[LinuxBIOS] bios of single devices

Richard Smith smithbone at gmail.com
Wed Aug 10 15:36:07 CEST 2005


> if every single hardware device or item of a PC-System, like hard-disk or
> cd/dvd-rom drive or pci-card or floppy,
> has an own little bios-chip,

Not exactly. What those devices will have is a microcontroller or an
ASIC that is programmed to do the work.  It's not a BIOS per say but
it is code.  Most of the time its not user accessible unless you
happen to know some special things about the device.
  
> the ques is, for example, if PC beginner simply
> plug off such a device, while PC is switched on,
> and "damage" this device,
> that probably this device is not damaged,
> cause you only need to re-flash this
> little BIOS-chip or hardware-TM-brand-chip of concerning device ???

Nope.  Most of your failures are from manufacturing flaws in the
solder reflow process.  A joint that is not completely soldered will
over time fail and cease to conduct or conduct so little that the
device no longer works.
The reason this happens on power on/off most of the time is thermal
expansion.  Every time a device is powered on or off the joint goes
through a thermal cycle.  A bad joint only has N cycles before its no
good.  So when thermal cycles > N the device fails or starts doing
flaky things.
I'd guess in most cases the higher resistance of the joint causes a
slight shift in the timing of a signal and the device starts doing
flaky stuff.

Its really hard to find these joints in a manufacturing line.  In a
lot of cases the flaws are only evident via X-Ray of the joint. 
Current cost margins for devices do not allow the Mfg to 100% X-Ray
inspect all the boards.  So they work for a bit and get the process
tuned out until the process line yields a set of boards with a %
defect rate that's below what some exec or marketing person had deemed
acceptable.

Then they just start cranking out parts with spot checks here and there.

In most cases these flaws will pass the "test" and "burn in" phases
with no problems since the N level may equate to a week or so of use. 
Some times the temperature extremes and shock its exposed to in
shipping will be enough to push it over the edge.

When a device does have flashable firmware and the Mfg provides an
upgrade to "Fix" some issues a lot of those cases are where they have
enough field failures to identify where the timing is "marginal" and
apply better software to compensate.  In the case of devices with an
FPGA in them they can directly change the timing by adding buffers or
latches.

Hope this helps.

-- 
Richard A. Smith




More information about the coreboot mailing list