Can anyone confirm or reject the impression I am getting? I have two circuit boards (in inverter welders) with malfunctions, and each unfortunately has an MCU. All diagnostic procedures (tracing and checking signal paths) lead me to pins of 64 or 44 (respectively) pin MCUs.
I am thwarted by my inability to know what signals I ought to see at these most of these pins (apart from power supply and earth pins) because their functions are programmed into the MCU by the equipment manufacturer.
My hope was that it is possible to download out of the MCU its programme and work out what should be happening, but my Google searches suggest to me that this is in practice impossible.
How do we use the diagnostic skills and techniques we have managed with for years on all sorts of circuit boards if there is this impenetrable device on the board?
To illustrate the difficulty, one of the welders (TIG process) only activates its HF (which is supposed to initiate the welding arc) for a very brief pulse, whereas is should persist until the arc is established. This pulse emanates from the MCU, which should be responding to something either sensing the drop in voltage as the arc establishes, or the presence of an arc current. Both are sensed and their signal paths lead in plausibly fault-free ways to the MCU, but without knowing just what the MCU is supposed to expect before halting the HF, I do not see what to investigate. For all I know, it might be sensing some entirely different fault condition, and halting the HF as a protection measure. But how does one proceed logically?
Any thoughts?
I am thwarted by my inability to know what signals I ought to see at these most of these pins (apart from power supply and earth pins) because their functions are programmed into the MCU by the equipment manufacturer.
My hope was that it is possible to download out of the MCU its programme and work out what should be happening, but my Google searches suggest to me that this is in practice impossible.
How do we use the diagnostic skills and techniques we have managed with for years on all sorts of circuit boards if there is this impenetrable device on the board?
To illustrate the difficulty, one of the welders (TIG process) only activates its HF (which is supposed to initiate the welding arc) for a very brief pulse, whereas is should persist until the arc is established. This pulse emanates from the MCU, which should be responding to something either sensing the drop in voltage as the arc establishes, or the presence of an arc current. Both are sensed and their signal paths lead in plausibly fault-free ways to the MCU, but without knowing just what the MCU is supposed to expect before halting the HF, I do not see what to investigate. For all I know, it might be sensing some entirely different fault condition, and halting the HF as a protection measure. But how does one proceed logically?
Any thoughts?
I am certainly no expert at fault finding.
In this case I would -
1 - clean the board (to ensure no muck was causing a leakage path). Also clean any component glue that can become conductive as it decays.
2 - look for dry joints.
3 - look for dodgy elco's.
4 - measure what voltages I could, supply voltages etc.
5 - Google for a manual or info from other users.
I suspect you've already done all of the above.
6 - at this stage I've exhausted my skillset and I'd be looking for a replacement board on Ebay.
Programming MCUs is indeed a one way process, you cannot extract the program.
If nothing else, I've bumped your post.
In this case I would -
1 - clean the board (to ensure no muck was causing a leakage path). Also clean any component glue that can become conductive as it decays.
2 - look for dry joints.
3 - look for dodgy elco's.
4 - measure what voltages I could, supply voltages etc.
5 - Google for a manual or info from other users.
I suspect you've already done all of the above.
6 - at this stage I've exhausted my skillset and I'd be looking for a replacement board on Ebay.
Programming MCUs is indeed a one way process, you cannot extract the program.
If nothing else, I've bumped your post.
My technique would be similar to russc's really. Typically I would use a scope and go around every single pin on the chip and see if what is present is either correct or (using your word 🙂) plausible. Yu are looking for anything that looks unexpected. Data pins with data at correct logic levels for example and not noisy. Clean supplies. Is the reset generator logic level clean at all times. You can not interrogate the data but usually if data is present and the correct levels then its usually correct in my experience.
Try and identify all the inputs to the chip and see what is happening there. Use the scope for every measurement whether DC or AC.
Always work on the assumption that the IC is in reality unlikely to be at fault.
Try and identify all the inputs to the chip and see what is happening there. Use the scope for every measurement whether DC or AC.
Always work on the assumption that the IC is in reality unlikely to be at fault.
Thank you both for your helpful replies. (a) It is good to know for sure that the programme of the MCU is inaccessible. One thing to stop trying! (b) I have hoped I can assume that the MCU is not the faulty component, and it is good to be able to upgrade this hope to something more like an assumption.
russc: Yes, I have done those things, and had a little help from the manufacturer. But they, understandably, prefer to think in terms of replacement boards. As it happens it is regarded by them as obsolete, and I think only a few thousand were built, and I never see anything to do with that welder on ebay, so I am stuck with pursuing component-level diagnosis and repair (of which I am glad because I think it always ought to be possible).
Mooly: Yes, I have made daunting charts of each pin's behaviour under various conditions (e.g. with a dummy load simulating arc present or not present), and tried to make sense of them. And it shows me which pins can be ruled out as suspects. I have traced numerous possibly relevant signal paths, writing notes on everything I do. I feel fairly sure that if I could concentrate twice as hard as I can, holding more in my brain than I can, I would be able to work something out. Perhaps taking a rest from it, then looking through my notes, and having another think might be the best thing now.
Anyway, thank you both.
russc: Yes, I have done those things, and had a little help from the manufacturer. But they, understandably, prefer to think in terms of replacement boards. As it happens it is regarded by them as obsolete, and I think only a few thousand were built, and I never see anything to do with that welder on ebay, so I am stuck with pursuing component-level diagnosis and repair (of which I am glad because I think it always ought to be possible).
Mooly: Yes, I have made daunting charts of each pin's behaviour under various conditions (e.g. with a dummy load simulating arc present or not present), and tried to make sense of them. And it shows me which pins can be ruled out as suspects. I have traced numerous possibly relevant signal paths, writing notes on everything I do. I feel fairly sure that if I could concentrate twice as hard as I can, holding more in my brain than I can, I would be able to work something out. Perhaps taking a rest from it, then looking through my notes, and having another think might be the best thing now.
Anyway, thank you both.
It sounds like you are doing all the right things and then some 🙂
On consumer gear (audio/TV etc) a more common issue in situations like this can be any separate memory chip (typically an 8 pin chip) that the uP accesses and addresses. Unfortunately when you reach a point of having checked as much as you can then you are forced into swapping the uP if only to eliminate it. Generally the uP is not normally at fault though.
On consumer gear (audio/TV etc) a more common issue in situations like this can be any separate memory chip (typically an 8 pin chip) that the uP accesses and addresses. Unfortunately when you reach a point of having checked as much as you can then you are forced into swapping the uP if only to eliminate it. Generally the uP is not normally at fault though.
Again, reassuring to know that the processor or controller is best regarded as beyond suspicion.
I still have the back-up strategy of making the welder useable by just adding a pulse-widening circuit (a 2 or 4 NAND gate IC and a few components) so that the HF runs for say a second. A crude solution. But I don't like knowing that there is a hidden fault swept under the carpet.
My oscilloscope is an old Telequipment (D43 I think, cannot see the nameplate without moving a lot of things), and I am now thinking of connecting its external trigger to trigger on the front edge of the HF pulse, one beam to the pulse and prod all the MCU pins with the other beam to try to find out which pins behave as if they have something to do with halting the HF. Then work back from any such pins. (This pulse is about 12ms long, but HF should keep going for even a few seconds if necessary to get the arc started, perhaps while the torch is being moved into position.)
I might have drawn the entire circuit diagram for this welder by the time I have finished!
I still have the back-up strategy of making the welder useable by just adding a pulse-widening circuit (a 2 or 4 NAND gate IC and a few components) so that the HF runs for say a second. A crude solution. But I don't like knowing that there is a hidden fault swept under the carpet.
My oscilloscope is an old Telequipment (D43 I think, cannot see the nameplate without moving a lot of things), and I am now thinking of connecting its external trigger to trigger on the front edge of the HF pulse, one beam to the pulse and prod all the MCU pins with the other beam to try to find out which pins behave as if they have something to do with halting the HF. Then work back from any such pins. (This pulse is about 12ms long, but HF should keep going for even a few seconds if necessary to get the arc started, perhaps while the torch is being moved into position.)
I might have drawn the entire circuit diagram for this welder by the time I have finished!
Yes, but that doesn't mean its never at fault, just that it really is way down the list.
Good luck 🙂
That's almost always true. But I just re-read news from a couple years ago. Cisco used an Intel chip and it was failing in 18 months. Essentially "all" of a certain network router-box were going to die within a year. It seems Intel had pushed performance/chip just a little bit too far.
Yes, quite. Nothing is 100% reliable, except that you can rely upon it eventually failing. There must be 1000 components in the welder (surface mount mainly), and every one except the MCU is susceptible to thorough checking and easy replacement, so it makes sense to leave the MCU until last.
I bought the welder as needing repair, and after quite a lot of investigation found the fault. Its root cause was a surface mount capacitor setting the frequency of a PWM IC, and its failure mode became clear when I unsoldered it from the board using two soldering irons, one for each end. It came away in two pieces. But its effect had been to operate the inverter at the best part of 1MHz, compared with the 80kHz intended. This had damaged the MOSFET driving the inverter, and overheated associated snubber components. Having put all this right (with considerable help over component values etc. from the manufacturer, Fimer) the welder worked perfectly for a year, but then developed the current problem. It leaves me with a degree of suspicion of SMD components. The rigid mounting seems to me fundamentally unsound given the inevitable differential expansion of board and component materials. On the other hand, they would not be so widely used if they were not adequately reliable. But for how long?
I bought the welder as needing repair, and after quite a lot of investigation found the fault. Its root cause was a surface mount capacitor setting the frequency of a PWM IC, and its failure mode became clear when I unsoldered it from the board using two soldering irons, one for each end. It came away in two pieces. But its effect had been to operate the inverter at the best part of 1MHz, compared with the 80kHz intended. This had damaged the MOSFET driving the inverter, and overheated associated snubber components. Having put all this right (with considerable help over component values etc. from the manufacturer, Fimer) the welder worked perfectly for a year, but then developed the current problem. It leaves me with a degree of suspicion of SMD components. The rigid mounting seems to me fundamentally unsound given the inevitable differential expansion of board and component materials. On the other hand, they would not be so widely used if they were not adequately reliable. But for how long?
I've seen articles in years gone by on that subject. I think it was Electronics world and would be 😱 around twenty years ago.
I am probably worrying about nothing, then. I can't say I have actually come across more than very few such failures. Generally the measured values after a decade or more of use surprise me by their accuracy. I remember when I used to salvage all my components from abandoned televisions etc, and the (in)accuracy of values surprised me in the other direction.
Anyway, back to work now.
Anyway, back to work now.
Hi . Some ideas . This welder has only one welding mode ,which is not working? What i know about such things ,is that they have two feedbacks - voltage and current. Current is get from shunt or current sensing transformer, voltage is from output with resistive divider ,and processor expects to get info from both . If voltage feedback not works , it may give protection mode. Also most of welders have separate ic for pwm control , where those feedback goes . Try to analyse ic's on pcb , and their function ,and try to test of they doing its job . Also frequently fails power parts like mosfets ,but if you saying hf is present for some time ,its not your case.
Hello, Ximikas.
Thank you for your knowledgeable ideas.
The welder is Fimer TIG 165 AC/DC. When set to AC there is not only the initiation pulse of HF, but also a very brief pulse (a single spark) at or just after each voltage reversal. In the AC mode this ignites (and sustains) the arc anyway, despite the inadequate initiation pulse.
So the problem only manifests itself in DC mode (which I use much more often).
I am interested in your information that both a fall in voltage (from open-circuit to arc voltage. i.e. from say 80v to say 15v) as the arc is established, and also detection of a non-zero current, are required to tell the controller to cease the HF. I had supposed that it would be one or the other, but I have traced circuitry (and checked that it functions correctly) from a current transformer in the inverter output all the way to the MCU; and also from a voltage tapping on the inverter's rectifier as far as, and into, the controller board, but not yet all the way to the MCU. So this is my next task; the last stage into the MCU. (I got stuck when disentangling tracks among a couple of op-amps, and was doubting that I was doing something relevant to the problem, but now I will put in more effort and should be able to sort it out.)
The IC doing the PWM control is the well-known UC3524, and its frequency control (timing) capacitor was the initially failed component. It has worked perfectly since having a replacement capacitor. I don't think anything is wrong with any power components. I can get the DC arc established by touching the tungsten on to the work (a highly unsatisfactory method), and everything seems to work perfectly until I next want to start the arc.
Thanks again. I will post again when I have traced and checked the volts feedback all the way to the MCU.
(P.S. I think that the design and manufacturing quality of the welder are both very good. The HF board took quite a lot of analysing. It has an IGBT working from an oscillator, dumping charge from a capacitor into the HF coupling transformer, so producing a series of sparks, which continue during the initiation pulse. But on AC the HF output is directed via one or other of a pair of MOSFETS into one or other (respectively) of the pair of windings (single turn) in opposite directions on the coupling transformer. I assume that this is so that the HF spark has the same polarity as the relevant half-cycle of the AC output. Sophistication! Nothing in the welder becomes more than mildly warm whatever I have used it for.)
Thank you for your knowledgeable ideas.
The welder is Fimer TIG 165 AC/DC. When set to AC there is not only the initiation pulse of HF, but also a very brief pulse (a single spark) at or just after each voltage reversal. In the AC mode this ignites (and sustains) the arc anyway, despite the inadequate initiation pulse.
So the problem only manifests itself in DC mode (which I use much more often).
I am interested in your information that both a fall in voltage (from open-circuit to arc voltage. i.e. from say 80v to say 15v) as the arc is established, and also detection of a non-zero current, are required to tell the controller to cease the HF. I had supposed that it would be one or the other, but I have traced circuitry (and checked that it functions correctly) from a current transformer in the inverter output all the way to the MCU; and also from a voltage tapping on the inverter's rectifier as far as, and into, the controller board, but not yet all the way to the MCU. So this is my next task; the last stage into the MCU. (I got stuck when disentangling tracks among a couple of op-amps, and was doubting that I was doing something relevant to the problem, but now I will put in more effort and should be able to sort it out.)
The IC doing the PWM control is the well-known UC3524, and its frequency control (timing) capacitor was the initially failed component. It has worked perfectly since having a replacement capacitor. I don't think anything is wrong with any power components. I can get the DC arc established by touching the tungsten on to the work (a highly unsatisfactory method), and everything seems to work perfectly until I next want to start the arc.
Thanks again. I will post again when I have traced and checked the volts feedback all the way to the MCU.
(P.S. I think that the design and manufacturing quality of the welder are both very good. The HF board took quite a lot of analysing. It has an IGBT working from an oscillator, dumping charge from a capacitor into the HF coupling transformer, so producing a series of sparks, which continue during the initiation pulse. But on AC the HF output is directed via one or other of a pair of MOSFETS into one or other (respectively) of the pair of windings (single turn) in opposite directions on the coupling transformer. I assume that this is so that the HF spark has the same polarity as the relevant half-cycle of the AC output. Sophistication! Nothing in the welder becomes more than mildly warm whatever I have used it for.)
I would like to clarify - how dc or ac is seperated / switched?
I think there should be diode rectifier from HF power transformer ,producing DC voltage ,which goes to bridge mosfets or igbt ,most likely mosfets , like bridge amplifier ,andtwo of them should be open constantly when DC mode, or all operating in AC mode , giving square wave at output . You can study pdf of 3524 ,and probably you find in your welder too ,that pin 2 gets reference voltage, and pin1 gets voltage from resistive divider, DC output .But not fact that both opamp of 3524 are used in your welder to control voltage and limit current . For current limit cpu sets reference voltage to one of opamps of 3524,according to your settings ( current knob ?) . So this creates an another idea - test without cpu ,cut traces or desolder something ,maybe jumper or resistor , add your potentiometer ,to set voltage at output ,like in power supply . SG3524 easy works in output voltage feedback mode ,keeping some dc voltage after inverter, until there is current detected . Then current feedback begins operating ,by limiting current to some value .But there may be different methods to accomplish this .Until you try to weld ,inverter should produce idle dc 36-72 volts ,kinda of that . Did you tried to check with oscilloscope what you have at output in AC and DC modes ? AC mode should give both polarities ,so DMM set to DC , should display zero DC in AC mode .If bridge mosfets fail ,at least one , you will get DC , and that may cpu see too ,and turn off inverter . Had a chance to repair something like yours ,but china model , unknown brand , one of modes not worked ,cant remember which one ,failed part was simple cd4xxx cmos logic ic ,controlling optocouplers ,whose controlled mosfets bridge .First you need to understand topology and check seperate blocks / parts of it .So power inverter should be possible to convert to power supply with voltage and control current ,for ensuring everything is ok . More measurements ( voltages ,pulses ) would help to identify whats wrong .
I think there should be diode rectifier from HF power transformer ,producing DC voltage ,which goes to bridge mosfets or igbt ,most likely mosfets , like bridge amplifier ,andtwo of them should be open constantly when DC mode, or all operating in AC mode , giving square wave at output . You can study pdf of 3524 ,and probably you find in your welder too ,that pin 2 gets reference voltage, and pin1 gets voltage from resistive divider, DC output .But not fact that both opamp of 3524 are used in your welder to control voltage and limit current . For current limit cpu sets reference voltage to one of opamps of 3524,according to your settings ( current knob ?) . So this creates an another idea - test without cpu ,cut traces or desolder something ,maybe jumper or resistor , add your potentiometer ,to set voltage at output ,like in power supply . SG3524 easy works in output voltage feedback mode ,keeping some dc voltage after inverter, until there is current detected . Then current feedback begins operating ,by limiting current to some value .But there may be different methods to accomplish this .Until you try to weld ,inverter should produce idle dc 36-72 volts ,kinda of that . Did you tried to check with oscilloscope what you have at output in AC and DC modes ? AC mode should give both polarities ,so DMM set to DC , should display zero DC in AC mode .If bridge mosfets fail ,at least one , you will get DC , and that may cpu see too ,and turn off inverter . Had a chance to repair something like yours ,but china model , unknown brand , one of modes not worked ,cant remember which one ,failed part was simple cd4xxx cmos logic ic ,controlling optocouplers ,whose controlled mosfets bridge .First you need to understand topology and check seperate blocks / parts of it .So power inverter should be possible to convert to power supply with voltage and control current ,for ensuring everything is ok . More measurements ( voltages ,pulses ) would help to identify whats wrong .
Hello, Ximikas,
Thanks for the effort you are putting in, going through the main modules of this sort of welder. As you say, the configuration is: Mains in, mains rectifier, big smoothing capacitors (about 340v dc), high frequency inverter (about 80kHz,), transformer, rectifier (about 80v unloaded), then bridge inverter composed of 4 devices (unexplored so far) 2 on, 2 off for DC welding, or all switching for square wave AC welding.
Since welding in both AC and DC modes remains perfectly functional once the arc is established I doubt that there is any serious (disabling) fault in any of the main (power or control) circuitry. In DC mode the current control, slope up time, slope down time, pulse operation etc are all exactly as they should be, and in AC mode, current, frequency and balance (wave asymmetry) all work correctly. The only fault the welder displays is the too-brief HF pulse.
Your information that the pulse is meant to be controlled by detection of both voltage and current was very valuable to me. I had wondered why there are two current-detecting components in the power chain; one, the current transformer I mentioned earlier on the inverter output to its rectifier, and the other, a Hall effect device on the DC output of the inverter's rectifier. I assume now (though I have learned again and again over the years to question every assumption) that the current transformer is there to serve the HF control, and the Hall effect device is to control the welding current (via, as you say, the 3524).
The HF board (a separate board) receives two signals from the control board. One carries the initiation pulse (which I regard as the problem), and the other carries the AC square wave waveform (when the welder is in AC mode). Measurements and waveforms detected on the HF board lead me to believe that the HF board is acting correctly on these signals, so I am concentrating my effort on determining why the HF initiation pulse is terminated after about 12ms, rather than waiting until the arc is established.
Two more observations: (1) Lift-arc (which I do not like or use) does not work either. It obviously depends also on current and voltage measurement and control, but is not implemented in the HF board. (2) More significantly, I have noticed an output from the current transformer even when the welder is not loaded. I had put this down (in my mind) to capacitance here and there and perhaps an inevitable slight leakage, and so rather ignored it. But, given your suggestion that current is a determinant of the HF pulse duration, perhaps this is indicative of a problem. When the 3524 drove the inverter at nearly 1MHz. a good deal of strain will have been placed on at least the inverter and its rectifier. Certainly the inverter driver MOSFET suffered, and so did the snubber components (a resistor glowing red and a diode subsequently leaking). The horrible oscillations in the inverter will have been received by its rectifier. Might the rectifier have suffered, and now leaks slightly, giving the false indication that a welding current is established?
I could try temporarily disabling the current transformer output, and seeing whether that results in the HF pulse length increasing, but I must first check that the current transformer signal does not affect anything else in particular the 3524. Anything which disturbs a PWM controller is risky, easily leading to blown driver semiconductors.
Must get back to looking at the welder again, now armed with new ideas. Thank you.
Thanks for the effort you are putting in, going through the main modules of this sort of welder. As you say, the configuration is: Mains in, mains rectifier, big smoothing capacitors (about 340v dc), high frequency inverter (about 80kHz,), transformer, rectifier (about 80v unloaded), then bridge inverter composed of 4 devices (unexplored so far) 2 on, 2 off for DC welding, or all switching for square wave AC welding.
Since welding in both AC and DC modes remains perfectly functional once the arc is established I doubt that there is any serious (disabling) fault in any of the main (power or control) circuitry. In DC mode the current control, slope up time, slope down time, pulse operation etc are all exactly as they should be, and in AC mode, current, frequency and balance (wave asymmetry) all work correctly. The only fault the welder displays is the too-brief HF pulse.
Your information that the pulse is meant to be controlled by detection of both voltage and current was very valuable to me. I had wondered why there are two current-detecting components in the power chain; one, the current transformer I mentioned earlier on the inverter output to its rectifier, and the other, a Hall effect device on the DC output of the inverter's rectifier. I assume now (though I have learned again and again over the years to question every assumption) that the current transformer is there to serve the HF control, and the Hall effect device is to control the welding current (via, as you say, the 3524).
The HF board (a separate board) receives two signals from the control board. One carries the initiation pulse (which I regard as the problem), and the other carries the AC square wave waveform (when the welder is in AC mode). Measurements and waveforms detected on the HF board lead me to believe that the HF board is acting correctly on these signals, so I am concentrating my effort on determining why the HF initiation pulse is terminated after about 12ms, rather than waiting until the arc is established.
Two more observations: (1) Lift-arc (which I do not like or use) does not work either. It obviously depends also on current and voltage measurement and control, but is not implemented in the HF board. (2) More significantly, I have noticed an output from the current transformer even when the welder is not loaded. I had put this down (in my mind) to capacitance here and there and perhaps an inevitable slight leakage, and so rather ignored it. But, given your suggestion that current is a determinant of the HF pulse duration, perhaps this is indicative of a problem. When the 3524 drove the inverter at nearly 1MHz. a good deal of strain will have been placed on at least the inverter and its rectifier. Certainly the inverter driver MOSFET suffered, and so did the snubber components (a resistor glowing red and a diode subsequently leaking). The horrible oscillations in the inverter will have been received by its rectifier. Might the rectifier have suffered, and now leaks slightly, giving the false indication that a welding current is established?
I could try temporarily disabling the current transformer output, and seeing whether that results in the HF pulse length increasing, but I must first check that the current transformer signal does not affect anything else in particular the 3524. Anything which disturbs a PWM controller is risky, easily leading to blown driver semiconductors.
Must get back to looking at the welder again, now armed with new ideas. Thank you.
Measure current transformer output winding resistance ,maybe its faulty. Current transformer is in series with primary or secondary winding of main transformer? Mostly current transformer must have load , resistor , maybe after small diode bridge . Instaed of disabling CT , increase its load ,or reduce amplification , and check if pulse length increases. But as you saying in ac mode all functioning as it should ,and current controls with knob works , its not your case. So there is another current sensor , need to determine then , which one is responsible for what. Try to measure output voltage from hall sensor when idle and when welding,if possible ,and adjusting current knob . Sensors are needed for some reason , and who knows , how their operation affects cpu operation. Logically thinking , inverter must supply dc to electrode ,until you begin welding , touch and make contact , and current begins to flow , and voltage feedback mode changes to current mode . Did you call that ignition pulses ? I'm not that much familiar with welding technologies and theories of operation.
But even if extract code success , you would get not c code ...,but pure assembler . And to determine , which part of code will enable output, need to know all info about that cpu . There would be many jumps from one location to another ,returning back then . Maybe someone had success of converting extracted machine instructions back to easier understandable c code .
The last two comments leave me convinced that extracting the programme from a MCU is, in practice at least, not something to attempt. I quite see that 'uncompiling' (to get from machine code to a higher level language) would be a tall order. Something I read a week or two ago suggested that it would be quicker to write code from scratch than to try to do anything with code extracted from a device.
Ximikas: I see that I have not explained the HF function. TIG welding uses a tungsten electrode which is not consumed by the welding process. An arc, in an atmosphere of (usually) argon, is initiated between the electrode and the work, and like a flame it produces a molten pool. More metal can be added to the pool by feeding in a rod. Initiating the arc can most crudely be achieved by touching or scratching the electrode on to the work, but a far more pleasant and satisfactory method is to create a series of high voltage sparks. These ionise the gas in the gap (a few mm), and the powerful welding current then passes through the gap in the form of an arc. This high voltage spark is described (misleadingly) as HF (high frequency) because, traditionally, the spark was created by connecting a capacitor to a high voltage source (typically a mains transformer delivering several thousand volts), and having a spark gap across the capacitor. The capacitor voltage would rise until it broke across the spark gap, and the spark would discharge the capacitor abruptly. This sharply defined discharge current pulse was passed through a single turn primary of the coupling transformer, which transferred the pulse to the welding electrode. Having been discharged, the capacitor would rapidly re-charge, and repeat the cycle. This repetition occurred at a fairly high frequency, hence the term HF. These days the spark is often produced far more elegantly by discharging the capacitor with a semiconductor (e.g. a thyristor), and not at a particularly high frequency. (Near the end of my second-to-last post I mistakenly identified that semiconductor as a MOSFET.)
I have just been looking at the circuitry from the current transformer (which is in the secondary of the main transformer, on its way to the rectifier), and I will not interfere with it because I see that it also connects to the current limiting pin of the 3524. I think that it serves to put an upper limit on the output of the inverter to protect it in the event of failure of the welding current control. After some more studying of the circuitry I will try under power and see how the current transformer output varies, and how low it goes. I might disconnect the rectifier output and see whether it then goes lower.
Ximikas: I see that I have not explained the HF function. TIG welding uses a tungsten electrode which is not consumed by the welding process. An arc, in an atmosphere of (usually) argon, is initiated between the electrode and the work, and like a flame it produces a molten pool. More metal can be added to the pool by feeding in a rod. Initiating the arc can most crudely be achieved by touching or scratching the electrode on to the work, but a far more pleasant and satisfactory method is to create a series of high voltage sparks. These ionise the gas in the gap (a few mm), and the powerful welding current then passes through the gap in the form of an arc. This high voltage spark is described (misleadingly) as HF (high frequency) because, traditionally, the spark was created by connecting a capacitor to a high voltage source (typically a mains transformer delivering several thousand volts), and having a spark gap across the capacitor. The capacitor voltage would rise until it broke across the spark gap, and the spark would discharge the capacitor abruptly. This sharply defined discharge current pulse was passed through a single turn primary of the coupling transformer, which transferred the pulse to the welding electrode. Having been discharged, the capacitor would rapidly re-charge, and repeat the cycle. This repetition occurred at a fairly high frequency, hence the term HF. These days the spark is often produced far more elegantly by discharging the capacitor with a semiconductor (e.g. a thyristor), and not at a particularly high frequency. (Near the end of my second-to-last post I mistakenly identified that semiconductor as a MOSFET.)
I have just been looking at the circuitry from the current transformer (which is in the secondary of the main transformer, on its way to the rectifier), and I will not interfere with it because I see that it also connects to the current limiting pin of the 3524. I think that it serves to put an upper limit on the output of the inverter to protect it in the event of failure of the welding current control. After some more studying of the circuitry I will try under power and see how the current transformer output varies, and how low it goes. I might disconnect the rectifier output and see whether it then goes lower.
So you saying there should be also high voltage pulses, but how they are formed ? Main transformer have step down , primary more windings than secondary, so it can't create hv . There must be small transformer in series somewhere , with upping config . So if pulses are too early terminated , maybe again ,some overcurrent or overvoltage occurs ,and cpu terminates them. Something similar is done in one kind of gas lamp electronic ballasts, until lamp ignites ,its getting 1-2kv for some time ,with pauses , and after some retries cpu will stop trying ,thinking lamp is faulty ,and after lamp replacement , need to remove mains for some time to reset cpu. So those ballasts have hv transformer ,small size , maybe in welder its present too.