Fullscreen HTML5Med Res (??MB)HTML4

Nicholas Vinen Adjustable brightness and ambient auto-dimming using an LDR Adjustable pattern cycle time Mostly pre-assembled; can be up and running in under an hour Power supply: 12V DC recommended <at> 1-2A (operating range: 6-16V DC) SC7535 Kit ($80) Includes a pre-assembled PCB with nearly all parts fitted, except for IC1, REG2 etc (see the parts list) 24cm tall and wide white star-shaped PCB 80 onboard bright WS2812B RGB LEDs in a star/circle pattern 12 different LED light patterns, each with four possible colour palettes Manual or auto-cycling patterns and palettes Jazz up your Christmas tree (or just about anything else) with this luminous RGB LED Star. It has 80 “NeoPixel” LEDs that create an array of dazzling, colourful patterns. You can choose which patterns and colour schemes you like, adjust the brightness and more. It’s quick and easy to build, too! siliconchip.com.au Australia's electronics magazine December 2025  41 Y ou can watch a video of some of the available patterns and colour schemes at siliconchip. au/Video/RGBStar It has been a few years since we’ve published a Christmas ornament project, despite those generally being very popular. Partly it’s because people were still building our previous designs. However, while making more kits for the November 2020 RGB LED Christmas Star, we had two realisations. Firstly, it had been five years since we published that design. Secondly, it’s a lot of work to build, but people obviously think it’s worthwhile as they continue to order kits for it. It uses 30 individual RGB LEDs, with four pins to solder on each, plus 90 current-limiting resistors, 13 driver ICs, more than 20 bypass capacitors and some other sundry parts. Assembling the board would take most of a day. That got us thinking: was there an easier way? Our first thought was WS2812B “NeoPixel” LEDs. These devices have GND and Vcc pins plus serial data input and output pins (four in total). The Vcc and GND pins are all joined in parallel, while the output of one LED goes to the input of another, forming a daisy chain (you may well have seen these on RGB LED strips). That means you only need one pin on a microcontroller to drive many – possibly even hundreds – of these devices. They’re bright and can display one of 16,777,215 different colours. By writing clever software on a microcontroller, we could generate all sorts of cool patterns using a string of these devices. So what’s not to like? Our concerns were that they are SMD-only and, if you buy them individually and put a lot on a board, the cost can add up. However, there is a solution to that: get the PCB manufacturer to solder the LEDs for you. That saves a lot of work and gives a neat, professional result. Also, because they have access to huge quantities directly from the manufacturer, it can be cheaper than buying individual parts and soldering them yourself. Down the rabbit hole So we set to work designing a PCB for this, with four questions immediately popping up. What shape, size and colour should the board be, and how many NeoPixel LEDs should we put on it? We liked the star concept, but weren’t sold on the five-pointed star used in the November 2020 project we mentioned earlier. Partly that was out of a desire to do something different, but also, five-pointed stars don’t look right to us. Of course, real stars don’t have points; they are distant, bright spheres that should look like a point source. It’s the optical imaging system (a telescope, or perhaps our eyes) that makes them appear to have points. Usually, those points are arranged symmetrically, meaning there are an even number of them. Fig.1 shows an image from the James Webb Space Telescope where you can see that its optics generate six lines that appear to emerge from the star’s point source. To cut a long story short, we decided that a more realistic and cool-looking star would have four long and four short points, with the short points offset by 45°. We also decided to make the board white, for two reasons. Firstly, to differentiate it from the earlier star, which was black, and secondly, because the original WS2812B LEDs came in white plastic packages. (Black ones are now available, but let’s put them aside for another day.) As for the size, we were able to design a board with that shape that wouldn’t be too expensive to manufacture by making it 238mm tall and 238mm wide. The trick is to rotate the design by 45° so that it fits inside a 168 × 168mm square, as that’s the basis on which the manufacturers charge (238mm ÷ 168mm ≈ √2). Having drawn the board shape, we arranged a string of LEDs around the edge of the star shape, plus two rings inside, ending up with 80 LEDs in total. That seemed like a reasonable number to generate some interesting patterns. So the mechanical side of things was sorted out, and it was time to turn to the electronics! Power delivery and control Fig.1: stars look like more than just points because of the optics viewing them (eg, a telescope or our eyes). The lines that appear to radiate from them are normally symmetrical, meaning an even number, and some are longer than others. Hence the shape of our Star, with four long points and four short ones. At full brightness, set to produce white light, each LED draws around 50mA. Multiply that by the 80 LEDs and we can see that we need to deliver around 4A at 5V to run them all. That’s a substantial amount of power: 20W. Of course, this isn’t a torch, so we would never actually drive them all white at once. The actual average power required would be a more modest 10W or so; 2A at 5V. And that’s assuming you would run it at full brightness, which, to get a bit technical, is eye-searingly bright. Most people would run it close to 5W of total LED power. Still, the 10W figure could be delivered by a reasonably efficient 2A buck (step-down) regulator running from a higher voltage, like 12V. The same 5V supply (or a separate one) could then run a microcontroller to generate the patterns. Besides the LEDs, the micro and power supply, we would also need a decent number of bypass capacitors spread around the board. The WS2812B LEDs control brightness Australia's electronics magazine siliconchip.com.au 42 Silicon Chip using pulse-width modulation (PWM), meaning there will be constant switching spikes distributed around the board. So we’ll want a few ceramic and perhaps tantalum polymer capacitors to keep the 5V rail nice and stable. All that would be left would be a few buttons and perhaps trimpots to do things like adjust brightness, change the patterns, set up pattern auto-­cycling and such. Add an LDR to monitor the ambient light level for auto-dimming, and the design was complete. Getting it assembled As mentioned earlier, our plan was to order some prototype PCBs and have the manufacturer fit the WS2812B LEDs for us. We would then add the power supply, microcontroller and a few other bits and pieces to finish it off. The power supply and microcontroller could go on the back of the board, so they wouldn’t mar its appearance, meaning only one side of the board needed to be pre-populated. We figured we could also get them to place the bypass capacitors for the WS2812Bs on the same side, as they would not ruin the appearance as long as we placed them symmetrically. But we’d leave fitting all the other parts for ourselves (or someone else building this later). As we went through this process, we discovered that the WS2812B LEDs require “special handling” because they have a high moisture-­sensitivity level (MSL5). Automated/contract PCB assembly is usually done using solder reflow, where solder paste is melted in an infrared (IR) oven, or wave soldering, where a wave of molten solder passes over the surface of the board and some sticks to the exposed pads and pins. In both cases, the components rapidly heat up from room temperature to about 250°C over a few minutes. They sit at the high temperature for just long enough to let all the solder melt and reflow, then they are cooled back down to close to ambient temperature. The whole process takes about 4-6 minutes. If any components on the board contain moisture (water), that water will flash boil and rapidly expand. The result can be exploding components – not what you want. So they have to be dry before you start; either removed from a hermetically sealed package just before soldering, or baked in an oven at a lower temperature (90-125°C siliconchip.com.au Parts List – RGB LED Star Ornament 1 white 168 × 168mm double-sided star-shaped PCB coded 16112251 1 12V DC 1A+ (2A recommended) power supply × 1 2-pin header or right-angle header (CON1; optional) • 1 5-pin header or right-angle header (CON2; optional, for ICSP) × 1 3-pin header or right-angle header (CON3; optional, for testing) × 2 6A 18mW (120W <at> 100MHz) SMD M3216/1206 ferrite beads (FB1, FB2) [Tai-Tech HCB3216KF-121T60] 1 2.7A 33μH shielded SMD inductor, 12 × 12mm (L1) [Sunltech SLH1204S330MTT] 1 GL5528 20kW-1MW LDR (LDR1) • 3 6 × 3mm two-pin SMD tactile pushbutton switches with white actuators (S1-S3) 2 10kW Bourns TC33X-2-103E SMD trimpots (VR1, VR2) Semiconductors 1 PIC16F18126-I/SL 8-bit 14-pin microcontroller programmed with 1611225A.HEX, SOIC-14 (IC1) • 1 LDL1117S50R or AMS1117-5 low-dropout 5V linear regulator, SOT-223 (REG1) 1 AP5002 20V 2A 500kHz integrated buck regulator, SOIC-8 (REG2) • 80 WS2812B-V5 serial RGB “NeoPixel” LEDs, SMD 5050 (LED1-LED80) 1 AO3400A 30V 5.8A N-channel Mosfet, SOT-23 (Q1) 1 BZX84C5V6 5.6V ±1% 250mW SMD zener diode, SOT-23 (ZD1) 1 BZG05C-5V6 5.6V ±6% 1.25W SMD zener diode, DO-214AC (ZD2) 1 B340A or S3A 40V 3A SMD schottky diode, DO-214AC (D1) Capacitors (all 50V SMD X7R MLCC, M3216/1206 size unless noted) 1 220μF 25V polymer aluminium electrolytic capacitor, 6.3×6mm SMD [Shengyang SM227M025E0600] 6 220μF 6.3V polymer tantalum electrolytic capacitors, SMB case [Panasonic 6TPE220MAZB] 12 22μF 25V X5R 1 22μF 10V SMA tantalum [Kyocera AVX TAJA226M010RNJ] 31 100nF 1 4.7nF 1 1nF Resistors (all ±1% SMD M3216/1206 size) 2 100kW 1 10kW 2 6.8kW 1 1.3kW 1 1kW 1 330W RGB LED Star Kit (SC7535, $80 + P&P): comes with a pre-assembled PCB with all parts fitted except those marked with • or ×. The parts marked • are included in the kit but must be fitted by the constructor. for 24-48 hours) to drive out all the moisture before soldering. So, we had to pay a little extra to get the WS2812B LEDs for our prototype boards baked. That meant we needed to choose the ‘standard’ assembly service rather than the ‘economic’ one. The standard service also allows you to put components on both sides of the board. Hmm. It was starting to look like we might as well get virtually the whole thing assembled! One of the parts we chose to use, the buck regulator controller, is no longer being manufactured (we’ll explain why we chose it a bit later). We have hundreds, but didn’t feel like sending them to the manufacturer, so we decided to solder it ourselves. It’s only an 8-pin device in a relatively large Australia's electronics magazine SOIC package, so not difficult for us or any other constructor to add. We also left the PIC16 chip that produces the patterns off the assembled board. It’s also in an SOIC package (with 14 pins, though) and we figured we could supply programmed chips that constructors could easily fit to the board. We also didn’t have them put any headers on it, since we figured people might have different preferred arrangements for wiring up the power supply. Other than that, though, all our prototype boards – and the ones we’ll supply to readers – come with almost all the parts fitted. So assembly is quick and easy, despite the large number of LEDs! This also helps to keep the total cost to build it down, since the December 2025  43 manufacturer can source these parts in bulk from their partner warehouse. As a result, we can supply a kit that includes the mostly assembled board, programmed microcontroller and switch-mode chip, ready to assemble, for around $80 + P&P. That is somewhat more than the $45 we charge for the November 2020 RGB LED Star kit, but considering the time savings in not having to assemble the whole thing yourself from well over one hundred parts, the larger PCB, larger number of LEDs, better patterns and brightness, it ends up being a pretty good deal. But for those intrepid constructors, the blank PCB will be available separately, if you want to assemble the whole project yourself. Circuit details The full circuit of the RGB LED Star is shown in Fig.2. By showing only some of the long chain of LEDs, we manage to keep it relatively simple. It can be broken into three main blocks: the LEDs, the microcontroller and the power supply. Microcontroller IC1 drives the chain of LEDs (LED1LED80) via its RA2 digital output and a 330W series resistor, as recommended in WS2812B data sheets to reduce overshoot and ringing from trace inductance and limit fault current into the first LED. Each LED’s output drives the subsequent LED input until LED80, which is loaded with a resistor to ground to reduce the chance of any ringing or EMI from an unterminated output pin. A total of 42 bypass capacitors scattered around the board, in three different values, provide local bypassing for these 80 LEDs. In total, they can draw up to 4A (!), and they dim the LEDs using PWM, so we want to ensure they see a low 5V supply source impedance. Test header CON3 gives us a way to drive the LEDs before IC1 is soldered to the board, should we need that. Most constructors will not need to fit it. The user controls are three pushbuttons, S1-S3, and two trimpots, VR1 & VR2; the buttons connect to digital inputs RC0-RC2 (pins 10-8). The software in IC1 enables a weak pull-up current on these pins so they are normally held high, at around 5V. When a button is pressed, it pulls that pin to 0V, transitioning to a digital low, so the software can sense that. Debouncing is performed in software. When VR1 & VR2 are rotated, they vary the voltages at pins 3 & 2 of IC1 between 0V (fully anti-clockwise) and 5V (fully clockwise). IC1 uses its internal analog-to-digital converter (ADC) to convert these voltages into numbers between 0 and 4095, to control the overall brightness and the duration of each different pattern, respectively. Another analog input, ANC5 at pin 5, connects to the junction of a light-dependent resistor (LDR1) and a fixed 100kW resistor. The voltage at this pin will be higher The front of the LED Star includes two buttons to change patterns. 44 Silicon Chip in bright ambient lighting conditions and closer to 0V in darkness. Again, the ADC is used to sample this voltage to provide auto-dimming, so the LEDs are not so bright at night, but bright enough to see clearly during the day. At 100% brightness, the RGB LED star is eye-searing! So we definitely need both manual and automatic brightness control. Five-pin header CON2 can be used to program IC1 in-circuit. We used this during development, but since kits/boards will come with a pre-­ programmed microcontroller, you won’t need to fit it unless you want to design your own patterns or otherwise make changes to the firmware. Power supply Power comes in via a two-pin header (CON1) or soldered wires shown at upper left. Mosfet Q1 and zener diode ZD1 provide reverse polarity protection without dropping much voltage. If the polarity is correct, Q1’s gate is pulled high, switching it on and providing a low-resistance path between the negative conductor and ground. If the polarity is reversed, Q1’s gate is pulled negative. It remains off, and its intrinsic body diode is reverse-­ biased, so no current can flow. The Mosfet is rated at 30V, and ZD1 clamps its gate voltage at a safe level, so nothing will happen with a negative voltage up to -30V applied to the circuit. The maximum positive voltage is limited to 18V by REG1. One 220μF aluminium polymer electrolytic capacitor and two 22μF ceramics provide bulk storage and bypassing for the input of REG1, an LDL1117 5V low-dropout regulator. This is a true low-dropout regulator, and it will provide a regulated 5V rail for the microcontroller as long as there’s at least 5.5V at its input. It also has a 22μF tantalum output filter capacitor, required for stability. Its 5V output is only used to power the microcontroller and connected components like the LDR and trimpots. This means that the switching noise from the LEDs doing their PWM brightness control won’t affect the microcontroller’s analog measurements. The remainder of the power supply is a DC/DC converter that supplies the high-current 5V rail for the LEDs. It is based around REG2, an AP5002 ‘2A siliconchip.com.au Fig.2: the PIC16F18126 microcontroller at lower left controls the 80 LEDs by sending serial data from its RA2 digital output. The LEDs have plenty of bypass capacitors of various sizes to provide them with the peak current they draw during operation. The power supply at the top includes reverse-polarity protection (Q1/ZD1), a linear regulator to power the micro (REG1) and a switch-mode step-down buck regulator (REG2) to power the LEDs from a ~12V DC source. buck’ converter. This chip is obsolete now, but we like it so much that we bought several hundred, so we will supply them with boards/kits (or separately, if you really need one for some other reason). Some of the reasons we are still using this is that it is easy to solder, coming in an 8-pin SOIC package with no thermal pad underneath; it has a useful range of voltages (up to 20V) and currents (it says 2A but can actually deliver 4A or more in some cases), can go to 100% duty cycle (meaning it’s a ‘low dropout’ buck regulator!) siliconchip.com.au and it’s up to 90% efficient. We also find that it ‘just works’. One of its nice features is an external compensation network that requires just two or three components. This is one of the keys to its stability in a wide range of situations. We’ve used it in a few projects before, such as the Simple 1.2-20V 1.5A Switching Regulator (February 2012; siliconchip.au/ Article/774) and the CLASSiC DAC (February-April 2013; siliconchip.au/ Series/63). Now, while in theory the LEDs could draw up to about 4A if they were all Australia's electronics magazine at 100% brightness and set to white; this is not a torch, so that’s unlikely to happen. In general, this regulator will see a load below 2A <at> 5V; perhaps a little higher in extreme cases. So, the internal switch current limit of 3.5A and L1’s rating of 2.7A are not really of concern (especially since L1’s current rating is saturation-based, not thermal). Since we’re recommending a 12V DC input and we’re producing a regulated 5V (more like 4.9V, actually) for the LEDs, REG2 is operating near its sweet spot, around 90% efficiency (its December 2025  45 Fig.3: the star has been rotated 45° to fit on the page; the upper-left corner is intended to be the top. All the components you see here come pre-soldered to the board except for LDR1, in the centre. Some holes at lower- right allow you to attach it to a stick or the top of a Christmas tree. 2A headline rating is more of a ‘worstcase scenario’, thankfully). We have ferrite beads at the input and output of the DC/DC converter section to try to reduce the amount of switching hash radiated from either end. FB2 is also convenient in that if you run into problems, you can remove it to disconnect the output of the DC/ DC converter from the LEDs, allowing you to test them in isolation. Zener diode ZD2 is not strictly necessary; it’s a ‘belts and braces’ protection measure. Should there be positive spikes from the output of REG2 for some reason (eg, the load suddenly goes from 4A to 50mA and it can’t switch off fast enough), ZD2 will 46 Silicon Chip conduct once the LED supply exceeds about 5.5V, limiting the supply rail to a safe level of about 6V in the short term. PCB design The top side of the assembled board is shown in Fig.3. The only part you need is add here is the LDR at the centre. The first LED in the chain, LED1, is towards upper left. You can follow the snaking path of the data from there clockwise around the star, back to LED56 just below LED1, then to the circle formed by LED57-LED72, and finally the smaller concentric circle formed by LED73-LED80. Note how there is a polymer tantalum capacitor on each of the longer Australia's electronics magazine ‘arms’ of the star, plus two in the middle, for distributed bypassing. The eight points of the star also feature a 22μF bypass capacitor, plus numerous 100nF capacitors scattered throughout so that all LEDs have a low source impedance. Most of the copper on the top of the board is a +5V distribution plane, with the underside copper being GND. Hence, there are pairs of vias near the GND end of most components to connect them to the ground plane. The star is intended to be rotated so that LED5 is at the top. There are holes at each long point (eg, for hanging it), plus some extra holes on the bottom one so that a stick or similar can be siliconchip.com.au Fig.4: the control and power supply circuitry is located in the centre of the back of the Star. The DC/DC converter delivers 4.9V DC at up to several amps to the centre of the board, where it’s conducted out to the LEDs arrayed around it by copper planes; the front of the board has the +5V (4.9V) plane while the GND plane is on the back side. attached for holding it up (shown with M2 machine screws in them in some of the photos). Turning to the underside (Fig.4), you can see that the power supply and control components are clustered in the centre, divided by fivepin header CON2. The control components are above and to the right of CON2, while the power supply is below and left. The power connections (CON1) are near the bottom of the star, so any attached wires can hang down behind it. A slight revision We made just a couple of minor changes to the design between the siliconchip.com.au prototype and the final version. Firstly, we added a 220μF 25V low-ESR aluminium polymer capacitor across the input of the linear regulator, REG1, visible on the left in Fig.4. This is to overcome any lead inductance from the power supply, suppress ringing and provide a local charge reservoir for the switch-mode regulator. While the prototype worked without it, we felt that it was worthwhile adding, especially since the part only costs about 10¢. We also added another 22μF ceramic capacitor in parallel with it for higher-frequency stabilisation of the input supply. Finally, we decided to replace the boring old AMS1117-5 regulator Australia's electronics magazine (REG1) on the prototype with the more modern LDL1117S50R. This is a direct drop-in replacement, as it’s in the same SOT-223 package and has the same pinout. However, it is a true low-dropout regulator that can withstand a higher input voltage (18V vs 15V), with a lower quiescent current (0.25mA vs 5mA), so we thought it was a worthwhile upgrade (the required tantalum output capacitor upgrade costs more than the regulator). Construction When you receive the board, it should have all the top-side components fitted besides LDR1, and all the underside components fitted besides December 2025  47 the three pin headers (all of which are optional), IC1 and REG2. It will also have a pair of ‘rails’ attached to it, which were used to hold it during the assembly process (see the photo below). There is a series of holes drilled between the rails and the PCB edges (‘mouse bites’). These allow you to easily snap the rail off by flexing the junction back and forth a few times. The breaks should be fairly clean, but if you want to, you can clean them up with a file. Just make sure you don’t breathe the resulting fibreglass dust (eg, do it outdoors and ideally while wearing a mask, or at least with the wind blowing it away from you). Testing Our prototypes worked straight away, so really the easiest way to proceed is to solder LDR1, IC1 and REG2 to the board, making sure IC1 & REG2 are orientated correctly, then apply power and check that it lights up and start producing some nice LED patterns. However, if you want to test it more methodically, you can (as we did with our prototypes initially, just to be safe). Firstly, you can solder CON3 to the board. If you don’t want it visible on the front of the Star, you can either solder wires to the pads on the rear, or remove it later. Connect CON3’s terminal marked + to the output of a bench supply. Short the The LED Star has some M2-sized mounting holes, for if you want to attach it to something solid. other two pins together and connect them to the negative output/ground. Set the current limit to 100mA, the voltage to 5V and apply power. Our board drew 40mA in this condition, exactly what was expected when powering 80 LEDs with 0.5mA quiescent current each. This verifies that there are no short circuits or faulty components among the LEDs or their bypass capacitors. Next, solder to the pads of CON1. You can use a right-angle header on the rear of the board if you have the type of right-angle header you can solder from the same side, such as a polarised right-angle header. Alternatively, solder the wires to the board. Apply 7-12V DC and you should see a current draw of around 5mA if you have an early board with the AMS1117, or 0.5mA if you have the LDL1117 regulator. Use a multimeter to check the voltage between pads 1 & 14 of IC1. You should get a reading close to 5V. Now remove power and solder REG2 to the board (remember what we said earlier about getting its orientation right!). One side is quite close to D1, but we managed to solder the pins on that side without bridges by bringing the iron in from above. Good news: pins 5 & 6 are connected electrically, as are pins 7 & 8, so as long as you don’t get a bridge across the middle two pins on that side, it’ll be fine! Place a jumper on CON3 between the middle and ground pins (ie, not on the + side). This holds the input to LED1 low. Now apply 6-12V via CON1 and measure the voltage across the pins on either side of CON3, being careful not to short anything. You should measure close to 4.9V DC, and the current draw should be around 50mA. If all is good, switch it off and solder IC1 to its pads, again checking its orientation carefully. This should be easy as there aren’t any components too close to its pins, but check for solder bridges and if you find any, remove them with flux paste and solder wick. Because parts are soldered to both sides of the PCB, it needs to be manufactured with rails, as shown here. They are used to hold it in place during soldering. When you receive it, you can flex them back and forth a few times carefully and they will snap off. If you want to clean up the rough edges left behind, it’s easy to do with a file, but don’t breathe the dust. 48 Silicon Chip Australia's electronics magazine siliconchip.com.au The finished underside of the RGB LED star is shown on the right, with a close-up shown inset. This inset photo shows the missing parts that you need to fit yourself, but are included in the kit. If IC1 hasn’t been programmed, you can attach CON2 and a PIC programmer and upload the firmware (1611225A.HEX). But since you most likely bought an assembled board, it will come with a pre-programmed chip, so you are ready to power it up properly using a high-current (capable of at least 1A) 12V DC supply at CON1 and you should be rewarded with colourful patterns. Using it By default, it will cycle through all 12 possible patterns and four colour palettes at an interval determined by the setting of trimpot VR2. Fully anti-clockwise will make it cycle roughly once per second, while fully clockwise will give you around five minutes per pattern. VR1 will adjust the overall brightness, and the light level on LDR1 will also affect the brightness. You’ll need a jeweller’s slotted screwdriver or a similarly slim tool to rotate the small screws of the two trimpots. Note that there are no stops; they rotate through 360° but they only work properly over about ¾ of that travel. So if you get flickering when adjusting VR1, or odd behaviour when adjusting VR2, you’re probably off the track. Pressing S1 briefly will switch to manual pattern selection mode. Each further press of S1 will cycle to the next pattern. Once pattern 12 is reached, pressing S1 will go back to pattern 1. siliconchip.com.au Similarly, pressing S2 briefly will switch to manual palette selection mode. Each further press of S2 will cycle to the next palette. Once the fourth palette is reached, pressing S2 will go back to the first. The auto-cycling works mostly independently for patterns and palettes, meaning you can have both change automatically periodically, or just one or the other, or neither. To resume auto-­ cycling the pattern, hold down S1 for a couple of seconds. To resume auto-­ cycling the palette, hold down S2 for a couple of seconds. If both are set to cycle, when it’s time to cycle, the unit will select the next enabled pattern with a random palette. That way, you get to see different patterns with different palettes. You can disable some patterns or palettes if you don’t like them. To disable the current pattern (regardless of whether it’s auto-cycling or manually selected), hold S2, then quickly press and release S1, then release S2. It will go to the next enabled pattern, and the last pattern will be skipped from now on. You can’t re-enable individual patterns since you can’t select them, but you can re-enable them all by holding down S2, then pressing and holding S1 for a few seconds, and then releasing S1 before S2. Similarly, to disable the current palette, hold S1, then quickly press Australia's electronics magazine S2, release S2 and then S1. It will go to the next enabled palette, and the last palette will be skipped from now on. You can’t disable the last pattern or palette, though. To re-enable all palettes, hold down S1, then hold down S2 for a few seconds, then release S2, then S1. There is a third switch on the back of the board labelled S3. It can be used to change the LDR setpoints or disable automatic brightness control. Usually, VR1 sets the maximum brightness, but it will be reduced if the LDR senses a low ambient light level. By default, it will be reduced to a very low setting, around 5% of maximum, in complete darkness. To change that, adjust VR1 to get the minimum brightness you want, then press S3 briefly. You can disable auto brightness adjustment by holding S3 for a couple of seconds; then only VR1 sets the brightness. Setting the minimum re-enables it. You may want to temporarily disable auto brightness when adjusting VR1 to set the minimum since it’ll remove the effect of ambient light while you are making the setting. If you find that too much of the light from the LEDs shines on the LDR, making its brightness control unstable or ineffective, you could shrink a short section of black heatshrink tubing around it to shield it from light from the sides. That way, it should only react to ambient or reflected light. We December 2025  49 didn’t need to do that, but since the LDR calibration is automatic, it must be exposed to both light and dark before it will sense properly. Firmware operation The firmware is written in C and split into five files: • neopixel.c is the main program that includes pin configuration, peripheral configuration, button, potentiometer and LDR sensing, user interface logic, LED update code and the main loop. • LEDs.c contains 320 bytes of data, stored in flash, on the locations of the 80 LEDs on the physical PCB for use by the effects code. • rgb.c contains helper functions that generate the four colour palettes and perform RGB colour mixing. • trig.c contains an integer sine table and sine/cosine functions for trigonometric effects. • effects.c contains the logic to implement the 12 separate patterns that can be displayed on the LEDs. Some effects are based on the locations of the LEDs in the chain, some on their Cartesian (X/Y) coordinates on the PCB, and some on their polar (distance/angle coordinates) on the PCB. For example, effect #1 is a gradient that snakes its way along the chain of LEDs from LED1 to LED80. It is a nice-looking effect but the simplest to implement. Effect #3 is coloured circles that radiate from the centre of the star, out the points, and then end. It uses the polar coordinates, comparing the distance of the LED from the centre of the board to the current radius of the circle. Effect #5 is a colour spectrum that rotates around the display like a spinning wheel. It is similar to effect #1, but it uses the polar angle of the LED rather than its position in the chain. Effect #10 is coloured bubbles that grow from random points within the star and then burst. It calculates the distance of an LED from the centre of a bubble using the formula distance = √(x1 – x2)2 + (y1 – y2)2. The effects are double buffered, meaning that the processor is calculating the next state of the LEDs to show (into buffer x) simultaneously with transmitting the last update to the LED string (from buffer y). The buffers are swapped on each run through the SC main loop. 50 Silicon Chip Updating Neopixels using the CLC hardware We got this idea from Microchip Application Note AN1606 “Using the Configurable Logic Cell (CLC) to Interface a PIC16F1509 and WS2811 LED Driver”. Unfortunately, the CLC peripheral in modern PICs is configured quite differently from the PIC16F1509, so the code in that App Note is no longer very useful. Also, we realised it’s unnecessarily complex – using two CLCs when the job can be done with just one (possibly due to improvements in the CLCs since it was written). The difficulty in driving WS2812B chips is that they use a somewhat unique scheme that encodes 0s and 1s into different-length positive pulses. Interestingly, different versions of the WS2812B, even within the same manufacturer, have different specifications for what is required. As shown in Fig.a, the specific version we are using (V5) requires: Zero bit: 220-380ns high, 580-1000ns low One bit: 580-1000ns high, 580-1000ns low Reset frame: low for at least 280μs Note how this scheme is difficult to encode with a system that emits bits at consistent intervals. A zero bit lasts for 800-1380ns and a one bit lasts for 11602000ns. So, to have consistent bit intervals, the interval must be between 1160ns (862kHz) and 1380ns (725kHz). If you select an interval in the middle of, say, 1250ns (800kHz), that means you could have a one bit with a 50% duty cycle, ie, 625ns high and 625ns low. The high time for the zero bit would need to be at least 250ns to avoid the low time exceeding the 1000ns threshold, giving you a range of 250-380ns. We use the MSSP peripheral to generate an 800kHz SPI stream as the basis for this signal. This gives us an SCK signal that’s high for 625ns and low for 625ns, and an SDO signal that’s either high or low for 1250ns depending on the value of the bit being transmitted. We feed both of these to a CLC cell, along with a synchronised PWM waveform that has a high time of around half the SCK high-time at ~312.5ns – see Fig.b. Fig.a: there are quite a few WS2812B variants, and they all seem to have slightly different timing requirements (sometimes compatible with each other, sometimes not). We’re using the current V5 variant; its timing is shown at upper right here, with the data formats (common to all variants) below. The timings in the lower part of the diagram are nominal. Australia's electronics magazine siliconchip.com.au Fig.b: we’re using the MSSP serial peripheral, CCP PWM generator and CLC configurable logic cell in the PIC to generate the WS2812B signal with minimal CPU overhead. Writing a byte to a buffer triggers the MSSP to generate eight bits of data. We combine its clock (SCK), serial data output (SDO) and synchronised PWM signal with the CLC to make the orange waveform the LEDs expect. That time of year is nearly here... CHRISTMAS Spice up your festive season with eight LED decorations! Luckily for us, it’s possible to synchronise the MSSP to a PWM peripheral by setting the SSPM bits in the SSPxCON1 register to a value of 3 (SPI Host Mode: Clock = TMR2 output/2). TMR2 is also the default clock source for a Capture/Compare/PWM (CCP) Module, which can be set in PWM mode. So we can synchronise the leading edge of the PWM pulses with the leading edge of the SCK pulses. We can then use the CLC to perform the simple logic operation (SDO & SCK) | PWM. The SDO & SCK operation produces 625ns positive pulses if the data bit is a one or no pulse if the data bit is a zero. By ORing this with the PWM pulses, which overlap with the start of the SCK pulses, we either get a 625ns pulse if the data bit is a one or a ~300ns pulse if it’s a zero – see Fig.b. Unfortunately, there’s no good way to stream multiple bytes of data out using the MSSP module; it only has a single buffer register (SSPxBUF) and you can only initiate a transfer by writing a byte to it when the peripheral is not already transmitting data. We solve this by setting up a timer (Timer 4) to run from the main oscillator with a period of 82, which is just a tiny bit longer than it takes the MSSP peripheral to transmit one byte. By initiating a series of byte transfers in the interrupt service routine (ISR) triggered by Timer 4, we create an effectively uninterrupted stream of data to update all the LEDs with minimal processor overhead. Excluding the setup code, that means we only need the following code to update all 80 RGB LEDs. ## Additional code required to update all LEDs void __interrupt() isr(void) { SSP1BUF = *RGBBufPtr; if( ++RGBBufPtr == RGBBufEnd ) PIE2bits.TMR4IE = 0; PIR2bits.TMR4IF = 0; } static void StartTransmission (unsigned char* start, unsigned char* end) { RGBBufPtr = start; RGBBufEnd = end; while( PIE2bits.TMR4IE ) ; TMR2 = 0; TMR4 = 0; PIE2bits.TMR4IE = 1; } This is the basic code used but it lacks the 280μs reset timer after each transmission, implemented with a separate timer peripheral. siliconchip.com.au Australia's electronics magazine Tiny LED Xmas Tree 54 x 41mm PCB SC5181 – $2.50 Tiny LED Cap 55 x 57mm PCB SC5687 – $3.00 Tiny LED Stocking 41 x 83mm PCB SC5688 – $3.00 Tiny LED Reindeer 91 x 98mm PCB SC5689 – $3.00 Tiny LED Bauble 52.5 x 45.5mm SC5690 – $3.00 Tiny LED Sleigh 80 x 92mm PCB SC5691 – $3.00 Tiny LED Star 57 x 54mm PCB SC5692 – $3.00 Tiny LED Cane 84 x 60mm PCB SC5693 – $3.00 We also sell a kit containing all required components for just $15 per board ➟ SC5579 December 2025  51