Sudden Death Answer

[gesper]

Advertisement (gone after registration)

For a $5,000 camera, you would think they could test every individual transistor if they know it can kill a camera after it has already gone through post-assembly QC (which should otherwise catch fatal problems before they get out the door).

[Riley]

the fix then involves finding out how extensive the fitment of this component is

then by deduction how many cameras are affected

locate if you can the serial numbers of that stream

 

since it would be a batch bought part that may not be easy

 

if the minimum voltage can be controlled by the firmware, thats an obvious fix

if not, its going to be difficult

[Riley]

For a $5,000 camera, you would think they could test every individual transistor if they know it can kill a camera after it has already gone through post-assembly QC (which should otherwise catch fatal problems before they get out the door).

 

im sure they do, and they have

but if a part is issued that is below spec

and that leads that component to fail

that can be hard to detect, and doesn't necessarily happen at the garden gate

[barjohn]

If, as stated, it turns out to be a transistor that is failing and causing the sudden death issue, whether it is because the parts delivered were not to specification or the part specified was not adequate the only likely fix is a replacement of the part. I wonder whether Leica will want to recall and replace the part on so many cameras now that they may be looking at 10K or more cameras. They has a hard time processing less than 1500 before.

[gesper]

im sure they do, and they have

but if a part is issued that is below spec

and that leads that component to fail

that can be hard to detect, and doesn't necessarily happen at the garden gate

 

What I meant was that according to their message, the suspect transistors are tested by random sample when they come in, and some bad ones didn't get tested. If they know this specific transistor is causing problems, they should test 100% of the transistors before they install them rather than a random sample.

[gesper]

If, as stated, it turns out to be a transistor that is failing and causing the sudden death issue, whether it is because the parts delivered were not to specification or the part specified was not adequate the only likely fix is a replacement of the part. I wonder whether Leica will want to recall and replace the part on so many cameras now that they may be looking at 10K or more cameras. They has a hard time processing less than 1500 before.

Since they were doing a random sample and the tested transistors were passing, the actually failures should be a small subset of all cameras manufactured. I would expect Leica will wait and fix the ones that fail rather than recall all of the cameras to replace perfectly good transistors. Bad news for those whose cameras fail, but I would just as soon not suffer another trip back to Solms if there is a chance mine won't fail.

[marknorton]

Advertisement (gone after registration)

It's not practicable to test components like transistors - which, depending on the type, cost less than $0.03 in quantity. They arrive at Leica's subcontractor, 1000 or 5000 to a reel ready for installation in a board stuffing machine or, more exactly, a surface mount component placing machine.

 

The economics of electronic production are such that it costs more to put the component on the board than it does to buy it.

 

The hole here looks to be that the automated board testing is not exercising this part of the circuitry.

[Guest tummydoc]

Being trained in science is sometimes a curse because most of the time I have to reject the most comforting "explanations".

 

Unfortunately there is no justification to the conclusion from random samples having passed a test, that faulty transistors are only lurking in a portion of the total number of cameras built to-date. Virtually all M8s that died--like my first two--were working for varying lengths of time before they malfunctioned. There is nothing to rule out that they are all defective and will malfunction at some point. Moreover, just because replacing the faulty transistor brings the camera back to operation does not prove that replacement is "the" cure. It bears a thorough investigation to determine beyond doubt that there is not some other fault that's frying the transistor.

 

I trust the technical people at Leica implicitly to understand this, and hope they are not overruled by the marketing people into declaring the war won prematurely, because if they are wrong it will be a senseless blow to their reputability.

[barjohn]

Random sample testing would reveal some level of failed transistors, or if the sample size was too small, may not reveal any even though there were some. However, even if a failed transistor was missed it would likely have shown up later when the unit was tested as a whole or when the sub-assembly was tested. It is unlikely that the random testing would have led to the discovery of components with marginal tolerance unless their testing includes testing to the margins. It is unlikely that a bad lot was involved because random testing to the margins would have caught at least a few of them and led to the discovery of the bad lot and action to fix the problem. My guess is the engineers underspecified the required transistor under some stress condition they did not anticipate and there are basically two fixes, 1) make sure the conditions do not arise and 2) replace the component with a higher rated component. There is a third posibility and that is that some other circuit is causing the condition that frys the transistor (basically my fisrt one but more specific).

[Riley]

What I meant was that according to their message, the suspect transistors are tested by random sample when they come in, and some bad ones didn't get tested. If they know this specific transistor is causing problems, they should test 100% of the transistors before they install them rather than a random sample.

 

just not possible, nor might it prove worthwhile

what they probably do is measure a sample

even that wont tell them much until a component fails