April 12th, 2007...on going, updated May 05, 2007- see end
Yesterday, I received my TerraLUX MiniStar 5 Model TLE-6EX, a replacement for incandescent bulbs, utilizing the Seoul P4. Made for 3 to 6 C/D cell flashlights.
I was pretty excited when I saw these, since they have the higher efficiency Seoul P4 LEDs in them.
I had no problems installing one into a 6D Cell Maglite, it was pretty simple.
So, what does it look like?
I did some lux measurements, against five of my MagLED units. I had one 2D MagLED and one 3D MagLED that was dimmer, one 3D MagLED that was equal brightness, and two MagLED units that were brighter(3D and 4D). I was very surprised to find this, since the Seoul P4 LED is a higher efficiency LED, than the Luxeon found in the MagLED. I will get back to this later on.
Due to the somewhat milky silicone in the P4, it helps hide some die artifacts in the beam, but tint warriors will need to beware that there are tint issues in the Seoul P4 beam, and artifact warriors will note there are bond wire artifacts in the Seoul P4. These can be smoothed out a bit, if one purchases a bumpy reflector retrofit, or purchases a diffuser lens, both options will decrease the throw a bit.
Before I get too far, the heatsinking concept to get the heat from the LED slug to the MagLite screw post is very robust, and one can easily tell, as the "heatsink" gets *quite* warm during operation.
A picture of the much better heatsinking is found below, and is much better than the MagLED design:
While testing the unit out, and shinning it around the house, after about half an hour, I found the LED had now become loose from the heatsink. This was not the case in the beginning. With the Seoul being an extra heat sensitive LED, when it came loose, the tint shifted baby blue.
Notice how there is no epoxy still bonded to the base of the emitter.
Here is a better shot of the epoxy still on the "heatsink":
I did not see this as that big of a deal, as I can easily repair it, but one has to bump it down a notch for this defect, which exhibited the problem in such short order.
Since it needed repairing, this is a great point to look at the internals, so lets proceed.
Unsoldering the two wires from each side of the Seoul P4 LED, will allow one to simply remove the PR lamp base from the heatsink. Looking down inside:
Taking a look at the module, one finds a 2 ohm sense resistor (causes high losses), a Coilcraft shielded inductor 10uH LPS4xxx series ( http://www.coilcraft.com/pdfs/lps4018.pdf )(I forgot to measure the inductor dimensions to determine exactly which one in the series), and a ceramic bypass capacitor.
Turning the board over, one finds the National Semiconductor LM3485 Hysteretic P-ch FET buck controller http://www.national.com/ds.cgi/LM/LM3485.pdf, a Fairchild Semiconductor FDN338P P-ch MOSFET http://www.fairchildsemi.com/ds/FD/FDN338P.pdf, and Fairchild's SS14 Shottky Diode SS14 http://www.fairchildsemi.com/ds/SS/SS14.pdf.
Now for some measurements on how well the converter operates.
(before measurements, I repaired the bad epoxy bond, removing the old material and using AS thermal epoxy)
First we have the output wattage going to the Seoul P4 vs. input voltage:
Then we have the Efficiency of the converter vs. input voltage. There is a reason for the low efficiency, which we will get to shortly, but basically, 30% to nearly 50% of the heat in the module is caused by the converter plus the losses in the very high value of resistance, sense resistor. Once the module is operating, one is looking at an efficiency ranging from 55% to almost 68%.
Here we have the Watts Input vs. Watts Output in Blue, and in Pink I've shown the loss contribution of the high sense resistor value that was utilized:
Next, I've taken and shown the loss contribution of the high sense resistor value above, so below is shown how much the efficiency could be improved without the high value of sense resistor:
It appears that the designer chose to take the low cost and simple approach to the current sense, and due to the high losses of the 2 ohm resistor, a large, higher wattage resistor had to be utilized. One could have utilized a *much* lower sense resistor value, and put an op-amp with associated components in the same space occupied by the higher power sense resistor. But this would have cost 0.15 to 0.50 dollars more. Techniques for utilizing an op-amp or a current sense amplifier, and links to application notes are shown on my power supply reference page.
(FYI, for measurements above and below, I repaired the bad epoxy bond, removing the old material and using AS thermal epoxy)
Measuring temperatures, installed in a 6D MagLite, within the first minute, the heatsink hits 50.1 degrees C (122.18 degrees F).
In two minutes, I read 68.1C (154.58 degrees F).
In three minutes, I read 74.5C (166.1 degrees F).
At five minutes, I read 91.8C (197.24 degrees F).
At roughly 10 minutes, I read 107.2C (224.96 degrees F).
These high temperatures cause a loss of lumens from the Seoul P4, which are being produced by the low efficiency converter design, plus the heat produced by the Seoul P4.
Anyhow, that is it for the night, hope you enjoyed the information.
I decided to add a direct link to the manufacturer's recommended circuit for the switcher chip's recommended HB LED implementation, showing one possible implementation with schematic and BOM for 350-1400mA constant current, which uses one tiny op-amp (and two tiny resistors and a tiny cap)(a current sense amp would get rid of the discrete resistors but costs a tad more):
350-1400mA reference design and eval board, June 2006
It is the weekend, and I decided to instrument things up. This time I waited until the TerraLux heatsink was fully cooled off, before starting the test. Due to the fairly high thermal resistance or "insulation" to ambient, the part takes a surprisingly long time to fully cool down (about an hour). Measurements for Lux and Temperature were made simultaneously at the same time.
Please keep in mind, this test is done without the protective plastic bezel lens, and actual temperatures would be even higher yet, than shown here in this test.
Here are a couple of pictures of the measurement point, with the thermocouple embedded in thermal compound, with the tip of the thermocouple tip in contact with the TerraLux heatsink:
Here are the resulting temperature and Lux @ 1m data graphs. I took considerable care in focusing and aiming the MagLite to get peak Lux readings. Surprisingly, the heatsink of the TerraLUX Seoul P4 actually gets hotter than the boiling point of water, indicated in the red tinted background of the graph.
During the temperature plot, I noted that the Seoul P4 LED in the TerraLUX module, tint shifted from a WO bin, to a cold powder blue-white tint. (the phosphor used in the Seoul P4 is extra sensitive to temperatures, unlike the Luxeon and CREE phosphor-or causes extra tint shift with temperature).
Due to what I see here, I have to mention that temperatures reached, with the in-efficient converter (due to sense resistor) which produces a large additional chunk of heat, I do have some worries for the longevity of the Seoul P4, which already drops in efficiency early in it's life cycle due to the phosphor Seoul utilized (unique), even when you hold the slug down at 30 degrees F, and the fact that heat reduces the lifetime of an LED, I am left to conclude that the situation here is definitely not optimal and somewhat concerning.
The low efficiency caused by the large sense resistor value adds considerable extra heat, and heat has a big impact on the Lumen output of the Seoul P4. Due to the phosphor used in the Seoul P4, the heat impacts the light output of their phosphor, which is extra heat sensitive, and when you loose the light contribution of the phosphor, you get a shift in the blue direction. The Seoul P4 already has an issue with a rather early, rapid drop in Lumen output by 12% (which stabilizes at -12%), which also results in the tint shifting blue, and the shift of a couple of bins is documented in the Seoul P4 datasheet on page 4. The extra heat will only compound these issues and accelerate it.
However, after a little bit, one can always replace the degraded LED, possibly with a better performing device, when it gets too objectionable.
Had TerraLUX spent a few more cents in the design, to greatly increase the converter efficiency, the benefits would have been multifold, reduction in heat produced, more light output, less LED tint shift, longer LED lifetime, longer runtimes, and likely- a more reliable solution. IMHO, they should take a look at the epoxy issue, an epoxy that is designed for these very high temperatures might hang on to the LED a little better, and prevent the LED from popping loose during use- proper surface preparation may also be of help.
Like a number of "mass market" designs, there are too many trade-offs for my personal tastes. I'm considering modifying mine, with the op-amp current circuit and low sense resistor value, right off the application note for the National Semiconductor LM3485 chip that I mentioned earlier, or utilizing a current sense amp, and in both cases using a low value current sense resistor (0603 sized- much smaller resistor) which then gives you room for the parts. The current sense amp is easier to haywire in (gets rid of a couple of resistors, making the mod yet even easier to do). This would remove a whopping amount of heat produced in the module, improving a multitude of things at the same time.
Here is the first runtime test I did:
If you look carefully, you can see where the module all of a sudden stared ramping up in temperature, and blew up when the "heatsink" part of the module reached about 254°F. Keep in mind that during my testing I had no lens on the MagLite, likely real world temperatures would be a bit warmer. Unfortunately, the smoke from the module messed up a brand new MagLite reflector in the 6D. Needless to say, I was not a happy camper.
When I opened up the module, I found only one failed part. It was the the MOSFET which failed, Fairchild Semiconductor P/N FDN338P (it is the SOT-23 package part).
I replaced the MOSFET, and everything is working normally again. I used a better MOSFET (it was even a much smaller device than the one utilized), which lowered the heatsink temperatures, and raised the efficiency a bit.
So, I've got it repaired and everything is working fine. Likely, with the excessive temperatures found in this module, and the low efficiencies of the switcher implementation they used, it is very likely that the MOSFET is likely being pushed outside of it's SOA. Schottky diodes get pretty damn leaky at these sorts of temperatures, and that doesn't help out the poor MOSFET much. Also, the MOSFET is mounted on the opposite side from the 2 ohm 2W sense resistor, which is producing roughly 33% of the heat in the TerraLux. This high value resistor makes *plenty* of heat, which isn't going to help the MOSFET much at all either.
The runtime test was repeated, and this time it made it through the test, with the better MOSFET. Likely, your runtimes will be a tad shorter in a 6D MagLite. I've done two plots, so you can better see the beginning and overall performance:
Next up, when I get a chance, I will be replacing the high value sense resistor with the standard National Semiconductor current sense circuit. Stay tuned.
Well, I finally had a chance to pull out the 2 ohm sense resistor that is in series with the Seoul P4 for sensing current Friday night.
I basically removed the huge 2 watt 2 ohm sense resistor, and replaced it with a much smaller 0603 resistor, with a value of 0.1 ohms. Took an old high side current sense amp chip that I had kicking around in one of my old junk boxes, hooked a 10k ohm resistor on it's output back to the FB pin of the LM3458, and hooked it's inputs across the sense resistor. I then took a 220pf from the + output that goes to the Seoul, and wired this directly to the FB pin of the LM3485 chip. I also added a 0.47uf capacitor on the output, basically across the Seoul P4. (the stock TerraLux has no output cap). Had found a larger value, I would have utilized that. I then put it all back into the PR base.
Some quick tests showed a dramatic improvement in efficiency. See graphs below:
I am currently doing a runtime test that is doing extremely well, and much improved, I'll post the datagraphs when the system gets done recording them.
I decided to post the first two hours of the runtime test (lux improvement and temperature reduction) comparision before I head off to bed for the night:
The unit has passed more than a 50% improvement in runtime, as I write this. At the same time, the improvement in efficiency has dropped the module temperature considerably, which also increases the Seoul P4 efficiency- resulting in roughly a 25% increase in lumen output. Alkaline cells are quite load sensitive, and if you reduce the load, you are able to utilize more of the energy within the batteries. Since the circuit is more efficient, it needs less power in, to produce the same output power. This lightens the load on the cells, with less going to waste, and it looks like this reduction in load from the higher efficiency current sense mod I did, is resulting in dramatic runtime increases.
It will be very interesting to see at what point the light drops out.
I've finished the runtime plot, see below. Depending on which way you prefer to compute percentages, and where you decide it drops off, the runtime increased by 53% to 55%, and also resulted in a roughly a 25% increase in average lumens at the same time. Not too bad. It is quite likely this could be optimized more, I just slapped this mod into the TerraLUX with old spare parts I happened to have lying around at home from old projects/hobby stuff. Examples would be better MOSFET, better schottky, better inductor, different current sense amp that allows lower resistance resistor, seeing if the circuit behaves better with the current sense amp power pins on the output side vs. input power pins, and running simulations on various changes to get it a little bit optimized. Of note, you will see the lux output going up as temperature drops, this highlights the sensitivity of the Seoul Phosphor to temperatures, as the Seoul P4 shifts blue as it gets warm, due to the phosphor efficiency dropping as it heats up, resulting in less green/yellow/red being produced by the phosphor. Around 38 minutes into the runtime, the nice white Seoul P4 shifts a strong baby blue color- I don't believe anyone on earth would still consider it white with a straight face.
This leads to another item. Since the Seoul P4 output drops (or "shifts") so much with temperature, one may want to consider tweaking on the current level a bit, as you may get more Lux or Lumens at a lower current level, due to the MiniStar/Seoul running cooler. It all depends on where the trade-off kicks in. The drawback, is this would lower the peak output spike at the beginning. What would be best really depends how the end user utilizes their flashlight.
BTW, I wanted to mention that MattK at Battery Junction was nice enough to send a replacement for this module, it arrived today! Thanks a bunch Matt!
I've decided to do a runtime test on the 100% stock module I received yesterday, it is in progress now.
Meanwhile, I've gotten a multitude of requests from folks for pictures of the mod I slapped together. Here is the stock photo of one side:
Here is the modified photo of the same side. You will note that I have a big honkin' SOT-23 here, one can always use the much smaller SC70 package and save a ton of room (example of current sense amp a SC70 package is the MAX4073). Also, I only had large resistors and capacitors kicking around the house, nothing like the 0402 parts unfortunately. Of course, if you put the traces on the board, you could get rid of a lot of those haywires and things would look much neater. Once things ar laid out properly, the missing lossy 2 ohm sense resistor creates a boatload of room for the four additional parts.
Here is the original converter chip side:
Here is the modified board, unfortunately, the MOSFET had fried as mentioned earlier, and burned the board, so it doesn't look as nice as it could. The MOSFET here is the FDJ129P, which has a thermal pad, which is great for pulling the heat out of the MOSFET die. I have it flipped over and deadbugged. Not the optimum choice of parts, but this old leftover part worked great.
I decided to get some scope waveforms. Here is the voltage present on the Seoul P4 used on this TerraLUX module:
You will notice there is a wide band at the top, so here I've put the scope on AC coupling, and as you can see, there is 84mV of ripple. This ripple would be worse, had I omitted the 0.47uF output capacitor.
Now, most folks know that when this CREE die used in the Seoul P4, when it heats up, the die voltage drops by about 3mV per degree C. So, above we had 3.4V, and in the scope capture below, you will see that the voltage has dropped to 2.94V. 3.4V - 2.94V = 0.46V, and 0.46V/3mV= 153 degrees C rise above room temp, so this indicates the die temperature is roughly 178 degrees C, well above the maximum rated temperature for the die. This is why the Seoul P4 phosphor stops working well (being extra sensitive to heat), which reduces the yellow, green, red output of the phosphor, and the color shifts quite blue.
Below you will see the ripple voltage in green, and this ripple voltage results in a small ripple in the output light, which is shown in yellow below:
I will see about getting a schematic together for the modified circuit, if anyone is interested. Please keep in mind, that no component choice optimization has been done, no simulations of the circuit, nor step load tests for loop response characterization.
I finished the runtime plots on the unmodified, not repaired, new replacement part that Battery Junction sent me. I got a bump in the light output during the runtime plot, I am unsure as to what caused that, but nothing was disturbed in the setup. Likely it is something that happened internally to the converter or something. I also fixed a few time offsets on the graphs between the different runs/data.
I'm not sure if I've mentioned it yet, but the Seoul P4 also has rather serious tint shift issue with time that they mention on their datasheet. I stumbled upon this when doing a long term lumen maintenance test, and found a 12% loss in lumens in only 1000 hours, and I was scratching my head on this. I'd very aggressively held the slug below 30 degrees C for the test, but this didn't help the Seoul P4 lumen depreciation issue with their new phosphor they have. After reading an article on the Seoul P4 phosphor, and the note about the shift problem, and a light went off upstairs. As white light shifts towards blue, with the same amount of power, the lumens decreases. I couldn't wait for the day to end to go home and check out the tint. After work I hurried home, took a look, and found the tint had shifted cooler, or more blue! Bingo! This explained the rapid Seoul lumen depreciation! On a side note, the Seoul P4 phosphor is also extra heat sensitive compared to what other LED makers utilize.
So, I took an LED and ran the test shown below. Please keep in mind, this is on a sample of only one part- results may vary. It is one of the best performing Seoul P4 parts I had tested on my Seoul page.
I would be amiss if I did not qualify the above comment. It is likely that this depreciation would be a non-issue for low cost flashlights, since they were low cost in the first place, looking at our throw away society mentality. Typically one doesn't EDC something like this, just tosses them in a drawer or toolbox for emergencies- which they are more than adequate for.
I do a decent amount of mine exploration, mainly to look for rock/mineral specimens, and often I hike in the mountains to rather remote areas, and quite often I find myself hiking back in the dark. I also do a lot of testing during the day, in the dark, where I often have to adjust or record data. Other folks like in the professional security business, or police may see similar flashlight usage patterns. Anyhow, in my case, it is not uncommon for me to go thru 6 or more lithium batteries in a week. Though it is quite difficult or impossible to see a 12% lumen depreciation, and once you include the additional lumen loss due typical heating in a flashlight, these sorts of things happen to add up, especially if the converter utilized in the flashlight is not heatsinked, and or you stack on an additional 15-20% loss for a sapphire lens vs. an AR coated lens. Each of these things in and of themselves is not a big issue, but when you take it all together at the system level, they can really add up- next thing you know, only 40% of the rated lumens is making it out of the flashlight. However, for me, a 15-20% loss is huge, as I can back the converter down and get a extra +20% runtime or more (due to the lighter load on the battery, which makes it more efficient also. If you do not have a protective bezel ring, and find yourself somehow bashing your lens a lot and cracking it, the heavy penalty for sapphire may be worth it.
However, there are some higher end lights, where you find folks who are *quite* serious about their flashlights- and I am not overstating things here, I promise. There are people that actually pay 200.00 for a flashlight! Why? Performance. Here you find folks who will upgrade the LEDs in their flashlights with the latest high output low Vf bin LED, many who are quite critical of the LED tints, and I've seen folks pay 10x for an LED that has the latest in high output, low Vf, and white tint. They buy five dollar lenses to replace the stock lens window on their MagLites! ( http://www.flashlightlens.com/item--Maglite.-Lenses--Maglite.html ) Why? It is more scratch resistant, and the clearer lens plus AR coating can give you back another 10% of the losses in a flashlight. They pay extra for that +90% efficient converter, where you generate considerably less heat generated, which reduces the lumens lost by heat and considerably increases the runtime on a given set of cells. You find special low resistance switches/springs/contacts, special aluminum body coatings which enhance the durability of flashlight finishes, and all sorts of other cool things.
There has been a lot of traffic on a few forums about LED tint. What new people fail to understand is that LEDs are binned according to tint. Low end flashlight makers just buy whatever bins, they are much cheaper that way, and they can save 60 cents on a 3 dollar Philips LumiLEDs Luxeon III or K2. It is difficult to control tint, and you get a variety of tints off the production line, and companies have been buying binned LEDs from companies like HP (which was Agilent and is now Avago), since the early 1970's.
One forum I used to frequent, tint was quite important to many end-users, yet there were folks of influence who poo-poo'd the importance of tint, coining terms like white wall hunters. Yet, just a year before, the very same folks were pushing hard for green tinted white LEDs, as they felt they were *much* superior for color rendering. They even had long threads on how much superior green-white tints were for color rendering. Later, you'd see the same folks pushing yellow white for superior color rendering. So is tint important or not? Anyhow, whatever floats your boat, it is your money, and the right to demand what you prefer, and if they choose not to offer what you want, take your money somewhere else. Don't let people be-little you for what YOU want/prefer.
The point on the tint thing, is that you will get phases on the forums where folks say Brand W LED is blue-white, Brand X LED is pink-white, Brand Y is puke green-white, or Brand Z is yucky yellow. Keep in mind that LEDs are binned by color, and the manufacturers get them by reel, of a particular tint. If they choose to save 60 cents for best white and brightness, thats their choice, but it isn't that it doesn't matter at all. I've seen white Luxeons all over the color gamut, the same with Nichias, OSRAMs, and others. What is really going on, is the manufacturer will get like 2,000 luxeons on a reel, and they choose to use a bin that wasn't WO or whatever, and you have the first two thousand flashlights constucted with these. So of course they are all bluish-white or whatever. It isn't the LED, it is the bin that they got when they purchased their reel, or paid 60 cents more for, like you'd often find on a 200 dollar flashlight. However, from what I've seen, manufactures often attempt to squeeze every cent of profit out of a product they can, even in a high dollar flashlight, and react almost violently when the end user wants/prefers something that is actually white- what the heck, I paid a premium 200 dollars for 30 dollars of parts, you can at least give me something that resembles white! When a maker starts to be-little the end user over things like this publically, one has to really wonder what in the world is going on, besides trying to bias the forum to what they prefer. The group mob mentality can get into the mix here, and make for a rather unpleasant experience for the poor soul who publically mentions some taboo subject that the maker of the light deemed so. IMHO, this is extreme bias coupled mob mentality, since the fan boys will jump in and rip someone up, to brown nose that flashlight maker, and get in with them tighter- the old buddy club effect at work here. In retrospect, it reminds me a lot of politicians and lawyers.
In fact, there is a company (before they combined with another company) that used to offer guaranteed white tint, and charge an extra 20 dollars or whatever it was, HDS Systems. They were very successful, end users even paid the very considerable increase in price to get a nice white product, and this was on a sub 200 dollar product!
The Seoul P4 adds yet another dimension into the whole issue. Many got spoiled with LEDs from LumiLEDs, OSRAM, and CREE, which have a much lower shift with heat and with time. However, the Seoul P4 is a part that will shift tint and drop in lumen output on you as you use it, with heat and with time. The amount of shift can be lessened by keeping the drive current to the Seoul P4 under 350mA, and also keeping the temperatures as low as possible by heatsinking the LED well. This includes things like using copper or aluminum in the contruction of the flashlight, thermal transfer goo, wide thermal transfer areas, thermal vias in boards, thermal spreading materials, dark flashlight finishes (high emissivity), and such, to help get the heat out through mechanisms like conduction, radiation, convection, and such. Avoiding things like Stainless Steel or Titanium can be very beneficial, since they have extremely high thermal resistances- But a person can also make the walls thicker with high thermal resistance metals like these, and reduce the penalty suffered. One can also take a hybrid metal construction approach. Thermal impedances to the environment can be surprisingly low, when you have a forced liquid cooling machine, like the human hand.
Anyhow, that is enough rambling for now.
Quite a few visitors emailed me and reminded me about posting the schematic of the mods I did. Before my email system crashes, I'd better get it put up here. LOL, please chill on the emails! You will find it below. Please keep in mind, this was just something I slapped together, I have not done any optimizations or simulations or load transient step response testing, so PLEASE assure you do this before designing with the idea. Humm, thats a good thought, lets consider this just a functional working idea, that is prime for optimization for use in a product- like an early proto of an idea. I have not included wiring for the other pins. Please note that TerraLUX elected not to implement the MOSFET protection circuit of the controller chip, thus Isense and ADJ threshold which is used to protect the MOSFET from blowing up if something goes wrong. If you are not as extremely spaced constrained, this may be a very good thing to think about, if you desire reliability of your converter. Either that, or move to 0402 parts, and you can easily fit everything in. To disable the protection circuitry, connect ADJ pin to ground and Isense to Vin.
I was again asked about the Seoul P4 issue, where the shinny metal surface inside the LED, turns brown- without the dome turning brown, about the dome turning brown without the reflector turning brown, and about the phosphor turning brown on the Seoul P4.
I have seen many pictures of this happening. I've mainly seen the phosphor turning brown, and not too many cases of the brown internal LED reflector, or the dome turning brown. I have seen plenty of Silicone coatings brown over time in my days, and the chemistry and the purity of the chemistry used in the silicone is of high importance to prevent this issue. There are special silicones that OSRAM, LumiLEDs, CREE, and others utilize that are specifically designed not to do this. These types of silicone costs pennies more, and they are specifically designed for this application. That said, just like when the dome is abused during mounting on the CREE part and it wiggles easier, the Seoul P4 silicone dome is sensitive to a number of chemicals. I'd take a look over the datasheets, application notes, environmental stuff, and assure that these chemicals are not used around the LED during mounting or flashlight assembly. You may also want to specifically look into things like o-ring materials, reflector plastics, and lubrication compounds, to make sure you have no outgassing materials that will damage the Seoul P4. You may also want to look at recommendations for the LumiLEDs K2 and Rebel, OSRAM Golden Dragon, and such, in case Seoul missed some chemical their silicone is extra sensitive to.
Even though the TerraLUX runs quite hot, and I have over 100 hours on one module, and 50 on another, I have seen no signs of Seoul P4 discoloring. Seoul has had a tendency in the past to change their devices, processes, and materials used in them from batch to batch- this is important to note when considering failure modes, or attempting to use them in applications where you need long term reliability.
Main site
Mirror:
Main site mirror