Inside low-cost LED ac bulbs

[Dave_H]

I thought to start a thread about what's inside ordinary low-cost LED ac bulbs, as I have taking the plastic tops off a number of them. This is usually not a complete teardown, and does not include glass bulbs, such as filament bulbs.

Majority of these use high-voltage linear regulation versus switching regulation, which reduces large components and eliminates switching noise.

Voltage measurements have been taken, with care as lethal voltages are usually exposed; not for those not sure what they're doing.

To start, here is inside an Ikea 40W (3.3W) 450 lumen LED bulb. Regular CDN price is $1.99 for a 2-pack, which must be a loss-leader. The 60W version costs $6.99 for 2-pack. No way could they manufacture, package, and sell this bulb for $1, at any profit, even for the material cost with two PCBs.

Unlike most low-cost ac bulbs, this one uses switching topology, using two inductors, one for the switcher and other for noise filter. The 4-pin switching control chip is marked PT4558EN (Powtech), not much detail found.

Each LED package has vf around 16v which indicates 6 series LEDs. Two die are visible so each is 3 LEDs. 10 packages are wired two parallel, five pairs in series.

The black component on one line input wire appears to be a fusible resistor (10 ohms).

Dave

 

[turbodog]

I see some caps there that look like they will rupture after some time... killing the bulb.

 

[jtr1962]

Majority of these use high-voltage linear regulation versus switching regulation, which reduces large components and eliminates switching noise.

I retrofitted some old halogen yard lights doing the same thing. Four COBs in series, each with a nominal Vf of ~37V, gave a total Vf of around 148V, give or take. I used a full-wave bridge, and large enough filter cap, to eliminate any ripple large enough to drop the unfiltered rectified AC below about 150V. An LM317 operating as a constant current source completed the mod. I figured overall the driver was close to 90% efficient, which is better than a lot of off-line switching regulators.

So long as the manufacturers use a large enough filter cap to keep any ripple in intensity under a few percent this is a good design decision. Alternately, you can use a capacitor fed full-wave bridge to limit current. This has the advantage of not needing to have the total LED Vf close to the rectified, filtered line voltage (~169VDC in the case of 120VAC). I modded a few fluorescent night lights using such a circuit:

Fluorescent to LED night light conversion

I've been cleaning my workshop lately ( one reason I'm not on CPF much last few days ), and came across an old flourescent night light: I remember it had died about a decade ago, and I put it aside in hopes of eventually doing something productive with it. Truth is I was ready to toss it but...

www.candlepowerforums.com

 

[turbodog]

There were some _halogen_ bulbs with this setup. Sylvania made them. They ran a very high pressure bulb inside a 1/4" thick outer shell to contain ruptures. They monkeyed around with the voltage in order to gain efficiency AND bulb life. Were not cheap...

Sylvania Capsylite High Efficacy with Diode Rectifier

 

[PhotonWrangler]

There were some _halogen_ bulbs with this setup. Sylvania made them. They ran a very high pressure bulb inside a 1/4" thick outer shell to contain ruptures. They monkeyed around with the voltage in order to gain efficiency AND bulb life. Were not cheap...

Sylvania Capsylite High Efficacy with Diode Rectifier

I remember buying one of those. A rated life of 3500 hours seems fairly decent for a halogen bulb.

 

[Dave_H]

I see some caps there that look like they will rupture after some time... killing the bulb.

Electrolytic caps are not necessarily going to rupture. Over time the electrolyte can dry out, internal resistance goes up, and they stop working. In Ikea bulb these are marked 130C. In a good design with properly rated components and reasonable usage, the caps may or may not be the failure mode.

Most of the (other) linear regulated designs have some size of electrolytic for filtering. One was only 4.7uF. Knowing the load current it's possible to estimate ripple. I did a simple simulation using "TINA-TI" tool.

Dave

 

[Dave_H]

Plastic bulbs were removed from two Dollar Tree ($1.25-$1.50) 100W equiv. (14W) LED bulbs, one being a few years old (one mfr.) and newer (different mfr.). Both claim identical specs on package and bulbs: 14W, 1500 lumens, 15k hours, 3000K, CRI not given.

Both use full-wave line rectification with some filtering (capacitor hidden under PCB); high-voltage linear current regulator ICs (8-pin SOICs), and LEDs stacked in series or series/parallel with total Vf slightly below the dc supply voltage, which the regulators can manage.

Image shows old (left) and newer (right) PCB tops. The old bulb has 24 LED packages wired in series as 12 parallel pairs. The new bulb has only 15 LED packages, all wired in series. LEDs/pkg can vary but overall vf is about the same.

IC regulators are marked as EG1000A; found datasheet for "JXY" CYT1000A. There are least a half-dozen vendors of these linear regulators, typically good up to 60mA. The 100W bulbs use two in parallel. Current is set by a single resistor, in most cases (but not here) two in parallel to get desired value.

What I don't like on PCBs is exposed copper, visible on old bulb, and a bit on new bulb. Possibly there is a clear coating.

Bulbs have some series resistance between socket tip and PCB, typically 10-20 ohms. Without full teardown can't be sure, but likely as with Ikea, a resistor or fusible resistor for protection and reduce load on the regulator.

Both bulbs run quite hot on (under?) their collars. This is expected, as more than half the power is going into heat.


Dave

BTW DT Canadian price for LED bulbs has now risen to $2.00.

 

[Dave_H]

Here is inside two more DT LED bulbs, both 40W (4W), round vanity (top) and chandelier (bottom); same design except size of PCB. Linear regulation is using KP18001AESP chip, nothing unusual. LED vf is about 26v, which means 9 series LEDs per package, total vf around 130v.

On the smaller bulb I progressively increased value of current set resistor, from 24 ohms (nominal) to as high as 2000 ohms. At that point, current is around 0.75mA (0.09W) which is low level but works great as a nightlight (though not nearly as bright as jtr's self-built light), and runs absolutely cool so should last a long time.

Dave

 

[Dave_H]

Here is inside of an IR remote RGB LED bulb "Simply Tech" brand, picked up cheap as part of a set from Value Village; plastic diffuser removed. It has six each blue, red, and green LEDs. Red and green appear to be phosphor-converted (green is actually "lime"). Vf of all colours is around 12v, so these are quads.

The low-voltage controller 8-pin SOIC is (of course) unmarked. The high-voltage regulator 8-pin SOIC is marked EV221P but have not found data for it. The 3-pin IR receiver at 6 o'clock is as usual.

I don't see any substantial filter capacitor on the dc output, not sure if there is one hiding below.

It's amazing how well RGB mixes in the plastic bulb to produce cool white.

Dave

 

[Dave_H]

Here is inside a Bluetooth speaker plus IR-remote RGB/W LED lamp having E26 base, part of the set earlier mentioned. Note 10 RGB LEDs interleaved with 10 (cool) white LEDs.

Cool white is not very bright, less than 40W equiv. Colours are OK, with some mixed tints via remote keys. Warm white is theoretically possible by RGB mixing but not supported.

SOIC control chip at the centre has a long part number which looks like a microcontroller family, but have found no data online yet. The little wiggly copper pattern above it is BT antenna on PCB. Note usual 3-pin IR receiver at lower left. Small SOIC at 1 o'clock is audio amplifier for the speaker.

Whole thing is powered by 5vdc. Inside the base (not visible) is a small line-powered 5v supply, probably 2A which would be common with 2.1A USB charging blocks.

I had no problem with BT connection, played back a voice podcast mp3, sounded pretty good but obviously not a high fidelity.

Dave

.

Dave

 

[JustAnOldFashionedLEDGuy]

They are not loosing money. In the volumes these are made with Chinese domestic ICs which are sold on much smaller margins than other semiconductor companies and with LEDs super cheap their costs are shockingly low.

 

[PhotonWrangler]

Here is inside a Bluetooth speaker plus IR-remote RGB/W LED lamp having E26 base, part of the set earlier mentioned. Note 10 RGB LEDs interleaved with 10 (cool) white LEDs.

Cool white is not very bright, less than 40W equiv. Colours are OK, with some mixed tints via remote keys. Warm white is theoretically possible by RGB mixing but not supported.

SOIC control chip at the centre has a long part number which looks like a microcontroller family, but have found no data online yet. The little wiggly copper pattern above it is BT antenna on PCB. Note usual 3-pin IR receiver at lower left. Small SOIC at 1 o'clock is audio amplifier for the speaker.

Whole thing is powered by 5vdc. Inside the base (not visible) is a small line-powered 5v supply, probably 2A which would be common with 2.1A USB charging blocks.

I had no problem with BT connection, played back a voice podcast mp3, sounded pretty good but obviously not a high fidelity.

Dave

View attachment 49079.

Dave

I'm guessing Q1 thru Q4 are the individual RGBW driver transistors?

 

[Dave_H]

I'm guessing Q1 thru Q4 are the individual RGBW driver transistors?

Yes I believe so, the control chip likely does not have the drive capability for all the LEDs. Hard to tell whether they are MOSFETs or bipolar transistors without further measurement.

Dave

 

[Dave_H]

They are not loosing money. In the volumes these are made with Chinese domestic ICs which are sold on much smaller margins than other semiconductor companies and with LEDs super cheap their costs are shockingly low.

I have difficulty believing that the Ikea 40W SOLHETTA bulbs can be manufactured, packaged, shipped, stocked, sold and inventoried for $1 each with any profit. It goes far beyond material cost of the semiconductor and other electronic parts. Reputable bulbs need to meet safety (UL/ULC etc.) and EMC requirements (FCC in US, ICES in Canada).

The 40W cost of $1.99/pair versus 60W at $6.99/pair (in Canada) should be a clue. The 60W is dimmible, but that does not account for such large price difference.


Dave

SOLHETTA LED bulb E26 450 lumen, globe opal white - IKEA CA

SOLHETTA LED bulb E26 450 lumen, globe opal white This energy-efficient and long-lasting LED light bulb has a lifetime of approximately 25,000 hours and spreads a pleasant warm-white light that allows the room's natural colours to come into their own. The light bulb is not dimmable.

www.ikea.com

 

[Dave_H]

Here are insides of two 115v A19 LED plant grow bulbs, similar but different.

At top is 9W Luminus bulb with linear regulation and 16 identical LEDs each having vf 8.5v for total of 137v. Light is pink. Not sure what's inside the LEDs, looks like phosphor-converted (PC), possibly all blue LEDs with red plus blue phosphors; or maybe all blue with red phosphor and blue pass-through? Large capacitor is 15uF/200v, don't know how flicker would affect plants. Regulator is "Bright Power" BP5116, current set to 44mA.

Bottom image is Globe 9W bulb using two blue LEDs, and six orange/red which appear to be PC. Combined light is also pink. One blue LED has Vf=8.4v (3 series), other is 17.8v (6 series). Orange Vf is 17.5v, grand total about 131v. Linear regulator is JW19813 in tab package, LED current set to 44mA. Capacitor below PCB is unknown.

Both bulbs have sizable resistance in the base, 27 and 32 ohms.

Claimed red/blue spectrum of bulbs is similar, from package.

I am able to emulate the pinkish tint at much lower intensity. with a low-power RGB bulb set to one of its fixed red+blue tints, but no idea how the plants would like it, certainly not optimized for that.


Dave

 

[JustAnOldFashionedLEDGuy]

Reputable bulbs need to meet safety (UL/ULC etc.) and EMC requirements (FCC in US, ICES in Canada).

Linear regulator so does not even need to meet any EMC, since no oscillator. UL is simple for a bulb like this, and the MFR may have an in-house lab to keep costs minimized, but they are also potentially amortizing over 10-20 millions bulbs (likely make for different end customers), so $10K USD, is $0.001.

What do you think the cost is of the PCBs, parts, and LEDs?

 

[Dave_H]

Linear regulator so does not even need to meet any EMC, since no oscillator. UL is simple for a bulb like this, and the MFR may have an in-house lab to keep costs minimized, but they are also potentially amortizing over 10-20 millions bulbs (likely make for different end customers), so $10K USD, is $0.001.

What do you think the cost is of the PCBs, parts, and LEDs?

Unless anyone knows otherwise, I believe that a bulb being linear versus switching is not going to exempt it from EMC requirements, although it makes it a lot easier to meet them.

In any case, the Ikea bulb uses switching (LED supply is 82vdc). Another smaller E12 Ikea bulb with 8 LEDs is similar (76vdc), but includes series resistors which lowers efficiency
somewhat (still 105 lumens/W).

Ikea could likely have made it cheaper using linear regulation, would eliminate two inductors and one electrolytic, possibly the mylar cap; but they did not. Bulb reaches 135 lumens/W which is no slouch, so I guess it's worth it.

Package and bulb show FCC spec, and Canadian ICES-005(B) which is required to be marked on packaging (but is not).

I've seen costed BOMs but not for this type of product, have some idea of volume pricing of passives (chip resistors and capacitors) and other components in high volumes.

Dollar Tree LED bulbs 40-100W equiv. used to sell for $1.25 each, then $1.50, now $2.00, which is not all price inflation; more likely a sign they could not sell that low and still make a profit. Currently they seem to be the only products up to $2. DT bulbs are unbranded, and they cut out middle distribution costs, which helps.


Dave

 

[Dave_H]

Linear regulator so does not even need to meet any EMC, since no oscillator.

You may be correct, not sure about FCC req't, but ICES-005 (section 1.21(a)) appears to exempt bulbs without "any active/switching electronic components". Leads me to wonder why a number of linear LED bulbs are so marked, including DT's and EcoSmart from Home Depot (opened one of these too ), if they do not need to meet the spec.

Dave

 

[JustAnOldFashionedLEDGuy]

Unless anyone knows otherwise, I believe that a bulb being linear versus switching is not going to exempt it from EMC requirements, although it makes it a lot easier to meet them.
they cut out middle distribution costs, which helps.


Dave

YOU are guessing, I am not .....

 

[JustAnOldFashionedLEDGuy]

I've seen costed BOMs but not for this type of product, have some idea of volume pricing of passives (chip resistors and capacitors) and other components in high volumes.

Dollar Tree LED bulbs 40-100W equiv. used to sell for $1.25 each, then $1.50, now $2.00, which is not all price inflation; more likely a sign they could not sell that low and still make a profit. Currently they seem to be the only products up to $2. DT bulbs are unbranded, and they cut out middle distribution costs, which helps.


Dave

Okay, tell me what you think an SMT resistor goes for? What about the switching regulator controller? How about a say 60-100 lumens LED?

During the pandemic, a lot of prices went up large amounts due to component shortages. They are now mainly back to normal except MOSFETs supply is still not great.