I was browsing through eBay LEDs a couple weeks ago and came across something pretty interesting… a new COB that runs directly from AC. In other words, driverless! You may have seen them either on eBay or Aliexpress, but if not, they look like this:
(sorry for the rough picture – it’s the result of adding a poor camera to my already-poor photography skills)
Being the first “AC” COB I’ve seen on eBay, I was pretty intrigued. Sure, I made a custom directly-rectified-from-AC LED bar (link to write-up) about a year ago, but this was the first low-cost ($15 CAD or about $11-12 USD) and fully integrated design I’ve seen. It was worth checking out!
I couldn’t find much in the way of details on the web when I ordered it, so I figured I’d do a short write-up with some details (power draw, observations, etc). If you’re interested in this thing, maybe something here will help you.
First a few physical notes:
- This is the 110v model. They’re less common than the 220v versions.
- The aluminum substrate is about 1mm thick.
- There are 6 rows of 48 LED chips -> 288 chips in total!
- It’s big – dimensions are about 190mm x 53mm (not exact, but within +/- 1mm).
- It has L and N pads for soldering the load and neutral to.
- Model number on the PCB is AC110V-53190-F17020 A0
- There are 6 smaller holes (about 3mm diameter), likely intended for screw mounting. I suspect the 2 larger holes (just under 4mm diameter) are meant for running wires through (trace clearance on the L end is only 2mm from outside edge of hole so mounting larger screws there probably isn’t a great idea).
A closer look at a couple chips
I wasn’t keen on cutting away the silicon from each and every chip. However, for those interested, here’s what I did find:
The chip shown is (3 lines) BP5132H / 16D57K / G29B – I didn’t peel the silicon from *all* of them but based on the size/layout there appear to be 15 of them. It’s a linear constant-current LED driver from BPSEMI (product sheet here).
The 2 largest chips (in the 1st picture of the write-up, bottom-rightish) look to be bridge rectifiers – I only peeled 1 but it had the +/-/~/~ labels you’d expect.
Many of the smaller covered chips I would guess are resistors or capacitors and are all of similar physical size – I didn’t dig here.
2 chips that are different in size from everything else but which I didn’t dig at are the ones located between the 2 bridge rectifiers – no idea what they are but maybe in the future someone a little more accurate than me with an exacto-knife will reveal the details.
I don’t have fancy light-testing equipment, but did check the wattage and temperature.
To temporarily mount it, I used a finned aluminum heatsink that came with a Mars Hydro light (184x117x8mm). Soldered some lamp cord leads, added a little thermal paste, some electrical tape to hold it in place, and plugged it in.
When plugged in, wattage at the wall ran from 180-185 watts. Yep, that’s not a typo. This is the first LED-anything I’ve bought from eBay/China that actually delivered more watts than advertised! Note that I only tested 1 of the units so far. Also note that reading through some Aliexpress reviews of 220v variants (which seemed somewhat similar to my 110v unit physically) people seemed to be hitting slightly just under the advertised wattage (~14_ for a 150w, 9_ for a 100w, etc). So this higher-than-expected-wattage could be a 110v thing, could be related to our home power actually being closer to 120v, or could be a one-off. Update: When our voltage dropped from ~120 AC to ~116v AC, I ran other test and it pulled 167-168 watts. So there definitely appears to be a correlation between the line voltage and the power drawn/output. After looking through the datasheet for the “Smart” integrated controller this would seem to make sense. I’d imagine that if your line voltage was exactly 110 volts you’d probably get pretty close to the advertised 150 watts.
Either way, I was pretty impressed that it was actually delivering some real wattage.
I checked the power factor. PF was 0.92.
When it comes to temperature though, things became a little more concerning: Within 30 seconds the heatsink had gone above 80˚C (above 176˚F) – my electrical tape wasn’t exactly clamping down hard either so you can bet the chips and components themselves were well over 100˚C. Now to be fair, this heatsink is meant to handle something like 140w as part of a forced-airflow system. Having it upside-down taking 180+ watts with no forced airflow isn’t terribly fair.
But suffice it to say, this LED really pumps out the heat and you’re probably going to need a custom heatsink with forced air cooling to deal with it. And unless you’ve build something massive that can handle passive cooling, if your fan dies, your LED probably will die very shortly thereafter… assuming it doesn’t do something exciting first like melt out some wire insulation (or melt off the solder and drop a live wire).
I’m betting longevity will be an issue unless I put together a system that does some heavy cooling. There are just so many components that could fail (at least 36 I could visually see + each of the 288 LED chips themselves), and there’s a lot of heat to deal with.
A few really cool things:
- Driverless (price) – consider that even a “cheap” 150w driver is currently around $70 CAD (~$50 USD). The benefit of skipping the driver is huge price-wise. This might be the first time you can get 150W worth of LED lighting for under $20 (plus heatsink and wiring costs).
- Driverless (space) – I personally hate having to make room in custom enclosures for a driver. Being driverless means more space for heatsinks and/or fans (which you’ll likely need).
- Simplicity – Just hook the thing up to AC. No fussing with driver specs, a zoo of wires, troubleshooting failure, etc. If it dies, you desolder the leads and solder a new one in.
A couple down-sides:
Can’t be underdriven – want higher a higher lumens/watt ratio, the ability to control light intensity, or the ability to lower power to keep temps down for longevity/cooling reasons? Too bad.Update: Did a few tests (which can be seen in the comments). Short version is that dimmer switches can work, and applying resistors to the AC end works to dim also (down to below 1 watt total consumption). I’d prefer it if a component-swap on the board itself could be used to set a certain current level instead, but until someone really maps the thing out, using an external dimmer or resistors may be the route to go for now.
- AC-safety – you’ve got mains voltage at the leads with no isolation. Thus you’re probably going to need to mount it in an enclosure with a panel/lens so that it can’t accidentally be touched or shoot flaming debris at something if it self-combusts. You’ll have to be dead-serious about grounding and probably want to use a GFCI for extra measure.
It would definitely be interesting to see a circuit diagram for the “Smart IC” chips utilized, and see if it’s the BP5132H chips or others that actually offer the “Smart IC” behavior. Update: Located the datasheet at
http://www.bpsemi.com/en/Data/BP5132H_EN_DS_Rev_0.9.pdf http://www.bpsemi.com/uploads/file/20161215114728_476.pdf for those interested! In looking through some of the 220V COBs that used SMD’s it appeared they were using a CYT3000B (datasheet for it can be found here and product sheet for CYT3000C can be found here for those curious). Perhaps if someone really in-the-know dissects one of these and gets a little more information it’ll be possible to modify resistors to tweak the power being used.
In any case, for the time being, I’m still pretty pleased with these. No doubt if they start taking off, we’ll start to see more of them – as it is there are 150/100/70/50/30w variants (plus 50/30/20w variants in the more common square package), but if they gain enough popularity I wouldn’t be surprised if we started seeing adjustable ones way down the line or something else moderately cool.
Just ran a few dimming tests. Used the 50w rectangular version since I didn't want the heatsink getting too warm.
1) Old wall dimmer (at least 30 years old, rotating knob) - the LED smoothly reduces in light output until a "cut-off" point where it'll suddenly drop out and not recover unless the knob is turned way back up to high power and then down again. You can't get the light low enough to look into it and make out all the individual LEDs. Guesstimating here, I'd say that compared to an incandescent it reached around the 1/2 way mark brightness-wise before it cut out, possibly a bit further.
2) 2-setting lamp - works exactly the same as a Feit Electric "dimmable" E27 bulb: it flickers quite badly at the low setting, works normally at the full setting.
3) Resistors - works well. Got it down to under 1 watt which was dim enough that I could identify each of the led chips while it was powered (2K to 4K Ohm range for that). For reference, 940 Ohms (2x470 Ohm resistors) is about the point where it swapped between reading 0 and 1 watt on the Kill-A-Watt meter. No noticable flicker, though at those really low brightness levels I doubt I'd be able to tell if there were unless it was really really bad. I didn't test smaller Ohm values because these were pretty low-dissipation resistors I was using and I was doing the old jam-them-in-the-outlet-hold-the-plug-against-the-leads-and-hope-nothing-slips-and-shorts method.
I'll update the write-up to mention the dimming possibility.
Could you perhaps hook up a scope to the power leads to check the current draw waveform of the module, or even just do some high speed photography of it in action? Maybe just take a picture with 1/120 shutter speed with the camera in motion to get some sense of what kind of waveform they are driving these LEDs with?
My first thought was that they used multiple steps of parallel/series connections of the led rows as the voltage ramps up from 0 to 1.44*120V on the AC line.
Spent a bit of time this morning cutting away a bunch of the silicon goop that covers all the components in the "square" versions of these (20/30/50W) which also use the BP5132H - the PCB itself does have a bridge rectifier, so there is certainly AC/DC rectification going on, and the larger rectangle version (150W) has this too.
I don't have a scope or high speed camera, so I'm kinda limited to observations by eye plus anything I can ascertain with a multimeter.
As to steps of parallel/series connections, I think some of the component-exposed 220v variants (CYT3000 I believe) do light up sections incrementally as the voltage rises, but they're based on different chips. I don't think the BP5132H chips are being implemented in a way that mimics that behavior, but I could be wrong.
I also bought 50 W driverless led COB on eBay.
I was expecting it is flickerless, but it is flickering ;-p
I tried to DIY a LED bulb using the driverless 50W COD.
My working desk got super bright!!!
I feel 50W COB is the same brightness as 500W halogen lamp.
I checked 50W 110V COB power consumption.
AC100V 0.3A 26W 31VA PF0.85 2200Lux(d=50cm)
AC110V 0.5A 51W 58VA PF0.88 3100Lux < =Nominal ope.
AC120V 0.6A 64W 70VA PF0.92 3500Lux
Work with flickerless circuit:
AC100V 0.8A 52W 79VA PF0.66 3700Lux
AC110V 1.0A 79W 118VA PF0.67 4300Lux
AC120V 1.3A 105W 152VA PF0.69 4700Lux <=Perhaps dangerous
I'm driving the COB at AC100V with flickerless circuit.
If you want to drive with more voltage, you need to pay attention to reliable cooling.
Because, Heating rapidly increases at high voltage.
120V is twice the power consumption from 100V.
What are your thoughts on running any of these integrated-driver COBs in series with a 470nF or 680nF X2 capacitor to reduce overall output, decrease heat, and hopefully increase lifespan?
So if heat's the big issue (and I agree on the heat output from these things... it's crazy), then definitely taking some steps to lower the output can be a good path to travel down. It also tends to be cheaper than buying a massive heatsink. That said, I haven't delved into adding caps to these things so I can't really offer much input in terms of capacitor values or anything.
As a final note, I'm not sure if you've come across these yet or not, but "Big Clive" did a couple YouTube videos of these LED COBs a few months after I wrote up this post. His videos are pretty informative if you're interested:
- https://www.youtube.com/watch?v=UGTXne_e554 Adding a capacitor to a driverless LED and other tests. (YouTube)
- https://www.youtube.com/watch?v=KKd2L9Exw0M Driverless 50W LED teardown and schematic. (YouTube)
Anyway, someone running at 130-150˚C probably shouldn't expect a huge lifespan and may have to keep an eye out for wire insulation melting (possibly solder too if the gauge used is too low and it ends up being a heat contributor resulting in solder getting a little... soft).
As far as the 150˚F you're getting into (~65.5˚C) it's probably just fine. I've got thermal switches hooked up to mine which I believe disconnect them at around 70˚C (thermal fuses too in case those fail) and I'm pretty close to that number. Cooler is obviously better (effic and lifespan), but at some point it becomes hard to justify spending piles of $ on heatsinks and cooling for these dirt cheap LEDS.
The 150w variant of the 6000k LED pulls around 185w from the wall, and delivers 21100 lux @ 0.75m measured in a black chamber.
The 3000k version pulls around 190w from the wall and delivers 18100 lux @ 0.75m in the same condition.
A few notes: Just because the device pulls more than the specified 150w doesn't mean it's outputting more light than you bought. If anything, the 30-40w is due to driver inefficiency, and this certainly scales with heat. The hotter the board, the less efficient it is.
Some other considerations... These get hot. Fast. The tradeoff is the simplicity of installation and cost. As temperature ramps, so does power draw (as in all LED devices). I'm assuming the individual LED dice are rated for around 1w, and the entire board is being underdriven at an assumed 120vac (or 220vac). Manually turning the voltage up to 130v increased power consumption by 40% and output by around 15%. It's safe to say the 115vac voltage range is perfect for these SmartIC COB boards.
I'm currently running 2 150w COB's together, (both warm and cool white) on a copper watercooled coldplate. It keeps my delta T (as measured on the aluminum COB substrate) under 10c compared to ambient.
Also, pls note that you should have 15 led strings, not 6 as mentioned. Each chip drives a 10w led string... E.g a 50w led has 5 chips sering 5 led strings, etc.
Also keep in mind you'd need wire thick enough to handle all that current. You probably don't want to daisy-chain via the solder joint either when parallel wiring - may need to split the parallel feeds via a junction box or some other sort of distribution block and run out to each light individually. At that point it probably makes sense to individually fuse each one too.
All this really depends on what's available to you, how you're looking to set things up, and the electrical codes (especially if setting up a permanent or semi-permanent installation...) in your area.
I have a few of the 50W 220V leds, identical to yours. Bought them from aliexpress.
I have 2 questions if you would be so kind to help me, because i really don't know anyone who could help me.
1. On my chips the N and L seem to be reversed. More precisely, I use a brown-blue cable and the chip only works if I connect blue to L and brown to N. And here the problem starts. I need to connect each chip to a PIR sensor. How do I do it right - in the PIR conenction box should I follow the same pattern and connect blue to L and brown to N? Or I connect them normally according to the color code? (the connection box inside the PIR sensor case is color coded so even the noobs get it right :) )
2. My first attempt was to connect the PIR according to the led markings and that is blue to L and brown to N. It worked a few times and then the led stopped working. It heated to around 50 degrees Celsius. There are no burn marks. It just won't light up even directly connected to the socket. Is there anything I can do to "revive" it? Like getting it to an electronist to replace some of the chips or something.
Thank you very much in advance!
I'm almost wondering if your LED assemblies are either defective or are designed differently. There's a bridge rectifier so it shouldn't really matter which is L and which is N.
Another possibility (for #1 and #2) would be a bad/cold solder joint. Because of the massive surface area of the backing/heatsink, the solder pads are really hard to solder. If you had a poor joint in 1 orientation (but not when trying the other) it's possible it may have misled you. I needed really high heat on the iron and had to be quick doing mine - even at that a couple of my joints were poor and needed to be redone.
As for the LED dying, assuming it's not a joint, I did have 1 (out of 9) that died a quick/early death. You could feasibly start testing components but since lighting the LED chips needs a high voltage source and since cutting into all the components to separate/test stuff is really messy it might just be easier to replace the thing - particularly since they're not terribly expensive to begin with (assuming it's a standard warm/cold white LED).
Really not sure on the PIR sensor - if it switches AC, can handle the voltage+current, and doesn't do anything fancy, I'd be surprised if the orientation from PIR -> LED matters.
One last note: 50 degrees C is kinda cold for these things unless you've got it attached to a really hefty heatsink, are running for really short durations, or are running at a low voltage. If none of those apply, I'd wonder if there might be something else going on I'm not thinking of here...
Thank you for your quick reply.
A little update: I tried a second chip, this time properly connecting blue to blue and brown to brown in the sensor box.
The sensor is a PIR combo with daylight sensor. Nothing fancy, it runs on AC 230V 50Hz and has a maximum power of 1200 W.
Indeed the soldering is hellish. On the not working chip I have one connection that keeps detaching. Do these chips have something that prevents them from lighting if the soldering is not proper? I knew nothing about this.
With the second setup the chip works almost perfectly, I only noticed a small delay when the sensor receivs and transmits the signals (1-2 seconds delay). But in about 1 minute of testing this chip's heatsink got so hot I couldn't touch it with my bare hand.
I don't have good heatsinks. I have passive ones, made from aluminum profile a little bigger than the chips. They are made according to the manufacturer's specifications but clearly those are ineffective. I googled for
passive heatsinks for similar chips and found some like the one you have in the images above. Only thicker. So for now I will look for a few of those and use a different light source.
Regarding the N L markings, my suspicion is that the chips are marked wrong. I searched similar products from other manufacturers and found that all rectangular 50W chips are marked like mine. Considering many chinese manufacturers just copy the design from one another, this might just be a mistake in the original prototype which was carried on over and over. So in your chip the L is to the left and in mine L is to the right. Maybe this info will be helpful to other future readers of your post.
It can get very hot any recommendations
Ideally I'd try to keep temperature as low as possible, but probably below 80C if you want the light to last a while. That's still hot enough to cause burns (not "safe" from that perspective), but the light itself should hopefully survive a bit longer. Note that I have a few of these and did have one die despite being kept below 75C so YMMV.
As for recommendations, an actively (fan) cooled heatsink would be ideal. You could also look around to see what resistor values people have used to reduce the current on these things. One other thing I've done in the past is used NC thermal switches attached to the heatsink which switched off at around 70C though obviously this only works if you're fine with the light shutting off on you when it gets past a certain temperature.
I don't want to get an engineering degree just so I can find the best light for my money. help!
Cheap versions similar to these should be able to be found at Amazon for under $100 USD, many claiming a higher wattage. They're likely not very power efficient compared to custom builds with Bridgelux/Cree/etc, but as long as they're kept cool they should do the job fine. If you're looking for something non-blurple, you may be spending a bit more.
Beyond that, anything that puts out huge amount of light should work: Just avoid the really small dinky stuff that plugs into a standard lightbulb outlet: most of those leave much to be desired.