thermalcam

thermalcam copied to clipboard

//this version is rewritten by james villeneuve (jamesdanielv). i am thankful to the MIT code included, and adafruit for the immense time savings they game me for providing functional code. it allowed me to work on efficiency methods. it is working and being tested on arudino ide 1.6.5. 1.8.5 works with default driver as long as spi burst is disabled. this has been fixed in https://github.com/jamesdanielv/thermalcam/tree/updates-(possibly-unstable) branch, and will be updated here after it passes all checks and latest updates are finnished.

prob best to get al files needed to run it without headache and download the thermal_cam.zip files. sometimes files seem to disapear? it could be me as well, anyways load the zip file seems to be best way to get it working. you will need to rename the extracted file folder to thermal_cam if it is not that by default

if you want output to be in deg c, then change line around line number 107 to #define temp_Fahrenheit false

adafruit has done some driver changes recently. so to test this first make sure display is working using graphics test, and make sure amg8833 is working using temp tests that show temp on terminal. make sure pinouts are set to be the same as in this sketch file. then i would run a zipped version of sketch that has all files inside of it, including drivers that i have tested. then if you want updated adafruit drivers replace them and troubleshoot.

make sure all pins that are defined are defined to display driver Adafruit_ST7735 tft = Adafruit_ST7735(TFT_CS, TFT_DC, TFT_RST); for example i tried an ili9341 adafruit driver that did not enable TFT_RTS pin Adafruit_ILI9341 tft = Adafruit_ILI9341(TFT_CS, TFT_DC,TFT_RST ); notice that 3 pins are defined. this allows for control of lcd for reset, TFT_CS for select enabled and TFT_DC, device mode)

updates will be posted first to here https://github.com/jamesdanielv/thermalcam/tree/updates-(possibly-unstable) to allow updates while keeping a stable buid.

in the future resolutions above interpolatemode of 2 will require use of included st77xx drivers.

The included adafruit_amg88xx.cpp file has a #define AMG88xx_PIXEL_MIRROR true. default is false. flips data order.

I designed a method of sub sampling that does not use floats, it works currently to 32x32 uno or equivilent and 64x64 (atmega or better) with st77xx display and arduino. spi mode,and pixel modes optimizations are more important on lower speed processors, however the more power does mean more resolution in the future, and ability to run several tasks together. the advantage will come for faster processors in the sub pixel method that does not require floats.

currently i have redesigned the sub sample system, and it will undergo further change possbily later. for now buffer needs to be about 1/4 size of final resolution, but all that is really needed is extra bytes interpolate side pixel information. for now it is easier to understand with full buffers. so eventually for example 64x64 will work on arduino with 2k of ram, even though an actual 64x64 buffer with 16 bit wide color values would normally be 8k, since output values are directly inerpolated to new values, 32x32 (2k 16 bit color) is max mem needed for 64x64 buffers. so this is why it is possible with atmega currently, even though buffers are not optimized they are 1/4 size! later on buffers will not be needed except for cache to send multiple writes to display at a time.

with a parallel display more resolution is possible, also i think arduino and st77xx display can go up to 64x64 sub sampling with little issue (not bad!) but currently as of 8_5 it does up to 32x32 (9_1_18 added:it now does up to 64x64) some of the optimizations are due to multiple rectangle writes at a time. i have posted an article on hackaday and to there blog as well. article on hackaday https://hackaday.io/project/160498-arduino-spi-up-to-30-times-more-performance https://www.thingiverse.com/thing:3017556 color table generator javascript page automates table builds https://youtu.be/tLwYMQjD0l4 video about project details //https://www.youtube.com/watch?v=PFyu2S1H0v0 info on color tables (some sound issues. sorry) //https://www.youtube.com/watch?v=3Lv6EnFy4k4 info about buffering //https://www.thingiverse.com/thing:3050327 a thermal camera pointer color generator file should be downloaded and opened in a browser. as of 7/22/18 most features can be changes from web page gui. for compatability with adafruit code, change #define spi_optimized_st77xx true to false, this allows standard libraries to be used with thermalcam.ino file. if use the code in this folder and includeds st77xx.ccp and st77xx.h spi bursts will be faster.

also subpixelcoloroptimized should be set to 0 if using default st77xx files, as they dont include 4 and 16 write rectanges at a time mode.

adafruit_amg88xx.ccp also has changes from default. it is only needed if sensor needs to be mirrored: #define AMG88xx_PIXEL_MIRROR true set to false

this code works out of box (for fast testing), but the best method is to rewrite the includes to use internal libraries, except for st77xx librarys which run slower than the extra ones i include in main directory. since display update is the most time intensive task (5000-7000 microseconds per pixel) (with my st77xx update it goes down to under 100 with same display!) i would imagine this code to scale performance across processor lines. much of improvement is on the bandwidth compression and ability to skip updating pixels where information does not change, until it does. some coding can be made more efficient, but the real issues are the display update speed. anything to reduce display writes increases performance. free memory has improved.

there also is an alternate library for st77xx, this includes a fillRectFast, in place of fillRect. this command includes a method of bursting spi, so while spi data being sent, other data is loaded, this way spi is always busy, even at fastest clock speed of 8mhz.

to enable faster method set spi_optimized_st77xx true however on 1.8.5 this method only works when sending more than one pixel at a time. ( possibly a bug) subpixelcoloroptimized 1 or greater it works fine still.

some known issues working on:

shift in left and right side of display.

1200ms processing time of 64x64 loop (seems high, even though this is only code that can do it on arduino)

code readability. need to clean up comments and seperate tasks into easily readable loops

removal of full buffers for sub sample after first sub sample.

feel the need to up resolution to 128x128