# Driving 100W COB From mains using discrete components I have on hand

I was wondering if I could realistically drive a 100w COB directly from mains using a bunch of discrete components that I already have on hand.

I have a bunch of BJTs MOSFETs and E cores and I want to build a circuit to run 1-4 COBs in series using a current sink directly from rectified mains voltage with no isolation. My reasoning is that they sell 50w COBs that plug directly into mains voltage on amazon that use very inefficient linear regulation, so why not make a switching regulator directly from mains and call it a day?

Here is the circuit I designed in LTSpice.

simulate this circuit – Schematic created using CircuitLab

Is there a better way to tighten up the current control, because when I simulate this circuit in LTSpice and I can't seem to get the current down to 3A, the lowest current I've seen was 3.6A which was like 125W (DEAD COB). If this idea isn't completely insane, how can I get the current down to 3A so that I can actually test this thing in reality where I wont have to risk killing 2 COBs in series?

Final Edit

The solution was R36, apparently the resistor was necessary to limit current enough to reset the latch cleanly, like in the case of a BJT RS flip-flop, where each transistor has a base resistor and the base of either transistor has to be pulled directly to ground for a period to enact function.

• Your design is irrelevant without design specs that include RdsOn, ESR, DCR, f and layout. You have far too much latency & complexity in BJT’s to control current with loss in GBW and thus PWM control Nov 10, 2020 at 7:05
• My microscope is out for cleaning. Please post a readable image. Nov 10, 2020 at 15:25
• @AnalogKid Maybe you should take that up with Circuit lab, I didn't make the image or perhaps click on the link and view the circuit properly? Nov 10, 2020 at 15:37
• Circuit is awfully complicated! You could use a hysteretic buck or a constant off-time, it would be simpler... Nov 10, 2020 at 17:25
• @bobflux The circuit is certainly a bowl of noodles, but with the COVID and all of the time I have on my hands, it's something to do. The circuit started out as a hysteretic version of this current sink, but the pulse at the gate looked really nasty so I had to add a flipflop to clean it up, then somehow the circuit evolved into this. Though when I actually build it, I'll probably use an LM339. Nov 11, 2020 at 6:06