Alright, I manage to make a movie for you mister @LesJones
Try to watch it fully, or skip when is not interesting. I am talking about many aspects of this construction I did.

 

- Mister @LesJones, your 12pcs x 1 watt LEDs for ring light for macro photography is very good idea. But that is good if you have money....and I will make it someday in the future, just "because", but only when I will be able to make money. Currently, for 8 years as an artist, I dont have anything. So, what I do here, I am trying to find the cheapest possible solution and with the highest luminosity possible, by scraping and re-use these light bulb P LEDs.
What I build already is very fragile. It is practical, it works ok, no problems there, but is very-very fragile. I am working these days to strengthen the ring, somehow. I have some bad ideas and some good ones. That's why I asked before about a viscous composition that rock hardens and also resist at 300*C. Also, 300*C is just better to have than having it around 200*C as some solders melts. So... still, if you can come up with such a solution for me, it will be awesome.
Here are my ideas so far, that I thought of these days:
(click on the image to enlarge and read it's content)

- As you could see from the video, the P LEDs I am using are not commercial ones in any way. My impression (I am not knowledgeable) that these leds are custom made, to fit a specification for an IC, like the one I showed. Or the more "common", the 8V ones which they probably started as something customized but now they are widely used. It is my best guess and my observation. About having a string or a matrix of leds inside... well, I know exactly what you are saying, but I see only 2 leds inside, when I under voltage a P LED. My point here is they are having a different way of calculus, and not as the normal little leds we are accustom with. And we must adapt to them. It sucks I know but is true.
My question here, still stands the same as before: - I need a cheap and DIY power supply for my 8V P LEDs , both categories, the Good category (the 11 P LEDs) and also for the Bad cateory (28 of them). My target is to learn how to make such a custom power supply (from you), and also I will actually use what I learn and put in practice, build some awesome stuff,unseen, unheard of.
Your help is with the hardening solution and the power supply for my weird ass leds. Haha.
As always, thank you so much for your kind and awesome support ! I really appreciate it.

 

Osram LED Lamp Socket: G10q/Cool White/4000 K/12 W Matt SubstiTUBE© T9 Em Replacement for 22 W N/A [Energy Class A+]

I got one of these from Amazon recently.
Fits reasonably well (original clips don't quite fit well enough)
One big advantage is the reduction in interference to equipment nearby that the old starter used to cause.

 

Now, I want to build a Power LED ring, to replace the round Neon.

Welcome.

You could use KiCAD or similar to lay out a PCB that represents a portion of an arc which has 2-3 LEDs on it, send the PCB order out to a China fab like JLCPCB or similar. You'd solder several of these together to form your circle at connecting pads on the face of the arc - but you'd need to make sure to have the arcs sized correctly. You could have traces coming from one of the LEDs out to pads to which you could solder supply wires to - just don't populate that one LED footprint when you assemble your disc. You can make it all more rigid by gluing the pcbs to a larger cut piece of material. Admittedly, if your project involves scavenging LEDs, designing PCBs and sending them off for fabrication probably isn't your thing. It's always a starting point for getting into electronics though - we have such amazing access for DIY these days.

That said, if you're trying to upgrade the lamp and get a nice result, don't Frankenstein something from old bulbs. Certainly not as a first go at something connected to mains and where you need it to fit into a certain enclosure.

IIRC, Cap Reactance is probably a taboo topic here, but at the point that you're dealing with a neon (though I think you really mean fluorescent) tube, your project is already exposed to mains. Exercise proper caution - we're all 2 wires and an electrical outlet and a lapse of judgement from electrocution anyway, but don't be stupid about it, including in the finished result.

I have a much larger magnifying worklamp which originally had a tube in it - and the tube worked just fine, but the design was so old that I had to hold a switch in the START position to turn the lamp on. Removing the tube and making a carrier board and mounting several 1W LEDs each to a separate home-etched PCB and wiring to a commodity LED driver PCB improved the lamp immensely since it could now simply be switched on from a toggle on my bench. I have thought about redoing it to extend it to have the present fixed output LEDs interleaved with a separate circuit (running on a separately switched dimmer, below), and possibly a UV lamp (for some inspection tasks). Each set of LEDs would be driven by a separate driver, but I have plenty of space to work with in the housing - it doesn't look like you have nearly so much -- which raises the question - where are your electronics going to safely reside within the lamp?

You can make light with much lower current and more LEDs, use the reactance to limit current, and a rectifier bridge and large cap to buffer the rectified voltage (which is what your first drawing is). So long as the current it suitably limited, it doesn't matter if you have 3 LEDs or 30 of them in series, so long as to the total Vf is below your supply, and not above it. Lower current LEDs spread out over a larger surface area will also result in more even heat dissipation. You can mail order SMD LEDs from China for cheap, and have a bunch of new devices with a reasonably consistent spec to one another (which is important if you want consistent light output).

I designed a little PCB with PWM dimming, adjustable current limit, and a variety of power input and regulation (for the control IC), which I've used for a while successfully, and recently paired with a small mains PSU based on a TI UCC28881 off-line switcher (which is an awesome little IC as long as your needs fall within it's output capabilities). As with a Capacitive Reactance circuit, this does not provide isolation - but it is a control IC with a datasheet and is used in real-world designs - it is a VOLTAGE regulated device, though not a current one - the PSU sits in front of a driver which limits the current. I'm also more than capable of getting my 160mA from it ( a figure not stated based on your LED measurements, but rather some "6V" 7020 SMD LEDs I have, which are actually 6.4V dual junction, 1W LEDs). A plus with this setup is since my driver PCB has dimming, I can adjust the brightness at time of use, which is nice for a task lamp - sometimes you want some light, other times you want a lot. The square board on the RHS here is the PSU, and the backside of the PCB has two DC capacitors (though one is rated 400V), which are buffering, not Cap Reactance.

The LED driver PCB and the LED carrier (for the 4x 6V LEDs were ordered in separate orders from JLCPCB, with v-scoring and laid out to maximize PCB yields - 150 of the driver PCBs and 120 of the LED carriers (which have vias through to separate silkscreened-over heat spreaders on the opposite side, and holes to allow for screwing or riveting the PCB to an Aluminium bracket). This is more copies than you'd want for your project. If you made full-on rings (if 100mm OD is sufficient), you could fab 2, 5, or 10 for pretty cheap. I have no troubles driving the pictured PCB (which is about 15mm wide by 50mm long) to 4W, and the Al strip it is pop-riveted against gets mildly - not uncomfortably - warm:

If one were clever about it, one could design a ring PCB with connected traces to a circular (or even rectangular) driver PCB situated in the middle cutout, of the ring, and be able to complete the cutout with a coping saw and solder lead wires between the driver PCB and the ring for assembly, enabling the LED and driver to be ordered as a single design (the cheap PCB houses are sticklers for charging extra for separate designs), and you wouldn't be wasting a big chunk of realestate for the void in the middle of the ring.

When you get down to it, driving your LEDs at something LESS than their maximum will pay dividends for the longevity of your lamp.

 

I truly don't know what to search for these types of P LEDs, to find their datasheets !!!
I'm !VERY! curious if you can find any datasheet.

I don't know about the specific LEDs you are working with, but LEDsupply.com sells several high-power LEDs and accessories. You can probably fine a similar one. They also have the data sheets that you can view or download. Typically, the high-power LEDs are available mounted on a small aluminum PCB as either a single LED or as 3 LEDs wired in series. Most high-power LEDs run around 3.1 Volts and is dependent on color and drive current. But don't drive an LED with a constant voltage; use a current source to get the maximum life and consistency (you only need to make sure you have enough voltage available under the worst-case conditions so that you do not "starve" the LED).

 

Interesting points and questions mister @splud, and I thank you for your interest.

...but I have plenty of space to work with in the housing - it doesn't look like you have nearly so much -- which raises the question - where are your electronics going to safely reside within the lamp?...

I made a video and I post the video exactly in this page, about my project and how I made my circular lamp, but I'm not quite 100% happy about my project, even if it's working - excellent I may add.
Here is a screenshot with the enclosure, and you are right it is very small: 6.5cmx5.5cm (2.5inx2.1in for our american friends here). That round gray donut is made from cement !!! chinesium style.

...That said, if you're trying to upgrade the lamp and get a nice result, don't Frankenstein something from old bulbs...

About that, I will use this answer to actually correct myself with the 'good' 8V led in the movie. If you've seen the movie already, at 34:20, I am testing a "good" led from a board that was burned. It was flickering in the movie. For some reason, in my previeous experiments, it didnt flicker at all, or I just didnt pay attention. I tested more other good leds like that, also from burned boards, and they all started to flicker after the heat really get into them, is my best guess. I lowered the voltage from my testing PSU, until I reach 6V for the 'good' led, and he is Not flickering anymore. And the light is a bit less powerful than it was at 8V, but still remains very powerful and very usable. THEN, to actually confirm to myself what the hell is the voltage for a good -brand new- power led, I opened a new light bulb, I took out everything from it and I started to read all the values I could from the electronic board and from the leds themselves. It turned out on a led bulb that holds x30 leds, 1one good -brand new- power led is working with 6V@100mA. So... I got smarter because I started this story with the burned power leds, from the wrong direction, like you said, frankenstaining the dead leds. So my 'good' led was overpowered at 8V. The 'bad' leds are totally and only absolutly working from 8V up @ 6mA per led. There was my mistake. But we have to start somewhere...
As I stated in the movie, I used a brand new led bulb, a bit weird construction, because it is using those 32V leds X8pcs per board. Now!!! I took out another led bulb, the same as the first one with X8pcs, and I am trying my best to construct a more DURABLE holding ring than the experimental thing I did with the first one. Here the experimental/prototype ring, still in use and no problems, tested on long time duration usage, working like nobodis bizniz.

I designed a little PCB with PWM dimming, adjustable current limit, and a variety of power input and regulation (for the control IC), which I've used for a while successfully, and recently paired with a small mains PSU based on a TI UCC28881 off-line switcher (which is an awesome little IC as long as your needs fall within it's output capabilities).

I looked into your UCC28881 pdf. It is a led driver like my BP9916D IC for x30 leds board.
(please zoom in the image to read the small values text)

Aaaa... so let me understand you better; In your project, you used your IC UCC28881 to drive only those 4 leds? And that is your full lamp? I like the part that you actually build it from components and using your brain. All I did was to remodel the board to fit my new ring shape, using the existing driver circuit. Im really not an electronist, haha, and I know it.

- Can you show me your circuit diagram with all the values in it? Please, and thank you !

 

IF the data sheet is the correct one for the IC on your board then from the datasheet and the value of R3 and R4 the LED current will be 15.3 mA. Using all the existing components is the best solution as we know the circuit was deigned to drive the LEDs correctly. You main problem now is providing heat sinking for the LEDs. I don't know where you can obtain aluminium substrate printed circuit board to make the circular PCB to fit your lamp.

Les.

Home Depot sells sheet aluminum. You will need to cut it to size. I use crazy glue to attach them to the aluminum. I've been using 3 watt LED's and running them at 1/2 power. The slight diminished light output is well worth the heat savings.

 

Home Depot sells sheet aluminum. You will need to cut it to size. I use crazy glue to attach them to the aluminum. I've been using 3 watt LED's and running them at 1/2 power. The slight diminished light output is well worth the heat savings.

Be aware that if the ground (well, either terminal, but it should be expected to be ground) of the LEDs has electrical conductivity to the base, you can't share a single heat sink ground plane for all of them. Shouldn't be an issue with LEDs scavenged from a single pcb-on-Al, but for example, the little 4-LED PCB pictured in my prior post - there are vias from each LED to a copper region on the backside of the PCB, and those are SEPARATE for each LED, there's solder mask over the regions on the backside, so when riveted to a piece of Al, they're still electrically separate.

I can confirm the superglue works for affixing large power LEDs to a larger heatsink - there's an electronics store in the Silicon Valley, Anchor Electronics, that has a 10W LED affixed to a piece of hardware store metal using superglue, and that's been there for years.

Heat savings isn't just about the heat coming off of the lamp, even if you can sink it well - under-driving the LEDs will result in longer service life.

I'll be in that store 3 days from now, I'll check and see if that's still there - photo is from ... 4+ years ago.

 

An update:
Now, I made the stronger version of the ring light !!!
I worked on it around 1 week, with breaks.
Please watch the 15min video in which I make an actual test of strength at 06:44

Yah, this is way better than the fragile thing I did before, Now I am happy.
The thing with this project it is very cheap to make, but time consuming and patience and manual work.
I was very lucky to find these unusual led (32V@1.2mA /led) pack of 8 leds per light bulb. That saved my ass.
I am unsure I can make it from a 30 leds disc. I think it is close to impossible, because you must have some space to cut between the leds and also the corresponding pads.
As you can observe, I like hackable things. Haha.
But mister @splud here built a circuit to power 4 leds. And my intention is the same, to build a power supply for a couple of leds from the 30leds disc, if not for all of them if I can cut them out all. I will be very happy to power 1 led from the mains. Right now I can power 1 led from my variable power supply and figure out the working voltage of the component.
But my goal is to make a constant mains supply. My idea is to use a triac-diac circuit. But I have to look into bta16-600b datasheet to get some power information.

 

I was very lucky to find these unusual led (32V@1.2mA /led) pack of 8 leds per light bulb. That saved my ass.
I am unsure I can make it from a 30 leds disc. I think it is close to impossible, because you must have some space to cut between the leds and also the corresponding pads.

I question how ever you are measuring the LED voltage. 32Vf is ... high for a single small SMD LED. See the image of the 10W LED, which would have been a 3x3 junction matrix) would have been 9.6Vf or so, 3x3.2V, and a higher current because internally, each series of 3 junctions is parallel to two others (in that particular LED) - it's really basically 3 strings of 1W LEDs. And even that isn't 32V.

32V @ 1.2mA = 0.0384W That's less an a bog stock 20mA white 5mm LED: 3.2 * 0.020 = 0.0644 W Not exactly "POWER LED" territory.

End-to-end of your string of 8 series LEDs would be approaching 30V. are you measuring across each LED, or the whole string? And, whatever you do, if you're connected to an non-isolated supply, exercise care when taking the measurements.

But mister @splud here built a circuit to power 4 leds.

Well, those PCBs have LEDs that are 2 junctions each, so it's more like 8 LEDs equivalent, and the circuit is capable of driving more or less, depending basically on input voltage (higher for more LEDs - the arrangement I posted is regulated to about 26V before the LED driver), and the feedback resistor that senses the current in the LED driver.

I still need to compose a reply to your earlier message.

 

I made a video and I post the video exactly in this page, about my project and how I made my circular lamp, but I'm not quite 100% happy about my project, even if it's working - excellent I may add.

Yea, saw the video on the second page of replies after I'd posted my commentary, though I must admit, even when I did see it, I didn't watch the whole thing. 30+ minutes is a bit much to dedicate to someone elses' project.

6.5cmx5.5cm (2.5inx2.1in for our american friends here). That round gray donut is made from cement !!! chinesium style.

Actually, concrete filled plastic "doughnut" forms are commonplace for things such as floor standing lamps, it's inexpensive formable ballast. The odd thing with your lamp is how it all just seems to be glopped on a plate, like dried frosting.

As I stated in the movie, I used a brand new led bulb, a bit weird construction, because it is using those 32V leds X8pcs per board. Now!!! I

Are you arriving at this 32V figure by dividing the measured input voltage to the whole circuit, or actually measuring across an individual LED (which would show you the Vf - forward voltage DROP - across that LED)?

I looked into your UCC28881 pdf. It is a led driver like my BP9916D IC for x30 leds board.

It is not. The UCC28881 is an "off-line switcher" - is is for providing a regulated output VOLTAGE. You could use it in a toaster (for control logic, not the elements), a doorbell, etc. LEDs are CURRENT driven devices (they need a minimum voltage to function, but above that, properly, you limit the CURRENT, not the VOLTAGE). If your supply has an inhernet current limitation - it cannot provide more than so many mA (such as many coincell batteries), then you can have something that appears to operate from a higher voltage without current limiting, but the limit is really imposed by the supply being UNABLE to provide more.

Aaaa... so let me understand you better; In your project, you used your IC UCC28881 to drive only those 4 leds? And that is your full lamp?

No, the UCC28881 is a HV switcher, capable of providing regulated VOLTAGE from mains-level voltages (after a bridge rectifier). The IC is rated for 700V, though in practice, you're likely to see a max of around 340VDC in (from 240VAC in, and obviously much less for 120V US residential mains), plus or minus a bit depending upon your utility - so the input cap I use is 400V rated. This is the power supply. However, in other builds, I've used a LiIon or LiFePO4 cell or a wallwart to provide power , depending upon the LED configuration I'm driving - but this UCC28881 board in a prior post, is capable of driving my configuration of 8 LED junctions (1/2W each, so each "LED" of 2 junctions is 1W) without the need for an external power brick - the small PCB fits inside whatever enclosure with the LED driver easily enough, and it's a LOT more current than you'll obtain from Capacitive Reactance unless you resort to a very large (and expensive) capacitor (but again, cap droppers are, I think taboo talk here). I have at least a second iteration of the PCB for the UCC28881 board to make, to add a MOV, fuse, use different sized capacitors (smaller, capable of being folded over with their body parallel to the PCB instead of perpendicular to it), resulting in a lower profile, to add a routed channel between HV and LV outputs, and to reposition the output pins so they're more compatible with being bridged to the LED driver PCB in a certain orientation (via a shared pin header rather than wires).

The LED driver which is pictured as a separate PCB (the black one) regulates CURRENT, and, in that particular configuration, sports a PWM dimmer (which is an optional configuration - if I don't populate the parts, the driver just drives at 100% of the set current). That PCB design has evolved over time, from a small PCB that piggybacked onto a commodity PIR module and was just on/off, to what it is now, with PWM dimming and multiple power connection and regulation options, so I can make use of it however I need to. There's even two tapered footprints for the current feedback resistor, allowing for using different sized resistor packages, and, if necessary a pair of them to "tune" to a more precise resistance based on available parts (two 1R5 resistors to achieve 0R75 for instance). I can also PWM from a µC input, allowing for dynamic dimming without using a potentiometer, like "soft" start if you will, or a separate non-contact proximity sensor input to switch between set brightness levels.

The PCB is by no means some stellar feat of design. I just needed something and wanted it to work in a variety of scenarios. I approach many problems with the mindset of designing something that can be adapted to multiple uses, rather than just the immediate need. All the moreso when sending off for PCB fabrication, and I might not need/want multiple PCBs for the immediate design need.

By raising the driver board, I wasn't intending to drive you to THAT approach, just that, using a driver - esp with dimming capability, IMO greatly expands the utility of the lamp, as maximum light output is often not ideal - sometimes you need all the light you can get, other times, not as much. Also, experimenting in a more standalone lamp, separate from the specific ring/magnifier setup you want to retrofit allows you to focus on the electrical nature of the design, get that sorted, then deal with the mechanicals.

- Can you show me your circuit diagram with all the values in it? Please, and thank you !

I'm not quite up to sharing the design at this time. Sans the dimmer and the multiple regulation and power connection options, the circuit is basically equivalent to that from the QX9920 data sheet.

 

...I question how ever you are measuring the LED voltage. 32Vf is ... high for a single small SMD LED.....
...are you measuring across each LED, or the whole string?....

My deduction is the most basic one possible. I am getting the 32V across 1 single led. Watch from this point (26m) in my video: like 4 minutes, until minute 30, where I present how I powered 1 led and also how I measure it.

.... the circuit is capable of driving more or less, depending basically on input voltage (higher for more LEDs - the arrangement I posted is regulated to about 26V before the LED driver), and the feedback resistor that senses the current in the LED driver.

I really need to see your circuit schematic !
This is why I open up the 30 leds light bulb and I look inside it, took all the measurements from the good working circuit there, in the hope I would be able to replicate the thing. It is over my comprehension to be totally honest. I kind of understand it, but not totally. I think the coil(s) (there are 2) are pumping up the voltage somehow. This circuit im talking about is in #26 answer.
Thank you for your interest !

 

@splud You updated your last answer AFTER my last post. And the website did not notify me about it. I had luck to refresh the page and look into it a little more up stream.
Excelent explanations !!!! I really like them all. Because I learn a thing or two from them.
Ok, I understand your point of sharing only the project and not the circuit in it, and it is ok. I'm glad you joined the cause here !!!
So...what I propose is, to collaborate into making a new and better (hopefully) circuit led driver or regulated voltage output (how you call it). Me with my imagination (artist) and you with the math (i imagine an engineer).
What I showed here already, with my custom power led ring project, was my interpretation of how to make it work, no matter how. It didnt matter, I just wanted it work. And it did, 2 times even !! Hehe. In a sense I cheated, because I didnt built anything, I just reassembled something already made into other shape and change it's destination. BUT... I did tried before this "genius" idea of reshaping, to build my own simplified driver for the leds. I used first the simplest [LED to 240V resistor current limiter] :

check my answer: #9 where I exemplify in detail what I used and did.
In summary, all those R1 I put in parallel smoked up, because immense power drop over R1.
I tested first, and smoked them hard enough, and after that I start to calculate the circuit and I realized I need a bigger power resistor.
@LesJones briliantly answered at #11 :



then I tried the capacitive dropper:

It works! but the light is supper dimmed down. It's good if you want a very dimmed signal led. It is providing under 10mA current per led, whatever the voltage that led requires, in my case 6V (not 8V as I corrected after awhile) for a Good scraped led; (and 8V for a burned scraped led, but still working).
Still, @LesJones briliantly answered at #11 :



But this capacitor calculation and practice is a bit over my ears, I never had any experience with it so im very new to this approach.
I like it very much since it looks like a door to a possible solution (in my eyes), but Im not good enough to try it (Yet !).
Im also suspecting it will have to be a fix capacitor of 1.27uF and not an electrolytic one.
I also have 100 fix capacitors brand new of 100nF@630V and theoretically if I will put them in parallel, to reach 1.3uF, I will need to use 13 caps. A bit chunky, but it will prove the theory, at least for my pretty eyes. It's a thing I consider to do !

My other options are now to try, like I already specified in #1 post:
(A)
" ...using Medium Power Transistors in parallel to reach that 38.4W limit for only 1 P LED..."
BD139 NPN 80V 1.5A 12.5W

From my calculations I got 38.4W over R1, but I trust more mister @LesJones calculations, hehe.
- If you compare this //Transistor circuit with [LED to 240V resistor current limiter] you will see they are the same circuit, I only changed R1 with a more elaborate circuit. My idea here with this circuit was to be able to manage all that huge power over R1 spread over a couple more stronger components than my fragile smoked resistors in paralel forming the R1 value.
Also, for my //Transistor circuit, I read an article, also here on AAC, and every transistor needs a < or = with 1ohm resistor (@0.25W?) on its collector . It's called "ballast" resistor, like a negative feedback. Because transistors in // like in my circuit, tend to get hot or/and burned because some very high instability in them. So that resistor in their collector will optimize them to work normally, and not get crazy.
(B)
using a SCR circuit (also specified in #1 post)

(C)
using a single High Power transistor like: 2sk2654. I have only 1 transistor in my stock, and it is big. I love it actually. But I also want to see him in action. It's properties: 900V 8A 150W N-MOSFET. I didn't made any circuit for this one but is an idea on the table.

-Both of these circuits I describe here, (A) and (B) are just ideas to try. I never try them out. I just thought of them as good candidates that potentially could drive all that power over R1. Feel VERY FREE to correct me at this point and bring your own ideas to the table, or continue my thoughts I already exposed with your possibly brilliant and more correct circuit ideas. If you can or if you desire, of course.
- If not, then I will still continue with what I got so far and smoke some transistors this time. Haha.
- Again, thank you for your last answer !!! I really like how you explained things there.

 

View attachment 245308

But this capacitor calculation and practice is a bit over my ears, I never had any experience with it so im very new to this approach.

As you need more and more current for a project, the capacitor grows ever larger. The type of capacitor which would be used here is a poly film cap, X2 or Y2 rated. A few hundred nF are easy enough to scavenge (say 330nF), but 1µF and up they start to get big, and you don't encounter them often. Some 2.2µF ones I have here are about 3.2x2x2 cm in size. The film caps though can handle AC, whereas an ElCap is polarized - bad things would happen if you dropped one of them onto AC.

the ESP wifi modules are something like 300mA at wifi association time. To supply that (and mind you, nothing for some other load the device might be driving, and there's a need to rectify and regulate as well), given a 240V 50Hz supply as much of the world relies upon, you would need: 1/(2π * 50Hz * (240V / 0.300A) ) = 0.0000039788 F (4µF). That's about what you'd need for driving a series of typical 1W LEDs, whether it be just one, or 60 or more of them.


Also, you should account for capacitor tolerance (see the datasheet for the part). ±20% is fairly common - so it could be over or under (IME, more often than not, under). In most cases, you might rather have an over-rated capacitor, but with reactance, that means you could be driving more current than desired, which is where things start to fail in a bad way.

In the circuits where you have a bridge rectifier, it should not be necessary to have an antiparallel diode. You would use this if you were driving the LED from mains, but the more popular/efficient thing to do is use antiparallel LEDs (often employed for LED light strings running from mains).

Burning off several orders of magnitude more energy than is used by the LEDs is not an efficient way to drive LEDs. Take your experiment with the resistors as enough of an foray into that. I suggest you to abandon the BJTs in parallel approach - it is unlikely to pan out much better than the resistor approach.

 

I took a new series of measurements today.
- I measured and labeled A,B,C,D,E,F,G,H,I,J,K,L all my burned "Good" leds. I got some fluctuating current while measuring, that's why on some you see multiple values, like 16,23,24mA. Probably bad connections, but I am usually connect everything the same way, or the heat up in the led, or just because it is 'touched'.
- After that, I took another measurement from a Brand New led, from an Aluminium disc board with x30Leds on it. I measured like 10 brand new leds from this x30leds disc board and I got the same results from each, with very minor fluctuations. I got this time 5.5V@20mA/led , a different value than my 6V@100mA as I got last time. Why? Because I get SF results many times in my tests and experiments. But this SF results, never put me down completely. Or maybe I got the measurement from another "good" burned x30 leds board.

I also re-read this entire page, https://en.wikipedia.org/wiki/Capacitive_power_supply ,which is quite short reading, and I don't agree with the circuit inside it. I've seen a ton of functional circuits without that limiting voltage bullshit stage they present in there. That is misleading and discouraging. It is what I think.
But other than that, it confirms everything we talked here and I will take a look more seriously over the capacitance calculation formula, to use it and apply it myself. Hopefully with good results. Will see. One interesting note, they are using 48 LEDs in that example, because the current is so low 3.1 V/20 mA/20000 mcd /LED. Definitely these PLeds are very different between them. It's not 1 single type, but they differ greatly for the Vf they start to work.

 

I also re-read this entire page, https://en.wikipedia.org/wiki/Capacitive_power_supply ,which is quite short reading, and I don't agree with the circuit inside it. I've seen a ton of functional circuits without that limiting voltage bullshit stage they present in there. That is misleading and discouraging. It is what I think.

I won't hold up that page as some stellar example of Wikipedia, but I don't see anything glaringly wrong with the math or the example diagram. The page is missing a flashing DANGER banner, and the example reference just drops a 1.5µF reference that isn't associated with the schematic at the top nor the example lamp to the right of the text. Poor editing by someone, but the math looks fine. If you feel it is misleading, you might need to back away from mains until you get a better grasp of what is going on there.

One interesting note, they are using 48 LEDs in that example, because the current is so low 3.1 V/20 mA/20000 mcd /LED.

48 * 3.1 = 148V (or if 3.2Vf, 153.6), so up close to the voltage of a rectified 120V AC mains(240V split phase, as seen in the US). Lots of LEDs for more light given the small amount of light emitted by each LED, but all in series (and below the rectified voltage), so it's one string of low current (before their example turns to making parallel strings, which complicates matters because each string doesn't have separate current regulation - once one of them fluctuates, the others start taking more current, and the LEDs suffer a cascade failure).

Note, you needn't run the LEDs all the way up to near the mains voltage. you could just as easily drive a single LED from a 20mA current limited supply - but that isn't much light. The same current, with all that voltage potential can drive 1 or 48 (or with higher voltage, more) LEDs in series.

For Power LEDs, you're best sticking with an LED driver. Seriously, the X2 rated capacitors with the necessary voltage and capacitance ratings quickly start to exceed the size of a compact PSU.

 

48 * 3.1 = 148V (or if 3.2Vf, 153.6), so up close to the voltage of a rectified 120V AC mains(240V split phase, as seen in the US).

I neglected to explain the rectified voltage - the peak DC voltage seen from a rectified AC source is AC Vrms * sqrt(2). So 120V typical US resi (the "240V split phase" - we do have 240V, but lights and televisions and coffee machines and the sort run from Line-Neutral) is 120 * 1.414 = 169.680. 170V for all intents and purposes. Voltage can run a bit high, or low (you've probably seen US voltage referred to as "110" sometimes). Do that math with 110VAC and refer to the quoted figures.

You're probably thinking "but I have 230 or 240VAC, why would the US voltage apply to an LED light sold here?"

If the number of LEDs * Vf is no more than the lowest supply voltage anywhere they may sell the bulb, the same PCB can be used worldwide. If the design uses a cap dropper, they're just populating with a different value for the cap and selling two models of the lamp which are 98% identical, and if it's an active IC, you're good to go, one bulb anywhere (except possible base design differences).

 

Thanks for your answers.
I managed to lit 5 burned "good" leds with a capacitive dropper that holds 3x100nF caps in paralel. Then I measured the current through all the leds and it was 17mA, so I decided to add more capacitors in paralel. On the schematic you can see point A and point B. There I inserted the ampermeter.
Looking over my previous tests for each led, named from A to L,

I could clearly see they should run fine at 30mA per each 6V led, so I've added caps and measured until I reached 27mA, in total 5 caps of 100nF.

But... it started to fail. And I realized it way too late, until ALL the leds burned. While I was testing the thing, the first 2 or 3 switch on/off , everything seemed to work fine. Then the first led burned off. After a couple of switching on/off, all they got burned. Then I realized I must have some spikes somewhere so I inserted that resistor of 150R on the A-B point, but it was already too late.
Life Lesson: Always presume there are spikes in the circuit and add counter measures for them. I think a diode in paralel with the electrolitic cap, pointing from - toward + rail. Like the ESD diodes. Or, some spikes smoothing circuit. I think an RC ? - i think it is called snubber? Im not that good with it, I only heard a couple things about it,but absolutly no practice.
I tested some of these leds, and outside of this circuit, using my variable PSU, and I got the led open barely i think at 6V BUT at 250mA !!!! so it is definitely (almost) fried. Aaah, the pain in the soul after them. Good experiment leds though. Eh.... And they worked quite good before the spikes. Damn those spikes.
You can see the cold white led on the white switcher it is lit, and the switcher is live when I took the picture, but all the P LEDs in the end of the circuit are off.
Or, because they were burned "Good" leds, they worked a little more, and thats it, they give up because they were dying anyway? It may be a possible cause.
I also think maybe you are right, to not try to revive the dead, by frankenstaining these leds. Very good word btw ! But I know for sure I push things beyond the limits, especially in my experiments, and not calculating the proverbial "shit" anything before I test. Somewhere in my soul I still think it is possible to make them work, but not with my knowledge and not with my limited experience. Its a challenge for someone more capable. I really think it is possible to reuse them. I failed for now, but I will learn to do better next time. If I remember anything from what I did...because I tend to forget the thousands of permutations I make. Well... will see.
I still have the burned "bad" leds and I have like 20 something of them, lighting very dimly but not as a power led, but more like a normal 3mm led. I'll have to test those as well.
These burned "good" leds I wanted to make a lamp with them, practically. But I'll probably take a good bulb and reshape it as I did with the circle before.
The burned "bad" leds... because I have so many, I want to make a board with them. A very simple luminous board. To put 2 piece of paper over it and transfer drawings between papers. That's the plan.

 

By far the simplest trick for getting a nice off-line LED drive circuit is to buy an off-the-shelf LED fixture (so stupidly cheap these days) and just ditch the LEDs and use the driver on your own LED array.

This video explains in detail how they work.

 

Thanks for your answers.
I managed to lit 5 burned "good" leds with a capacitive dropper that holds 3x100nF caps in paralel. Then I measured the current through all the leds and it was 17mA, so I decided to add more capacitors in paralel. On the schematic you can see point A and point B. There I inserted the ampermeter.
Looking over my previous tests for each led, named from A to L,
View attachment 245719
I could clearly see they should run fine at 30mA per each 6V led, so I've added caps and measured until I reached 27mA, in total 5 caps of 100nF.
View attachment 245717View attachment 245718
But... it started to fail. And I realized it way too late, until ALL the leds burned. While I was testing the thing, the first 2 or 3 switch on/off , everything seemed to work fine. Then the first led burned off. After a couple of switching on/off, all they got burned. Then I realized I must have some spikes somewhere so I inserted that resistor of 150R on the A-B point, but it was already too late.
Life Lesson: Always presume there are spikes in the circuit and add counter measures for them. I think a diode in paralel with the electrolitic cap, pointing from - toward + rail. Like the ESD diodes. Or, some spikes smoothing circuit. I think an RC ? - i think it is called snubber? Im not that good with it, I only heard a couple things about it,but absolutly no practice.
I tested some of these leds, and outside of this circuit, using my variable PSU, and I got the led open barely i think at 6V BUT at 250mA !!!! so it is definitely (almost) fried. Aaah, the pain in the soul after them. Good experiment leds though. Eh.... And they worked quite good before the spikes. Damn those spikes.
You can see the cold white led on the white switcher it is lit, and the switcher is live when I took the picture, but all the P LEDs in the end of the circuit are off.
Or, because they were burned "Good" leds, they worked a little more, and thats it, they give up because they were dying anyway? It may be a possible cause.
I also think maybe you are right, to not try to revive the dead, by frankenstaining these leds. Very good word btw ! But I know for sure I push things beyond the limits, especially in my experiments, and not calculating the proverbial "shit" anything before I test. Somewhere in my soul I still think it is possible to make them work, but not with my knowledge and not with my limited experience. Its a challenge for someone more capable. I really think it is possible to reuse them. I failed for now, but I will learn to do better next time. If I remember anything from what I did...because I tend to forget the thousands of permutations I make. Well... will see.
I still have the burned "bad" leds and I have like 20 something of them, lighting very dimly but not as a power led, but more like a normal 3mm led. I'll have to test those as well.
These burned "good" leds I wanted to make a lamp with them, practically. But I'll probably take a good bulb and reshape it as I did with the circle before.
The burned "bad" leds... because I have so many, I want to make a board with them. A very simple luminous board. To put 2 piece of paper over it and transfer drawings between papers. That's the plan.

When you plug that circuit in, if the AC line happens to be at or near the peak, you have a massive inrush current for a few microseconds while the capacitors charge.
This current is sure to roast your LEDs.

The series resistance must be sized to limit the peak to less than the rated max LED current.
I am not sure of what your line voltage is, but for example:

120 VAC rms is around 170 V peak, with your 150-ohm resistor, the peak current is 170/150 = 1.133 Amps.
Of course, if you are at 230 VAC... much worse.