Puny Induction Heater That Can Act As a Terrible Vape

I designed from scratch a battery-powered (3 or 4 LiPo cells connected in series) induction heater. Instead of using popular Mazilli ZVS flyback driver, I decided to base the design on the H-bridge (2x IRF520 & 2x IRF9520) that is controlled by the ATtiny13A microcontroller, 4 NAND logic gates, and 2 discrete MOSFETs. Frequency and heating power is selected with the use of two potentiometers.

Control of power transistors is extremely simple, and alternately connects one of the coil’s terminals to supply voltage and the other one to the ground, and vice versa. PWM signal with frequency of 100 Hz is generated. When it is low, both of coil’s terminal are connected to the supply voltage. Circuit contains a toggle switch that disables voltage regulator and 5V components. When no signals are generated by the microcontroller (main coil driving frequency with 50% duty cycle & 100 Hz PWM), both of the coil’s terminals are connected to the supply voltage.

When current is flowing through the coil in one direction, after a fixed interval H-bridge is switched to the alternative configuration, that is supposed to reverse the direction of the current. For some time current will flow through the coil in the same direction, as energy is stored in the magnetic field, preventing instantaneous changes of the current flowing through the inductor. LR circuit can be used to model this behavior. Regenerative braking is based on switching transistors placed between power supply and inductors in a way that directs noninstantaneously changing current to flow into power supply. I do not know how current behaves in real life, as I do not own equipment that would allow precise measurements.

This induction heater is quite weak and inefficient. Power transistors heat up like crazy and I was only able to make portions of paperclips glow red. Nevertheless circuit works and possibly can be tweaked and improved (by making current flow more similar to simple sine wave or introducing zero voltage switching for instance). Only tools, materials, and components already present in my house were used in this project, nothing was bought specifically for it.

I installed both small ceramic and electrolytic capacitors parallel to the H-bridge, which might not be required, as circuit is powered by large battery (but if power supply would be blocking reverse current, then this capacitors are important). Interesting video about inadequacy of using small and cheap ceramic capacitors in filtering voltage ripples can be found here (as I said, I only used components that I have already owned). Another video from which you can learn about certain ceramic capacitors that decrease their capacitance with voltage increase, and other interesting topics.

I made mistakes while working on this project. I was hopping to achieve PWM signal frequency close to 100 Hz, but I forgot to change value of F_CPU macro, so it is closer to 800 Hz (ADC is overclocked too, but high resolution was not important). I managed to burn one of the transistors while testing new coil with low amount of turns on a fully charged 3S LiPo battery with duty cycle of PWM signal set to 100%. I could not make any repairs yet. I have made some modified PCB designs however. First of them has changed orientation and position of power MOSFETs that allows mounting of hetsinks on them (transistors might be mounted higher than usually due to resistors being placed next to them, use heatsink silicone grease that will not “intrude into a semiconductor device”). If you are going to use battery power, you may also want to measure supply voltage and reduce duty cycle of PWM signal when battery is fully charged. I created PCB design that accommodates appropriate voltage divider that connects to the RESET/ADC0/PB5 of the microcontroller and has transistors that cutoff this divider when it is not in use (if you do not care about additional power drain, you can just connect R1 directly to +BATT). I have not written appropriate program yet. I might post further hardware updates.

I initially planned to use 3S 1300 mAh LiPo battery as main source of power, but 4S LiPo will work better, because the same amount of current can be supplied to the coil operating at higher voltage. Power supplies with output voltage of 11 - 17 V providing 2 A or more can also be used.

You can download source code, schematic, PCB, etc. for this project here:

induction_heater_all_files_v1.zip, mirror, SHA256: 768ff1861491b7296c31c7251e9f749537f9bb1b9666b9122d743853532042ca

Parts and materials:

• ATTINY13A-PU microcontroller (8-PDIP)
• 8 pin DIP IC Socket (300 mil body width, 0.1 inches [2.54 mm] pin pitch)
• CD4011BE quad NAND gate (14-PDIP)
• 14 pin DIP IC Socket (300 mil body width, 0.1 inches [2.54 mm] pin pitch)
• 78L05 voltage regulator (TO-92)
• 2x BS170 transistor (TO-92)
• 2x IRF520 transistor (TO-220AB)
• 2x IRF9520 transistor (TO-220AB)
• 2x 6x6mm tactile switch buttons (THT)
• 3x 1k ohm trimpot 6mm
• 3x 10k ohm 1/4W resistor
• 6x 100 ohm 1/4W resistor
• 2x 1M ohm 1/4W resistor
• 4x 100nF ceramic capacitors
• 100uF electrolytic capacitors
• 10uF electrolytic capacitors
• miniature toggle switch SPDT ON/ON (main body size : 13 x 10mm)
• double sided copper clad board (45.72 x 93.98 mm at least; if you want to make one of modified PCBs, you need board that has size of 45.72 x 101.60 mm or more)
• few pieces of copper wire (UTP cable can be a great source of wires)
• enameled copper wire, diameter 0.4 mm
• pipe with outside diameter of 6 mm made from glass or other nonconductive and temperature resistant material that will become bobbin of the coil (mine was a part of one hitter [cigarette holder])
• relatively thin insulated braided conductor that can fit into holes in PCB (some connectors that can mate with battery or power supply may be nice, but I just bent tinned conductors multiple times)

Tools:

• diagonal cutter
• pliers
• flat-bladed screwdriver
• tweezers
• utility knife
• file
• center punch
• hammer
• cyanoacrylate glue
• 2000 grit dry/wet sandpaper
• paper towels
• 4x 0.8mm drill bit
• drill pres or rotary tool
• sodium persulfate
• plastic container and plastic tool that can be used to take PCB out of etching solution
• brown packing tape
• multimeter
• soldering iron or gun
• solder
• soldering flux (I used RMA class, flux gel intended for SMT assembly and repairs, that came in 1.4 cm^3 syringe)
• desoldering wire
• helping hand with crocodile clips
• pencil
• laser printer
• glossy paper
• clothes iron
• cream cleaner
• acetone
• rubbing alcohol
• permanent maker
• AVR programmer (standalone programmer like USBasp or you can use ArduinoISP)
• breadboard and jumper wires that will be used to program microcontroller outside of PCB (or any other tool that can achieve this goal)
• 3S or 4S LiPo battery with ~1000 mAh cells and appropriate charger, alternatively power supplies with output voltage of 11 - 17 V providing 2 A or more can also be used

Construction of this device requires use of moderately dangerous tools and may cause harm or damage to property. You build and use described device at your own risk.

NOTE: Following description concerns basic circuit without heatsinks on the power MOSFETs. Alternative variants can be found inside induction_heater_all_files_v1.zip and require larger copper clad board (at least 45.72 x 101.60 mm). Amount of connections on the front side is quite small, so you can use wires instead of troubling yourself with double sided copper clad board (make sure that thicker tracks are replaced by decent wires). All of the holes lie on a 2.54 mm (100 mils) grid (within a certain tolerance, buttons are especially nonconformant), so prefboard could be used to make this project. I used different clothes iron than the last time I was making double sided PCB, and toner spread too much, so I recommend that you don’t use files with small drill holes if you can’t be sure that high resolution can’t be achieved (like I did).

You need to print mirror image of F.Cu (front side) and normal image of B.Cu (back side) on glossy paper using laser printer (without any toner saving settings on). External dimensions of printed images should be 45.72 x 93.98 mm (or as close as you can get).

Cut PCB to the size of printed image, you can add few mm to each side of PCB if you like. Piece of paper used as template that guided pencil was not a glossy paper and was not used during thermotransfer process. I personally like to do it by making a deep row along the whole length of a laminate with a utility knife (you need to cut along the whole length a few times), then repeating the process from the other side. When the rows are deep enough, whole laminate breaks in half easily. You need to perform the process of breaking laminate two times, because you need to have right length and width of the resulting piece. Smaller pieces of laminate can be broken of with the use of pliers (make sure not to scratch copper too much, use protective layer of, for example, paper between pliers and the PCB). Now you should smooth the edges of the resulting board piece with the file.

Next, you will need to clean copper layers using wetted fine sandpaper, then remove particles left by sandpaper with cream cleaner (you can also use washing up liquid or soap). Then clean it with rubbing alcohol. After that you should be very careful not to touch copper with your fingers.

Now its time to cut sheet with mirror image of B.Cu to a more manageable size (leave few cm around the external rectangle) and to put it on top of the clothes iron (toner up). Your clothes iron must not be equipped with safety feature that turns it of when it is in “wrong” orientation, or you need to reset it from time to time. You can hold iron between you thighs, but be very careful that a soleplate is constantly up and does not touch anything. Then, place PCB on top of glossy paper (cleaned side faces toner) and turn iron on (use full power). After short while paper should stick to PCB. You can use piece of cloth or a towel to push the board against the paper and move paper sticking to PCB a little. Wait at least few minutes, until paper will change color to yellow. Unfortunately, you need to determine right time to stop transfer process experimentally, so in case image on the copper has very bad quality, you will need to clean toner with acetone, sand and wash board again and start the whole process from the beginning. I actually kept PCB on top of the soleplate of cheap travel iron for 15 minutes, but it might have been too long, as toner spread too much. To much movement might also be to blame here.

When you think toner transfer is completed, put PCB with paper to water (you can add cream cleaner or washing up liquid) for 20 minutes. Next, rub paper from PCB. If there are places where toner didn’t stick to copper, use permanent marker to replace the toner.

Now you need to mark the centers of four empty spaces in the corners of PCB with a punch. Later those centers will be drilled, and the resulting holes used to align both sides of PCB.

Next, you need to cover front side of laminate with brown packing tape. Mix fresh water with sodium persulfate and put PCB in the etching solution. Try to keep solution at 40°C. You may put plastic container on top of radiator or other heat source. Electric oven could be used to keep constant temperature. From time to time mix solution in the container. Wait for uncovered copper to completely dissolve. When it is done remove PCB from the solution and rinse it in water. Peel packing tape. At this point you may start removing any short circuits with utility knife.

Now, drill four aligning holes using 0.8mm drill (you may want to make initial pilot hole with needle or something similar). Then, drill corresponding holes through the paper with the image of F.Cu using the same 0.8mm drill. When this is done, sand and clean front of PCB. Then put the board on top of flat surface (cleaned copper on top), cover it with glossy paper holding image of F.Cu (toner down) and put four 0.8mm drills into the holes (round part down), to keep paper and the laminate aligned. Now you should gently touch the paper with the tip of hot clothes iron for a short while, so that paper and the PCB stick to each other. Then, remove drills, place iron between your thighs and place paper with the laminate on top of the iron and repeat the procedure of transferring toner. Later soak paper in water in order to remove it and replace any missing toner with permanent marker.

Now you need to cover back side of PCB with packing tape. Then etch back side the same way as you did back side, peel tape, remove the toner with acetone (decent nail polish remover should contain fair amount of it), and start hunting for short circuits.

You also need to drill rest of the holes in the PCB. All holes can be made using 0.8mm drill. Use wetted fine sandpaper to smooth surface after drilling.

NOTE: If you want to make one of modified PCBs you need to have 4 TO-220 heatsinks (the ones with width of 15 mm should be perfect) with appropriate bolts, nuts and thermal interface (more on that in introduction). Battery voltage sensing will require additional regular 100 Ω and 100 kΩ resistors, 2 preferentially high-precision resistors with values of 6.2 kΩ and 15 kΩ, and 2 transistors: BS170 and BC557.

You may start by covering all copper with solder (perform the operation on surface already covered in flux). If after this operation excessive amount of solder is present at some points, remove it with desoldering wire. If any tracks were dissolved in etching solution, replace them with thin wires.

Check if no short circuits are present and electrical connections are conducting, simple multimeter with continuity tester should be adequate. Repeat this at later stages of device assembly.

Vias can be completed by placing wire (for example unused length of component lead) inside the hole, bending it in right way, soldering it from both sides of PCB (final bending may be easier if wire is already soldered at one end), and then cutting unnecessary part.

Insert component leads into appropriate holes. Cut unnecessary parts of the leads, and solder them to the copper pads at the back side.

To mount SPDT toggle switch that is not supposed to be mounded directly to the PCB I soldered wires (leads cut from other components) to the switch leads at 90 degree angle and then connected them to the PCB. Two additional wires (they were extracted from UTP cable) were mounted around other side of the switch. First one end of one of the wire was soldered to PCB, then made to pin the switch with some force by pulling it through other hole in PCB. Other end was soldered when wire was tensioned. Few drops of cyanoacrylate glue were placed at the edges of switch's plastic case.

I strongly recommend soldering sockets for multi-lead ICs to the PCB, instead of soldering ICs directly, as it will allow for easy exchange of them (I managed to destroy CD4011 by supplying voltage of around 19V; one of the sockets used by me was slightly longer than necessary, I removed 2 pieces of metal from it). ATmega13A must be flashed with right program in order to generate necessary signals.

Two braided conductors were stripped from insulation at both ends. On one side only minimal length of insulation was removed that would allow bar conductor to go through hole in PCB and be soldered. On the another end much larger length of insulation was removed, allowing for multiple bends to be made. Exposed conductors were latter tinned. I bent longer exposed parts of conductors multiple times, forming multiple layers of conductor, that can snugly fit into T-plug (Deans) female connectors. Opposite ends were pushed into inner holes in the PCB of J1 and soldered. Into outer holes short piece of another wire was pushed and soldered, that secured main conductors, through which large current will flow.

Connect ATtiny13A to your favorite programmer, you can use a breadboard and jumper wires to do it. Open your favorite AVR dev tool and write one of .hex files to microcontroller FLASH memory (I only tested bugged version). Keep default fuse bits (H:FF, L:6A).

I used USBasp and AVRDUDE, so after correctly connecting VCC, GND, RESET, SCK, MISO, MOSI pins of ATtiny13 to programmer I only needed to execute one simple command to upload one of .hex files:

avrdude -c usbasp -p t13 -B 8 -U flash:w:induction_bug-800hz.hex

or

avrdude -c usbasp -p t13 -B 8 -U flash:w:induction_100hz.hex

When programming was done, both DIP ICs could be inserted into appropriate sockets.

Measure required amount of 0.4 mm diameter magnet wire that will used to create a coil that has a length of 12 mm, and is wound onto bobbin with outside diameter of 6 mm. I found that 2.7 meters of wire (around 20 cm of the length was used to make connections or was not used due to last turn being in the wrong place) is required to achieve decent performance with supply voltage of over 16 V. To make induction heater work better with lower voltages I tried making coil out of 1.4 m of wire, but using fully charged 3S LiPo with PWM set to 100% was enough to fry one of power MOSFETs. I also experimented with longer and shorter wires, but they were useless. Photos in this Step were made when 1.4 meters of wire were transformed into coil.

Start wounding wire around one end of a pipe/bobbin, try to achieve tight helical shape. After few turns are made, secure them in place with a drop or two of cyanoacrylate glue (if you are afraid it might make your fingers sticky, use gloves). Continue producing first layer of coil, and when you reached desired length of 12 mm, secure last turns with glue (you can use toothpick or other similar object to spread glue more evenly). Start making another layer on top of the previous one, progress in opposite direction to the one that was used in previous layer. Secure wire with glue when required. Make as many layers as are needed. If last layer cannot be fully filled and tight without last turn being in middle of the coil, make last layer sparse.

When main part of coil is made, cut unnecessary length of wire, and scratch insulation from straight parts of the wire. Later stiff leads cut from electrolytic capacitor were used to reinforce straight parts of the magnet wire and make extension that will go through holes in PCB. During this process leads were held by crocodile clips (one at a time), that were a part of a helping hand. At first leads were covered with flux and tinned, then long solder joints were made.

Before testing circuit with coil, you can use 2 antiparallel LEDs and resistor (let’s say 10 kΩ, depends on LEDs used) to see if everything is wired right (I actually noticed that one of the diodes light up faster than the other at low PWM, I am not sure if LEDs, MOSFETS or NANDs are to blame).

Finished coil was soldered to the PCB, unnecessary length of (capacitor) leads was removed.

Set both potentiometers half-way, connect battery or power supply, close SPDT switch (it is there to reduce power consumption when battery is connected and device is not in use) and press the FIRE!!! button. If everything was done well, current should be flowing through the coil in both directions.

You can now try heating objects inside coil. The only ones that I was able to make red hot were straightened-up paperclips (even achieving temperatures to low to result in glowing is enough to permanently change color of the paperclip). I found that best heating is achieved when PERIOD trimpot has slightly higher resistance between GND and wiper, than between wiper and +5V, corresponding to the voltage on the wiper being slightly higher than 2.5V, and H-bridge switching frequency lower than 38 kHz. When coil made from long wire is used with low supply voltage, you probably have to set PWM to 100% (5.0V on the wiper). In other cases, it might be beneficial to reduce heating power, and not fry MOSFETs (I encourage you to try one of the modifications that allows for mounting of heatsinks).

Interesting fact is that efficient heating frequency of slightly less than 38 kHz is different than resonant frequency of the circuit. When varying period/frequency I was able to find the “sweet spot” that was characterized by lowest power draw of the device, but it varied with different coils and was not necessarily good at heating metal. When testing a coil made from 2.7 meters of wire I was able to increase power draw significantly by placing multiple paperclips inside the coil (even some sparks appeared near the paperclips). I was not able to replicate this behavior with other coils. There was some change in power draw when paperclip was inserted, but it was not significant. Inserting multiple ones did not have any effect on the current.

The goal of this project was making small portable vaporizer that inductively heats stainless steel mesh onto which pharmacologically active substances were deposited. Unfortunately this mesh (it was cut out from a pipe screen) did not heat up very well, most likely due to being made from material that is discontinuous and non-ferromagnetic. Heating glass pipe externally via flame produced by a small torch or lighter worked much better (red glow could be achieved).

Straightened-up paperclips covered in chemicals extracted from Cannabis sativa plant (some of them may be regulated or illegal in some jurisdictions) worked better, but were far from ideal. Rate of vaporization was extremely small. Tip of of the paperclip was hard to heat up. Portion of substances was left unvaporized (possibly chemicals closest to the metal left some residue that acted as thermal insulator that prevented rest of the chemicals from heating up). Chemicals were extracted by combining isopropyl alcohol with dried inflorescences (both ingredients had low temperature because they were taken out of the freezer), and mild crushing of plant matter with fingers protected by glove for few minutes. Later liquid was strained through tissue paper installed in funnel made from part of plastic bottle, and majority of of isopropanol was evaporated (placing container with mixture on the other container with hot water speeds up the process). Then pipette was used to transfer highly concentrated liquid into a cap that was part of the mechanical pencil lead package (other narrow container can be used as well). Cap was secured in place by making a hole in piece of cardboard, sticking cap into it and taping cardboard to another piece of cardboard. Multiple straightened-up paperclips and preheated (their surface changed color due to high temperature exposure) were placed in the cap filled with mixture, and then after some time alcohol evaporated, leaving substances on the tips of paperclips and walls of the cap.

This induction heater would work much better as vaporizer if substrates were deposited on a different substrate. Thin-walled pipe made from ferromagnetic steel would work better than paperclip, allowing eddy currents to flow freely. Drilling holes or dents into pipe would make it more mesh-like, possibly making deposition process easier. Metal pipes could be loaded inside chambers of a cylinder, that resembles the one in a revolver. Similar concept was previously used in smoking pipes, other mechanisms resembling those in automatic/repeating firearms could be used as well. Cylinder may be made from weak material, and possibly it could be single use. Chambers would align with the bobbin, and a mechanism would insert the pipe into the coil and hold it there. This mechanism could consist of 2 two objects resembling tweezers, one located slightly behind the other and rotated by 90 degrees. Those “tweezers” would move along the axis that goes through the center of coil and one of the chambers. Parts of those “tweezers” would be made from spring steel, allowing them to extend when they are near the pipe and hold pipe from the inside. Inside out tips of the “tweezers” could be made out of ceramic or other nonconductive and nonferromagnetic material. When “tweezers” would move away from pipe, they would encounter narrowly spaced rollers that would force them to contract and stop gripping the pipe. Some locking mechanism would be needed, possibly consisting of a single screw. Permanent magnet could be used to prevent pipe from slipping inside the coil and stem (extension of the bobbin, possibly replaceable, and being positioned inside coil, allowing for easy cleaning of any condensed substances).