CoB LED filament analysis

This still doesn't make sense, but you piqued me enough to make me do my own measurement. I don't see any phase shift between voltage and current, although the power factor is certainly far from unity. You can see my results here:

OK then. What is wrong with my interpretation? There is a sudden drop in the voltage waveform which implies a sudden increase in the current drawn. The current waveform confirms this. The slope in the voltage waveform suggests a capacitive component. There must be some hysteresis, which is the most puzzling element.

With 60 Hz excitation, there will be no time lags. Nothing will be sudden.

Instantaneous current will be about

I = (Vline-Vleds) / R

where R is the sum of the two end resistances. No time factor, no hysteresis.

This needs a proper measurement. DC with a DVM would be best, since the AC scope measurements aren't working well.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Iron-core current transformers get very nonlinear at low currents. The magnetic domains get sticky. You showed a gigantic CT.

Was the curve on the left done at DC?

Why is your voltage waveform flat-topped? At 100 volts?

At 30 mA, your 1K resistor drops 30 volts. So the "Voltage" connector doesn't represent the voltage seen by the DUT.

And does the "Voltage" connector go straight into a scope channel? No probe?

Your current waveform does have about the right shape and timing. No time lags.

The x-rayed COB seems to have a 4-diode bridge rectifier, two resistors, and one string of LEDs. That is more cost effective than having two strings, twice as many LEDs, each on half the time.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

The current flow through an LED is a non-linear - exponential - function of the voltage across it.

Essentially no current flows until you get close to about 3V across each LED. The series resistance then kicks in and holds the voltage across the LEDs to a bit over 3V per device.

There's no capacitative component. As soon as the voltage across the string drops below about 3V per LED the LEDs stop conducting. Spice it.

Few are as deeply flawed as Jim Thompson was. He once claimed to have denounced me to the FBI for anti-American attitudes. This was long after I'd been cleared to "most secret" by Australian military security, which had got me into US Army ECOM in 1970 when it was at Fort Monmouth New Jersey. Not the kind of denunciation which would have made him look good to the FBI.

That didn't worry me. What got up my nose was that John Larkin forwarded my e-mail directly to Jim Thompson.

And it wasn't "a job". John had said that he needed skilled help, and I'd done enough remote consulting by then to know that it could work, though it often fell over because it's hard to get enough information into an e-mail to inform proper collaboration.

In the event John didn't take the offer seriously and exploited it to score brownie points with Jim Thompson.

The character insights it gave still inform my behavior. Tony Williams had thought better of John Larkin than I did, and up to that point I'd cut John more slack than he turned out to deserve.

And if you don't use them carefully they can blow up on you.

No, it's the same data. I just plot I versus V. It's a bit noisy; I should have filtered it a bit.

True, the voltage recording is distorted by the superimposed current curve. That also explains the flat top. I could use a smaller current sense resistor, or compensate for its voltge drop.

I used regular 1:10 scope probes.

True. I expected my strings to be similar to Don's, but they weren't.

Exposing in some detail what I did lends people the ability to really understand what's going on and to propose improvements. Don's measurements leave too much room for mystery.

Jeroen Belleman

Not if they are used sensibly. How you managed to screw up your circuit by building it on a solderless bread-broad is anybody's guess.

Not that I'd ever use one. You do have to know what you are connecting to what, and anything that hides the connections or can introduce an unexpected connection that you can't see has got to be a bad idea.

Strictly speaking, a filament is something long and thin. The fact that the little bits of circuit boards with added light-emitting diodes get plugged into lamps originally designed around tungsten filament lamps doesn't make them "filaments"

Remembering Bob Pease isn't conjuring him up.

It has been a diabolical exhibition of confused thinking. If you find that entertaining, your attitudes need work. Piglet and John May did eventually introduce some sanity, but it took quite a while.

There's probably a respectable circuit to do the job, but wasting energy in current limiting resistors won't be part of it.

You probably need a rectifier with power factor correction to set up a more or less steady DC voltage across a capacitor and a switching driver to turn that into a well-controlled more or less constant current through a string of illumination LEDs.

There do seem to be some fast GaN switches around that could run at few MHz, and control a switched current though a nickel-zinc cored inductor.

Ideally you'd want a printed winding, so an EFD core or the like.

I've yet to see anything like it.

I replaced the flat dome ceiling lamps (there were four of them) in my flat a few years ago with LED-based units, but just bought four of them and paid an electrician to put them up - drilling the necessary mounting holes in a reinforced concrete ceiling needs more strength than I've now got. It took ages when I was younger. The lamps weren't designed to be dismantled and I didn't try.

The LEDs do seem to be well-spread out - the surface brightness is pretty uniform.

I'd fooled around with LED replacements for the linear quartz halogen lamps in the originals, but none of them proved remotely reliable, losing output rapidly over a few months as individual LEDs conked out.

The new lamps don't seem to be losing brightness.

But yes, the plastic solderless things are awful.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Bravo! If I remember correctly, you use a dental drill? My own freehand dremel on copper isn't nearly as straight. Perhaps my new milling table [1] will help. My media tends towards universal board. As an artiste associate, appreciate my abstract art arrangement:

Note how voltage and current perfectly align after the solderless protoboard is swapped out and replaced by my abstract art arrangement:

There's little, if any, wriggle room for the phase anomaly to caused by anything other than stray capacitance from the solderless breadboard.

Note.

Danke,

How about parallel and/or series R? Gunk between breadboard contacts or oxidized/loose contacts?

Ed

Yes! All of the above plausibly contribute to capacitive current curve corruption caused by my real world breadboard /system/, so to speak. You get it. It's critical to understand the difference between the idealized stray capacitance of 2, 3, or 5 pF, as mentioned in TROUBLSHOOTING ANALOG CIRCUITS by Pease and my breadboard system's real world stray capacitance. Although idealized components make theory easier to comprehend, it can be a mistake to assume idealized values in the real world. With this very project, for instance.

Danke,

I use a Drenel with a carbide dental burr.

That still doesn't look right. What's the value of the series shunt resistor?

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

One explanation is that the solderless breadboard wasn't connecting things. They tend to not do that.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

OK, there's a fundamental misunderstanding. The green component on the left hand side of my abstract art arrangement, entitled PROBE EMBRACES LAMP, is a fuse. Because you can never be too careful with 1:1 isolation transformer secondary Line voltages. The big red clamp on the right side is a current probe. (One of my goals is to learn how to use it.) Its P2 < P1 decal is oriented to point from a higher isolated Line to a lower isolated Line. And now that you mention it, the probe's current curve is relatively fuzzy. Because it's being pushed to its limit. If I remember correctly, a 100 ohm series shunt resistor paints a sharper trace. I'll drop the probe and swap-in a series shunt soon and share its sharper current curve.

Danke,

Do you mean you were using the current probe without its required shunt (not 'series shunt') load resistor? That would explain most of the problems, including phase shift etc.

Arie

Yes, that's precisely what happened. At the very beginning, when this topic was still embedded in another thread, I openly asked if anyone knew how to use the i1000s Fluke current probe. Yet you're the first person to offer any insight into its operation. Fluke's own user guide is almost as useless as AI in this regard. AI keeps circling around the same old set of websites using different words to regurgitate its useless operating instructions with each iteration.

Perhaps now that the key phrase "required load resistor" is known, my search results will return a useful operator guide. Thank you for your insight.

Danke,

Yes, that's precisely what happened. At the very beginning, when this topic was still embedded in another thread, I openly asked if anyone knew how to use the i1000s Fluke current probe. Yet you're the first person to offer any insight into its operation. Fluke's own user guide is almost as useless as AI in this regard. AI keeps circling around the same old set of websites using different words to regurgitate its useless operating instructions with each iteration.

Perhaps now that the key phrase "required load resistor" is known, my search results will return a useful operator guide. Thank you for your insight.

And, by the way, there's yet another fundamental misunderstanding. In place of a full wave rectifier, the CoB LED utilizes two half-wave rectifiers, one at each end. Per piglet's xray interpretation, each half-wave rectifier consists of two diodes.

Danke,