LEDs failing intermittently in a LED matrix

[Axk]

Hello everyone!

I've a simple project with a handmade 16x24 LED matix soldered on a grid of bare wires on 75x50cm wooden frame and a driver board with a LED driver chip (HT1632) and a high side PNP driver for each of the 16 commons and a low side NPN dirver for each of the 24 rows + atmega 328P. The LED matrix is connected to the driver board by 40 wires (24 for rows and 16 for commons) about a meter long each.

The issue is that a small percentage of the LEDs (red wide angle through-hole LEDs) in the matrix (e.g. 7 leds out of all 16x24 LEDs) would not light up and this is intermittent with some of the LEDs not working then starting to work and others filing. Touching a failed LED would often (not always) bring it back to life, but then it may stop working later again. Applying voltage to a failed LED from a battery through a resistor would also most of the time bring it back. Sometimes a LED would not fail completely but get dimmer or flicker.

I'm testing this primarily with all LEDs in the matrix on.

#1 I though about bad solder joints and checked that and they appear to be fine.

#2 I also tried soldering a 100nF capacitor across a faild LED (also replaced the LEDs with new ones before adding the capacitor in case they are somehow permanently damaged) to remove the voltage spikes at the time of switching which I suppose are because of the wire inductance but then it would eventually fail gain with one of the LEDs which had a capacitor, but I'm not giving up on this yet, going to try to put a capacitor from each common to ground (by adding an additional ground wire to the matrix)

#3 I also tried upping the supply voltage from the initial 5V to 6V but that didn't fix it

#4 Tried removing a failing LED from the matrix and PWMing it separately from Arduino and it worked fine that way, but this doesn't show anything because the issue is intermittent and the same LED could have started working

#5 Tried turning on one row at a time in case the high side PNP doesn't provide enough current for all 24 LEDs in a column and some of them don't turn on - same result, some LEDs are failing


Schematic of a single common and row of the driver circuit (there are 24 identical rows and 16 identical commons):



A scope capture of a failing LED vs a working one (the yellow trace #1 is a failing LED):



A screenshot of the LED driver's datasheet with the diagram of the mode I'm using:

[sleemanj]

I think you are pushing circa 100mA through the LED? 

That's about 5 times more than a typical 5mm LED can handle constant-state for longevity, pulse current may be higher, but that's infrequent pulse current.

Replace the dicky leds, and reduce the current.

[Ian.M]

Also, mechanically stressing the legs can micro-crack the encapsulation and cause the internal bonding wire connection to the LED die to become unreliable.  You should always support the leads where they exit the body when bending them - either with a metal or plastic plate with a closely fitted slot for each lead, or by holding the lead with needle nose pliers between the desired bend location and the body, and *NOT* holding the body or other lead.  When wrapping a lead round a terminal or wire, you need to support the lead you are wrapping, avoiding pressure on the body or other lead. 

Physical handling damage to LEDs often results in total or near-total failure, intermittents, or even LEDs flashing on their own as the heat from conducting current causes the plastic encapsulation to expand disturbing the bonding wire's contact with the die or pin if that joint has cracked due to the damage.  It typically turns on, heats up, cuts off, cools down and turns back on, with a flash rate that can be anywhere between several times a second and a few times a minute.

[Axk]

sleemanj,

The measured total current is 800mA, 800/24=33mA so I think I'm fine on current. (Theoretically the LED current should have been (3.3-0.6)/42=64mA, but it is limited by the high side dirvers (I should probably lower the resistance on the high side drivers to increase the current)

Each LED is only on 16th of the time, so I suppose this warrants a somewhat higher current.

Ian.M,

This must be my problem because I bent the LEDs' legs at 90 desgree angles and I didn't follow the correct procedure you describe, so I suppose this mechanical damage is my issue.

Will try to locate all the LEDs that are failing and replace them with new LEDs with the leg bending done properly.

Thank you both for your help!

[mariush]

It's probably unrelated to your problem, but have you accounted for the resistance of the wire  ... you have meters of probably thin wire, which will have a bit of resistance so you may have some voltage drop.

Maybe it would make more sense to use something like 12v or 24v power supply instead of 6v and move some of the stuff closer to the leds to reduce the length of the wires?

Another suggestion ... you could buy some prototyping boards from eBay and solder the leds to the prototyping boards and use solder to thicken or tie several "traces" together for thicker wires.

For example, here's a 10 pack of 20cm by 8.5 cm pcbs for 6.5 $ : http://goo.gl/l2zBaw

You can cut the edges with some small saw to get a repetitive pattern and you're left with probably 18cm by 7.5-8cm pcb which you can tile with the other pcbs and you won't have thin wires going between leds and in theory it would be more reliable.

[Axk]

mariush,

Thanks for the suggestions.

I may do this in the future, but this would be an entire redesign of what I have at the moment.

I think the wires are thick enough for the currents in this design and for the resistance to be negligible a couple of ohms at most.

I don't have much headroom with supply voltage in this design to not overload the 3.3 linear regulator for the micro, the led controller and bluetooth module, but from what I see (SPICE and the real circuit) 5v is enough and convenient to power the circuit from a regular USB charger supply.

I realize this matrix construction is not very reliable, it was my original intention to have the matrix look similar to those christmas lights they hang in windows, so that it is a transparent grid, not an opaque one like with PCBs. I woldn't build a second one because it is labor intensive with stripping all the wires and soldering and isolating all the LEDs.

[Someone]

The measured total current is 800mA, 800/24=33mA so I think I'm fine on current. (Theoretically the LED current should have been (3.3-0.6)/42=64mA, but it is limited by the high side dirvers (I should probably lower the resistance on the high side drivers to increase the current)

Each LED is only on 16th of the time, so I suppose this warrants a somewhat higher current.

This does not work, you'll need to refer to the specific data sheet for the LED. For instance:
https://docs.broadcom.com/docs/AV02-1558EN
Average Forward Current, 25mA
Peak Forward Current, 90mA

There are LEDs designed for higher peak currents but they are speciality devices.

[Andy Watson]

Also, mechanically stressing the legs can micro-crack the encapsulation and cause the internal bonding wire connection to the LED die to become unreliable.

Also, avoid cutting both legs of the LED simultaneously with wire cutters. If you watch the action of the cut you will see that the metal of the leg is extruded longitudinally as the width is squeezed - this action can put considerable mechanical stress on the legs if both legs are gripped simultaneously.

[Axk]

With the failing LEDs replaced with new ones with the legs bent properly the problem appears to be solved, all LEDs in the matrix work fine now.

Thank you everyone for your help!

[Audioguru]

Were the LEDs cheap ones from ebay or alibaba? Then they are probably factory rejects and of course they will fail soon.

[Axk]

Audioguru, from ebay. Maybe they are. Work for now. Will see what happens in the future.

[Ian.M]

If yoiu are using cheap leaded LEDs for an art project it may be worth setting up to do a 24H 'burn in' soak test at 90% of max rated IF before installation.  Set up series strings of 10 LEDs each string with a current limiting resistor, all strings paralleled on a solderless breadboard.  If any string starts acting funny or goes out, bridge across each LED in turn with a known good one till the rest of the string start working, pull the bad one, and extend the test for that string by the time since you previously checked them.  I figure it would be possible to test 240 LEDs at a time in a typical three panel breadboard . . . .