They probably dice out to the edges as far as they dare.
I expect incoming parts to have PPM incoming quality, and I'm going to notice, and blackball, any supplier that hits hundreds of PPM. I'm guessing that no semiconductor could be manufactured to PPM quality without testing every one.
John
As you scale up production volume, you scale up all the manufacturing equipment. Why not?
A machine that tests 20-40K diodes per hour might cost $30K. Run that
24/7 and do the math.Semiconductor test is a big business.
-------------------------------
It is, and that?s why they are not going to waste money on things they don't need. Doesn't matter how you scale it up as it's always a fixed percentage of the total. If your selling components at 1c that can become a significant extra cost factor.
Basic silicon wafers have defects. Then you diffuse it to make the junctions, and hope it's perfect and uniform across the whole wafer. Dicing creates more defects. Then you have to solder the diced diode down to nailhead leads on both sides, or solder one side to a lead frame and wirebond the other. Then mold and cure the epoxy case around it.
--------------------------------
The defects are generally well known. The outer sides of the wafer have a higher degree. The people that actually do this for a living know much much more about it than you.
Another thing you seem to fail to understand is that of repeatability. If something is highly repeatable you can statistically get a degree of confidence of what to expect. This is enough. If 99.9999% of components produced are fine then out of 10M that is just 10 bad components.
Is it worth spending several million dollars just to determine 10 bad components?
-------------------------------------- Do you think that can be done to PPM quality levels, without testing?
I don't.
I don't think any semiconductor, from diode to FPGA, can be sold without testing every one. I don't know the test yield of, say, power diodes, but it sure isn't low PPM. Things like processors and FPGAs are sold with test yields below 25%. Some chips, like flash and drams, are built with redundant stuff, tested, and bad blocks swapped out for good ones; otherwise yield might be 1%.
---------------------------
lol, an FPGA is a lot different than a diode. The number of failure modes of an FPGA is much much greater than that of a diode.
Even so, an FPGA is not tested for every failure mode. Tests are ran that make sure it works 99.9% of what it is expected to do.
What sort of electronics are you designing lately?
-----
Care to sign an NDA?
Well, scratch that... I don't trust you either way... not that your competent to do much with it...
So your answer is none?
Thought so.
Well, they didn't test the Hubble telescope until it was put in orbit, did they?
He is confused by "extensive testing" and normal production testing. He doesn't understand that sample testing involves environmental, mechanical, life tests, etc, while production testing (100%) provides those "lots" that the samples are pulled from. They can't perform 100% extensive testing on the production line, but they can perform 100% certain tests to ensure a good product.
Cheers, John
I expected as much. Or as little, actually.
John
b...
ly
mum
if
les
needed
ndon't
ge
cant
t ah much
fs of
tthe
s e .I've seen a transistor production line, but not diode. Every device was tested, and each wafer produced a wide variety of transistor types according to what specs each device fitted. There was nothing like consistency of specs achieved, and wafers normally had rejects unsuitable for anything. Diodes of course are simpler, and time will have improved the process to some extent. But it would take a very big leap to produce on-wafer components so consistent that they could be sold without testing them all.
NT
Probably impossible, certainly too expensive.
John
ty
ing.
to
-- Cheers, James Arthur
"Only the Paranoid Survive"
- Andy Grove
John
[snip]
I understand that Stretto, Ian Field, AlwaysWrong and John Larkin were part of the study group ;-) ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
"You're" not "your". Of couse, we couldn't expect any better from an ignorant troll like you.
Stretto, Are you just an ignorant troll? The methodology is simple. EVERY part is tested for basic parameters, but a FEW are tested for EVERY parameter. The basic test assures that every part works, and the extensive test assures that all parameters are met.
What a maroon... ;-)
Charlie
20
=3D
ility
esting.
uct to
Speak of the devil...
"New York Times columnist Thomas L. Friedman ... [wrote] [l]et tired old companies that do commodity manufacturing die if they have to. If Washington really wants to create jobs, he wrote, it should back startups.
Friedman is wrong. Startups are a wonderful thing, but they cannot by themselves increase tech employment. Equally important is what comes after that mythical moment of creation in the garage, as technology goes from prototype to mass production." --Andy Grove
-- Cheers, James Arthur
testing.
to
I wonder if Friedman has ever had a real job.
he left out "in Maylasia"
John
Yes thay did. But they tested it wrong. How do you do SPC when the entire lifetime production is one piece?
Good book: The Hubble Wars.
Actually, there were two mirrors. I think Kodak made the backup mirror. It was right, but they didn't fly that one. Astoundingly, the two mirrors were tested with different equipment.
John
but
And not cross-tested, even thought the two test setups existed. The Perkin-Elmer mirror was smoother, so it was expected to produce less scattered light, but of course it's the wrong shape.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics Electro-optics Photonics Analog Electronics 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
I remember one situation where a failed thermistor, during accelerated life test, ended up *lowering* the SPQL. Their excuse was that it was evidence that the test was meaningful. Of course the failure was mechanical, likely someone pulled the leads off the thing when loading up the chamber after a readout.
.
In my experience mislabeled parts are about 100 times more common than defective parts.
Mislabeling is in many ways dang close to defective. But in many cases it's obvious. If you ordered a sump pump and got delivered a nuclear power plant, you would probably notice, huh? :-)
ll).
Because of MAXIM supply crunches there's a huge market in counterfeit stuff.
None of these issues are particularly new. Look up "tube washers".
Tim.
If you don't use Maxim parts, you won't have problems with counterfeit Maxim parts!
John
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.