Fullscreen HTML5Med Res (??MB)HTML4

Dual Mini LED Dice This article is a blend of the old and the new. It’s similar to our May 1994 Dual LED Dice design but has been updated to use more modern parts (still with discrete logic) and runs from a 3V coin cell. As we have used mostly SMDs (on the larger side), it will easily fit in your pocket. Project by Nicholas Vinen T his small board ‘rolls’ two sixsided ‘dice’ each time you activate it, giving you a pair of random numbers in two different colours. It’s small and light at just 60 × 28 × 15mm so it’s convenient to use. You could even build two or three for games that require more than two dice to be rolled. I am sure there are plenty of dice apps on smartphones these days, but there’s something pleasing about a design based on old-fashioned discrete logic with Das Blinkenlights (in this case, 14 LEDs). I also think it’s interesting to have such simple circuitry that does a useful job and only draws a few milliamps. It even switches itself off automatically, so the small coin cell should last a long time. One thing I did to make it a little more interesting is add the option of triggering the dice roll with a vibration sensor. That way, you can shake it to roll! That part is optional, but it’s pretty fun as you can just pick it up and quickly get some random numbers. Coming up snake eyes My original plan was to shamelessly copy borrow the May 1994 project circuit by Darren Yates, change the parts to make it run from a lower voltage and redesign the PCB to be smaller. However, I quickly ran into a problem. He had used four 4000-series logic ICs: two 4015 dual 4-bit shift registers and two 4093 quad schmitt-trigger NAND gates. The shift registers kept track of the state of the dice and also did some of the ‘decoding’ to drive the LEDs (more on that later). The NAND gates formed the oscillators to ‘roll’ the dice & performed some logic to always keep the shift registers in valid states. There are direct equivalent 74-series logic chips to the 4093 NAND gates, such as the 74HC132, which would run from a 3V supply. However, I could not find any such equivalent of the 4015 dual 4-bit shift registers, at least, not at a reasonable price. There are 74-series shift registers but single eight-bit shift registers seem to be much more common/popular than dual 4-bit types. So, while it might be possible to base a new circuit off the old one, it would make building it quite expensive, which I thought was against the spirit of the project. I wanted to have cheap kits, under $20 each, to make it a fun device that you can build several of if you are so inclined. So, back to the drawing board then, to come up with an equivalent circuit using more modern (readily available and inexpensive) parts. While I was revising the design, I thought I would see if I could come up with a way to do it with fewer chips. Spoiler: my design uses just three to do the same job (with two spare logic gates!). But before we get to that, let’s look at the problem I had to solve. Of course, I considered using a PIC to do this, but what’s the fun in that? The software would be trivial and the resulting board would be tiny and pretty cheap to build, but it would be a ‘black box’. Keeping with the discrete logic means that anyone can understand how the circuit works. Rolling the bones There are basically four things a battery-powered circuit needs to do to emulate rolling two dice: 1. Switch on when the button is pressed (and switch itself off some time later, so you can’t forget) 2. Trigger two oscillators when the button is pressed, each The SMD and through-hole LED prototypes. Both versions have the option of a white or black PCB. The black PCB is showing the dice rolls four & three, while the white PCB shows six & one (although it’s only faintly visible due to the camera flash). 44 Silicon Chip Australia's electronics magazine siliconchip.com.au of which increments the number on a die, going through the sequence 1, 2, 3, 4, 5, 6, 1... with decreasing frequency, so it eventually stops on two numbers. 3. Keep track of what number each die is currently showing. 4. Convert that number (1-6) into a pattern of LEDs akin to the dots on the face of a traditional die. For #1, I decided on a trick we’ve used a few times in previous projects: a Mosfet with a capacitor and parallel resistor between gate and source, plus a second resistor (and in this case, diode) to pull the gate up when the button is pressed. The RC time constant of the first two components sets the maximum time the Mosfet will remain on, powering the circuit, before switching itself off. The advantage of this approach is its simplicity and low cost, requiring just one small Mosfet (as the circuit’s current requirements are low) and a few passives. The disadvantage is that it switches off by slowly lowering the supply voltage. That means the dice LEDs fade out rather than just switching off, but I can’t see the harm in that. That gives you a bit of warning that it’s going to switch off! For #2, I copied the design from the May 1994 circuit, where the same button that switches the unit on also charges up a pair of capacitors that control two oscillators using schmitt-trigger NAND gates as inverters. The voltage on the capacitor affects the oscillator rate, so they slow down and then stop when you release the button. I made one change here; the original circuit used two capacitors of identical values and relied on the fact that no two capacitors will be exactly the same value to cause the oscillators to desynchronise, so you don’t get the same numbers on each die. I found that did not work too well – in one test, I had 20 rolls in a row where both dice gave the same number! I think this was because the oscillators were close enough that feedback through the power supply was locking them together. Using two different values for the capacitors fixed that, and I think it’s more pleasing that the ‘dice’ stop at different times; just like real ones. #3 and #4 are where the designs really differ, and this is where I was able to save one IC. Unlike the 1994 design, where state-keeping and decoding were somewhat muddled siliconchip.com.au together, my design keeps them mostly separate. To keep track of the state of each die, instead of using shift registers, I am using the registers in a dual 4-bit counter IC, the common 74HC393. Normally each counter will go from 0 to 15 and then back to 0, repeating forever, with the counter incrementing on each clock pulse. However, we want it to roll back to zero after five, so it has six discrete states. We achieve that by creating a ‘crude’ AND gate for each counter out of a dual common anode schottky diode. We connect the cathodes to the O1 and O2 outputs, the anode to the CLR input and pull the CLR input up with a resistor. Fig.1 shows how the O1 and O2 outputs both go high for the first time when the counter reaches 6. It is at this point that the diode stops conducting, allowing the pull-up resistor to assert the clear input, causing the counter to reset to zero. When it resets, O1 and O2 go low, so clear is immediately de-­ asserted. This causes the counter to go 0, 1, 2, 3, 4, 5, 0, 1, 2, 3... LED driving So we have our die states and we can roll them, but how do we drive the LEDs? I spent a couple of hours pondering how to convert the O0-O2 outputs of each counter to the six required LED states that are shown in Fig.1, trying to find the absolute minimum of low-cost logic to do it. The logic required can be minimised by driving some LEDs from one end, with others driven at both ends. By driving the LED from both ends, we effectively get a ‘free gate’, because it will only light in one of the four possible states of a pair of digital outputs. It will light with the anode pulled high and the cathode low. In two other states (low/low and high/high), there is no voltage across the LED. In the fourth, it is reverse-biased and will not conduct (at least, not with the meagre 3V we are applying). Complicating things a bit is the fact that our counter doesn’t go from 1 to 6, but from 0 to 5. I considered that we don’t necessarily need the numbers to come in order; as long as all are present and equally likely. However, I figured out a way for them to occur in order, so I kept it that way. The problem with mapping counter values of 0-5 to die face numbers of 1-6 Australia's electronics magazine is that the O1 and O2 outputs change on the counter transitions from 1 to 2, 3 to 4 and 5 to 0 on the counter. That would correspond to die face values of 2 & 3, 4 & 5 and 6 & 1, when those are actually the most similar states (two of them differing only by the state of the middle LED). Instead, I decided that a counter value of zero should show six on the die face, with the other five values (1-5) mapping to those same values on the die face. That makes decoding much easier, but keeps the numbers in sequence (6, 1, 2, 3, 4, 5, 6, 1, 2...). Circuit details Having decided on that, we can immediately sort out the central LED. It is always lit for odd numbers but never for even numbers. As shown in Fig.1, O0 is always high for odd die case numbers and low for even ones, so we just need to connect the O0 output to the middle LED’s anode (via a resistor) and connect its cathode to ground, as shown in Fig.2, and it will light at the right times. Next, let’s consider the two diagonal LEDs that will light initially to show two, then three, remaining on for four, five and six. We could chose either diagonal pair but I have opted for LED2 (upper left) and LED3 (lower right), as per Fig.1. The only die face number Fig.1: how the three binary counter outputs O0-O2 correspond to the counter value and die faces. August 2024  45 where they are off is one; they are on for the five remaining possibilities. The logic required to detect a one from the O0-O2 outputs is O0 AND NOT (O1 OR O2), which gives 1 for a die face of one and 0 for everything else. That can be rewritten as O0 AND (O1 NOR O2). NAND ICs are more common than AND, but that’s OK because using one instead just inverts the result, meaning we get a result of 0 for a die face of one from IC2b/IC1c. We therefore connect this gate output (pin 8 of IC1c) to the anodes of LED2 & LED3, connect their cathodes to ground, and they will light for any die face state but one. So far, besides the counter IC, we just need two NOR gates and two NAND gates for both dice. Two-input logic ICs usually have four gates each, so with one NOR and one NAND IC, we have two of each gate type left. We want to use two NAND gates for our oscillators, leaving us with just two NOR gates. Is that enough to drive the remaining four LEDs? Actually, we don’t need any more logic gates; we’ve already performed all the logic we need! The other two diagonal LEDs, LED6 & LED7, need to light for die face states of four, five and six. That’s the same set of states as for LED2 & LED3, except the ones where the O1 output is high (two and three). Therefore, all we need to do is connect the anodes of LED6 & LED7 to the same point as LED2 and LED3 and connect their cathodes to the O1 output. LED6 and LED7 will therefore light when LED2 and LED3 are, except when the O1 output is high. In that case, both ends of LED6 & LED7 will be at the same voltage. Therefore, LED6 and LED7 are off for values of 1, 2 & 3 and on for 4-6. Finally, we have the middle LEDs on either side, LED4 & LED5. They only come on to show six, when all three digital outputs of the counter, O0-O2, are low. We already have a NOR gate (IC2b) combining outputs O1 and O2; its output will be high only for two die face values, one and six. So all we need to do is eliminate one. So we connect the NOR gate output (pin 4 of IC2b) to the anodes of LED4 Fig.2: the circuit is based on one dual 4-bit counter, four schmit-trigger NAND gates, two NOR gates and a few other bits. 46 Silicon Chip Australia's electronics magazine siliconchip.com.au & LED5, and join their cathodes to the O0 output. They will only light when the anode is high (states one & six) and cathode is low (states two, four and six). Therefore, they only light up for six. That’s it – all LEDs are lit at the appropriate times, and we have two NOR gates to spare! I couldn’t think of anything useful to do with them; I suppose they could have been used to buffer some LEDs, so the NAND gate didn’t need to drive so many, but I found it easier to leave them unused and tie their inputs to GND. Power supply and oscillators 22μF capacitor C6 is usually charged up to the full cell voltage, so Q1’s gatesource voltage is 0V and it remains off. The circuit’s ground is therefore disconnected from the bottom end of the cell, and the circuit is not powered. In case of any leakage, C6 is kept charged by the 10MW resistor between Q1’s gate and source terminals. When the contacts of S1 (tactile pushbutton) or S2 (vibration switch) close, current can flow from the positive terminal of the 3V cell via the two schottky diodes in D5 to two places. One of those current paths flows through a 1kW resistor to discharge C6, raising Q1’s gate voltage to around 3V and switching it (and thus the rest of the circuit) on. Once the switch is released, the 10MW resistor slowly recharges C6, eventually switching Q1 and the rest of the circuit off after about a minute. 22μF capacitor C7 is discharged at the same time, via a second 1kW resistor, but this one charges more quickly, via a 100kW resistor to ground. This produces the voltage that varies the oscillator speed from fast to slow, then stopped, to simulate the dice roll. That voltage starts high when the switch is pressed and then drops. It is applied to the inputs of both schmitt-trigger inverters (IC1a and IC1d) via 1MW resistors, charging up the 47nF & 68nF capacitors from those points to ground. Once those capacitors charge to a certain point, the output of the inverter goes low, discharging the capacitor quickly via D3 or D4. The cycle then repeats. We only use parallel diodes for D3 and D4 so that we can use identical diode parts throughout the circuit, making component sourcing and construction easier. siliconchip.com.au As the ‘dice roll’ voltage drops, the charge current through 1MW resistor drops, so it takes more and more time for the oscillator cycle to complete. When the voltage from C7 drops below the negative-going threshold of schmitt-trigger inverters IC1a and IC1d, they can no longer oscillate, and the dice display remains static until a switch contact closes or the unit powers down. The reason we have both C6 and C7 is that we want C6 to charge more slowly, so the unit stays on for a while, but C7 charges fast so the dice roll completes within a couple of seconds. There is a 22μF capacitor across the coin cell to improve its surge current capability, plus a 22μF bypass capacitor for IC1 and 100nF bypasses for IC2 and IC3. A high-value bypass capacitor is used for IC1 because we don’t want voltage variations due to different LEDs lighting to affect the oscillators too much, or that could bias the dice rolls (increasing the chance of them stopping on certain numbers). The different value oscillator capacitors (47nF & 68nF) ensure the oscillators run at different rates, so there is no relationship between the number shown on the two dice. LED colours You could use the same colour of LED for both die faces but we think it’s helpful to have them be different colours. For example, if two people need to roll one die, you can assign them each a colour and roll them together. Still, it’s up to you. We chose blue and red because they both have a high efficiency and give similar brightness with 1kW current-limiting resistors running from a 3V supply. The red LEDs do draw a little more current, as they have a lower forward voltage, but both are pretty economical on power. The blue LEDs are quite bright at about 0.5mA while the red LEDs are similar at around 1.2mA. We tried green LEDs and they barely lit up with the 1kW series resistors running from 3V. We considered lower resistor values but that would put quite a strain on the button cell. Another colour that could work well is white. Yellow or amber LEDs might work well if they are high-efficiency types. A note on vibration sensors One of the biggest challenges during Australia's electronics magazine the development of this project was finding vibration sensors that actually worked! One of the most common such devices is the SW-18010P, which we have used before. However, it turns out that there a lot of dud/fake/counterfeit/ badly made devices around sold under the name SW-18010P. So you have to make sure you get them from a reputable supplier. The first lot of SW-18010Ps we got were complete duds, despite getting them from a supplier who had sent us good parts previously. They have tiny writing on the black body, as shown below. While they showed some signs of life, you had to shake them at Earthquake Magnitude 10 level to get any sort of switch closure and it seemed very inconsistent. So in the bin they went. Thinking that maybe the SW-18010P was not a good part to use, we looked for alternatives and found several likely ones: the SW-200D, SW-420D and SW-520D, all described as “highly sensitive vibration switches”. We duly purchased some of each, and were shocked upon receiving them to find that they were all tilt sensors, not vibration sensors! It’s easy to tell that because you can hear a ball rolling inside them when you tip them, and they have a high resistance in some orientations and a low resistance in others, even when static. So they clearly were not suitable. Finally, we found a seller online who actually supplied us with SW-18010P sensors that worked. As you can see in the photo, they have a slightly lighter body and larger writing. DigiKey also sells the AdaFruit version of this part (Cat 1528-2158ND) which would be a good option if you need to buy it yourself. Still, our kits will come with parts that we’ve checked and found to be working, so if you build this from a kit, Two different SW-18010P vibration sensors we purchased. We found that the ones with smaller writing on the side were highly unreliable! The ones shown at the top work much better. August 2024  47 The underside of the SMD (left) and through-hole (right) versions of the Dual Mini LED Dice use the same components. There is also a Nylon screw used to secure the coin cell, to reduce the risk of a child getting a hold of it. you shouldn’t have to worry too much about the sensor being functional. By the way, the less-sensitive SW-18015 and SW-18020 devices are probably no good because even the SW-18010P is barely sensitive enough (you have to give it a pretty firm shake to activate it). While the vibration sensor makes it a very fun device to use, it it is a bit of a gimmick. Even though you have to shake it fairly hard to get a good roll, accidental triggering is still a problem. For example, if you transport it in a car, it will roll the dice if you go over a pothole or big bump in the road. If you keep it in a pocket, it could be triggered while you walk, wearing down the battery. If you’re playing a game and depend on a good roll, we suggest you use the pushbutton to roll the dice as it seems to give better randomisation. Still, as long as you make sure you give it a good shake, it seems to work well enough, and it certainly will wake it up from sleep reliably. Construction Despite the design being mostly SMD based, we’ve chosen to use 3mm through-hole LEDs as we think they look more like the coloured dimples on a die face and, as their lenses project above the tops of the SMDs, they stop the other components on the board from detracting from the LED display. We have produced an alternative PCB (coded 08103242) that uses SMA/ M3216/1206 (imperial) sized SMD LEDs instead, for any constructors who might prefer the slimmer result. We could have designed a PCB to accept both but then we think it wouldn’t have looked as good when using the 3mm through-hole LEDs. Both PCBs measure 59.5 × 26mm and the overlay diagrams are shown in Figs.3 & 4. Components mount on both sides of the PCB. The top side mainly has the LEDs and their current-­limiting 48 Silicon Chip resistors, while all the ICs and the battery are on the other side. Once assembled, the whole thing can be encapsulated in a length of clear heatshrink tubing for protection. We recommend that you start by mounting all the SMDs on the side of the PCB with the LEDs (the ‘front’). The resistors will be labelled with codes like those shown in the parts list; you may need a magnifier to see them. The capacitors will not be labelled so don’t get them mixed up once you remove them from their packages. There are various ways to solder these components but the way we assembled the prototype was to put a little solder on one pad then, holding the part with tweezers, slide it into place while heating that solder. We removed the iron and let it solidify once the part was centred on its pads. We then checked its alignment and, if it was off, reheated the solder and gently repositioned the part with tweezers. Once it was nicely centred and flat on the board, we soldered the opposite pad, ensuring we added enough solder for it to flow onto and adhere to both the pad and part. We then waited for that to solidify, added a tiny bit of flux paste to the initial joint and heated it with the iron tip to reflow it. Repeat until all the passives are in place on the top side. Next, mount Mosfet Q1 (SOT-23) towards lower left. Use a similar technique but this time there are three pins to solder. Follow with the other SOT23 package devices on the top side, diodes D3 through D5. If you are building the board with SMD LEDs, fit them next. Don’t get the different colours mixed up or it will look odd; use all the same colour LEDs for each die face. Ensure the cathodes are orientated as shown for LED1 in Fig.4. You can check this with a DMM set on diode test mode. Carefully touch the probes to the LED leads. When it lights up, the red probe is on the anode and the black probe on the cathode. Now is a good time to clean any flux residue off this side of the board with isopropyl alcohol, methylated spirits or (ideally) a specialised flux cleaning formula. After that, flip the board over. Parts on the other side The three ICs mount on this side. All three are in similar 14-pin SOIC packages, so don’t get them mixed up, and make very sure that you identify pin 1 and locate it as shown on the underside overlay. It’s difficult to remove and refit an SMD IC unless you have a hot air station! Use a similar technique as before, tacking one pin in place and checking that all the pins are aligned over their pads before soldering the other corner pins, then the remainder. You can add a little flux paste along both rows of pins and drag solder them, or just touch a soldering iron loaded with a little solder to each pin and the flux should draw it onto the pin and pad. Don’t be too concerned if you accidentally bridge two pins. Once all pins are soldered, check for bridges and, if you find any, add more flux paste to those pins and use a piece of Both versions of the LED Dice (shown at actual size) can be covered with heatshrink for protection. You can then use it via the pushbutton, or by shaking it (if you have mounted the vibration sensor). Australia's electronics magazine siliconchip.com.au solder-wicking braid to draw off the excess solder. Once all three ICs have been soldered, clean off the flux residue and check that all the solder joints are good with a magnifier, and verify there are no bridges. You can then fit the two remaining diodes on this side, then the three 100nF capacitors and two 100kW resistors. Clean off any new flux residue, then flip the board back over and solder the tactile pushbutton in place. Try to get it straight so it looks neat. If using the through-hole LEDs, now is a good time to solder them in. With the board right-side up (the side that the LEDs sit on), the anodes (longer leads) all go towards the top, and the flat side of the lenses to the bottom. Insert each LED fully, then solder and trim the leads once you are sure it is sitting flat on the PCB. Return to the underside of the board and tin one of the rectangular cell holder pads near the edge. Add a smear of flux paste onto both of those rectangular pads. Rest the cell holder in place and make sure the entry side is facing the edge of the board (if you’re unsure of the correct orientation, consult our photos). Add a bit more flux paste on top of the two tabs that rest on the PCB. Once you’re sure it’s lined up correctly, gently press it down and feed solder onto one of the tabs. You may need to turn your iron up due to the thermal mass of the metal holder. If you will be using the vibration switch, leave it off for now as it’s easier to test the circuit without it. Testing If you have a current-limited bench supply, you can set it to 3V/50mA and connect it using clip leads. Clamp the red alligator clip to the metal shell of the cell holder but make sure it isn’t touching any other components or tracks on the PCB. Carefully clip the black one to the edge of the PCB near the cell holder so it contacts the round pad under the holder but nothing else. Switch the supply on. If you don’t have that, you can just use a lithium coin cell. They can’t deliver a lot of current and it’s easy to temporarily slip one into the side of the cell holder, making enough contact to power the circuit but allowing you to quickly pull it out if something seems wrong. Note that if you use a coin cell, the siliconchip.com.au Figs.3 & 4: we didn’t find the coin cell shorted the adjacent LED leads, but make sure you trim them close to the PCB. For the SMD version, the LED cathodes all go towards the bottom of the PCB. circuit will take a little while (probably 60s) to settle. The LEDs may be dim at first but should get brighter, assuming you are using a fresh cell. When power is applied, you should Australia's electronics magazine see the LEDs immediately light up and the dice roll. If that doesn’t happen, or the circuit draws more than 20mA, switch it off check for incorrectly placed or soldered components. August 2024  49 Parts List – Mini LED Dice 1 double-sided PCB coded 08103241, 59.5 × 26mm ● 1 SMD 20mm coin cell holder (BAT1) 1 CR2032 lithium coin cell 1 2-pin SMD tactile pushbutton (S1) 1 SW-18010P vibration sensing switch (S2) (optional) 1 75mm length of 30-40mm diameter clear heatshrink tubing 1 M2 × 6mm Nylon panhead machine screw 1 M2 Nylon hex nut Semiconductors 1 74HC132 schmitt-trigger quad 2-input NAND gate CMOS IC, SOIC-14 (IC1) 1 74HC02 quad 2-input NOR gate CMOS IC, SOIC-14 (IC2) 1 74HC393 dual 4-bit binary counter CMOS IC, SOIC-14 (IC3) 1 AO3400 30V 5.8A N-channel logic-level Mosfet or equivalent, SOT-23 (Q1) 7 blue 3mm high-brightness diffused lens LEDs (LED1-LED7) ● 7 red 3mm high-brightness diffused lens LEDs (LED11-LED17) ● 5 BAT54A dual common-anode schottky diodes, SOT-23 (D1-D5) Capacitors (all SMD M3216/1206 size 50V X7R unless noted) 4 22μF 6.3V 2 100nF 1 68nF 1 47nF Resistors (all SMD M3216/1206 size 1% unless noted) 1 10MW (code 106 or 1005) 2 10kW (code 103 or 1002) 2 1MW (code 105 or 1004) 16 1kW (code 102 or 1001) 3 100kW (code 104 or 1003) Substitutions for SMD LED version (replaces the parts marked with ●) 1 double-sided PCB coded 08103242, 59.5 × 26mm (instead of PCB coded 08103241) 7 blue 3mm high-brightness SMD M3216/1206/SMA size LEDs (LED1-LED7) 7 red 3mm high-brightness SMD M3216/1206/SMA size LEDs (LED11-LED17) 1 Mini LED Dice kit with through-hole LEDs (SC6849; $17.50) 2 Mini LED Dice kit with SMD LEDs (SC6961; $17.50) Each kit includes everything in the parts list, except the cell. Price does not include postage. A common cause of faults is a bridge between IC pins that’s near the body of the IC, making it hard to spot. Assuming it’s working, check that both dice show valid numbers (refer to Fig.1). Press S1 and roll the dice, then check again that the states are valid. Repeat until you have seen all six numbers on both dice. If any of the dice don’t look right, check if it’s because one or more LEDs are not lighting. If so, they might be connected backwards, have bad solder joins or (in rare cases) be duds. If all the LEDs are lighting but some of the patterns are wrong, check for solder bridges between IC pins or between components. About 30 seconds after pressing S1, you should notice the LEDs fading out. Typically the blue ones will fade out and switch off before the red ones due to their higher forward voltages. After about 90 seconds, the LEDs should be off and the circuit is in a low-power state. Pressing S1 should switch it back on and roll the dice again. 50 Silicon Chip Note that it’s possible to get a short roll with a short press of S1. The results should still be random, but if you want to be sure, hold it down for a half a second or so rather than just pressing it. Final assembly If you are fitting the vibration sensor, remove the cell and bend its leads at right-angles to fit the PCB pads. We suggest doing this with two pairs of fine-nosed pliers to avoid damaging the sensor by applying too much force to the lead where it enters the sensor body. Lay it over the rectangle in the top-left corner of the board, solder it in place and trim the leads. Now reinsert the cell and shake the board. It should switch on and roll the dice. They should roll every time you shake it. Insert a short Nylon M2 machine screw through the small hole in the PCB, with the head next to the coin cell, and add a Nylon hex nut on the back. Do it up tightly, then trim off the Australia's electronics magazine excess screw shaft length with side cutters. While it is almost impossible for children to remove the coin cell (due to the holder’s tightness), it provides an extra layer of safety against especially keen toddlers. Finally, slip a length of ~35mm diameter clear heatshrink tubing over the whole assembly, shrink it down (try to spread the heat out, rather than heating just one area) and trim the ends. That will protect it from moisture, dust, shorting against anything metal etc. To change the cell, cut it off and shrink on a new piece. You can use the board without the heatshrink tubing but be aware that, as parts of the circuit operate at fairly high impedances to improve the battery life, your skin resistance (which can be well under 100kW) can mess with its operation. So it’s better to sleeve it. I noticed when I encapsulated the prototype, because the board got quite hot, it activated and the dice started rolling really fast. It went back to normal when it cooled down. I put this down to increased leakage through the Mosfet due to heat, providing enough current for the circuit to run, along with changes to the schmitt-­ trigger thresholds affecting the oscillator speed. Also, if you are using the vibration sensor, its operation could be affected if it is squeezed too tight. I noticed a slight reduction in sensitivity but that could probably be fixed by adding a small slit in the tubing near the sensor to relieve the pressure on it. Alternatively, try to avoid shrinking the tubing fully in that area. Conclusion We aren’t sure whether the randomness of our Dual SMD LED Dice is sufficiently good to run a tournament, but it should be fine for casual game playing and it’s a conversation piece compared to regular dice. It also demonstrates what you can achieve with some very simple digital logic! If using the vibration sensor, it probably isn’t a good idea to keep it in a bag, a pocket or a vehicle as it might use up its battery quite quickly. SC Coin Cell Precautions Even though we have added protections such as the locking screw, it is best to make sure that children do not use this device unattended. siliconchip.com.au