Addressable LED hell

[cbowen]

Hey all. So, I've been using addressable LED's (WS2812b and SK6812) for a while in strip and breadboard-friendly form. I made my first prototype PCB's last week and started on trying my hand at soldering the SMD 6812's. Results have been pretty bad. Everything checks out on the PCB with a meter on all connections. I haven't been able to get more than two in a row working consistently. I started working with them as soon as I unsealed them. My iron was turned down. Just a quick dab of solder and get out, no melted plastic. These things are making me nuts.

Anyway, I started looking online and am finding out how delicate these things are. Really low reflow temperature, failure rates. etc. I think I'm about done with this experiment. So then I started thinking about other form factors. Through-hole isn't an option because there will be components on the opposite side. The breadboard-friendly ones though (seen here https://www.adafruit.com/product/1558?gclid=Cj0KCQjw2v7mBRC1ARIsAAiw34_7LFtj5KBFgzBI7Ko_i0QQKyPMl4eV25bOCzR2U8cMyCRUfZslQdIaAoalEALw_wcB) gave me an idea.

This is probably going to sound ridiculous, but I've never had a failure on these. What if I made footprints for these and soldered them on a bigger board? Advantages include: the LED's are already soldered on, they include the decoupling caps, and it would reduce board complexity because they're 6-pin instead of 4-pin. I figure, lay them on the footprint, fill the hole with solder, and profit? Board on board love? I'm an idiot, aren't I?

[james_s]

Hard to really say, have you had much practice soldering surface mount parts? I've not actually tried soldering one of those specific LEDs, are they in fact known to be fragile?

[cbowen]

My experience echoes a lot of what I've seen online. VERY sensitive to heat, some colors work, some stop, lights coming on with no instructions at all, etc.

I can solder the SMT caps and resistors seemingly fine. I guess for this prototype phase, I just want to get past this part to show proof of concept. If it ends up going into production and pick-and-place, it probably won't be a problem then.

[cbowen]

The iron is as low as it goes, the dial says 200'. I'm using 1.0mm 60/40, 2% flux.

Is the breadboard LED a ridiculous idea?

[Kalvin]

Have you seen this video? The soldering iron needs to have a very fine tip ie. only very little heat will transfer to the LED. Also, using soldering paste seems to be the way to go.

[Buriedcode]

I've soldered a few, not many, but I didn't have any problems.  I realise that doesn't mean much, but it could be a bad batch of devices, or your soldering technique/tip/iron.  Your temp seems far too low, which can often means one increases the time heating the pad/part for soldering, which can increase the temperature of the device.   Roughly a second is about all one needs to solder the standard 5x5mm 4 pad versions per pad.  With a low tip temperature I can't imagine that ever working.  So, whilst it may sound counter-intuitive - turn up the iron temp.  What iron/tip are you using?

Do you have a hot air station?  Although I'm fairly sure it will end up with the device being heated for longer, you could always attempt reflow with solder paste and heating the underside of the board.

Buying them in strips/packs and soldering those carrier boards onto your boards isn't such a silly idea, but you may have to glue them down and solder them to the main board with wires.  The ones in the link you provided could be done with wire plain tinned copper wire (TH parts cut offs?) and if that does the job, with greater reliability then why not?  But I would be curious as to whether or not this is how you're doing it or the devices themselves (entirely possible given their size).

[Mr. Scram]

Dave once did a kit with similar LEDs in the form of a ring led lamp. He used hot air to solder them on.

[artag]

I agree they're fragile.

I haven't been able to repair strips by replacing LEDs (they fail pretty often too. In fact, they're a bit crap).

I've resorted to repairing strips by cutting out a short length and replacing with more strip. I still have to be very careful, or just soldering the strips together kills the LEDs nearest the iron.

[artag]

You might also consider this type. Normal resin packaging instead of the white/gel package - probably a lot more reliable.

https://www.adafruit.com/product/1938

You did say not though-hole. Maybe lay the leads flat on the pcb, then bend the lens upwards ?

I've tried to solder through-hole pcbs together by filling the hole with solder, didn't find it very reliable. sometimes the bead of solder stays in the hole and leaves a film of flux between the boards. It's better to have actual wire. Or maybe solder paste between them instead of down the hole.

[janoc]

Your problem may not be heat but ESD - these LEDs are notorious for ESD sensitivity.

[cbowen]

Thanks everyone. There's a lot of good takeaways from this thread.

 - The multi-pad SMD video makes it look pretty easy. I'll try the paste syringe setup and try again.
 - I'll only handle with tweezers and not touch them.
 - I love the term "carrier board". It makes it sound so professional instead of a copout because you're not capable.

It's going to take a few days for the paste to arrive. Thanks again everyone.

[cbowen]

Oh, one other thing. This package makes it seem like the end of the world if I don't mount these within 12 hours of opening the package. Will I have to prep these again after the paste arrives?

[thinkfat]

Did you try leaded or lead-free solder? If these are really so critical to the temperature profile, maybe you'll have to get a low-melting-point alloy with a high amount of bismuth?

[cbowen]

So the paste arrived today. Super easy process, same result. So on every board, the first LED has worked, and second has intermittent non-existent functionality. Taking data off the distribution block and touching any of the LED's except the first results in small, dim flashes every now and then. Every connection tests out on voltage and continuity. So with all that consistency, I think this could be a power screwup.

In efforts to keep this a 2-layer board, I brought power from the terminal block down with almost each LED getting it's own trace. Ground however, took it's main trace to the last LED and daisy chains back to the first LED. I'm just going to go ahead and guess this is some catastrophically terrible practice. Please don't make me post a picture of the PCB screenshot, it will be so embarrassing..

[Raj]

Never had a problem like this, unless I ain't using variable power (like an inlne fan regulator)/ temperature regulated iron.
Ensure that your tip is grounded.
Have thick traces, so they wick away the heat as fast as possible.
Don't use lead free solder...they need higher temps...use leaded ones.
use lots of flux

[tszaboo]

I had a product with a WS2818 (or whatever) on it. It was designed by a junior engineer who then left the company before I took over the department. First thing that had to go. It was easily the most fragile component, causing errors in something like 5% of the time.
Board perfectly assembled, the BGA CPU runnign on it perfecly, dead 20 cent LED. I replaced it with a proper LED sold (probably not made) in the west, a TI I2C LED controller next to it. Sure it was more expensive, but the trade-off was 1% more expensive board for 5% of faulty boards. Ah yeah, and the LED had a tiny 8 bit microcontroller next to it, because the CPU running Linux couldn't handle the timing requirements. With it's own firmware and everything. Stupidest idea ever.

So dont place these POS LEDs on anything other than LED strips. It is just not worth it.

[jeremy]

I’ve used a ton of these LEDs without issue, and I have even cooked them until they are yellow and they are still fine 

They do however need quite good decoupling because they use pretty fast edge rates, and each LED has a re-driver for the serial data. Should have a 100n cap on *every* LED. It can also be pretty surprising how much power they use; I have tested some to be 60mA per LED when showing white, add a few together and you’ve got a good recipe for brownouts. So you should also have wide traces powering them, as much as you can cope with.

With that said, I much prefer the APA102C LEDs; they perform the same job, but are synchronous (ie use a clock as well as data) and are SPI compatible, so you can DMA data to them pretty easily.

Edit: also no, moisture doesn’t matter for your purposes

[Mr. Scram]

I had a product with a WS2818 (or whatever) on it. It was designed by a junior engineer who then left the company before I took over the department. First thing that had to go. It was easily the most fragile component, causing errors in something like 5% of the time.
Board perfectly assembled, the BGA CPU runnign on it perfecly, dead 20 cent LED. I replaced it with a proper LED sold (probably not made) in the west, a TI I2C LED controller next to it. Sure it was more expensive, but the trade-off was 1% more expensive board for 5% of faulty boards. Ah yeah, and the LED had a tiny 8 bit microcontroller next to it, because the CPU running Linux couldn't handle the timing requirements. With it's own firmware and everything. Stupidest idea ever.

So dont place these POS LEDs on anything other than LED strips. It is just not worth it.

It surprises me none of the real players has released a version yet. It seems plenty of people would rather use a part with an appropriate pedigree.

[janoc]

These parts are fairly fragile (ESD kills it, the driver seems latchup-prone, etc.) and some are outright dodgy, see e.g.:

https://www.eevblog.com/forum/projects/ws2812b-leds-reliability/

(some good tips there)

or:

https://wp.josh.com/2016/10/29/a-quick-test-for-crappy-ws2812b-neopixels/

Basically, if you want reliability, avoid these unless you have a 100% reliable supplier ...

[cbowen]

Thanks everyone for their input. I really, really appreciate it. I'm going to stop wasting everyone's time at this point, and redo the boards with carrier board footprints, just for proof of concept. I'm thinking place the holes over the pads, fill the hole with paste, iron needle tip in, and go?

@Buriedcode, you recommended using a wire. I'm not sure how that would work with surface mount pads. I'm afraid if the wire would be filling the hole, it would prevent the solder from making it down to the pad and having a good connection.

@Artag, Yes, I was planning on using the through hole variety where the plans would normally call for side-mount LED's to try to make the POC easier. Unfortunately, in the above case, I can't use them in this spot because there's components on the other side.

@Blueskull, that is very kind of you to offer your soldering skills.

@Jeremy, I've seen the Dotstars. So more reliable on the data end and smoother animations?

@Mr. Scram, I would happily pay $1 a piece for something reliable. : )

[Buriedcode]

@Buriedcode, you recommended using a wire. I'm not sure how that would work with surface mount pads. I'm afraid if the wire would be filling the hole, it would prevent the solder from making it down to the pad and having a good connection.
...

Ahh yes sorry, I meant to fix the "carrier" boards to the main board, not individual devices.

[chris_leyson]

Can't say I've had any problems hand soldering WS2812B's. Each WS2812B does need a 100n decoupling cap as jeremy pointed out. Also, I found that the first LED in the string usually needs a series resistor of a few hundred Ohms connected to it's data input, I tested a few Adafruit 8x8 Neopixel panels and without the series resistor only the first LED in the string would work and even then it was a bit random. 330 Ohms fixed it but I didn't get around to having a good look with a scope.

[cbowen]

yeah, all that is in place. Eh, screw it. I'll just post the PCB.

[Buriedcode]

Thats a cool shaped PCB.  One thing.. well a few things actually..

If you rotate each device 180 degree's then each output will flow to the next devices input.  Also, why not just two fat traces for VCC and GND running parallel with all the devices? That also makes decoupling caps easier since they are just across the rails.  Right now it looks like you're splitting up the ground or power rail into lots of thin traces. 

With bags of space there you could even have a ground plane - although I'd advise against it if you're worried about these things over heating during soldering. But you could still have two wide rails for the power lines and have a very short thinner trace coming off those to the device as a sort of thermal - still very low impedance, but without the heatsink.

[Liam]

Send a photo of what you got. It will be clearer what you have a problem.

My blog: https://wisepick.org/best-electric-toothbrush-for-kids/