Addressable LED's ?

oi18ct · Posted: Fri Jul 01, 2022 5:59 pm Post subject: Addressable LED's ?

Hello AVR/Bascom people. I am knee deep in developing an industrial grade control product using Addressable LED's from Inolux that are quite similar to the WS2812 or neopixel in control.

As I continue with prototypes in this project, I am experiencing failures in manufacturing related to thermal reflow even within the parameters recommended by the manufacturer. I have solved this for the most part, but it's making me question the reliability of this type of LED technology for long term use in Industrial environments where conditions are not always predictable.

SO... I am wondering if anyone in the forum has used these types of LED's in Industrial or aggressive environments. I do not see the likes of Osram, Cree or Lumileds making them, nor is there an automotive grade variant. I am quite used to high yields and long life from Osram and others.

Thanks in advance for any feedback! I am looking forward to posting my completed product running Bascom on a AVR32DB128 soon!

albertsm · Joined: 09 Apr 2004 Posts: 6198 Location: Holland

I have a customer that is using chips with external leds. there you have the option to use proper leds. i only dont recall the name of the chip. but it was using similar protocol and timing.
in this case external leds were used because more light was required.

good luck with your project !

i.dobson · Posted: Sat Jul 02, 2022 7:33 am Post subject:

Hi,

I had a project using WS2812 and production was a real problem. Even though the correct reflow profile was used (Professional PCB manufacturer) we had a 10-20% failour rate where at least one LED wouldn't work per PCB (9 LEDs per PCB). Manually desoldering and soldering a new LED solved the problem most of the time, but cost us alot more due to manual interventions.

For the prototypes I manually soldered the LEDs using a lead solder and a very low temperature, even there I had a failour rate of about 5%.

In the end we managed to create the PCB's that we needed for production but is was really hard work.

looking at the LED's that died under a microscope I could see that most of the time the bonding wires where brocken.

Regards
Ian Dobson

SZTRAD · Joined: 30 Dec 2019 Posts: 165

Hi
if it's not too heat stressed, use Sn42Bi52 paste. Don't forget to adjust the temperature profile. This paste has a melting temperature of 141°C . So really to places where the temperatures of the switched chip will not exceed half of this temperature.
If you're using standard lead-free paste, you're exceeding the chip's allowed parameters. The profile for Sn96,5% Ag3% Cu0,5% is given at about 220°C, the max temperature of WS2812 is 200°C.
Standard Sn62% Pb36% Ag2% paste has a temperature of approx 180°C therefore you have a lower failure rate.

RS

oi18ct · Posted: Sun Jul 03, 2022 4:37 pm Post subject:

Excellent information, thank you all. I'll share some more of my experiences in developing this product (it's been going on for quite some time).

I.Dobson: In my early evaluations (~3 years ago), I saw the same thing- failure at the bonding wires on WS2812. On that alone, I decided the World Semi product would not do well in potentially high vibration environments (Industrial machinery like stamping presses and assembly machines). More manufacturers started making these type of LED's, and the Inolux people came around one day promoting their product. My initial tests were favorable and I have not seen a bonding wire failure yet. Unfortunately, I had to put things on hold for about a year and go make some $ to pay for all of this- when I came back to it, the same reel of LED's had been open for a year and the failures are now moisture related (see pdf images). Baking them at ~150C for 48 hours seems to have helped significantly.

MCS/Mark: I am glad to know you have put these into industrial applications. I am modestly concerned these LED's are for hobby projects only... asking the factory about that is fairly pointless as their interests are selling product. Internet searches produce endless projects that are unrelated to industrial applications (automatic cat bowl feeders, lava lamps, etc). Since these networked LED's allow for a single point of failure, I really want to have a grip on their proper manufacturing & application. A lot of effort has gone into the power supply as a result.

SZTRAD: Thank you!!! You are confirming what my production efforts have revealed to date. Our standard lead free solder is Indium 3.2, and it's worked well on everything for 13 years. Except these LED's... When the LED's were fresh out of packaging, I made good boards at 230C peak with Indium 3.2, however, that wasn't a big enough sampling & luck was probably on my side. I have made better boards with this same paste at 217C peak, however, the solder fillets are not ideal. I like the idea of a Bismuth Paste at lower temperatures, but I read it can be quite brittle- which is concerning for applications of our product in potentially high vibration machines. I read about Indium Durafuse LT as a formulation that is low melt with high impact tolerance. I have an inquiry in with them and it seems like a reasonable next step to try that out, especially taking into account your feedback.

It would seem like the additional process of an optically clear conformal coating will add mechanical strength to the Bi based solder joints.

I have attached a pdf with images of the LED's. I left out images run at 240C because the damage is so obvious it's comical.

many thanks for your input!

SZTRAD · Joined: 30 Dec 2019 Posts: 165

Hi
the temperature profile is correct for the original chip. It's just a question of what is original today.
What you have in the last pictures is the humidity. Didn't you have the parts unpacked for a long time before fitting?

albertsm · Joined: 09 Apr 2004 Posts: 6198 Location: Holland

please notice that i refer to the WS2811 which is a chip without led. it uses external leds. it is reliable and you can chose the leds yourself. but it uses more space and will also cost more.

what is real and fake is hard to tell. when there is a demand for items, counterfeit pop up. 30 euro razor blades i understand but some times even low budget stuff is 'cloned'.
you can request a document of origin so you know it is a genuine product. mouser, farnell, etc. request those documents from their sources.

you could make some device that will vibrate and shake to test the pcb.
there are also newer SPI based leds, these are real shift registers without timing problems. dont recall the brand or name. but you should be able to find them.

oi18ct · Posted: Sun Jul 03, 2022 9:33 pm Post subject:

thanks for the clarification. I will look into the spi command type LED's.