Re: 2 Fried G210's in a single day :-(

Alan,

Thanks. Quality control is a tricky item to tame reliably. The
failure percentage floor (if you do everything else perfectly) is
pretty much set by individual component reliability.

Look at it this way. Each drive has 136 surface mounted parts and 23
thru-hole parts for a total of 159 parts. There are 477 solder joints
per drive to connect those parts.

That means in 12 months over 1 million parts were assembled using
3.25 million solder joints.

Out of those, 22 parts and/or joints failed that were clearly our
fault. That is a failure rate of 22 out of 4.25 million or 1 failure
per 200,000 parts/joints. If all 159 drive failures were so
attributed, then the rate would be 1 out of 27,000 parts/joints.

An argument for simplicty is if the drive had 27,000 parts/joints in
it, then every single drive would fail.

Most problems are caught during microscope inspection of the drive
panels (12 drives per panel) as they come out of the conveyor oven.
The failure rate per drive is about 10%, consisting of solder
bridges, tombstoning of parts and parts that have flipped upside down
during SMT placement. These errors are corrected using SMT rework
equipment under a microscope.

After the thru-hole parts are mounted and the drives are de-
panelized, they undergo initial electrical power-up. The failure rate
here is about 3%, consisting of primarily of individual component
failure. Again, these drives are reworked and retested. If they still
don't work, (<<1%) they are trashed.

The drives are then mounted on the plate, connector plugs are
installed and the cover is put on. Then they are run thru a maximum
electrical stress test stand. For G201/210s this consists of a 100VDC
power supply and a 7A load. About 2% fail this test and are trashed.

These tests catch well over 99% of all problem drives. Some of the
ones that get away (the 22) would have been caught by a 24 hour burn-
in cycle. This however is not economical (cost vs. benefit) for the
price range drive we manufacture. It would require at least 50 drives
running continuously while consuming 250W each (load and drive
dissipation). That's 12,500W 24 hours a day with no certainty that
every bad drive would be caught.

Mariss

--- In CAD_CAM_EDM_DRO@y..., beer@s... wrote:
> On 8 May, CAD_CAM_EDM_DRO@y... wrote:
>
> > User induced or undeterminable failures were 19* drives out of
4,680
> > OEM units shipped, or 0.4%. Non-user induced failures (our
fault)
> > were 12 units, for a failure rate of 0.25%. This is a much more
> > tolerable number.
>
> Mariss, thanks for this information. I've always been a bit
curious.
>
> These would be great numbers for something as low-tech as a toaster
> oven.
>
> Considering the technology inside a Gecko, the terrible task we put
them
> to and the wide range of end user electronics experience, these are,
> to my mind, astonishingly good numbers.
>
> Congratulations.
>
> Alan
>
> --
>
> Alan Rothenbush | The Spartans do not ask the number
of the
> Academic Computing Services | enemy, only where they are.
> Simon Fraser University |
> Burnaby, B.C., Canada | Agix of
Sparta