Uploaded January 2026 | Updated November 2026, 3 hours ago
IDK what I was doing lol but I YET AGAIN didn't set the supply right which I'm just noticing now.. But if I actually turn it to the whole 24V input it will run 19V into the resistor load which brings it to ~70 watts out at 100 watts in (approx ~106W in at 24V to 19V at ~5 ohms resistance). So it gets down to ~70% at that point.
I'm done with this one now so I can show how it turned out. Not going to complain about the throughput when I just rando-wound a transformer and don't feel like winding another one. Eventually I'll try to push more power from another one that will probably use a split secondary with 2-diode full wave rectification and I'll use some actual ferrite transformer core I can wind more easily. Maybe 3D print some bobbins so I can have several test windings. If the idea is more power in that regard there isn't much point using this type of topology unless I make it a high voltage converter. Otherwise it needs to be fairly high voltage AC in before rectification where something like a half bridge push pull makes more sense. If I'm using AC in then I'm not going to waste all this time with isolation in mind, I'll just use a transformer. I already have a ~700W AC to DC converter that does 60VDC out so maybe it would be cool trying to make one that does more using some bricks.
If you just want a basic feedback loop it seems like the TL431 is stable enough by just tossing it in like I did. TI does say it's stable as-is but that's the chip by itself in a random test. So I guess the point of the cap would be trying to create a shift (delay) and then fine tune it with the ESR to shift it back toward a relative lead. I'm not exactly sure what was really dropping out when I went below 5V output at first but the opto LED had to be pretty close to the ~1mA region at that point. So in that regard even if I've limited the output to below 5V, I wouldn't want to try pushing too much current from that or I will start running into the same problems again if it starts dropping. And yea you're not crazy it's a 12K and not a 10K I'm biasing the TL431 cathode with. I should probably drop that closer to 5K. Adding an output inductor is also probably a good idea so if I can find like a small 100uH I'll place that in between the rectifiers and output caps (I didn't add any output LC filter cause it screwed with my feedback loop). With 1nF drain caps you could see my transitions were around 1us. Adding 47 ohms for RC's reduced some mhz ringing I had at my notches but I ended up with about 220pf and 22 ohms to get rid of it at about 500ns transitions. The main peaks are something I just couldn't get rid of without running close to max Vout.
I probably wouldn't rebuild this the same if I had to do it over again but it seems like you can buy something similar for around 30 bucks. Like a Meanwell SD-50C-24 or something. That would include better EMI suppression and a guaranteed 2A output at 24V. IDK if mine would do that but I could probably get it to since I can just lower the oscillator frequency and dramatically increase the power. If I had to buy every single component then it's like screw it I'll just buy the whole supply.
IDK what I was doing lol but I YET AGAIN didn't set the supply right which I'm just noticing now.. But if I actually turn it to the whole 24V input it will run 19V into the resistor load which brings it to ~70 watts out at 100 watts in (approx ~106W in at 24V to 19V at ~5 ohms resistance). So it gets down to ~70% at that point.
I'm done with this one now so I can show how it turned out. Not going to complain about the throughput when I just rando-wound a transformer and don't feel like winding another one. Eventually I'll try to push more power from another one that will probably use a split secondary with 2-diode full wave rectification and I'll use some actual ferrite transformer core I can wind more easily. Maybe 3D print some bobbins so I can have several test windings. If the idea is more power in that regard there isn't much point using this type of topology unless I make it a high voltage converter. Otherwise it needs to be fairly high voltage AC in before rectification where something like a half bridge push pull makes more sense. If I'm using AC in then I'm not going to waste all this time with isolation in mind, I'll just use a transformer. I already have a ~700W AC to DC converter that does 60VDC out so maybe it would be cool trying to make one that does more using some bricks.
If you just want a basic feedback loop it seems like the TL431 is stable enough by just tossing it in like I did. TI does say it's stable as-is but that's the chip by itself in a random test. So I guess the point of the cap would be trying to create a shift (delay) and then fine tune it with the ESR to shift it back toward a relative lead. I'm not exactly sure what was really dropping out when I went below 5V output at first but the opto LED had to be pretty close to the ~1mA region at that point. So in that regard even if I've limited the output to below 5V, I wouldn't want to try pushing too much current from that or I will start running into the same problems again if it starts dropping. And yea you're not crazy it's a 12K and not a 10K I'm biasing the TL431 cathode with. I should probably drop that closer to 5K. Adding an output inductor is also probably a good idea so if I can find like a small 100uH I'll place that in between the rectifiers and output caps (I didn't add any output LC filter cause it screwed with my feedback loop). With 1nF drain caps you could see my transitions were around 1us. Adding 47 ohms for RC's reduced some mhz ringing I had at my notches but I ended up with about 220pf and 22 ohms to get rid of it at about 500ns transitions. The main peaks are something I just couldn't get rid of without running close to max Vout.
I probably wouldn't rebuild this the same if I had to do it over again but it seems like you can buy something similar for around 30 bucks. Like a Meanwell SD-50C-24 or something. That would include better EMI suppression and a guaranteed 2A output at 24V. IDK if mine would do that but I could probably get it to since I can just lower the oscillator frequency and dramatically increase the power. If I had to buy every single component then it's like screw it I'll just buy the whole supply.










