[Swansea Hackspace] LEDs failed in my World Globe nightlight

[Brian J Hoskins]

Hello Richard.

*** SUDDEN DEATH ***

Q: Did this fault occur suddenly or has it degraded to the point that
you describe over a period of days/weeks/months?

*** SOME INITIAL OBSERVATIONS ***

>From the photos you've provided it looks like the PCB has an A.C. input
which is rectified by D1, D2, D3, D4.  I guess the large electrolytic is
a reservoir cap.

There are three transistors, so I am guessing that these supply outputs
for the three sets of LEDs (RGB).

There is a daughter board which is not shown very clearly in your
photos, but perhaps this is what generates the control to the three
transistors and is therefore responsible for rotating through the range
of colours.

Q: Are the LEDs connected in series or parallel?
If series, then a well designed circuit will supply a constant current
to each string.
A badly designed series circuit will supply a constant voltage to each
string, and this can result in one (or more) of the LEDs taking a larger
share of the available potential difference than the rest.  It will have
a very bright, and very short life.  When it dies, it takes the rest of
the chain down with it.

Still, I would not like to bet my house on an LED failure here.  The
size of my bet would depend on whether this happened suddenly, or
gradually, and whether one colour died before the other.

*** SOME DIAGNOSIS HINTS ***

First, pick a 'bad' LED and measure it.  A green one would be a better
choice since it'll have a forward voltage drop within the range of
pretty much any standard multimeter's diode test.  Using the diode test,
measure the forward voltage drop.  A standard green LED will drop around
2.2V at a decent If, but it'll measure lower than that on a DMM diode check.
If it measures in the ball-park, the LED is probably fine.  You should
see it light dimly as well.
Measuring a blue one is possible if your DMM's diode check function can
provide sufficient voltage.

If you find that the LED measures fine, repeat for all the rest of the
LEDs.  You might find one bad LED in the group - that probably means
they are connected in series and you just found the culprit.

If none of the LEDs are holding their hand up to say "I'm faulty!" then
I would turn my attention to the transistors and the daughter board.

*** LED POLARITY ***

You asked some questions about LED polarity.  Here are the answers:

* The shorter lead is (almost always) the cathode.
* LEDs usually have a flat side, and I can see that yours do.  This is
(almost always) the cathode.
* Inside the LED you'll see two electrodes.  One is noticeably larger
than the other.  The larger one is (almost always) the cathode.
* You can see that the flat side of the LED lines up with the silkscreen
marking on your PCB.


*** SOD THIS.  I'LL MAKE MY OWN! ***

A few years back I bought myself a colour changing LED cube from B&Q.  I
wasn't very satisfied with the limited range of patterns.  Also there
was no control to select specific patterns at all, and the transitions
between colours were... shall we say... "sub-optimum".

So I thought, fine, I'll take it apart and see if I can modify it!

I took it apart and, to my dismay, I discovered that it amounted to just
a single LED!!! No controller, no PCB, nothing.  It was one of those RGB
LEDs with a standard colour pattern built into it.

So I thought, fine, I'll throw this LED in the bin and make my own
controller!

So that's what I did.  Got myself a high power RGB LED, an 8-pin PIC
with three PWM outputs on it, and wrote some code to output colour
patterns with nice transitions.  Turning it on puts it into a random
mode where it runs a pattern to conclusion, maybe repeats it a few
times, and then selects a new pattern.  Pressing a button (which I
mounted into the bottom of the cube) runs a specific pattern and
repeated presses selects through the different patterns.

After that I thought, fine, would be good if I made a PC based pattern
generator and had a bootloader on the PIC to transfer the generated
patterns over.  But I never got that far.

So the point of me telling you all this is: could be an excuse to roll
your own RGB controller!



+--------------------------------------------+
| Brian J Hoskins BSc MIET                   |
| Electronics Engineer & Computer Programmer |
+------+-------------------------------------+
| WEB: | brianhoskins.uk                     |
| GIT: | github.com/bh4017                   |
| PGP: | keybase.io/bjh                      |
+------+-------------------------------------+





On 03/09/17 16:13, Richard Morgan wrote:
> Hi Hackspace List,
> I have a ‘Discovery Channel’ globe nightlight that was bought years ago.
> 
> It rotates through a range of coloured light that illuminates from
> inside the globe. It’s now only showing RED light which is not ideal for
> my Son’s nightlight.
> 
> I’ve opened it up and the problem appears to be the green(?) and blue
> LEDs are failing. Some slight illumination from a couple of the  non-red
> LEDs as it cycles through the colours and strong illumination from the
> reds. 
> 
> I’ve included some images of the PCB with the LEDs on it. I’d like to
> remove the current non-red LEDs and replace with new ones to see if this
> fixes it….but…..
> 
>   * How do I know which side to solder in the long leg of the LED to the
>     PCB? There are markings on the PCB where the LEDs solder in.
>   * Should I also replace the red LEDs at the same time?
> 
> How plausible is it that the non-red LEDs are failing at the same time
> due to the LEDs themselves failing? I could go through and replace the
> LEDs and find no benefit as something else is amiss.
> 
> How else might I troubleshoot this problem and what else could it be?
> 
> All help gratefully received,
> 
> Thanks,
> 
> Richard
> 
> 
> 
> 
> _______________________________________________
> Hackspace mailing list
> Hackspace at swansea.hackspace.org.uk
> http://swansea.hackspace.org.uk/mailman/listinfo/hackspace
> 

-- 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0x76760E4A.asc
Type: application/pgp-keys
Size: 4668 bytes
Desc: not available
URL: <http://swansea.hackspace.org.uk/pipermail/hackspace/attachments/20170905/fc76994d/attachment.bin>