SteveRay, there are previous threads on this. Right now I am behind on my CPF messages. My apologies.
Here's an update:
We have made several revisions to our design to reduce the instance of this occuring. Previous revisions concentrated on the how the PCB was grounded. But the problem persisted even though we have a good grounding system now.
Personally, I think the most recent problem is caused by a odd power supply stall. The switch on the chip gets stuck in the on position. As a result, it continues to draw power but does not make enough to fire the LED since it is not switching. Our circuit has not changed in quite a while so I think the chips we are using have changed in some small way. The fix I came up with involves adding another part to make the circuit more tolerant of chip variations.
It is possible it also gets stuck in the off posistion, but I have yet to observe that.
The chip manufacturer is not to blame because their parts meet the specification they published. It is my fault for using their chip in such a small housing which forces me to use minimal parts. However, their part it the best for our application in my opinion. Fortunately, I can add yet another part (with difficulty and extra expense) into the housing without changing the size of the Arc-AAA. I will not increase the price of the AAA or AA because of this. My contract with you is to produce a unit that works.
One of the reasons we have been out of the AAA and AA lately is because I pulled all of our stock (very expensive). Testing of the new design so far seems to fix every instance of the problem we observed. This was done by taking unglued electronic assemblies and placing them in a head and turning the light on an off until it failed to start. Sometimes this would take dozens of turns. We then took that board out, made a change and repeated the test. In all cases, the units then worked as they should without stalls. This is a simplfied description since it took trial and error to find the correct part value that would cause the unit to start each time. Like I said, it is an odd problem in the first place.
Unfortunately, I cannot go back and fix any units that has been fully assembled because the heads are glued. Taking them apart destroys more than you save. I will probably liquidate them on ebay someday.
Any rate, I have been working variations to this problem for about 6 months. I haven't always been able to devote as much time to is as I would like because of the Arc4 and other obligations (I really need to hire another engineer). But I have a fix that has been performing well in our tests. We have gone ahead and put it to manufacturing and those have been shipping now for about 2 weeks. So far, they seem to continue to work fine once they get in the customer's hands.
I think the units we shipped in December are the bulk of the problem units. If you have one that is acting up, please let us replace it with the new design.
And yes, I would prefer we fixed every problem before it went out the door. The reason the problem units passed QC is that we turned each unit on and off only about 4-5 times before it shipped. Now we take 10 from each batch, put them in the freezer, then in hot water and then turn them on and off for 50 cycles (by hand) and note what percentage fail to start even once. So far that test has produced 100% light up on all of the new units tested. There is a possibility that a similiar version of this problem may appear or that a defective unit may escape the new tests but I think we have at least significantly improved reliability.
Peter