Jim, you are really wasting your time with this "analysis"
of Carl's sentiments!
<there is nothing wrong with education>
Apparently something is very wrong. Maybe with your
own education?
You seem to be even incapable to comprehend what Carl
is talking about. See:
<I think they need the solder to hold the components in place and make a good electrical connection. <G>>
Unless you are really joking and desperate to stick
with your agenda, you are showing off your total
ignorance here. When board design
people are talking about amount of solder joints,
they mean first that every such joint is a inhomogenuity
along the transmission line, hence it introduces a
source of reflection. The reflections from multiplicity
of spatially distributed sources tend to superpose
toghether and may form unpredictable signal
level at the bus receiver, causing system errors
time to time.
Second, when designers are talking about multiple
solder joints in a ultra-high frequency design,
they also think about variance in solder mass
soaked into the pad. It affects the local capacitance
of the joint, and therefore contributes to uncertainty
in the overall signal. To reduce the unwanted effects,
manufacturer has to tighten up control process. Everything
needs to be better controlled - preheat and soldering
temperature, viscosity of the solder compound and flux.
All this add the cost, and not the 4- or else-layered
boards or the number of traces as you primitively tend
to think:
<By the way bidirectional busses reduce the number of traces requirred on the MOBO.>
You must be joking again here. You probably are not
aware that to make a bus bi-directional, the point
of connection must have both receiver and transmitter
attached, and be enabled in alternative manner. It is
hardly possible to provide the same impedance for
active and inactive transmitter, therefore the
bi-directional busses have inherent impedance mismatch
at every connection point. The impedance also changes
dynamically as the device transmits or receives data.
It again introduces inhomogenuities and hence
the unpredictable reflections.
Once again: the problem with the Rambus "technology"
is their attempt to combine several mutually exclusive
things - ultra-high frequency and long _multi-drop_ bus.
The keyword here is "multi". When you have only two
agents - transiver and receiver, you have only few
problems. If you have many bi-directional agents,
the problems escalates as a power of their number.
One way to solve the "problem" is to tighten device
specifications to almost zero to match the
single working laboratory sample. This escalates
the cost of manufacturing, or what exactly we see
in reality. Zero-tight specifications is an indication
of lack of understanding in both design tolerances
and mass manufacturing capabilities.
To save your face I would recommend you to say that
you were just goofing around here and was not
serious. Good? |
