My comments on Tony Frediksson's post are indented among his words.
It is indeed the case that testing in one representative host meets the
requirements. But the statement, "...of the PWBA manufacturer's choice so as to
ensure compliance of the PWBA with the entire population of hosts in which it is
intended to be installed" is absurd.
If the PWBA manufacturer is smart, he will indeed pick a representative
type of chassis. Otherwise, people who want to use his board will
reject it as not passing in their implementation. More on this later.
Often times, the host chosen by the PWBA manufacturer is the quietest device
that the manufacturer can find that will allow the card to pass and is not
representative of a typical host. This may allow the PWBA manufacturer to
eliminate reasonable port filtering, minimize pcb layers, use unshielded
connectors, avoid fixing poor layout on the card, all in the interest of
cost reduction and time to market.
Too true -- but this is all taken care of when WE fail them.
Then, when the PWBA is installed in a real typical host, the thing fails
emissions limits by wide margins. I can think of one video card that will
remain nameless that is very popular. I have tested the thing in a variety of
host systems with a variety of video displays. The card has an FCC Class B ID
and CE Marking and is marketed as Class B. I have never seen it pass Class B in
any combination of hosts, and have failed it for Class A on several. I know of
ethernet cards with single and multiple ethernet ports that have this problem
too.
We could swap horror stories! But the bottom line is... if they don't
pass, we can't use them. I've worked with more than a few vendors to
help them bring products they wish to sell up to a standard that would
allow my employer to use them.
Likewise with 15" video displays. They are typically such a cheap commodity
that most no longer use even a full shield on the video amp much less any
enclosure shielding. Tested on a variety of video cards and in a variety of
systems, they often fail Class B limits. I find two problems typically: poor
video cable shield termination, and poor shielding of video emissions emanating
from the cathode drivers in the displays. It is obvious to me that the 15"
color display manufacturers are choosing quiet video sources to qualify these
products. I conclude this assuming that they are staying within the law by
manufacturing products that are typical of the original test sample.
There are a number of dodges I've seen manufacturers use. One of the
slickest is software which created a page of fairly large "H's" on the
screen -- and kept the same size no matter what video resolution was
selected. Need I tell that this was not the way I tested it? I
remember, too, one manufacturer produced equipment with extremely high
performance -- but certified it to a level less than it could achieve.
That's not the way I tested it, though, and I imagine a firm desiring a
high level of performance would insist on having the ability to make use
of it.
Then there are the keyboards and mouses that seem to display curious anomalies
(to be kind under ESD, EFT, radiated immunity and conducted immunity testing
that are CE marked. They have essentially no enclosure shielding, little or no
cable shielding and no common mode noise suppression, cheap PCB layout, and are
CE marked. Again, I have tested certain models on a variety of hosts to draw
this conclusion.
Pointing device tests are one of the areas for which the standard may
not be really adequate. How do you exercise a mouse when you can't get
a conductive close to it? The same is true for keyboards. It takes work.
>From a technical point of view, testing with a representative host/support
device to show compliance with other hosts is a total "crock." The standard
should be ammended to remove this bogus claim.
The question that needs to be asked here is, with what would it be
replaced? Would it be something we could actually _do_?
>From a business and practical perspective, it makes perfect sense. To do the
right thing technically (test on a typical host) would be way too big of a
burden for industry to bear since "typical" changes rapidly and would be
impractical to prove. Therefore, as much as I dislike it, the actual
requirement in the standard should remain the same.
This makes the point I brought out one paragraph ago.
Unfortunately, the burden falls on system integrators and importers trying to
qualify the myriad combinations of components requested by their aggressive
marketing/sales departments. But, nobody said it was an easy job!
Right! And it gives us an excuse to beat up marketing types in dark
alleys! (g)
Cortland Richmond
