Searching \ for '[EE] Production LED testing' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/displays.htm?key=led
Search entire site for: 'Production LED testing'.

Exact match. Not showing close matches.
PICList Thread
'[EE] Production LED testing'
2011\08\01@231709 by Vitaliy

face
flavicon
face
We have a device with four LEDs, and we'd like to make sure the LEDs work. Visual inspection at the factory, proved to be unreliable. Does anyone have experience with reasonably priced machine vision systems that can be configured to perform this sort of test?

All we want to know, is whether the LEDs turn on and off. Making sure they're the right color is a bonus.

Vitaliy

2011\08\02@005752 by Mohit (Lists)

picon face
I'm sure you must have thought of this, but anyway...

You could design a PIC based test device jig with photodiodes (or some kind of light sensors) that sense if the LEDs are lighting up. The photodiodes could be spaced just like the LEDs in you device and have something around them to block stray light.

For colour sensing, you could place a coloured filter (same colour as the LED under test) in front of the photodiode.

If everything is good, the PIC could light up a green LED for PASS and a red LED for FAIL.

HTH,
Mohit.

Vitaliy wrote:
> We have a device with four LEDs, and we'd like to make sure the LEDs work..
> Visual inspection at the factory, proved to be unreliable. Does anyone have
> experience with reasonably priced machine vision systems that can be
> configured to perform this sort of test?
>
> All we want to know, is whether the LEDs turn on and off. Making sure
> they're the right color is a bonus.
>
> Vitaliy
>

2011\08\02@043024 by alan.b.pearce

face picon face
> We have a device with four LEDs, and we'd like to make sure the LEDs work..
> Visual inspection at the factory, proved to be unreliable. Does anyone have
> experience with reasonably priced machine vision systems that can be
> configured to perform this sort of test?
>
> All we want to know, is whether the LEDs turn on and off. Making sure
> they're the right color is a bonus.

Webcam ?

Board should be in a jig to get it correctly placed under the camera. Software then checks to see if light appears at correct time within given region..

With a colour camera it 'should be possible' to 'just' take a snapshot JPG and process that to look at each colour plane for the same information to determine colour shown.


-- Scanned by iCritical.

2011\08\02@044844 by Mike Harrison

flavicon
face
On Tue, 2 Aug 2011 08:29:39 +0000, you wrote:

>> We have a device with four LEDs, and we'd like to make sure the LEDs work.
>> Visual inspection at the factory, proved to be unreliable. Does anyone have
>> experience with reasonably priced machine vision systems that can be
>> configured to perform this sort of test?
>>
>> All we want to know, is whether the LEDs turn on and off. Making sure
>> they're the right color is a bonus.
>
>Webcam ?
>
>Board should be in a jig to get it correctly placed under the camera. Software then checks to see if light appears at correct time within given region.
>
>With a colour camera it 'should be possible' to 'just' take a snapshot JPG and process that to look at each colour plane for the same information to determine colour shown.
>
>

A couple of alternative approaches, on the basis that if a LED's electrical characteristics are
right, it's highly unlikely that it  isn't lighting: a) Configure the circuit so the LED forward voltage could be measured using a spare ADC pin - this
could also discriminate incorrect colours due to different Vf
b) Measure the current draw of the device while the LEDs are operated in a known sequence - this
would catch opens and shorts.

2011\08\02@093414 by Kerry Wentworth

flavicon
face
Vitaliy wrote:
> We have a device with four LEDs, and we'd like to make sure the LEDs work..
> Visual inspection at the factory, proved to be unreliable. Does anyone have
> experience with reasonably priced machine vision systems that can be
> configured to perform this sort of test?
>
> All we want to know, is whether the LEDs turn on and off. Making sure
> they're the right color is a bonus.
>
> Vitaliy
>
>   An LED can be made to detect light.  Put a red LED next to any red LEDs on your unit, green next to green, etc.  Make sure that the detecting LEDs have colored lenses, even if the LEDs on the unit are clear.

Kerry

2011\08\02@094235 by Yigit Turgut

picon face
On Tue, Aug 2, 2011 at 6:29 AM,  <spam_OUTalan.b.pearceTakeThisOuTspamstfc.ac.uk> wrote:
>> We have a device with four LEDs, and we'd like to make sure the LEDs work.
>> Visual inspection at the factory, proved to be unreliable. Does anyone have
>> experience with reasonably priced machine vision systems that can be
>> configured to perform this sort of test?
>>
>> All we want to know, is whether the LEDs turn on and off. Making sure
>> they're the right color is a bonus.
>
> Webcam ?
>
> Board should be in a jig to get it correctly placed under the camera. Software then checks to see if light appears at correct time within given region.
>
> With a colour camera it 'should be possible' to 'just' take a snapshot JPG and process that to look at each colour plane for the same information to determine colour shown.
>
>
> --
> Scanned by iCritical.
>
>

2011\08\02@194516 by YES NOPE9

flavicon
face
Webcams ( such as the C901 ) will work with OpenCV ( under OS X , Windows and Linux ) to capture images ,
manipulate and analyze images.
If you are interested, I can offer more details.
99guspuppe

2011\08\03@014952 by Vitaliy

face
flavicon
face
Mohit wrote:
> I'm sure you must have thought of this, but anyway...
>
> You could design a PIC based test device jig with photodiodes (or some
> kind of light sensors) that sense if the LEDs are lighting up. The
> photodiodes could be spaced just like the LEDs in you device and have
> something around them to block stray light.
>
> For colour sensing, you could place a coloured filter (same colour as
> the LED under test) in front of the photodiode.
>
> If everything is good, the PIC could light up a green LED for PASS and a
> red LED for FAIL.

Yes, we've thought of this. Building an in-house test jig is plan B. An off-the-shelf solution doesn't have the uncertainty, and it seems to me that this sort of recurring problem necessitates the existence of a relatively inexpensive camera-based solution.

Vitaliy

2011\08\03@015811 by Vitaliy

face
flavicon
face
Mike Harrison wrote:
> A couple of alternative approaches, on the basis that if a LED's
> electrical characteristics are
> right, it's highly unlikely that it  isn't lighting:
> a) Configure the circuit so the LED forward voltage could be measured
> using a spare ADC pin - this
> could also discriminate incorrect colours due to different Vf

Sorry, no spare ADC pins.


> b) Measure the current draw of the device while the LEDs are operated in a
> known sequence - this
> would catch opens and shorts.

This sounds like a good idea, especially since we're measuring the current anyway. Unfortunately, one of the LEDs is wired permanently ("Power") and one is driven by a third party module (it blinks). It may still work, though.

Vitaliy

2011\08\03@015924 by Vitaliy

face
flavicon
face
Yigit Turgut wrote:
> If LED's are not continuously on (some sort of pwm or
> case-dependent-operation or LED's lid up individually for individual
> cases) taking snapshot of the led's won't work you need to capture
> lots of jpegs in time domain (movie) to compare results.

Yup.

2011\08\03@020749 by RussellMc

face picon face
I've been watching this. Enough good ideas so far without mine, but:

> Visual inspection at the factory, proved to be unreliable.

!
Sounds like some employee re-education self-criticism may be in order :-).

However, always on  power LED should be able to be visually checked
without excess effort.

Flashing LED should produce a well correlated varying current at flash
rate. Even if other currents obscure this a very simple flash rate
extraction algorithm should suffice.

If commandable under test, other two should be able to reflect
operation in variation of overall current draw.

BUT

What is LED failure rate?
Why?

Is it dead LEDs or soldering?

If dead LEDs, are you using antistatic protection and/or ESD safe LEDs?
LEDs are extremely ESD sensitive.
As supplied failure rate should be low.
If it's not low LEds are non quality or you are breaking them.

If it's a soldering issue, fix it.
(Easily said ...)


    Russel

2011\08\03@020914 by RussellMc

face picon face
>> If LED's are not continuously on (some sort of pwm or
>> case-dependent-operation or LED's lid up individually for individual
>> cases) taking snapshot of the led's won't work you need to capture
>> lots of jpegs in time domain (movie) to compare results.
>
> Yup.

Or integrate over a period > maximum off period.

Adding an LED coloured filter will help contrast.


      Russel

2011\08\03@021128 by Vitaliy

face
flavicon
face
Kerry Wentworth wrote:
>> We have a device with four LEDs, and we'd like to make sure the LEDs
>> work.
>> Visual inspection at the factory, proved to be unreliable. Does anyone
>> have
>> experience with reasonably priced machine vision systems that can be
>> configured to perform this sort of test?
>>
>> All we want to know, is whether the LEDs turn on and off. Making sure
>> they're the right color is a bonus.
>>
>> Vitaliy
>>
>>
> An LED can be made to detect light.  Put a red LED next to any red LEDs
> on your unit, green next to green, etc.  Make sure that the detecting
> LEDs have colored lenses, even if the LEDs on the unit are clear.

Yeah, I remembered reading an article back in the 1990s about LEDs being used as photodiodes, so we did a bit of experimentation and found that the S/N is too low for our application.

Vitaliy

2011\08\03@021623 by Vitaliy

face
flavicon
face
YES NOPE9 wrote:
> Webcams ( such as the C901 ) will work with OpenCV ( under OS X , Windows
> and Linux ) to capture images ,
> manipulate and analyze images.
> If you are interested, I can offer more details.

OpenCV looks interesting, but I'm looking for more of an off-the-shelf solution. It could work, if someone was willing to write the software for $1000. :)

Vitaliy

2011\08\03@041435 by M. Adam Davis

face picon face
Under some circumstances (a known firmware is running on the device,
the device is in a resting state) you can tell via total device
current consumption whether the LEDs are on or not.

It might not work for your device depending on how it uses current,
but depending on your device and how you test it, it might work well.

Vision QA systems are in wide use in the automotive industry, however
they could not be characterized as cheap. If you really are looking
for a sub $5,000 off the shelf system that will manage to do what you
need done, you're going to have a hard time finding it.  But if you
do, I know many people would would love to hear about it!

A google search of "machine vision" yields a lot of companies that
provide such systems.

-Adam

On Mon, Aug 1, 2011 at 11:16 PM, Vitaliy <.....piclistKILLspamspam@spam@maksimov.org> wrote:
> We have a device with four LEDs, and we'd like to make sure the LEDs work..
> Visual inspection at the factory, proved to be unreliable. Does anyone have
> experience with reasonably priced machine vision systems that can be
> configured to perform this sort of test?
>
> All we want to know, is whether the LEDs turn on and off. Making sure
> they're the right color is a bonus.
>
> Vitaliy
>
>

2011\08\03@045020 by RussellMc

face picon face
> If you really are looking
> for a sub $5,000 off the shelf system that will manage to do what you
> need done, you're going to have a hard time finding it.  But if you
> do, I know many people would would love to hear about it!

I'll be so audacious as to suggest that you might manage to achieve
what you want with a Canon A series camera and the open source Canon
"CHDK" software. Note that this does not run on the neewer Canon's -
for some reason Canon pulled the plug on what appeared to be a
brilliantly successful concept -  low levela cess to all camera
functions on all their A series point and shoot cameras.

You can eg  make decisions about "motion" occurring in selected
display  areas and much more. An LED changing state is "motion" to the
camera.

CHDK:

Download, manual, MUCH information, camera list, ...

         http://chdk.wikia.com/wiki/CHDK

Manual    http://chdk.wikia.com/wiki/CHDK_User_Manual

Basic features list http://chdk.wikia.com/wiki/Features

USB remote control using CHDK   http://chdk.wikia.com/wiki/USB_Remote_Cable

CHDK for dummies   http://chdk.wikia.com/wiki/CHDK_for_Dummies


==================================

OTHER:

Canon SDK:      This is officially available from Canon, only works on
their EOS DSLRs, may do what you want and is *NOT* what I am talking
about


http://www.canon.co.nz/en-nz/Support-Services/Support-News/Canon-SDK

GNU Canon camera utilities

         http://canoncam.sourceforge.net/

2011\08\03@050649 by cognitive h

picon face
How does it feel 46C today's weather in Phoenix? Sorry for off-topic.

Talking about image recognition, it would be nice to have an embedded
such system in a car. I'm almost sure the better more effective
approach yet to show up. Why not to pioneer the field? You could try
inventing it and developing it from the scratch. Think of it as
emulating human brain functionality. Current mobile CPUs are quite
capable and you could use FPGA later.

Regards.


On Wed, Aug 3, 2011 at 9:15 AM, Vitaliy <piclistspamKILLspammaksimov.org> wrote:
{Quote hidden}

>

2011\08\03@103554 by Carey Fisher

face picon face
On Wed, Aug 3, 2011 at 2:07 AM, RussellMc <.....apptechnzKILLspamspam.....gmail.com> wrote:

{Quote hidden}

the box". Maybe I haven't used enough of them.  I would assume they are good
when manufactured and get killed later - rough handling, ESD..

2011\08\03@112606 by RussellMc

face picon face
> > Yeah - what Russell said plus I don't think I've ever had a bad LED "out of
> the box". Maybe I haven't used enough of them.  I would assume they are good
> when manufactured and get killed later - rough handling, ESD...

I've seen a very few short circuit Chinese manufactured LEDs in
Chinese assembled equipment. But I'd not be surprised if these
happened to be due t lack or proper handling procedures.
Incidence  is perhaps in the 1:10000 range. May be more and I haven't
seen them before they were caught but probably not.



             Russell

2011\08\03@134609 by Vitaliy

face
flavicon
face
cognitive h wrote:
> How does it feel 46C today's weather in Phoenix? Sorry for off-topic.

I have no idea how it feels. It's 23C at the office, I don't go outside.


> Talking about image recognition, it would be nice to have an embedded
> such system in a car. I'm almost sure the better more effective
> approach yet to show up. Why not to pioneer the field?

Let's do it together. Our company will do the development, you'll provide the capital.

2011\08\03@134609 by Vitaliy

face
flavicon
face
RussellMc wrote:
>> Visual inspection at the factory, proved to be unreliable.
>
> !
> Sounds like some employee re-education self-criticism may be in order :-)..

They're not my employees.


> However, always on  power LED should be able to be visually checked
> without excess effort.

You'd think so.


> Flashing LED should produce a well correlated varying current at flash
> rate. Even if other currents obscure this a very simple flash rate
> extraction algorithm should suffice.
>
> If commandable under test, other two should be able to reflect
> operation in variation of overall current draw.

Yeah, the more I think about this idea, the more I like it.


> BUT
>
> What is LED failure rate?

I don't know, maybe 0.2% (I'm totally guessing). This device is not in production yet, I'm basing my estimate on the previous designs.


> Why?
>
> Is it dead LEDs or soldering?

Usually, it's wrong polarity. Sometimes, it's soldering.

Vitaliy

2011\08\03@190750 by Sergey Dryga

flavicon
face
Vitaliy <piclist <at> maksimov.org> writes:

> > An LED can be made to detect light.  Put a red LED next to any red LEDs
> > on your unit, green next to green, etc.  Make sure that the detecting
> > LEDs have colored lenses, even if the LEDs on the unit are clear.
>
> Yeah, I remembered reading an article back in the 1990s about LEDs being
> used as photodiodes, so we did a bit of experimentation and found that the
> S/N is too low for our application.
>
How about color sensor like TCS3200 from parallax?  Comes complete with lens, so
it should be "easy" to detect intensity and color at the same time.
Sergey Dryga
http://beaglerobotics.com


2011\08\03@212755 by Barry Gershenfeld

picon face
On Tue, Aug 2, 2011 at 10:57 PM, Vitaliy <EraseMEpiclistspam_OUTspamTakeThisOuTmaksimov.org> wrote:

> one of the LEDs is wired permanently ("Power") and
> one is driven by a third party module (it blinks). It may still work,
> though.
>

Obtain four optical sensors (LED's or something more suitable).  Put each
one behind a proper color filter.  All sensors can see all LED's (so no need
to "aim" or carefully line up unit under test).   Apply pulses and correlate
what is picked up by the sensors.  Sensor of appropriate color should
register highest amplitude of all.  Apply pulse to power lamp by switching
the power of the unit.  Apply flashing to the third party LED by...letting
it flash...watch its current for something to correlate against.  The last
two--do those in the firmware.  Use different pulse rates since you may need
to discriminate them from the "on" and "third party" ones.

Barr

2011\08\03@220136 by Vitaliy

face
flavicon
face
Barry Gershenfeld wrote:
>> one of the LEDs is wired permanently ("Power") and
>> one is driven by a third party module (it blinks). It may still work,
>> though.
>>
>
> Obtain four optical sensors (LED's or something more suitable).  Put each
> one behind a proper color filter.  All sensors can see all LED's (so no
> need
> to "aim" or carefully line up unit under test).   Apply pulses and
> correlate
> what is picked up by the sensors.  Sensor of appropriate color should
> register highest amplitude of all.  Apply pulse to power lamp by switching
> the power of the unit.  Apply flashing to the third party LED by...letting
> it flash...watch its current for something to correlate against.  The last
> two--do those in the firmware.  Use different pulse rates since you may
> need
> to discriminate them from the "on" and "third party" ones.

I think the color sensor module that Sergey suggested, would be perfect for this. We'll pursue Adam's current consumption method first.

Thanks!

Vitaliy

2011\08\04@064603 by alan.b.pearce

face picon face
> > Talking about image recognition, it would be nice to have an embedded
> > such system in a car. I'm almost sure the better more effective
> > approach yet to show up. Why not to pioneer the field?
>
> Let's do it together. Our company will do the development, you'll provide
> the capital.

In the UK (and I am sure in the US as well as many other countries represented on this list) the police have a very efficient system called ANPR (Automatic Number Plate Recognition) and a cop on his bike at the side of the road can be notified of a vehicle of interest (no tax on vehicle, no insurance on vehicle, no vehicle safety certificate, driver normally associated with vehicle may not have valid licence, etc), and will pull you over based on this.

So no capital and development needed ;))))))))



-- Scanned by iCritical.

2011\08\04@073018 by MarcoG

picon face

>>Yeah, I remembered reading an article back in the 1990s about LEDs being
>> used as photodiode..


is it this?   http://www.merl.com/publications/TR2003-035/


I tested a similar solution few years ago for an inexpensive  led-based colorimeter. Results wasn't so bad and may be improved with some digital filtering. With this technique I think that may be possible to make an inexpensive colour sensor using only an RGB led

regards,
Marco



{Original Message removed}

2011\08\04@163819 by Vitaliy

face
flavicon
face
MarcoG wrote:
>>>Yeah, I remembered reading an article back in the 1990s about LEDs being
>>> used as photodiode..
>
>
> is it this?   http://www.merl.com/publications/TR2003-035/

No, this paper is from 2003. The one I read was an article in a Russian magazine, the device was a wireless set top box/headset.

Vitaliy

2011\08\04@230923 by PICdude

flavicon
face
I had to do something similar-ish last year, to a bit more complex --  to capture the value of a a few multi-digit 7-segment display at  multiple points for calibration purposes.  Looked at several of these  options, and in the end I went with the quick, lowest-dev-cost,  bang-for-the-buck option -- I hired a high-schooler for a few hours  and equipped him with paper and pen.

Cheers,
-Neil.

More... (looser matching)
- Last day of these posts
- In 2011 , 2012 only
- Today
- New search...