First stop after the datasheet should always be the app notes:
https://www.ti.com/lit/an/slua143/slua143.pdfMy bad, I was in a hurry with posting this as I had something to do.
1. 12V-24V, from a 4S LiPo or a 24V PSU.
Ok, you'll need a regulator to deal with the high voltage range. The 3843 can handle voltage just fine, it's the MOSFET -- and drive power -- that can't.
Like on the quad LED supply I showed in the linked thread, the 3842s run around 18V and get rather hot, even into fairly modest size MOSFETs. And those are only about 50W/ch.
2. VCC is just the supply voltage, U1 creates +5V and U2 receives it.
3. Because the UC3843 current sense works with 1V internal comparator. The voltage from the sense resistor is too low. Picking a higher sense resistor value would result in huge power loss so I need an op-amp to amplify it. Same goes for VREF but 2.5V.
4. What do you recommend I use instead? I think I could get away with using the LM358 for the LED current feedback since it doesn't need to be fast at all.
I don't have an option offhand, but you'll need a low voltage (5V) RRIO or single-supply amp with GBW > 20MHz and input offset < 1mV. Probably this will be a CMOS type. Also acceptable is a high voltage (>=16V) single-supply type
"Single supply" means V_OL very near to, and V_ICM extending below, VEE, while the upper limit (and V_OH) doesn't matter. Distributors rarely track this (or accurately), you need to read datasheets to be sure.
5. Yes it's a tapped inductor.
6. I didn't really think of that, good observation. Would using a regular non-tapped inductor be possible with this voltage ratio?
It's not terrible. It does put more peak current into the rectifier and output filter caps.
7. Should I add ceramics?
Probably? Depends what mix of capacitors you have/need. Have you shopped for any yet?
8. I want it to be dimmable. I suppose I can make the gain of the op-amp adjustable. I copied parts of the circuit from this thead.
Variable gain..? Nah, that's hard and nonlinear, just vary the offset. Same as we discussed in that thread.
9. I don't really understand the purpose of this. Could you perhaps link to an application note or something?
https://www.ti.com/lit/an/slva662/slva662.pdfI don't get why all the 3843 datasheets/appnotes show R||C, it's stupid. Well, I know why they
all do, they're all copying the original (Unitrode). A resistor between COMP and VFB causes a "squishy" output -- poor regulation, i.e., output parameter depends on error amp output. That's literally the meaning of such a feedback resistor.
The capacitor breaks that path, giving excellent DC regulation, and the resistor provides phase margin.
10. Seems reasonable. I'll see if I can find the equation in the datasheet.
Yes I still have to add input caps.
Oops, VCC to GND is a silly mistake.
11. I intend to minimize parasitics as much as possible, I'll see if I need a snubber.
L1 design is critical, guessing it's hand wound unless it's just on the edge of what those like uh, Wurth has some dual flat-winding things probably about the right size, and should have adequately low leakage.
Selection of coupled inductors is quite poor, especially in large sizes. FYI, you'll probably want to avoid toroidal types, like, these have a ton of leakage:
https://www.digikey.com/en/products/detail/pulse-electronics-power/P0596NL/2266694note that they aren't even polite enough to provide a k or LL spec.
12. I think the voltage on the sense resistors is far too low for this. The resistor tolerance is too big. Should I just treat VFB and COMP as a regular op-amp and use the Gain=R2/R1+1 equation?
Well, use 1% resistors obviously, and I don't have a problem with tweaking say 0.5V up to 2.5V. Which is about half range (2.5V is about halfway between 0.5V and 5.0V) so the error amp sees 0.25V of range. That's not grotesque.
If you can't handle 3A at 0.5V = 1.5W, not even a percentage point against overall efficiency... you probably have bigger problems than this.
I have no idea what all this means, sorry. I prefer a simple solution. If there's an IC that does the job I'll gladly buy it.
Just follow along with a pad and paper and the description..?
I can never tell if my descriptions are actually terrible, or if I always end up talking to people with aphantasia. Text is a poor medium for circuits, granted, but surely there is some middle ground between those extremes.
Since I almost never get feedback on these matters, I tend to write these lengthy posts for my own edification, or in the hopes that someone does eventually find them useful (if not always the OP).
Anyway, they show the connection in the appnote, Fig.32, with the 2N2222, 15k and 470Ω. It's a lot of words for just three components.
13. I think this would be too complicated.
They also show the CT connection in the same figure, though it's complicated by the, heh that must be a flip-flop, so they're doing a full wave alternating thing. Just imagine that's not there and it's flyback into one half of the transformer, and the CT is in series with it as normal.
14. Yeah that's also something I worried about. If the output goes open circuit it would go beserk. Is there some way of combining voltage and current control onto VFB?
I might consider a newer controller yes, provided it makes my design easier. You got any good suggestions?
Yes! I wrote about both.
HTH,
Tim