Russell, Thanks a ton! That was comprehensive. And yes, street and flood lights are also one of the products that we are designing. Thanks once again, Mohit. RussellMc wrote: >> Seoul Semi. The company rep sent us about 250 HB LEDs of the Z-power >> series and of different colour temperatures in white and some R,G,Bs. So >> I can safely assume they are genuine. >> >>> Where were they made? >>> >> I think South Korea. http://www.acriche.com/en/company/map/map.asp >> >> >>> What certainty do you have about lumen maintenance lifetime and how do >>> > you know? > >> Their datasheets. >> > > Sounds almost safe :-). > Seoul are reasonably good. > Counterfeiting is rife - always be sure of your supply chain. > > This is useful > > Understanding power LED lifetime analysis > Philips 2007 > > http://www.philipslumileds.com/uploads/165/WP12-pdf > > Look at and understand the implications of fig 3 and fig 4. > Current and junction temperature are usually rolled together as in eg fig= 5 > and the point is lost. > Lifetime is independently affected by current AND temperature - few peopl= e > seem to appreciate this. > > The above paper and 2 more are found here > > http://www.philipslumileds.com/support/documentation/white-papers > > WP12 Understanding Power LED Lifetime > Analysis > 268 May 1, 2007 WP14 LEDs: Coming Soon to a Street Light Near > You > 397 Sep 8, 2008 WP15 Evaluating the Lifetime Behavior of LED > Systems > 1796 May 4, 2010 > > Much is common sense. But, not all. > > >> We have done quite a few light measurements experiments. >> > > Have you done anything like the lifetime test I suggested. > While that is crude it allows you to sort real from fake and good from ba= d. > Top suppliers datasheets can generally be depended on (including Seoul). = Any > others are suspect. I have seen data with lifetime curves AND the equipme= nt > on which the results were said to be obtained. The LEDs supplied in no w= ay > matched the claims. > > >> We would also like to check what the measurements are in different >> environmental conditions. >> > > Major manufacturers will supply test data at elevated temperature and > humidity combinations. > I have a range of data that I could dig up as examples - mainly from Nich= ia. > > >>> Higher humidity, within limits, can IMPROVE lifetimes in some cases. >>> > > >> Is this documented somewhere or did you find this out while testing your >> lights. >> > > Shown by actual test data. Nichia and probably others. It surprised me. > Keep on increasing T & H and you reach a point where things get worse. > > >> Primarily, our aim is to evaluate what kind of warranties we can offer >> on our products. The LEDs come with tens-of-thousand hours lifetime >> ratings, but the other components that go in are no where near the >> figure. An LED could last 10 years, but will a street light built out of >> them with a power supply, PFC etc. last half that many? That's something >> we'd like to find. >> > > Ultimately testing is *extremely* desirable BUT if you are serious about > volume products then you first need to be able to first do lifetime > calculations based on manufacturers data or on tests which derive that da= ta. > If the manufacturers cannot supply this information or you do not control > component supply well enough then your warranties will be meaningless. > Presumably you are acquainted with Weibull distributions > > http://en.wikipedia.org/wiki/Weibull_distribution > > http://www.weibull.com/LifeDataWeb/the_weibull_distribution.htm > > Parameters usually need to be empirically derived. > People like Philips do the hard work for you. > Other may well not. > > At a first thought, if you are building street lights (implied by your em= ail > but maybe not) major component items that come to mind apart from LEDs wi= ll > be capacitors, & cores for magnetics*. Resistors, PCBs, connectors, ... c= an > be specified at a quality that allows 10 year lifetimes without too much > effort. Environmental issues may lead you to either quality conformal > coatings and/or quality housings. Water intrusion, resistance to vibratio= n, > temperature, humidity ... are in the designers hands controllably enough = (do > not, unlike Toyota, allow your PCBs to be ABLE to break :-( ). > Encapsulations, if used, can have lifetime issues. > > * magnetics core lifetime depends on technology and power level. Iron cor= e > and ferrite may not be overly time critical. Powdered iron cores can be c= ost > effective but in power handling inductors MUST be designed thermally for > lifetime and MUST have a known reputable source. Low spec iron powder cor= es > can suffer from thermal degradation runaway with crippling results. > > Do not rely on adhesives for mechanical support and/or location unless yo= u > are certain of their long term performance, and even then it's preferable > not to. Many manufacturers provide support for heavier items (eg large > capacitors) using quick setting thermal adhesive - hot melt (!!!) or cera= mic > type cements or ... . These all seem to have a finite and short lifetime= .. > Hot melt adhesive is a fatally bad product to use. It has an adhesion > lifetime to many materials in the 6 months to 2 years range. Unless keyed= in > place it will fail to supply support and even then it's bad practice. > > Drive LEDs properly, understand worst case design and design conservative= ly. > Never rely on Vf matching for current sharing unless you can guarantee th= e > match. (I have some LEDs that are matched to within +/- 0.05 volt Vf in 9= 9% > or so of all samples. If you know this you may be able to use the fact. B= ut > far better is constant current drive not dependent on LED parameters. > > Unlikely to be an issue - but be aware of the tight ratio of Imax to I > > Semiconductors from reputable manufacturers can be expected to have long > lifetimes compared to some components above provided that they are opera= ted > at sensible junction temperatures under worst case ambient conditions. > Imitations may not only be out of spec but may suffer early failures in > adverse conditions due to sealing failures. > > Be aware of worst case internal ambient temperatures. "Wet" aluminum > electrolytics may have a shorter life at temperature when powered off tha= n > with voltage applied. So a light that sits in pleasant Indian summer > conditions powered down during daylight hours may reduce capacitor lifet= ime > more when turned off than turned on 10 years ~=3D 100,000 hours. For a 20= 00 > hour capacitor to last that long it needs to operate at a mean temperatur= e > of less than 10 x 2^(100000/2000) ~=3D 50 to 60 degrees below its rated > temperature. ie a 105C cap should run at a mean temperature of under 50C = and > an 85C cap at under 30C. If my comment about powered down lifetime being > shorter than powered up is considered to be true( some question the claim= ). > Then a 2000 hour 85C cap has little chance of surviving Indian conditions > for 10 years. Use of other technology capacitors is recommended where > possible and economic. > > Design for ripple current! If your cap has a ripple current lifetime > understand why and how to ensure it is met. Often two or more smaller cap= s > of the same net capacitance will have a higher total ripple current rati= ng > than a single larger one. > > Solid aluminum caps are greatly preferred over tantalum. They have simila= r > performance and price but are not subject to tantalum's exciting flames > smoke and explosion failure mode. > > If components depend for their longevity on load or current or voltage > sharing, ensure they really share. > > Some resistors "just fail" even when nowhere near ratings for voltage or > current or power or temperature. Investment in the small absolute extra c= ost > of reputable resistors will pay dividends in products aiming at super lon= g > life. > > Solar ultraviolet light will destroy your diffusers or lenses in well und= er > 10 years if you don't explicitly design things so that it doesn't and kno= w > why your product will last that long. Polycarbonate with the right > inhibitors will last double that. Polycarbonate with no UV inhibitors or = the > wrong sort or ones with too low a lifetime won't. Polycarbonate has nice > double UV susceptibility peaks and an inhibitor must either be broad enou= gh > to cover both OR use a combination of inhibitors that cover both > wavelengths. Thin section plastics (and lenses may well meet that > description) are susceptible to loss of inhibitor with time and a materia= l > must be chosen which works both well enough initially AND keeps on doing = so > - for 10 years if needs be. Special low volatility formulations are produ= ced > by reputable manufacturers. If you don't design and control the brand an= d > type and amount of inhibitor being used in your lens then it is almost > certainly wrong. > > Not so much an issue for streetlights, but some materials are much more > prone to insect or vermin attack than others. Termites will eat right > through some thin plastics. Rats like some PVCs (may depend on plasticise= rs > used). > > ROHS is probably a given. With high temperatures and long target lifetime= s > you may want to not get too creative with tight packing of PCBs. Tin whis= ker > growth is largely a problem for the super fine area (eg watches) but it i= s > going to be a problem in due course for some. (I have old computer equipm= ent > which the component CANS are growing whiskers :-).) I've never seen a rea= l > tin-whisker failure. Yet. YMMV. > > > > Russell McMahon > -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .