Search our partners

Find the latest content on electronic components. Datasheets.com
Please add exception to AdBlock for elektroda.pl.
If you watch the ads, you support portal and users.

reverse engineering of HI-END class soldering station

semselectronics 12 Jan 2021 14:55 513 1
  • #1
    Level 1  
    reverse engineering of HI-END class soldering station

    Introduction to the course

    In order to understand what kind of device we will be designing today, let's first briefly recall what soldering stations generally are, and how they differ from each other.

    The entire lower price segment of such equipment, as you might guess, is captured by Chinese brands, mostly copying the rather successful design of the Japanese hakko soldering iron. The principle of operation of both the original and numerous copies is very simple: a nichrome or thin-film heater transfers heat to a removable stinger, the temperature of which is controlled by a thermocouple or a thermistor built into the heater. This is a simple and inexpensive solution, but in Chinese copies, the quality can limp a little: the heater is slightly the wrong size, a little savings on the tip material, and as a result, foil is wrapped on the heater, the original Japanese stinger is ordered from abroad, the connector is changed to a more powerful one... in General, there is something to do.

    Somewhere in the middle of the cost scale are branded soldering stations of well-known Western brands. German ERSA, American Weller, Japanese Hakko, that's all. The principle of operation is essentially the same, but no collective farm is needed here, pleasant buns like a soft silicone cable that does not melt from the slightest touch of a soldering iron come out of the box, and ... Yes, actually not so many buns! Price? Corresponds to the level. 600$ will upset not only a modest home lover to spend evenings for hardware debugging, but even a medium-sized legal entity.

    However, the topic of today's article is not about that. I will tell you about the most real HI-END in the world of soldering stations, namely about induction soldering irons of the American company Metcal (under this brand they are now produced by OK International). In fact, there are several manufacturers of such devices, in addition to the aforementioned Metcal, I still know Thermaltronics, JBC, and even Hakko has a similar model. The principle of operation of the induction heater in such devices is very elegant:

    reverse engineering of HI-END class soldering station

    As you can see, there is no thermal sensor at all, the stinger core is made of copper with a coating of ferromagnetic material, which under the influence of a high-frequency (13.56 MHz) alternating magnetic field heats up, then at a certain temperature, called the Curie point, loses its magnetic properties, and, as a result, ceases to heat up further. When you touch the soldering point, the ferromagnetic element cools down slightly, and the power from the inductor immediately begins to transfer to the tip of the soldering iron. Stingers come with four fixed temperatures, of which only two are actually needed – for lead and lead-free soldering. That's all.

    OKI / Metcal produces several types of induction soldering stations of different prices and with different output power, but the order of amounts in the region of 800$ discourages any desire to touch the beautiful, no matter how beautiful it may be. Well, let's try to save some money?


    Let's formulate it as follows: using only open sources, perform virtual reverse engineering of the original MX-5200 device, and as a result, develop a single-channel source of sinusoidal RF voltage with a peak output power of 80 W, which repeats the functionality of the original soldering station as closely as possible.

    On the Internet, you can easily find a neatly drawn diagram from the Board stations of the previous generation Metcal MX-500. You can't directly use circuit solutions from here, since the output power of this device is only 40 watts, and it doesn't scales in a simple way. However, the old scheme will help us to understand the principles of the major components.

    So, in the document, we see:
    Quartz high-frequency generator with three resonant circuits at the output;
    Pulse step-down converter for powering the generator (1), with an output voltage varying in the range of 17-21 V;
    Feedback circuit that regulates the voltage of the step-down converter (2)depending on the voltage on one of the output resonant circuits of the generator (1);
    Protection unit that turns off the generator (1) when the inductor is disconnected;
    Transformer power supply with 53 V output voltage.

    Let's immediately estimate the General circuit design solutions. For powering the circuit, for example, a toroidal low-frequency transformer is perfect. Although ... we'd better use a resonant LLC Converter based on a rare HiperLCS chip manufactured by Power Integrations: I've been wanting to work with it for a long time. Step-down Converter used to adjust the output power also take a more modern one, see if you can really squeeze five amps out of a case the size of SO-8. But what kind of project is this without Arduino, sketch, and led? Add an STM32 microcontroller and a screen to display the current output power. For simplicity, we will measure the power on the RF generator power line, and take the efficiency into account in the software (or not). The case will take a suitable metal size, it will serve as both a screen and a radiator.

    For direct soldering, Amazon will purchase the Metcal MX-UK1 upgrade kit, which includes a stand and a soldering iron itself (this is essentially just a pen with a wire), as well as soldering cartridges themselves. Historically, it is more convenient for me to work with small parts with the so-called "hoof" (a cone truncated at 30°), and for soldering massive elements it is better to take something wider, more massive, and hotter, so here is my choice: Metcal SMTC-0167 for fine work, and Thermaltronics M7K100 for working with large elements. Yes, the cheaper Thermaltronics Stingers are also suitable.
    reverse engineering of HI-END class soldering station

    While the details are on the way, let's draw a flowchart of the device being designed. Here it is:
    reverse engineering of HI-END class soldering station

    It is very important to immediately say a few words about the feedback between the output of the RF generator and the control input of the step-down Converter. The fact is that after the sting has reached the operating temperature, the generator continues to produce a voltage of a fairly significant amplitude (about 100 V), and this power begins to dissipate on the active resistance of the inductor coil, which, due to the skin effect, is much larger than can be measured with a conventional multimeter. As a result, the tiny coil gets red hot and burns out. To prevent this from happening, the original stations use negative feedback, which reduces the supply voltage of the generator with an increase in the standing wave coefficient accompanying the change in the inductor impedance. The 40-watt version uses a fairly simple method from the patent. US4626767A, and the 80-watt version uses a more complex operating system with a current transformer.

    RF generator

    We will start designing the high-frequency part with the output resonant circuits. Let's take a look at this high-resolution snapshot:
    reverse engineering of HI-END class soldering station

    Here we see three coils wound on yellow toroidal cores, the number of turns – 4, 6 and 7, if you count from left to right. According to the classification of Amidon, yellow indicates a core made of atomized iron with a magnetic permeability of 8.5 (material No. 6). We estimate the size of the rings by measuring the size of the ring with a ruler on the screen, and the size of some known element, for example, the output transistor in the TO-247 case. Apparently, T130-6 magnetic lines are used here; in my opinion, this is some overkill – such large rings are designed for noticeably greater power. But I don't want to be too smart here: of course, I won't use the original American rings, instead I'll order cheap Chinese copies on AliExpress, let's see how they will work (spoiler: everything is fine with them). The calculated inductances were about 180, 400, and 540 NGN, respectively.

    In resonant circuits, capacitors are also attached to the inductors. Their capacities cannot be determined from the photo, but it is easy to find a post in which the pedantic mikeselectricstuff (the author of the previous video) shares his observations (highlighted in yellow):

    reverse engineering of HI-END class soldering station

    If you substitute these values in the spice model, you will notice that the resonant frequencies of the circuits are slightly shifted from 13.56 MHz. The fact is that the closer the frequency is to resonance, the lower the supply voltage needed for the RF generator, and the more current it consumes. In the original, a step-down Converter with a maximum current of 3A was used to power the output stage, so the developers slightly upset the output circuits so that they could increase the supply voltage and reduce the current consumed. We plan to use a five-amp microchip, but this current was also not enough to work in resonance, so we will slightly upset the circuits in the same way. We will select the exact values of capacitances experimentally, based on the maximum supply voltage of the output stage of 22 V and the maximum current consumption at the level of 4 A.

    I note that quite a large amount of power circulates inside the resonant circuits, which tends to be released into the environment in the form of heat. Therefore, in order to increase the quality factor for the coils, we use an enamel wire thicker than 1.25 mm, and we will put several capacitors in parallel.

    Choosing an output transistor is also a difficult topic. When replacing or disconnecting the stinger, the overvoltage can reach quite significant values (300-350 V), but in the original, the developer did not bother much with protection, and put a rather rare, fast and expensive RF transistor IXFH12N50F from IXYS with a maximum drain voltage of 500 V. Of course, we can't afford such a luxury. Let's take an ordinary 200-volt field-effect transistor STP19NF20 worth 1$, and connect a 150 V suppressor in parallel. Perfect! The limiter will slightly trim the tops of resonant emissions, preventing the circuits from swaying too much, and about 10 milliseconds after the load is lost, the protection will stop the generator.

    Due to the large input capacitance and high frequency, it is not possible to control the gate of the output transistor directly using a conventional driver. In the photo of the original Board, a frameless inductance is visible between the two power transistors. This is a well-used little trick: the inductance, together with the gate capacitance, forms a resonant circuit that recycles power in the gate circuit, resulting in a dramatic increase in the efficiency of the preamp. The same circuit simultaneously imposes an unobvious restriction on the output transistor model: its gate resistance must be minimal so that the q-factor of the circuit remains acceptable. Without going into too much detail, we will repeat the solution used by the manufacturer. The value of the inductance is selected according to the maximum efficiency of a real circuit by compressing / stretching the turns of the coil.

    Well, then the circuitry becomes more trivial. The preamp, made on a transistor with a low input capacitance of the IRF510, will be rocked by a dual MAX17602 driver , its speed characteristics are quite good. The MAX17600 or MAX17601 are even better, their outputs could be connected in parallel, but I didn't have such options available, so we'll work with what we have.

    We will set the desired frequency of generation with a quartz resonator. Unfortunately, I was also unable to find quartz at 13.56 MHz for the master oscillator. But it doesn't matter. Let's take a more common resonator at 27.12 MHz, and divide the frequency in two. This is where the microcontroller comes in handy, namely, one of its timers programmed accordingly. I also want to note that only quartz resonators operating on the first harmonic are suitable for direct connection to the MCU. Widespread Russian resonators at 27120 kHz, working on the third harmonic, can only be connected with a crutch in the form of an additional resonant circuit.


    After long and fruitless experiments with Chinese industrial products, it was decided to power the RF output stage from a step-down Converter on the TI TPS54560 chip. To avoid the occurrence of audible beats, we will set the internal generator frequency to approximately 450 kHz, away from the frequency range of the LLC Converter. There is also an option to do the opposite, synchronize the step-down Converter with the generator of the LLC Converter, but then laziness has already begun to make itself felt. We won't do that.

    The tps54560 Converter itself, despite its tiny size, has a fairly large output current, and sometimes it may seem that this is some hitherto unseen miracle in the fight for energy efficiency… But no-the chip needs really good cooling. The demo Board offered by Texas contains two large "earth" polygons with a thickness of 2oz on both sides, and for heat transfer between the layers, six vias are used, located directly under the belly of the chip (where it has a heat-removing contact). This arrangement makes it somewhat difficult to manufacture a printed circuit Board at home, so you probably have to order production in China.

    To power the driver and preamp, we take an unstable voltage of 12 V from the second winding of the LLC Converter. The current consumption of the remaining parts of the circuit will be very small, so for the five-volt controller and the backlight of the LCD screen, as part of import substitution, we will supply a linear stabilizer L7805, designed specifically for use in the national economy, and the 3.3 V line for the MCU will be provided by the ld2985 booger.

    The LLC Converter on the LCS708HG chip will lower the mains voltage to the required 30 and 12 volts.
    reverse engineering of HI-END class soldering station

    I am sure that many readers do not know what kind of animal this LLC Converter is, so I will focus on the principle of its operation in a little more detail. LLC is not exactly an abbreviation, these letters mean "inductance-inductance-capacitance", and, in short, describe the circuitry of connecting the primary winding of a transformer. The fact is that part of the magnetic field lines of the primary winding does not "catch" on the turns of the secondary, as a result of which the so – called scattering inductance is formed-parasitic inductance that is not able to transfer the accumulated energy to the secondary circuits. In conventional reverse-pass converters, this energy has to be dissipated on suppressors or snubber resistors, so transformers (or, more precisely, double-winding chokes) are usually designed in such a way as to reduce the dissipation induction to the lowest possible value. But everything changes when you design an LLC.

    In a resonant Converter, the scattering inductance together with the capacitor connected in series to the primary winding form an oscillatory circuit that performs two important tasks. First, it provides switching of the output high-voltage transistors of the Converter at close to zero voltage (the so-called Zero Voltage Switching mode), which radically reduces switching losses. And secondly, the energy stored in the unbound inductor is returned back to the circuit: now snubbers are not needed, and there is no energy loss either. In the an-55 document Power Integrations explains in detail how to design a transformer in such a way as to increase the scattering inductance (this is necessary to create the correct adjustment characteristic). I, for example, wound the primary and secondary windings away from each other, in two different sections:

    reverse engineering of HI-END class soldering station

    In General, the result of such circuit design refinements is the achievement of very decent efficiency, in particular, the LCS708HG microcircuit installed without a heat sink, with its very small size, provides an output power in the region of 200W! This is indeed an outstanding result, but it can only be achieved if you work exactly at the resonance frequency of the output circuit. And here we are ambushed.

    The fact is that the output voltage is regulated here by changing the frequency, not the duty cycle of the pulses, and this regulation is limited to a very narrow voltage range – about ±15%. Moreover, when the input voltage deviates from the nominal value, the conversion frequency shifts away from the resonance, and the switching of transistors inside the chip becomes "hard", with a loss of ZVS, which is accompanied by a significant heating of them. In fact, we can say that the Converter at the input needs an already stabilized voltage!

    In commercially manufactured products, an active power corrector (APFC) is turned on before the input of the Converter, which, in addition to the actual power correction, also maintains the output voltage approximately equal to 380-390 volts. However, our development is still Amateur, so we can safely close our eyes to a small joint in the form of sensitivity to the quality of mains power. Calculations show that, taking into account ripples on the buffer capacity, the input voltage range is approximately 230 V ± 10%, so if the network parameters do not go beyond GOST, then everything will work. Let's leave it like this for now.

    We will copy the rest of the Converter's circuit design from the Datasheet. Perhaps only the resonant capacitor, which at first glance seems to be a very simple element, needed some attention. And if you've ever wondered what the difference is between polypropylene and polyethylene terephthalate (polyester) capacitors, then you'll know the answer right now: the first tangent of the loss angle is ten times smaller. That is why an attempt to use a cheaper and more compact polyester K73-17 instead of the overall K78-2 (Yes, import substitution is also used here)is accompanied by interesting special effects: the capacitor heats up strongly and starts to crack suspiciously. Interesting.

    HiperLCS series chips require a separate 12 volt power supply. In order not to bother with an additional winding, rectifier and start-up chains, let's go, perhaps, along the most canonical path. We will take the required voltage from a separate miniature Converter on the LNK304 chip. Its key feature is a transformerless design, and only a factory-made penny choke is required from inductive elements. The maximum output current is not very large, on the order of hundreds of milliamps, but the minimum of details and simplicity of design are captivating (and the number of converters per square decimeter of surface begins to unnerve. More converters and Converters!)


    Well, there's just a little bit left. The original station has an LCD that shows something like power output for all the money paid. Let's do a similar thing: take the STM32F030 controller in the most minimal configuration (in the TSSOP-20 case), hang one ADC line to measure the supply voltage of the output stage of the RF generator, and another line to measure the current. In order not to break the "ground" circuit, we will place the resistive current sensor on the positive wire, and to convert the levels, we will use the ina138 microcircuit designed specifically for such cases. which was developed by Burr-Brown in its heyday. To display information, we use a 16x2 text OLED screen manufactured by WinStar. Well, that's actually all. Ah, well, one leg of the processor was left idle. Well, let the led blink. Don't ask me why.

    The controller firmware is written in C using STM32CubeMX and the free version of IAR Embedded Workbench. The program code is very trivial. The main loop for interrupting the system timer reads data from two ADC channels every 300 milliseconds, multiplies them, and displays them on the screen as power digits. At the bottom, the same power is visualized by a bar drawn with custom fonts. When the sting is disconnected, the interrupt handler stops the RF generator's master timer from the output of the load detector. In case the MCU freezes or crashes, hardware error handlers and a watchdog timer are added; CSS (Clock Security System) technology is also used in the firmware, which allows switching to an internal RC generator and restarting the microcontroller in case of attenuation of the main quartz resonator vibrations. The total amount of firmware is 10 Kbytes. I posted the source code of the firmware along with all the other project files on GitHub, the most curious can get acquainted (but do not expect something very interesting there).


    Chokes in the drain circuits of field-effect transistors and a current transformer in the feedback circuit are wound on rings of size K16x8x6 from ferrite of the M50VN brand. The joke about "modify with a file" will be very useful here: the Russian industry, it seems, still has not learned how to make ferrite rings with rounded edges. Enameled wire is suitable for a diameter of 0.6 mm, the number of turns is 15 for chokes, and 2x14 for a current transformer;
    The frameless coil is wound on a rim with a diameter of 5 mm with an enameled wire with a diameter of 0.6 mm. It contains 10 turns and has a length of approximately 10 mm.;
    For the production of a network transformer, we use a miniature W-shaped EFD25 core made of N87 material manufactured by Epcos. We will set the gap in the magnetic circuit by laying two layers of note paper on each side of the core (this is approximately 0.2 mm). For the primary (33 turns) and first secondary (2x6 turns) windings, we use a triple-insulated litzendrate of size 100/46 and 175/46, respectively (here the first number is the number of cores, the second is their thickness according to the American Wire Gauge table). The second secondary, 12-volt winding is two turns of ordinary MGTF wire.

    All calculated data for all elements of the LLC Converter, including the inductance of the transformer windings, are provided in the design file attached to the project, which can be opened using the PIXls Designer application. Also, just in case, I added to the project all the documentation used during the development of the electronic components used, filled in LTspice models of some parts of the circuit, and of course photos, where now without them.

    The result of the above development was the following electrical circuit diagram:

    reverse engineering of HI-END class soldering station

    he circuit diagram and layout of the printed circuit Board were drawn in the DipTrace package, and the Board drawings were converted to Gerber format for sending to the factory. The Board is separated exactly to the size of the used case, for shielding delicate low-current circuits, one layer is completely given over to the "ground". This layout greatly simplifies the production of the Board at home, since precession matching of photo templates is not required here: almost the entire reverse side of the Board can be filled with one solid polygon, and then chamfer the pin holes with a thick drill that does not require connection to the "ground".

    reverse engineering of HI-END class soldering station reverse engineering of HI-END class soldering station

    The RF generator of the design whistles decently into the air, the power elements are very hot, so the choice of housing material is self-evident: of course it will be aluminum. Choose from the catalog of the company Gainta approximately suitable for the size of the finished case G0476. A window for the OLED screen cut in the casing using the Dremel, the case itself will connect directly to the "ground" wire of the power cord along with the wire screen of a soldering iron and ground of the PCB.

    Unfortunately, the idea to connect a more contrasting OLED instead of LCD came to my mind after the order for the boards was sent to the factory. The input CMOS levels of the weh001602agpp5n00001 OLED screen produced by WinStar differ from the standard TTL levels of the LCD, so the feint of ears when +5V is applied to the display controller and its backlight, and logic signals are taken from a microprocessor powered by +3.3 V, does not roll here. Therefore, the screen had to be powered by wiring from the 3.3 V line.

    To reduce the level of interference, interference-canceling resistors with a nominal value of 390 Ohms are added to the "loop" connecting the Board and the screen, and the microcontroller is covered with a copper foil screen. During normal operation, a mating part is put on the programming connector, which attracts debugging pins directly to the" ground", and NRST-through a capacitor.

    In the end, the developed device took on a complete appearance:
    reverse engineering of HI-END class soldering station


    Well, now the most important thing for which all this was started: the feeling of working with the device. It feels as if you are working with a very powerful and very hot soldering iron, while holding a small and light tool in your hands. Is it worth the money and effort? It's hard to say. I'll leave this question open.
    Do you have a problem with Arduino? Ask question. Visit our forum Arduino.