Broken LED light bulbs ---> Summer circuits fun

[RoGeorge]

What goodies might be inside a LED light bulb?  Depends.



Older models even have some toroidal transformers winded with colored wires! 
I like the toroidal core with red, yellow and blue turns.  I think I'll investigate that first.

[Renate]

Hmm, are the LED "filaments" (the sticks) series or parallel?
A 5 V source and a resistor will tell you if they are parallel.
If not they'll need ~60 V to make them glow.

[ftg]

Those ferrite toroids are often usable substitutes for those made from Fair-Rite material 43 in radio circuits.
Perfect for the output transformer in some 1watt shortwave power amplifier.

[Whales]

The colourful toroid is gorgeous.

I can smell them from here.  Looks like the power supplies are on a mix of phenolic resin FR1 boards (brown) and CEM1 (white only on one side).  Aluminium PCBs for the LEDs.  Through hole parts galore.  Quite the microcosm of PCB manufacture.

[RoGeorge]

Hmm, are the LED "filaments" (the sticks) series or parallel?

The LED filaments turned out to be 25 series LEDs for this particular light bulb, it was a bulb with a normal size E27 socket, but the glass was about only half the size of a normal mains light bulb.  Each junction/LED looks like a tiny brick of transparent glass-like material, with some visible traces inside (visible under microscope).  They are glued on a white, ceramic-like material.  The whole white-ceramic ruler + the transparent LED bricks are then covered in a silicon-like consistency yellow/orange phosphor.  The orange phosphor equally covers both sides of the white ceramic ruler, so in the end it looks like a small stiff orange bar.

The LEDs are in fact UV LEDs covered in a 2700K warm white phosphor (for this particular LED bulb).  There were 4 bars inside a bulb, one is still good.  Normal voltage drop on a single junction is usually bigger than 3V, but the 25s1p LEDs bar can light up at only 64V (the max my PSU can give), well barely visible, but enough to count the LEDs under microscope, and to see the "blue" UV diffusing into the phosphor while turning into warm white.

I happen to have a couple of pics with the orange bars, taken under microscope from the last weekend, when I was making a LED phosphor transplant from the orange bar to some other torch LED, in order to turn the flashlight from cold white into a warm white torch:  https://www.eevblog.com/forum/chat/today-ive-operated-a-led-face-transplant/ 

Those ferrite toroids are often usable substitutes for those made from Fair-Rite material 43 in radio circuits.
Perfect for the output transformer in some 1watt shortwave power amplifier.

That's exactly what I was trying to find out, if they are of any use for RF, thanks for the info!

Also the miniature-transformer is made from 2 solid E pieces of ferrite-like material.  It was a surprise to see it's not made laminar sheets of FeSi, as I was wrongly assuming. 

I can smell them from here.

About the smell, at a closer look, turned out the miniature transformer-like inductance has some burnt turns on it, and it smells... 


That particular board with the 3 colors toroid was in fact from a CCFL light bulb, not from a LEDs bulb.  Found a page with a very close schematic.  Not identical, mine doesn't have a diac, but the rest is close enough, except the transistors are DK 13001:  http://www.pavouk.org/hw/lamp/en_index.html

[RoGeorge]

That little toroid with 3 color coils has about 100uH for 9 turns.  Core diameters out/in are 8mm/4.8mm, and it is 3mm thick.

I've tried to get magnetostrictive resonance, as described here:  https://www.pa3fwm.nl/technotes/tn11a.html

Tried that with no application in mind, only out of curiosity, just that I couldn't see any resonance anywhere between 10kHz and 10MHz.  No matter how hard I've tried to magnetize it, it doesn't seem to care.  Even with 9 turns and 3A, the core still doesn't remain magnetized.  Once the DC current is removed, the toroid works like new. 

I've also tried to magnetize it with a small neodymium magnet, but still nothing.  The inductance changes drastically when the permanent magnet touches the toroid, but once the magnet is removed, it all comes to normal.  No matter what, the core refuses to remain magnetized. 

I guess this toroid was intentionally made from a material with low magnetic remanence, so to not become permanently magnetized if the oscillator sees some DC bias during its normal working, dunno. 

[RoGeorge]

Meanwhile found out that the typical (cheap) self-oscillating driver for CCFL (Cold Cathode Fluorescent Lamp) is based on a saturating transformer.  That colored toroidal transformer was meant to saturate back and forth during normal oscillations.  The toroidal saturating transformer is the Tr (Ns1, Ns2, Np), while the other larger E+E ferrite coil is the Lr from this schematic:



The schematic is figure 3 from DOI: 10.1109/41.753763, "Steady-state analysis and simulation of a BJT self-oscillating ZVS-CV ballast driven by a saturable transformer" by Yueh-Ru Yang and Chern-Lin Chen, a paper that describes in detail this type of CCFL driver:  https://web.archive.org/web/20190220123144id_/http://pdfs.semanticscholar.org/2865/8ee498e041eb71e6b2423bb7a73baada2a97.pdf

I guess a saturating transformer implies that its core must be a low remanence material, so to not waste energy/heat the core by permanently magnetizing it back and forth at each oscillation.  This might explain why the colored toriodal transformer didn't show any permanent magnetisation, and therefore no magnetostrictive resonance in the experiments from yesterday. 

[Whales]

Ooh I built a ZVS driver like that for a half-length fluoro tube above my closet.  Ty for the paper.  I run it off 6x really flat AA batteries, to the point where some of them get reverse charged by the others.  It's a brilliant circuit.  I've definitely never zapped myself on it.

My version was a bit simpler with no silicon other than the two NPN transistors, based of Ludens' schematics https://ludens.cl/Electron/Fluolamp/fluolamp.html



I mentally categorise this as a joule thief (self oscillating, runs down to very low voltages, very efficient) with 3 changes:
1. Transformer instead of inductor (I build some joule thiefs with just this change)
2. Push-pull LC resonant instead of single transistor resonant flyback (I think this bit is responsible for better time-utilisation of the magnetic core?)
3. Supply voltage boosted by sort of semi-converting it to a constant current supply with an inductor that sees pulsed currents.

[Sensorcat]

RoGeorge,

what you have discovered here is the difference between soft magnetic materials (used as flux conductors in applications with varying fields) and hard magnetic materials (used as permanent magnets). Your observations and conclusions are correct, I just wanted to add how it is called.

[Benta]

Wow! What a collection. I'm sure you'll have a lot of fun with those..   

[RoGeorge]

Wow! What a collection. I'm sure you'll have a lot of fun with those..   

Absolutely, already spent days with just a single magnetic donut and 3 pieces of colored wires, and the subject keeps branching into more directions than I can follow:

Dog's dream comes true!
njalakomboya


 

- learned about the ZVS-CV (Zero Voltage Switching with Capped Voltage) and the trick with the saturated transformer in series with a non saturated inductance
- the trick of starting a CCFL, how at first the (non saturable) ballast coil Lr + the tube filaments + ignition capacitor Cig oscillate together on a higher frequency, heats the filaments, then the tube eventually starts, and now the tube appears as a low resistive load, in parallel with Cig, which dampens the starting capacitor Cig, and the series resonant Csr becomes the dominant capacitor in the tuned circuit, and that it what gives the switching frequency of the ZFS.  So, before ignition the resonance frequency is given mostly by Lr+Cig, and after ignition by Lr+Csr, and the switching between the two modes is made by the arch inside the CCFL tube itself! 
- that also branches into some side quest reading about neon light bulbs.  I like neon light bulbs because of how they glow, and with time collected a few of them, with the intent of trying some very simple neon experiments.  Always wanted to try that, particularly because the experiments are so simple.  Yet, because of the region of negative resistance all kind of funny circuits can be done with neon bulbs, from simply making them glow, to oscillators, or even to use their glow as a 1bit digital memory, and make counters with neon bulbs. 
- then the links to all kind of funny builds from others (spent a whole afternoon yesterday browsing the https://ludens.cl/ , thanks for the link), and so on

the difference between soft magnetic materials (used as flux conductors in applications with varying fields) and hard magnetic materials (used as permanent magnets)

After finding out that's a saturable transformer (by design), the immediate question was "how easy would it be to saturate it"?  So I've tried to measure the saturation current of the tricolor toroid.

In theory, if you apply a constant voltage on a coil, the current starts to raise linearly, by the formula U=L*dI/dt.  As long as the core is not yet saturated, the coil can "oppose" to the increase of current (L value is large).  But once the magnetic core is saturated, all the magnetic dipoles are aligned inside the material, and they can not oppose much to the field increase (the L suddenly becomes smaller after the saturation moment), and the rate of current increase should become faster after the core is saturated.

With a digital oscilloscope, that should be trivial to see:  put the tricolor toroid in series with a small R (say, 1\$\Omega\$), then suddenly connect it to a voltage power supply (say, 1V).  Something like this:



The black/red alligators on the left are from the power supply (set on 1V, with limiting at max 1A), and on the right side is the oscilloscope probe measuring the voltage drop on the resistor(s) (which voltage will be proportional with the instantaneous current through the coil).  In the picture, the red alligator is not yet connected to the coil.  When the connection will be made, the current will start to flow, the oscilloscope will trigger, and we will see the evolution of the current through the coil.

If the trigger mode is set to single (so to trigger only once, and memorize that), a probe that reads the voltage drop on the small series R (the "small" R is that bunch of resistors in the pic, totalising 1 ohm) will indicate the current through the toroid, and the oscilloscope capture should show a raising slope, the angle of the slope being inverse proportional with the toroid inductance.

After saturation, the current increase should be faster (a stiffer slope), or is it?



Yes, it is! 



There are 2 segments with different slopes.  The voltage was apply at t=2us.  Then there is an almost straight slope until t=10us or so, then a knee curve until about t=12us, then another segment of a straight line, with a stiffer slope until about t=18us, then the slope becomes flat, max current being limited by the power supply.  The "missing" voltage up to the total 1V that we fed to the series LR is what was dropped on all the contact resistance in the alligators, on the wires, and on the binding posts at the PSU front panel at 1A.

The saturation knee in the toroidal core can be seen on the oscilloscope somewhere after the first 100mV of linear increase.  100mV drop on 1 ohm means 100mA through the coil, so the saturation current for this core with 9 turns on it is about 100mA (it's the tricolor toroidal transformer from the picture in the opening post, just that I've removed the red and the yellow turns).

[Whales]

Nice.  It looks like the power supply was sinusoidal? 

Another way to do it is to saturate with DC and measure the effects on a tiny AC signal.  You can put a small AC signal into one winding with a series resistor to measure current, then saturate the core with lots of DC in a second winding (using an ordinary adjustable DC bench power supply).

[RoGeorge]

The power supply was DC, a Rigol DP832 PSU, set to output constant 1V DC (at max 1A).  The yellow trace is the transient evolution of the DC current through the coil.  That's the first 20us after the coil was suddenly connected to 1V constant DC voltage.  In fact the Y axes is in mV, but since the probe reads the voltage on a 1 ohm resistor (1 ohm is the pack of small resitor together, I didn't have 1 ohm, made 1 ohm by 10 resistors of 10 ohms each in parallel), the 7 divisions on the Y axis (700mV) correspond to 700mA through the coil.

It's a one event, the yellow trace/waveform is not repetitive.  I've connect the 1V DC wire manually, by hand, and the oscilloscope captured what happened in the first 22 microseonds after that, then it freezes the image (the scope was in single sweep mode), and I've saved the capture and annotated it later, on the PC .



- at first, the coil is not connected at all (that correspond to the first 2us in the screen, the first division on the time axis)
- then, at the moment where the green X shows, I've connected the 1V DC to the toroid
- from that moment on, the current through the toroidal coil keeps raising almost perfectly linearly for the next 8us or so (the trace is following the orange/mustard slope segment)
- then, somewhere in the curved knee zone pointed by the blue arrow, (between 8us and 12us or so) the magnetic core becomes saturated
- after the core was fully saturated, the current through the coil keeps growing, again almost perfectly linear, but faster, this time following the red slope segment
- then, at the end the current through the coil eventually settles at about 700mA DC, and that's the maximum it will reach for a permanent 1V DC applied.  Should be in theory 1.0A, not only 0.7A, the 'missing' 0.3A are because about 0.3V are lost as voltage drops on the powering wires and alligators to the power supply.

Then, I've manually disconnected the 1V DC, but that event was a couple of seconds later, not shown on the oscilloscope.

The measurement was made using DC instead of AC, because the measuring setup using only DC was simpler, and because the oscilloscope was able to take a snapshot of the transient response, immediately after the 1V DC constant voltage was applied.

From that DC transient capture (the first 20us or so of the yellow trace), plus the number of turns and the mechanical dimensions of the toroidal core, should be possible to calculate the magnetic properties of the toroid, and thus to deduce the type of the magnetic material, but I didn't try.

[temperance]

Just subscribing.

[tooki]

Hmm, are the LED "filaments" (the sticks) series or parallel?

The LED filaments turned out to be 25 series LEDs for this particular light bulb, it was a bulb with a normal size E27 socket, but the glass was about only half the size of a normal mains light bulb.  Each junction/LED looks like a tiny brick of transparent glass-like material, with some visible traces inside (visible under microscope).  They are glued on a white, ceramic-like material.  The whole white-ceramic ruler + the transparent LED bricks are then covered in a silicon-like consistency yellow/orange phosphor.  The orange phosphor equally covers both sides of the white ceramic ruler, so in the end it looks like a small stiff orange bar.

The LEDs are in fact UV LEDs covered in a 2700K warm white phosphor (for this particular LED bulb).  There were 4 bars inside a bulb, one is still good.  Normal voltage drop on a single junction is usually bigger than 3V, but the 25s1p LEDs bar can light up at only 64V (the max my PSU can give), well barely visible, but enough to count the LEDs under microscope, and to see the "blue" UV diffusing into the phosphor while turning into warm white.

Just FYI, they’re not UV LEDs, they’re just regular 450nm blue — the entire white LED ecosystem is built around 450nm LEDs, since they’re the peak efficiency, and so the entire phosphor ecosystem is also focused on 450nm light sources. There are some very rare exceptions that use UV, but not mass-market items like LED filaments.

[RoGeorge]

Indeed, blue LED, not ultra violet.  Thanks for pointing that out! 

Don't know where from I've got the UV idea, maybe it got stuck with me from the times when the blue LED was not yet invented, dunno  .  Anyway, found a datasheet from a bulk producer of such filament LEDs, they specify the LED junction is with InGaN, so 450nm blue, not UV.

Single datasheet for a filament LED:  http://www.runlite.cn/userfiles/js9jaw90nxmvc1372646905.pdf
Full line Runlite catalog 2019-2020:  http://www.runlite.cn/userfiles/file/20190823/20190823090650_99555.pdf

Also, did an eye comparison for the perceived color, by comparing the light from the naked junction of a white LED (stripped from its phosphor) with the light of these two LEDs in the picture:



The upper LED was bought as "ultraviolet 395-400nm", the lower one is 450nm blue LED, both types are from an untraceable vendor, so not very sure about their wavelength.  The light from the phosphor-stripped formerly-white LED junction is not in the picture, but its color was about the same as the color of the blue LED, so most probably 450nm, but I didn't try to measure any of these wavelength.


At this pointy it's very tempting to cobble a spectrometer, so to confirm the wavelength of those Aliexpress LEDs. 

Now, if I were to make a light spectrometer, still have some CDs and DVDs around, but no Blu-ray discs.  Their track to track distance would make a uniform diffraction grating, just that the reflections will be a little curved.



I wonder if the CMOS or the CCD sensor from a USB camera could be used instead of a diffraction grating.  That should be not only straight, but also a more precise/uniform grid.  Or maybe reflect the incoming light on the window of an EPROM, and read the reflected diffraction with an USB camera.

The grid array of memory cells in a vintage EPROM might make a perfect diffraction grating, and with zero effort! 

[RoGeorge]

Meanwhile moved from CFL bulbs to LED bulbs.  The boards shown in the first post (those for LED bulbs, not for CFL) are all quasi-resonant DC-DC converters (QRC), sometimes called valley-switching converter.  There is a Fairchild application note describing how QRC works, and has a design example too.  Will let a link here just for the docs:

AN4150 - FSQ0365RN, FSQ0265RN, FSQ0165RN, FSQ311 Green Mode Fairchild Power Switch (FPS™) for Quasi-Resonant Operation - Low EMI and High Efficiency
https://www.alldatasheet.es/datasheet-pdf/view/162475/FAIRCHILD/FSQ0265RN.html

Typical low power LED bulb schematic has a 3 pins SOT23 SMD IC regulator, and a small EE ferrite core of about 12\$\Omega\$ and 5mH or so, various brands and various part numbers, though they are all similar.  Attaching a Google translated datasheet, and a typical schematic, just for the docs.

[RoGeorge]

Leaving a link, just for the docs:

"The µTracer V4, a sequel to the µTracer V3?", by Ronald Dekker

The uTracer4 project explored the idea of using transformers to make a tube tester with extended specifications compared to the uTracer3.  After a lot of experiments the project was stopped for a number of technical reasons that are explained in section 15.

https://www.dos4ever.com/uTracerlog2/tubetester3_new.html

Might appear unrelated, but that link is a very good read for understanding (pulsed) transformers.