Critique my capacitive dropper design

[arcfour]

Hello everyone,

I'm making a capacitive dropper to power an LED COB at ~10mA. I've built these supplies before (and am aware of the risks posed by the lack of galvanic isolation), but I decided I wanted to try and make a very robust and safe design this time, so I'd like to solicit feedback on the design and component selection. This isn't a commercial design or anything, I am just a hobbyist. Here's the schematic:



And here's the BOM:

Component	Description	Model/Datasheet
F1	250V 200mA fast-blow	Bel Fuse MRF 200
C1	305V .33uF X2 capacitor	TDK B32932A3334K000
R1	500V 68R 1W Wirewound	TE EP1WS68RJ
R2, R3	250V 470k 1/4W	Yageo MFR-25FTE52-470K
DB1	800V 2A rectifier	Vishay 2W08G
D2	39V 600W TVS	Littelfuse TP6KE39CA
C2	63V 47uF Electrolytic	Wurth 860020773013
COB	34.2V 80mA LED module	Luminus CXM-3-30-90-36-AC40-F5-3

C1 is specifically rated for series mains usage (also, I'm in the US with 120V). DB1 has a surge current rating of a whopping 60A. D2 is well within its limits under a crude IEC 61000-4-5 1kV surge test, shunting much of the pulse away from everything else (C1 sees about 850V of the spike). It also should prevent C2 from turning into confetti, should the COB go open-circuit. The COB itself is incredibly bright at just 10mA with a prototype of this circuit, but is rated for 80mA, so it should stay nice and cool - even during transients and fault conditions I don't think it ever sees more than 20mA. I know the TVS standoff voltage is pretty close to the circuit voltage, but I'm not too concerned if it's acting like a crude zener and wasting <1mA of current, I'm really just concerned with it properly blowing the line fuse in case of a serious fault.

So, what do you guys think? Any room for improvement? I'm open to suggestions but from my simulations I'm confident this thing should be safe no matter what - I made sure there were no SPOFs and everything had a comfortable cushion to keep it within its design limitations. Heck, this thing seems to have better protection than any of the commercial designs I've seen...guess that's the benefit of not having to design it for mass market!

Thanks in advance.

Edit - changed the DB1 to the 800V version, not that it likely matters. I have both on my desk and mixed them up. Oops.

[ledtester]

Not criticizing, but I'm curious about your reasons for using a TVS over a Zener. Clearly zeners are cheaper which is one reason you see them in commercial products and in your case cost is not a concern.

I found this little blurb about using a TVS for regulation:

http://www.icchipword.com/article/What-is-A-TVS-Diode-Difference-between-Zener-Diode-and-TVS-Diode.html

Both TVS and Zener regulators can be used as voltage regulators, but the TVS tube Zener breakdown current is smaller. The voltage regulation greater than 10V is only 1mA. Relatively, the Zener diode breakdown current is much larger, but Zener Diode voltage regulation accuracy can be done relatively high.

[arcfour]

I was having trouble finding a zener that I was sure could handle the peak pulse power of a fault/transient. Then I realized I don't really care about voltage regulation since it's a constant-current source driving the LEDs, and that the primary function of the zener was to act as a protective device instead - thus, TVS, which are much easier to source with the appropriate ratings.

Edit: additionally, under normal conditions, there isn't really much extra current for the TVS to sink anyway (<1mA, if that), so I'm not too fussed.

[jonpaul]

R1 68 Ohm: Must be flameproof

R2, R3 : use 1/2 W carbon comp

D2  Not needed

C2 use only if LED must have pure DC, normally LED is fine for illumination with unfiltered DC.

Jon

[mikerj]

D2  Not needed

What do you think will happen if the COB goes open circuit?

C2 use only if LED must have pure DC, normally LED is fine for illumination with unfiltered DC.

Omit C2 if you enjoy flicker, otherwise keep it.

[Zero999]

C2 use only if LED must have pure DC, normally LED is fine for illumination with unfiltered DC.

Omit C2 if you enjoy flicker, otherwise keep it.

In addition to smoothing out the flicker, it also limits the current surge through the LED. Even with R1, the peak current through the LED will be over 2A, if the power were to be applied, during the peak mains voltage, which could destroy it. C2 will bypass the LED, when the power is first applied, so no current will flow through the LED, until it charges up.

[tunk]

I think usually there's a resistor in series with the LEDs and one in parallel with C2, e.g.:
https://www.techliminals.com/electronics/5-5w-led-light-bulb-schematic.php

[Zero999]

I think usually there's a resistor in series with the LEDs and one in parallel with C2, e.g.:
https://www.techliminals.com/electronics/5-5w-led-light-bulb-schematic.php

The additional series resistor, reduces flicker slightly, at the expense of slightly higher losses. The parallel resistor ensures the LED doesn't glow dimly, due to the tiny current flowing through the stray capacitance in the wiring and should also help the capacitor to protect the LED from current surges, if the LED is turned on and off fairly quickly.

[arcfour]

Great feedback guys, thanks!

I've replaced R1 with a 120 Ohm 1W flameproof resistor. I found that it helped reduce anything nasty even better than the 68 ohm one.

It's correct that C2 reduces current through the COB in case of a voltage spike or during inrush. However I had to add D2 as C2 + the COB alone was not enough current to blow the fuse should C1 fail short, but R1 would get *very* hot (as would the COB). It and C2 also take the brunt of anything nasty from the power lines that gets past C1 and R1, keeping the COB safe. Like I said, not strictly necessary but I wanted to make a safe and robust design that would last a long time, even if I could just throw something together from what I have on hand that would probably *work*.

I've seen those parallel resistors with C2 before on bigclive teardowns to prevent stray capacitance from keeping the COB on dimly. I'll play around with that and see if I like it. A series resistor is not required since the design is essentially a constant-current source providing 10mA.