[EE] "Sunlight" brand LEDs

[Neil]

Anyone have any mid-to-long-term experience with "Sunlight" brand
7-segment LED displays?
Looking at these for a long product as the Lite-On LEDs I've used to
date are in short supply, being discontinued.  And the ones I found from
a broker were counterfeits, which have cost me considerably.
Hence, looking for any reviews anyone can provide.

Cheers,
-Neil.


--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[RussellMc]

On Tue, 4 Aug 2020 at 11:41, Neil <[hidden email]> wrote:

> Anyone have any mid-to-long-term experience with "Sunlight" brand
> 7-segment LED displays?
> Looking at these for a long product as the Lite-On LEDs I've used to
> date are in short supply, being discontinued.  And the ones I found from
> a broker were counterfeits, which have cost me considerably.
> Hence, looking for any reviews anyone can provide.
>
> I have NO experience of them.
Chinese LEDs seem to generally be of substantially higher longevity in most
cases than when I used them in largish volume around 10 years ago.
However, if Lite-On fakes are problematic then other names also may be.
Depending on what the issues are and what time you have available you may
be able to carry out longevity testing yourself.
At 168 hours/week, ~ 700 hrs/month you may start to get
lifetime indications in a useful time period when operated at top end
currents.

Output often rises initially.

I placed LED strings in series with a constant current source (say 10 LEDs)
plus a switchable reference in series - usually shorted out.
Occasional output checks using a lashup "integrating cone" and light meter
and multiple strings allowed early degradation to easily be seen.


       Russell

_SNS _LED
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Neil]

The failures I had with the counterfeits were random segments failing
within weeks, months or more, and replacing a 12-pin through-hole
display on a very tight PCB was a very difficult task, so the boards had
to be replaced.

I recently got some gauges back for upgrades after 12 years (with the
original Lite-On LEDs)... still working beautifully!

I have some samples of these Sunlight LEDs, and testing them with an
actual multitplexing circuit as they would be used (rather then
continuous on).  But extensive testing will get arduous quickly... For
example 3 years of testing at a few hours per day can only be crammed
down to 136 days.  And if I set an "acceptable" failure threshold of 1%,
do I need to test 100 of these?  And these will be used in cars, so
continuous bench testing won't be anywhere near a representative test as
on/off with lots of thermal cycling.

Cheers,
-Neil.



On 8/3/2020 8:00 PM, RussellMc wrote:

> On Tue, 4 Aug 2020 at 11:41, Neil <[hidden email]> wrote:
>
>> Anyone have any mid-to-long-term experience with "Sunlight" brand
>> 7-segment LED displays?
>> Looking at these for a long product as the Lite-On LEDs I've used to
>> date are in short supply, being discontinued.  And the ones I found from
>> a broker were counterfeits, which have cost me considerably.
>> Hence, looking for any reviews anyone can provide.
>>
>> I have NO experience of them.
> Chinese LEDs seem to generally be of substantially higher longevity in most
> cases than when I used them in largish volume around 10 years ago.
> However, if Lite-On fakes are problematic then other names also may be.
> Depending on what the issues are and what time you have available you may
> be able to carry out longevity testing yourself.
> At 168 hours/week, ~ 700 hrs/month you may start to get
> lifetime indications in a useful time period when operated at top end
> currents.
>
> Output often rises initially.
>
> I placed LED strings in series with a constant current source (say 10 LEDs)
> plus a switchable reference in series - usually shorted out.
> Occasional output checks using a lashup "integrating cone" and light meter
> and multiple strings allowed early degradation to easily be seen.
>
>
>         Russell
>
> _SNS _LED

--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[RussellMc]

On Wed, 5 Aug 2020 at 02:16, Neil <[hidden email]> wrote:

> The failures I had with the counterfeits were random segments failing
> within weeks, months or more, and replacing a 12-pin through-hole
> display on a very tight PCB was a very difficult task, so the boards had
> to be replaced.
>

1. Possible source.
May be about as reputable as any of the Asian sources
https://lcsc.com/products

2. There are too many variables known only to you or not known at all to
offer anything like certain advice.
Depending on quantity involved, acceptability or possibility of redesign,
time available, knowledge of failure environments my approach MIGHT include
these semi random thoughts:

Any possible clues from environment? - temperature cycling or extremes,
vibration, hours of use, voltages,  ...

Soldering temperature or profile affected? Could torture test bad units to
see if relevant.

Are old "bad" versions available for comparison testing?

Are mechanical aspects a factor - twisting or differential cooling or
vibration or ... ?
(Not directly relevant, but as an example, small slices from PV panel
wafers have stress issues which many manufacturers seem unaware of).

Test rig?
High end currents?
On off cycling? Whole power supply square wave?
Hot cold cycling - maybe in and out of refrigerator every day or so while
running?
Vibration ? Doable "simply enough" if liable to be relevant.

I can envisage a module with N displays, mounted on a woofer to provide
vibration, cycled on and off grossly and powered on long enough wires to
allow it to be moved in and out of a refrigerator easily.
Mix of liteon (1 or 2), old bad units (few), Sunlight samples.
Not overly "controlled"- but at 3 hours/day usual operation (if on time is
relevant) that gives you 8 months per month.

ANY segment failures are probably too many.

Statistical rule of thumb is if you get no results in N samples then actual
figure is - hmm - mind says either < 1 in N/3 or < 1 in 3N. If that sounds
of relevance it can be checked.


  Russell



>
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[Neil]

On 8/4/2020 7:22 PM, RussellMc wrote:
> 1. Possible source.
> May be about as reputable as any of the Asian sources
> https://lcsc.com/products

LOL yeah... that's where I found these.
PS: I did find the company's website ( sunlightled.cn ) and several
things broken, so could not find English datasheets, but I requested them.

> 2. There are too many variables known only to you or not known at all to
> offer anything like certain advice.
Expected.

> Depending on quantity involved, acceptability or possibility of redesign,
> time available, knowledge of failure environments my approach MIGHT include
> these semi random thoughts:
>
> Any possible clues from environment? - temperature cycling or extremes,
> vibration, hours of use, voltages,  ...

Vehicle cockpit.  Hours can vary significantly.  Mostly trailer queens
(an hour an year?) to daily drivers (few hours a day every day), though
this may extend to commercial vehicles that'll do 8+ hours a day.


> Soldering temperature or profile affected? Could torture test bad units to
> see if relevant.
Temps are generally the lowest that would work, and we still use leaded
solder, whereas these parts should be able to handle lead-free solder.  
So I don't feel that's an issue.  I can tell you that we use red, blue,
orange and some green originals.  When Lite-On made the blues scarce, we
found that broker "stock" which had the (significant) issues.  Same
process, PCBs, environments, etc.


> Are old "bad" versions available for comparison testing?

Not sure if I have any stock of the bad ones on hand.


> Are mechanical aspects a factor - twisting or differential cooling or
> vibration or ... ?
> (Not directly relevant, but as an example, small slices from PV panel
> wafers have stress issues which many manufacturers seem unaware of).
Comparison with originals and other colors says that these are not factors.
I'll add that this is one of the original displays (running as a clock),
It was in my car for over a year at first, but since then on my desk (so
a much kinder environment) running continuously for > 10 years now...
http://orlandorobotbuilders.com/stuff/Con_Rod_Clock_01.jpg


> Test rig?
> High end currents?
> On off cycling? Whole power supply square wave?
> Hot cold cycling - maybe in and out of refrigerator every day or so while
> running?
> Vibration ? Doable "simply enough" if liable to be relevant.
In the early days, I used to do the refrigerator and toaster oven test
quite a bit, until I was confident all was great.
I've done vibration testing only minimally, but on a dashboard it's not
a major factor.
Not sure what you mean by "high end currents"

> I can envisage a module with N displays, mounted on a woofer to provide
> vibration, cycled on and off grossly and powered on long enough wires to
> allow it to be moved in and out of a refrigerator easily.
> Mix of liteon (1 or 2), old bad units (few), Sunlight samples.
> Not overly "controlled"- but at 3 hours/day usual operation (if on time is
> relevant) that gives you 8 months per month.
>
> ANY segment failures are probably too many.
I agree, but trying to be realistic here.  If the product failed for
this reason, I've been eating the cost, even well out of warranty
period, as it's nothing the customer did.  I'll stomach a very small
number of these.  Outside of that, it's not just the cost, the there's a
quality-image issue here.

> Statistical rule of thumb is if you get no results in N samples then actual
> figure is - hmm - mind says either < 1 in N/3 or < 1 in 3N. If that sounds
> of relevance it can be checked.
>
>
>    Russell
>

My usual method is to build a dozen or so protos and send them out to
people for testing in different environments.  Just hoping some feedback
from anyone who's used these might help me accelerate that a bit.

Cheers,
-Neil.

--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[RussellMc]

> Not sure what you mean by "high end currents"

Operate at top end of rated current - or maybe above.
May be of no relevance at all (of course).


            Russell

>
>
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist

[RussellMc]

>
> > Statistical rule of thumb is if you get no results in N samples then
> actual
> > figure is - hmm - mind says either < 1 in N/3 or < 1 in 3N. If that
> sounds
> > of relevance it can be checked.
>
> Just for (possible) interest:

https://wiki2.org/en/Rule_of_three_(statistics)

Summary: If you make N tests with negative result then there is 3/N
probability that there are no positive results at the 95% confidence level.

- 300 parachutes open OK ->

<1% chance of parachute failure at 95% confidence level.

Nobody shoots me on 1000 trips to town ->

< 0.1% chance that nobody will shoot me if I take a trip to town at 95%
confidence level.
(NB: prior trips were until end of last year!)

       R
--
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist