Gregory's ServerFound a "potential" solution for my high-voltage power supply - just use an off-the-shelf capacitor charger, then abuse a Power-over-Ethernet transformer and run it in reverse, turning it into a kilovolt generator... None of the vendor-specified transformers are rated above 1 kV so I initially did not consider this idea. But after some hunting, I found a PoE transformer with seemingly workable specs. #electronics
 41 comments
Status update: Abusing a PoE transformer for a high-voltage supply isn't exactly a good idea... Its low primary inductance (~3 μH) is indeed a problem, as expected - very little energy can be transferred in each flyback cycle, I got only 600 V max even when running it near saturation. #electronics Found another AC/DC transformer model to abuse, it finally reached 1000 V. ⚡ Time to upgrade the impulse generator for switching 1 kV as well. #electronics Bang-bang voltage regulation for the 1000 V power supply is working. The set point and hysteresis just need a little adjustment... #electronics Impulse generator has been upgraded, now it can switch 1000 volts. I also noticed that the measured ratio between capacitor and output peak voltage is within 0.05% of the theoretical calculation by coincidence... :blobcatsurprised: The math says 108.2%, I measured 108.256%, unbelievable. #electronics Success. My 1.2/50-8/20 μs impulse generator breadboard prototype now can generate 1000 V surges, and its waveform conforms to IEC 61000-4-5. ⚡ For better project reproducibility, I also replaced the solenoid with air-core toroid inductors using "Material #0" plastic cores. This design is ready for a real circuit board. #electronics Destructive surge testing of a 47 Ω through-hole resistor, using the new impulse current generator I've just built. Dielectric breakdown and arc flash occurred at 954 volts. The instantaneous power into the resistor was 19364 watts, or nearly 20 kW. ⚡💥 #electronics Time to transfer the impulse generator prototype to a real circuit board. It's really a deceptively simple circuit. Just a capacitor, a switch, an inductor and two resistors. But practical implementation problems are many... #electronics I really don't want to use a microcontroller in this high-voltage power supply. Pure analog control is entirely sufficient without the complexity of software or computers. Unfortunately, to meet 1.5 kV reinforced insulation safety requirement, it needs some very unusual and expensive optoisolators. To minimize the use of isolators, the most reasonable option is sending commands via a one-wire serial port with microcontrollers. #electronics Proposed block diagram of the revised high-voltage impulse generator. Now with 1500 V reinforced insulation to ensure the high voltage impulse will never be backfired to the operator. #electronics A little bit of progress on the IEC 61000-4-5 high-voltage surge generator. Now with a safety crowbar feature to short the capacitor bank to ground immediately. The special high-voltage FET is a default-on JFET, so the capacitors can be discharged even when all power and control signals are lost. #electronics Completed the first draft of circuit schematics (many obvious mistakes, including wrong power supply labels, but it's just a draft). Now moving to the next stage with more difficulties: circuit board design. Nearly every single component needs a custom PCB footprint. #electronics More PCB layout progress... The sheer size differences between the huge capacitors & inductors and the tiny DC-DC converters, gate driviers and microcontrollers is giving me a bit of cognitive dissonance... #electronics There must be a 24 mm creepage distance between the LV and HV side, leaving a large gap. I'll just spam some high voltage warnings here as placeholders... #electronics The original layout sucks and I just started over again. Now it looks slightly better... #electronics High-voltage section is fully routed. Traces are ugly since it's just a trial to verify component placement, to be redone later. I should be able to complete it by tomorrow. #electronics PCB design of my high-voltage surge tester is almost finished. #electronics Now the problem is whether I can drive this 128x160 screen from a microcontroller with only 7 KiB of ROM. It's possibly more than enough if you use some extreme space-saving techniques. But the point of this prototype is testing the analog electronics. It's NOT the time to be nerd-sniped into a pointless programming exercise. Perhaps it's a good idea to add a large EEPROM chip here, at least as a temporary solution during prototyping. #electronics The circuit board still needs more routing, but at the schematic level, this whole design is now completed. P.S: For a finished project, always remember to fill the title block section of the schematic with some words to make it look nicer, otherwise it's just an empty box... I always use a copyright notice if I don't have something else to show. #electronics Once again scrapped the finished circuit board layout and started over... The original component placement simply couldn't meet the high voltage electrical spacing requirements. Technically, there is no regulation on electrical spacing for non-safety, purely functional insulation. But I'd still like to leave a bit of margin here. #electronics My third circuit board layout attempt is almost finished. A component rearrangement solved the high voltage electrical spacing problem, it also made the circuit layout flow much better (instead of squeezing everything into whatever space that happened to be there). #electronics Number of unrouted connections: 0. Still needs some tuning, including electrical spacing and looking for a suitable banana connector. But it's basically done. #electronics Circuit board arrived. I found my PIC programmer is 5 V only and cannot be used on this 3.3 V chip, then I found a great project called Pickle which can do serial port bitbanging for a makeshift programmer. Just turned my LEDs on, ready to start firmware development... #electronics Wasted an hour trying to set up a timer on the PIC microcontroller. The frequency was always wrong by an order of magnitude, I kept fiddling with the clock generators, PLLs, prescalers, and fuses. Only an hour later I found the real problem... I simply forgot that a 8-bit timer has 256 counts, not 1 count. :blobcatwoozy: Just finished the high-voltage controller firmware in 500 lines of C. It's mostly just housekeeping tasks. The actual interesting part of the circuit board is the analog circuitry under control, which I finally can start testing tomorrow... The low-voltage controller also needs a firmware, but for now I can simulate that using a stub program on the computer... #electronics I just figured out the failure mode of the malfunctioned flyback chip. The flyback controller turns the transistor on, wait for inductor current to ramp up, and turns off the transistor only after the peak current comparator trips. If the power supply drops out after an overload, the original peak current target is never reached, now the transistor is permanently on! It's not that the gate drive malfunctioned and stuck high, it's actually working exactly as designed. #electronics I wonder if my fatal analog design problem can be "fixed in software" and suddenly had an idea... Instead of turning the flyback converter on continuously, I changed the firmware to run it in small 1-millisecond bursts per every 10 ms, allowing the 10 V isolated power supply to recover and recharge between pulses. Unbelievable, this hack worked! It successfully charged the capacitor bank to 1000 V... ⚡ #electronics Voltage regulation at 500 V is even better than my expectation. 2% within theoretical 536 V set point. #electronics Setting up the fail-safe crowbar and watchdog timer early on before debugging other parts of the high-voltage circuit turned out to be really worthwhile. Now I love how I can physically discharge the dangerous capacitor by just pressing Control-C in my Unix terminal emulator. #electronics Trap for young players... If you're disabling another interrupt within an interrupt, that disabled interrupt source may have already set its flag just before you've disabled it, causing a TACTOU situation. When the ISR returns, that supposedly-disabled IRQ is immediately invoked. I just spent an hour trying to debug a strange problem because of this race condition... :woozy_baa: #electronics Success. High-voltage supply and regulation are now fully functional. My fatally flawed analog power supply problem has been fixed in software. The burst-ON hack turned out to not "just a software hack" but kind of an optimal solution given the circuit components constraints. If I need a real hardware solution I'd just implement exactly the same control with RC timers and AND gates instead of an interrupt service routine. #electronics More progress on the high-voltage impulse generator. The huge capacitors, resistors and inductors for pulse shaping are now installed for testing. They mostly work. But this test uncovered a serious problem - each time the circuit fires, I can hear an arcing noise, the oscilloscope also shows a weird glitch, likely caused by arcing. The generator is supposed to destroy the device-under-test, not itself! Now my suspect is the isolated DC-DC module. Just ordered a better one for another try... #electronics Problem solved. The "arcing" sound during the impulse current discharge is NOT a fault. It's actually a common phenomenon in all high-current pulse circuits. The "snap" noise was caused by sudden physical deformations of the circuit board itself, due to Lorentz force from the intense current. #electronics Success! My first circuit board prototype of the IEC 61000-4-5 Combination Wave Generator (aka Lightning Surge Generator, aka Impulse Generator) is working after a month of development. This PCB has greatly improved safety and usability compared to the original perfboard. #electronics |
It still has problems. The primary inductance is too low, datasheet says this breaks LT3750's primary-side voltage regulation, the output will rise uncontrollably. I have an external secondary regulation so it's still going to work to a degree. Let's see if I can find a better transformer. If I cannot, for a quick and dirty test, I'll just add a gas discharge tube crowbar as the regulation failure protection... #electronics