[rescue] Slightly OT: Bad Cap Saga

[Curious George]

--- On Mon, 8/18/08, Geoffrey S. Mendelson <gsm at mendelson.com> wrote:

> On Sun, Aug 17, 2008 at 02:48:36PM -0700, Curious George
> wrote:
> 
> > I've never seen such a margin *published* by a vendor.  Instead,
> > WVDC was always the "rated specification" for the component
> > (e.g., unlike something like TTL that you *can* run on 7VDC
> > despite it being designed for a nominal 5V supply).  "Best
> > practices" always have you seriously derating the specs on
> > things like caps for exactly this reason.  (of course, the
> > things I design/build are intended for longer service lives
> > so I can't play fast and loose with choice of components  :-/ )
> 
> Published by a vendor, or told to you by the manufacturer
> if you asked.

I don;'t care *where* the number comes from -- as long as
the vendor stands behind it (and we have recourse as to
how to respond when/if this ever becomes untrue).
When we meet with Japanese vendors, it is very unnerving
trying to get them to give you specifics about design
limits.  They always want to know what we "want it to be".
Perhaps this is a cultural issue -- I think Americans are
used to working to specifications whereas the Japanese
seem to be geared to working to *your* specifications!

> Without meaning to sound racist, I doubt that if YOU asked
> the question,

Not sure why I should consider that "racist"... are you
oriental?

> you would get an answer. Things told over a meal with the
> guy down the
> street are very different to customers on the phone or in
> the conference room. 
> 
> I don't know about the Chinese, but I had a very
> expensive lesson in 
> "the deal is not really signed until we get drunk
> together" with the

The worse (IMHO) experience is sharing "big meals"
(since my decidedly American palate finds many of their
foodstuffs quite unappealing  :<  )  Getting drunk is
the same regardless of whether its Saki, Ouzo, Grappa,
or Budweiser!  :>
 
> president of a Korean company, you probably have 20-30
> products of
> in your home, or items containing parts made by them.
> 
> > I don't know how one could even *try* to design that way!
> 
> Oh come off it. You do it every day. I'm sure that you have design
> limitations and you try to minimize cost by staying at the edge of
> those limitiations. You just have a different understanding of those
> limitations.

Yes.  Exactly!  Now, what *are* the limitations in these cases
where you don't test product, can't rely on whether or not
your vendor has given you a product that conforms to a (dubious)
specification, etc.?  Please look at the context of my reply.

I *really* don't know how (without lots of historical
data to rely upon) to tell my boss "we can make these products
and expect no more than N% scrap (which we'll have to eat)
for a cost of $X".  I have *no* experience in that sort of
manufacturing environment.

When I started designing products, I used to factor in a 3%
warranty cost (i.e. 3% of DM+DL -- so the actual factor
was fairly less than this).  Over the years, experience has
taught me how to work that down to a fraction of a percent.

I don't have that experience when the components, labor,
assemblies, etc. I am given are of unknown quality.  Given
time (years?), I imagine I could apply the same statistical
processes to *learn* what those new factors would be -- but,
in your scenario, it seems likely that they could vary
widely from vendor (supplier) to vendor.

> > I.e., you would have to spend considerable effort characterizing
> > the parts you buy (you being the actual manufacturer) so you
> > could be sure the units would pass *your* outgoing inspection!
> > Or, are you saying that the original manufacturer (*not* the
> > company who's name is on the OUTSIDE of the piece of equipment)
> > makes no warranties to *its* customer (i.e., the company who
> > will ultimately sell to John Doe)?
> 
> Yes. The price for untested items is significant lower than
> the cost of tested ones. Statistical testing (testing every
> 5th, 100th, 1000th) raises the price less, but still costs as if a unit
> fails, you either toss the ones since the previous test, or test
> each one.

So, presumably, people (actuaries?) at these firms have developed
their own characterizations for *their* products using *their*
vendors/suppliers.  Are they wed to key suppliers?  Or, do
they take a (statistical) gamble each time they opt to
purchase a subassembly from some new/other firm?

> If you buy a consumer item, let's say a DVD player from
> Wal-Mart, you have no idea of where or how it was made, the working
> conditions and age of the workers who made it, and so on. You have
> no more desire to know about it than to know 
> that hambuger you ate for lunch used to walk around and moo. 

Of course!  But, if my role as consumer is one of VAR, OEM, etc.,
then I am *very* concerned about where that subassembly I just
purchased for use in my product came from!  And, the likelihood
of it being able to meet its specifications (performance and
reliability) necessary for *my* product to meet likewise!
(since *I* have to give a warranty as I am selling to an
"end user" -- or, VAR who will hold me accountable for same)

> If this DVD player dies in the first 90 days, you take it
> back to Wal-Mart
> and they trash it. Trashing 1 out of 1000 or whatever the
> expected rate
> is was included in their cost calculation. It's
> probably cheaper than 
> having a local technician look at it, let alone repair it,

Of course.  But, where did they get their numbers?
If you're WalMart, you probably have a deal with your vendor
whereby you just *tell* them how many you trashed and are
compensated.  *Or*, threaten to stop doing business with them
(and, since you have lots of clout, they comply!)

> and far cheaper
> than shipping it back to Long March and having them examine
> it to determine why it failed. 
> 
> If it fails afer 90 days, you buy another one. If you were
> upset by the short life, you buy another brand, or go to another store.
> If you do, I'll bet you never look far enough to find out they
> all say Long March DVD player company in small print on the
> circuit boards in Chinese, or even
> could read it if you looked.

Or, if you are like me, you find a product/brand that has given you
satisfactory performance and stick to that.  Conversely, avoid
products that give you *poor* performance (or, perhaps, poor
customer service, etc.)

> > Or, do these folks operate in an environment where they never 
> > have to "pay" for their mistakes (i.e., through repairs and/or
> > lost business due to bad reputation, etc.)
> 
> Do you know anyone that shops at Wal-Mart? Ask them what
> they think of the store, the service, the products, and so on.
> Does a negative opinion stop them?

Remember, all consumers are not John Q Public.

I approach a Taiwanese vendor to outsource some aspect -- perhaps
the entire product -- of a product.  I am a consumer/customer in
the eyes of said company.

By your comment, they don't test their work before they sell it
to me.  I pay real money for their products and then resell
or integrate them into my own products.  *I*, dealing with the
end user, give a warranty (since that is expected, even if
a silly 90 day wonder).

I then discover that this company has produced total crap!
I see 10% failure rates.  *I* eat that cost (or go belly up).

Now, I reask my question:

   "Or, do these folks operate in an environment
   where they never have to "pay" for their mistakes
   (i.e., through repairs and/or lost business due to
   bad reputation, etc.)"

When I go to said Taiwanese firm, do they just smile and
say, "So sorry..."?  Do they never have to pay for their
problems?  Sure, *I* will stop using them as a supplier
and will probably gladly recount my less than wonderful
experiences with them... are you saying life just goes
on for them without any repurcusions?

I.e., they *have* to somehow factor this experience into
their business model/pricing/methodology *or* just hope
there's another sucker qaiting behind me...