| Electronics > Projects, Designs, and Technical Stuff |
| High voltage feedback |
| << < (2/2) |
| T3sl4co1l:
A push-pull converter needs a choke just as much as any other. Without, it's a charge pump, there's no regulation, and the transistors dump fatally large current spikes. (Put in a shunt resistor and see!) Tim |
| MagicSmoker:
--- Quote from: timh2870 on May 15, 2019, 11:24:26 am ---A push pull converter "shouldn't" need a choke. ... As for compensation around the tl431, I don't know how to calculate those and can't seem to find what would be reasonable values to start with. --- End quote --- If you've been here long enough you learn to think twice before disagreeing with T3sl4co1l... and he is absolutely correct here that you need a choke in between the rectifiers and capacitor in any voltage-fed forward (ie - buck-derived) converter. Only the flyback, current-fed (boost-derived) forward and certain resonant topologies don't need an output inductor (but they do need an inductor, somewhere). The fundamental reason why is that the pulses of energy delivered by the switch need an inductor to integrate the current waveform and a capacitor to integrate the voltage waveform. |
| T3sl4co1l:
You do, by the way, see such circuits pretty often -- they're not impossible to start up. Usually what's done is a combination of very generous soft-starting (so the massive surge currents are brief at first), very low resistance switches (usually a big pile of IRFZ46Ns, for automotive (12V) input), and excessive leakage inductance in the transformer (which is usually a toroid, wound in no particularly special way). Which, putting those clues together, yep -- that's almost every automotive power amplifier out there. A DC-DC converter, turning 12V into +/-30V say, followed by either a class AB (linear) amplifier, or class D (switching) amplifier. The 30V rails are unregulated -- the converter runs at full duty cycle (typically a TL494 or similar, so 45-48% duty per switch), and the supplies are proportional to the input. They're also notorious for exploding. Dead power supplies are probably the most common failure of those amps. The better quality ones do actually have current sensing and a filter choke, and almost never fail -- they control one of the two things that destroys silicon, so it's never allowed to reach an explosive value in the first place. Proper design approach. (The remaining factor is excess voltage, which is a bit harder to deal with, but also not insurmountable.) Tim |
| MagicSmoker:
--- Quote from: T3sl4co1l on May 15, 2019, 02:22:44 pm ---You do, by the way, see such circuits pretty often -- they're not impossible to start up. Usually what's done is a combination of very generous soft-starting (so the massive surge currents are brief at first), very low resistance switches (usually a big pile of IRFZ46Ns, for automotive (12V) input), and excessive leakage inductance in the transformer (which is usually a toroid, wound in no particularly special way). Which, putting those clues together, yep -- that's almost every automotive power amplifier out there. ... --- End quote --- Indeed, and they likely also benefit from extra leakage inductance from what would usually be considered poor transformer design (intentional or not). |
| timh2870:
I don't mean to disagree with anybody; I guess I'm just too used to seeing Chinesium that it's hard to think about proper circuit design. This project was going to be a one or two-off at best, so I had no problem scavenging components. The transformer in the picture is an ee42 ferrite core that I boiled apart, then re-wrapped with 3:3:60. I was using a tl494 and two tc4420 drivers, but I couldn't get that over maybe 15-20 watts without the spikes on the primary side drowning out the clean square wave (and making the fets scorching hot). I found this gzf-1000w unit on aliexpress https://www.aliexpress.com/item/1000W-DC12-or-24V-to-AC0-220V-380V-high-frequency-inverter-DC-variable-AC-boost-module/33011023057.html and figured it was a good place to start and also simple enough to reverse engineer. It has an RC snubber under the transformer and uses an sg/ka3525 controller with the non-inverting input tied to vref/2, the inverting input pulled down via a 10k resistor, and the feedback as shown in the schematic above (100k/100n fed back to the inverting input) |
| Navigation |
| Message Index |
| Previous page |