Fullscreen HTML5Med Res (??MB)HTML4

A thermal infrared camera measures hot or cold spots compared to the surrounding area. This is extremely useful in diagnosing hot spots in electronic circuits, which may indicate a failing component or the need for a heatsink. They can be pricey, but not this one, a DIY version that’s easy to build. Pi Pico-based Thermal Camera IR thermal cameras have many uses beyond those listed above, such as checking for overheating mechanical bearings or identifying areas of heat loss in a building. Panasonic produces the AMG8833 Infrared Array Sensor (“Grid-EYE”) that detects IR emissions on a 64-pixel 8 × 8 array. It uses the I2C serial protocol, so it can easily interface with a Raspberry Pi Pico running the Pico­ Mite operating system. Objects emit infrared energy in proportion to their temperature; the higher the temperature, the more IR energy is emitted and the higher its frequency. For really hot objects, the frequency extends into the visible wavelengths, which is why hot objects are seen to glow. By measuring this energy, we can get a pretty good idea of the temperature. There are some pitfalls, which we will mention later. With the Grid-EYE sensor, each pixel has a viewing angle of approximately 7.5°, so the overall sensor has a viewing angle of 60° (7.5° × 8). Each pixel has a tolerance of ±2.5°C when operated within specification. We can minimise this error by calibrating the sensor, as described below. Also, there can be random operating ‘noise’ of up to ±2.5°C per pixel. To reduce this, the sensor is used in moving-­average mode, which averages two readings when the sensor is set up for a 10Hz frame rate or 20 readings when for a 1Hz frame rate. If the raw output of the Array Sensor is displayed directly on an LCD screen, it appears very ‘blocky’. Still, it can easily be upscaled using a 62 Silicon Chip technique called bilinear interpolation to give the appearance of many more data points. The PicoMite Thermal Camera can upscale by factors of two, four or nine. These factors were chosen as they make the best use of the screen width. Below the thermal image display is a text read-out showing the maximum, minimum and average temperatures and the current operating mode. As mentioned above, the Array Sensor can sample at 10 FPS (frames per second) or 1 FPS. The former is most suited to fast-changing subjects, while the latter better smooths out random noise in the sensor, giving a more stable and accurate output. Bilinear interpolation This involves drawing an imaginary straight line between two data points, then generating new data points in between that lie on that line. It’s a simple technique that produces a much smoother-looking result than the more basic ‘nearest neighbour’ technique that gives a blocky image. More complicated interpolation schemes like trilinear, bicubic, Lanczos or anisotropic interpolation involve considerably more processing (arithmetic) than bilinear. In this case, their advantages are minor; bilinear gets us most of the improvement compared to no filtering with very little processing. Object emissivity The ‘fly in the ointment’ for a thermal camera is that objects vary in emissivity. An ideal IR emitter is called a Australia's electronics magazine by Kenneth Horton ‘black body’ with 100% electromagnetic emission/absorption. Shiny objects like mirrors have an emissivity closer to 0%. If you point an IR thermometer or camera at them, you will measure the temperature of an object that the mirror is reflecting, not the mirror itself. Luckily for us, many electronic components are dark colours and will have an emissivity of 90%+, so a thermal camera will measure their temperature accurately. Human skin has an emissivity of 97-99.9%, so IR thermometers also work well for measuring our temperature. This isn’t a fatal flaw but be aware that the temperature measurements of metallic objects using this IR camera could be inaccurate. It isn’t just well-polished metal surfaces either; even rough, oxidised aluminium only has an emissivity of about 20%, with polished metal surfaces usually below 5%. A known work-around to measuring the temperature of shiny surfaces (eg stainless steel pipes) is to apply some matte painters tape, which has a better emissivity. For more information, see: https://w.wiki/6R6E Circuit details As shown in Fig.1, the hardware for the project is relatively straightforward, consisting of just three modules: the Infrared Array Sensor, a Raspberry Pi Pico running the PicoMite operating system (MMBasic) and a 1.8-inch SPI TFT LCD screen with a resolution siliconchip.com.au Fig.1: the Thermal Camera circuit is straightforward, with the IR sensor array (MOD1) communicating with the Raspberry Pi Pico over an I2C bus (SDA/SCL) and the LCD screen being driven over an SPI bus (CS, SCK & MOSI). The only other components are the pushbutton for changing modes (S1) and a 39W resistor to set the LCD backlight current. of 128 × 160 pixels and an ST7735 controller. The sensor array is connected to the Pico via an I2C interface, while communications with the LCD screen are over an SPI interface. The only passive components are a pushbutton to change modes and a 39W resistor to set the current at which the display backlight operates. The following Pico GPIO pins are used: • GP08: LCD data/control (D/C) • GP09: LCD chip select (CS) • GP10: LCD SPI clock (SCK) • GP11: LCD SPI data (MOSI) • GP15: LCD reset (RST) • GP18: pushbutton sensing • GP20: AMG8833 I2C data (SDA) • GP21: AMG8833 I2C clock (SCL) The double-sided PCB is a carrier for the three modules, the pushbutton and the resistor. The display runs from a 5V DC supply from the Pico. On the Pico board, this is stepped down by a regulator to 3.3V. That 3.3V runs the RP2040 microcontroller on the Pico and is also available off-board, where it is used to power the AMG8833 IR sensor array. The Array Sensor is available from the usual auction sites pre-mounted on a breakout board, and you can find the display on the same sites. There are some suggested links in the parts list. The prototype was powered via the USB port on the Pi Pico, but there are also pads on the PCB for an external 5V power supply. This way, the Thermal Camera can be powered by a battery. The pushbutton is connected so that siliconchip.com.au it pulls the GP18 pin to GND when it is pressed. The Pico has an internal pull-up current enabled on that pin, so its voltage is high when the button is not pressed and goes low when it’s pressed, allowing the digital input to sense the change. Software operation The basic flow of the program is as follows: 1. Initialise the PicoMite, LCD screen and IR sensor array 2. Restore the calibration data and last pushbutton settings 3. Load the colour spectrum from the table 4. Enter the main loop Read 64 pixels from the sensor and adjust with the calibration data b. Calculate the maximum, minimum and average temperatures c. Convert the absolute temperatures to points on the colour spectrum d. Interpolate the intermediate colour values for each row using bilinear interpolation e. Interpolate the intermediate colour values for each column using bilinear interpolation f. Update the display g. Check the pushbutton state h. Delay if necessary Repeat items a-h above indefinitely a. The rear of the enclosure (86 × 33.4 × 57.3mm) has a cutout for the AMG8833 IR sensor; you can also see a small cutout for the Pi Pico’s USB connector on the lip. July 2023  63 by Silicon Chip), they will be plated, and nothing else needs to be done. If you etch the board yourself, those nine vias need to be drilled and short wire links soldered between the top and bottom layers in each location. The LCD screen and switch mount on the underside of the PCB, while the Pi Pico and IR sensor are on the top. For convenience, the three modules can be mounted via socket strips rather than soldering them directly to the PCB. You can cut them from longer strips if you don’t have 6-pin and 8-pin sockets. The resistor can be mounted on either side of the board. The switch is a two-pin or three-pin SIL-type vertical pushbutton that solders directly to the PCB. Alternatively, there is a provision in the 3D-printed enclosure to mount other types of pushbutton below the LCD screen and The Raspberry Pi is mounted on pin headers in sockets to make it easy to replace. wire them up to the pads on the board using short connecting wires. LCD screen limitations good imagination), and the span from Once plugged into its socket, the IR Although the LCD screen is, in the- yellow through green to cyan seems sensor is secured to the board by two 20mm-long M2.5 machine screws and ory, a standard item, displays from particularly compressed! Also, the different suppliers have different char- display is extremely sensitive to the nuts with 3D-printed spacers (Fig.4) acteristics. One display tested had the viewing angle and must be viewed between the PCB and sensor. One of red and blue colours reversed, whilst head-on to get the full spectrum of the spacers for the IR sensor has a cutthe latest batch had random pixels at colours. Otherwise, adjacent colours out to fit around an SMD component next to the module's mounting hole. the bottom and right-hand side of the blend into each other. With the IR sensor array and LCD display. As a result, three constants are attached to the PCB, now is also a good defined to allow the program to be tai- Construction The Thermal Camera is built on a time to plug the Raspberry Pi Pico into lored to the attached display: double-sided PCB coded 04105231 its sockets. ' Set to false for RGB displays that measures 60 × 52.5mm. The comYou can print the custom-made and true for BGR displays ponents are mounted as shown in enclosure in two parts (body and lid), Const BGR_display = False Figs.2 & 3. shown in Fig.5. The STL 3D printer ' Some ST7735 displays have a There are nine vias on this board. files (available to download from pixel alignment problem! Try = 2 If you are using a commercially-­ siliconchip.com.au/Shop/6/202) are Const HRES_offset = 0 produced board (such as the one sold optimised for 0.2mm layer height, ' Some ST7735 displays have a pixel alignment problem! Try = 1 Const VRES_offset = 0 Despite the display supposedly having 65,536 colours, they can’t actually show that many. Firstly, RGB(247,251,247) is one step down from white but looks significantly dimmer. The difference between this and the next step down, RGB(239,247,239), is less noticeable, as is each subsequent step. For dimmer values, the less effect each step has. RGB(127,127,127) is very dim, and RGB(63,63,63) is almost black! Another, more technical way of saying this is that the display has a very high gamma value. As a result, it is difficult to get more than about 38 distinct colours across the spectrum (even with a 64 Silicon Chip Figs.2 & 3: components are mounted on both sides of the board. On one side are the Raspberry Pi Pico and IR sensor, both plugged in via header strips. The LCD screen, pushbutton and resistor are mounted on the other side, although the resistor can go on either side. Australia's electronics magazine siliconchip.com.au 0.4mm wall thickness and 100% fill. Note that the first layer of the screw holes is filled as it gives a more pleasing appearance to the front and back of the case – just drill them out after printing. The holes in the lid are countersunk under the top layer and are best cleared with an 8-10mm drill by hand. The PCB assembly is held in the case by four 25mm-long M3 machine screws and nuts, with 3D-printed spacers between the display and the PCB at the opposite end to the connector. It is necessary to insert the display into the case first, insert the machine screws from the front of the case, place the spacers over the screw shafts and then plug the PCB onto the display. Finally, secure it with the nuts. Loading the software It is assumed that readers are familiar with loading PicoMite software, which was described in the article on the PicoMite in the January 2022 issue (siliconchip.au/Article/15177). Briefly: 1. Download the PicoMite operating system from http://geoffg.net/picomite. html and unzip the file. 2. To load the operating system onto the PI Pico, plug the USB cable into a PC while holding down the white button. 3. The Pi Pico will appear as a USB drive. Copy/drag the file PicoMitexx. xx.xx.uf2 onto that drive. 4. Connect to the PicoMite’s USB serial port using your preferred serial terminal emulator (eg, TeraTerm or PuTTY). 5. Once connected, enter each of these commands in turn, but note that many of them reset the Pi Pico, so the USB connection is lost and will need to be restored before entering the next: The pushbutton is visible on the back of the PCB at upper left. A few different compatible types can be obtained. camera.bas” file into the PicoMite, again using your preferred serial terminal emulator or MMEdit. Use the “Autosave” or “XMODEM receive” commands, depending upon your preference. 8. If you’d prefer to skip most of the above sequence, you can download the “Thermal camera (RGB).uf2” or “Thermal camera (BGR).uf2” file from the Silicon Chip website and upload it in the third step above. That’s equivalent to running all the configuration commands and loading the BASIC code. The only difference between the two files is the expected LCD screen configuration, so if the displayed colours are wrong, load the other file. Operation The pushbutton has the following functions: • Short press (less than 1.5 seconds): cycles the display scaling factor through 1, 2, 4 and 9 • Long press (more than 1.5 seconds): toggles between 1 FPS and 10 FPS Fig.4: 3D-printed spacers are used rather than off-the-shelf types since we can make them exactly the right dimensions, and you can print them at the same time as the case. Note the cut-out in one to clear an SMD component near the mounting hole on the IR sensor module. OPTION RESET OPTION CPUSPEED 252000 OPTION SYSTEM SPI GP10,GP11,GP12 OPTION LCDPANEL ST7735,RP,GP8,GP15,GP9 OPTION SYSTEM I2C GP20,GP21 6. The following two commands are optional; the first shows you what you have configured, while the second lets you verify that the LCD screen is working: OPTION LIST GUI TEST LCDPANEL 7. Finally, load the “Thermal siliconchip.com.au Fig.5: the 3D printed enclosure base and lid. The holes do not go all the way through because it gives a neater result to drill the thin panels after printing the case than print the case with the holes. Australia's electronics magazine July 2023  65 Parts List – Raspberry Pi Thermal Camera 1 double-sided PCB coded 04105231, 60 × 52.5mm 1 Raspberry Pi Pico 1 3D-printed enclosure (body & lid) 4 3D-printed spacers 1 AMG8833 Grid-EYE IR sensor array breakout board module with pin order VIN, GND, SCL, SDA, INT & ADO [AliExpress www.aliexpress.com/item/33012193094.html] 1 1.8-inch 128×160 pixel SPI LCD TFT screen with ST7735 controller [Tempero Systems TS-S006; eBay; AliExpress www.aliexpress.com/ item/1005003797803015.html {1.8 inch option}] 1 SPDT momentary PCB-mounting subminiature pushbutton switch (S1) [Altronics S1493 or APEM TP32P0] 1 39W 5% ¼W axial resistor 2 20-pin headers, 2.54mm pitch 2 20-pin header sockets, 2.54mm pitch 1 8-pin header socket, 2.54mm pitch 1 6-pin header socket, 2.54mm pitch 2 M2.5 × 20mm panhead machine screws and hex nuts 4 M3 × 25mm panhead machine screws and hex nuts 4 No.2 × 6mm countersunk head self-tapping screws • Very long press (more than 10 seconds): enters calibration mode Note that the frame rates are sensor refresh times, not screen refresh times. At 10 frames per second, the screen update time is longer than 1/10th of a second for scale factors 4 and 9. At scale factor 4, the display will be updated approximately every 220 ms and, at scale factor 9, every 700ms. This is because the bilinear calculations take some time to complete for higher scaling factors. In calibration mode, the sensor is set to 1 FPS, and 10 readings are taken over a 10-second period. These are then averaged, and the correction factors for each pixel are stored in non-volatile memory. Good results are obtained by holding the sensor perfectly still 2-3cm from a white sheet of paper. If the button is pressed during calibration, calibration is abandoned, and the correction factors are cleared. We recommend letting the sensor stabilise for at least one minute with power on before performing calibration. Software tweaks In the software, the constant “Fahrenheit” can be set to “true” to display temperatures in Fahrenheit rather than centigrade/Celsius. The constant “Minimum_span” sets the minimum temperature span for the display when there is little temperature variation across the It’s critical you purchase a module with the same output pin layout as the one shown above. whole display. This prevents wildly varying colours for minimal temperature changes. Lower values make the display more sensitive when there is an almost uniform temperature gradient. Speeding up the refresh rate The latest version of the Pico­ Mite firmware (“PicoMiteV5.07.06. uf2”) allows the CPU speed to be increased from the old maximum speed of 252000 to 378000 with the command: Option CPUspeed 378000 This means that, with a scale factor of 4, the display will be updated approximately every 165ms rather than 220ms and, at scale factor 9, every 520ms rather than 700ms. However, note that this is ‘overclocking’ the RP2040 processor and it’s possible that it won’t work on every board or under all conditions. Still, most Pico boards should be capable of running SC at this speed. Dual-Channel Breadboard Power Supply Our Dual-Channel Breadboard PSU features two independent channels each delivering 0-14V <at> 0-1A. It runs from 7-15V DC or USB 5V DC, and plugs straight into the power rails of a breadboard, making it ideal for prototyping. Photo shows both the Breadboard PSU and optional Display Adaptor (with 20x4 LCD) assembled. Both articles in the December 2022 issue – siliconchip.au/Series/401 SC6571 ($40 + post): Breadboard PSU Complete Kit SC6572 ($50 + post): Breadboard PSU Display Adaptor Kit 66 Silicon Chip Australia's electronics magazine siliconchip.com.au