff
end function
 
 
 'return a 2'th complement 11 bit value (remove hamming code) from MLX90641 register n
function get_reg_11_2(n) as integer
 get_reg_11_2 = read_reg(n) and &h7ff
 if get_reg_11_2>1023 then inc get_reg_11_2,-2048
end function
 
 
 'return a 16 bit 2'th complement value from MLX90641 register n
function get_reg_16_2(n) as integer
 get_reg_16_2 = read_reg(n)
 if get_reg_16_2>32767 then inc get_reg_16_2,-65536
end function
 
 
function read_reg(n) as integer
 local a%(1))
 if I2Cport=2 then
   i2c2 write SA, 1, 2, n\256, n and 255
   i2c2 read SA,0,2,a%()
 else
   i2c write SA, 1, 2, n\256, n and 255
   i2c read SA,0,2,a%()
 end if
 read_reg=a%(1)+256*a%(0)  
end function
 
 ' ------------------- write chip functions -----------------------------
 'this functions writes a 16 bit value v to MLX90641 register n
 
 
 'write 16 bit value to MLX90641 registers n, MSB first
sub set_reg n,v
 if I2Cport=2 then
   i2c2 write SA,0,4,n\256,n and 255,v\256,v and 255
 else
   i2c write SA,0,4,n\256,n and 255,v\256,v and 255
 end if
end sub
 
 
 '-----------------------matrices for math -----------------------------
 'this sections fills the multiplication matrics that do the up-scaling
 
sub fill_bd_4
 local i
 'matrix b(61,15) for scaling up 2x horizontal, column 61 is not programmed
 for i=0 to 15 : b(4*i,i)=1 : next
 for i=0 to 14 : b(4*i+2,i)=0.5 : b(4*i+2,i+1)=0.5 : next
 for i=0 to 14 : b(4*i+1,i)=0.75 : b(4*i+1,i+1)=0.25 : next
 for i=0 to 14 : b(4*i+3,i)=0.25 : b(4*i+3,i+1)=0.75 : next
 
 if debug then math m_print b():print
 
 'matrix d(11,44) for scaling up 2x vertical
 for i=0 to 11 : d(i,4*i)=1 : next
 for i=0 to 10 : d(i,4*i+2)=0.5 : d(i+1,4*i+2)=0.5 : next
 for i=0 to 10 : d(i,4*i+1)=0.75 : d(i+1,4*i+1)=0.25 : next
 for i=0 to 10 : d(i,4*i+3)=0.25 : d(i+1,4*i+3)=0.75 : next
 
 if debug then math m_print d()

end sub

Edited 2024-06-06 00:08 by Volhout
PicomiteVGA PETSCII ROBOTS
 
     Page 1 of 4    
Print this page
© JAQ Software 2024