Barbarians - they break what they don't understand.
Witness the actions of the Democrat party towards the economy over
the last forty years.
"With Age comes Wisdom. Although more often, Age travels alone."
Tearing stuff apart is the greatest way to learn about how things work.
I was just thinking about other things that are just overly complex for no
reason and remembered server rail kits from Sun and especially the ones
from Sun designed by Fujitsu. They're supposed to just be rails that allow
a server to slide in and out of a 19" rack. Pretty simple, like glides for
a desk drawer. HP has it figured out, Dell took years too, and almost got
it right, but not Sun/Oracle/Fujitsu. I've still never figured out what
all the extra pieces are for, even with the installation book, and I've
not come across anybody else that has either. As to why rails need to be
highly asymmetrical from left to right is mind boggling. Even with ball
bearings, they're harder to operate than metal on metal sliders, are prone
to just falling apart and require special alignment jigs for installation,
even into industry standard racks. Plus, with no matter what you do,
you're going to get grease all over your hands. Here's a personal message
to anybody involved in those products - "you're a complete idiot".
Have you ever personally participated in the design of new products?
You start by defining the requirements, or rather debating them until
you're too tired to argue, then distribute the work among your
personnel, come up with a separate solution to each requirement,
prototype and test them individually and then together, and finally
try to combine the elements that seem to need no further redesign to
serve multiple functions and reduce tooling, fabrication and assembly
cost while management pesters you to release it to production NOW to
beat the competition to market. They are obsessed with the name
recognition and sales momentum that comes with being first, and know
that the engineers would love to keep playing with it.
All the while realizing that you may be out of a job when it's
complete, unless your performance gets you nominated to the next new
product design team, if there is one.
At the prototype stage having each part serve a single function is an
advantage when it needs to be reworked. Combining and simplifying them
later is time-consuming and non-essential.
Any competent draftsman can design complexity, simplicity requires
Or, as they used to say at GM, "Any damned fool can design a
carburettor for a Rolls-Royce. It takes a genius to design one for a
Or the tongue-in-cheek motto applied to Mercedes-Benz: "Never use two
parts to do a job when you can get away with three." <g>
The trouble with Chevy is they would order the tooling and test
station for that carb well in advance and then keep calling with "Oh,
by the way..." changes as they refined it.
We larded the test station for their 1970's analog ABS controller with
jumpers so they could change the test parameters themselves. While I
enjoyed flying first-class I didn't at all like Flint MI. For some
reason there weren't many passengers on those flights.
GM's project engineer was a Ph.D. from India with absolutely no
practical hands-on experience. He wanted the ramp-down curve of wheel
sensor speed accurate to 8 decimal places because that's what his
calculator gave him. No one had taught him that resistors have
Prior to emission controls the only electronic device in a car was the
radio, which they bought. The new electrical engineers they hired
faced a steep leaning curve to adapt to the contamination and
fighter-plane-like range of temperatures in an engine compartment.
You can buy 'as marked' but they are only that value at a specific
temperature and they aren't cheap. Considering that the first standard
was 50%, .01% is a 5000% improvement. I've bought reels of .o1% on Ebay
for well under one cent per part.
Anyone wanting to run for any political office in the US should have to
have a DD214, and a honorable discharge.
GM solved its problem by changing from analog op amps and comparators
to digital microcontrollers in the ABS module. I heard that the analog
ABS design had come from someone who created it in his garage and then
dropped dead before he had properly documented it.
The bigger accuracy issue in building high-precision automated test
equipment, where I used the 0.01% resistors, was actually capacitance.
While dual-slope A/D converters may automatically compensate for
temperature drift, the storage caps in the input Sample-and-Holds
don't. They matter because they are very small to track fast and the
FET that isolates the cap from the signal to maintain a steady voltage
during the conversion also couples unwanted charge into the cap from
the voltage swing on its gate. We could couple a trimmed compensating
charge switched by the gate control into the sample cap but the
tempcos weren't matched since the compensating cap was a small patch
on the circuit board.
The larger capacitance error source was dielectric absorption, when a
polar capacitor dielectric acts slightly like a chemical battery and
takes a short while to fully charge or discharge after the voltage
changes. When you have only 5mS to set up and make a measurement it
becomes a serious concern. We had to have W.L.Gore make custom
Teflon-wrapped reed relays to cut the dielectric absorption in the
relay matrix that configures each test down so it didn't degrade
When I was working QA at a manpack radio mfgr, we got a new HP LCR
meter. I was amazed at how much capacitance and/or inductance were
added to leads during tests just by getting near them or touching
them. I can't imagine designing a multilayer PCB, with all those
stray waves invading other spaces. Ditto ATE.
Layered directly on the board, or soldered on?
I'll bet that was fun to figure out, even without the cycle-time
Teflon sheds electrons? Or is it just a good insulator? (Or are those
the same question? ;)
Give me the luxuries of life.
I can live without the necessities.
For a 6 layer board you put the power planes on layers 2 and 5 and set
the preferred directions on signal layers 3 and 4 at right angles to
each other, which is standard practice anyway to make autorouting
simpler. Then each trace has a constant capacitance to a power plane
and minimal coupling to the ones that cross it, and thus a uniform
characteristic impedance. If the logic family uses multiple voltages
like ECL you may have to go to 8 layers, but I could normally build a
densely packed, low-noise board with 6.
The solid planes need to be arranged symmetrically or the board will
warp since it's pressed together hot and the copper and glass-epoxy
shrink at different rates as it cools. I had to salvage a board
another tech designed with an outer ground plane on only one side
which had dished so much it would float.
When you see an unused area on a board that is crosshatched it's to
equalize the thermal contraction during manufacture.
If the power and ground planes are properly bypassed either can serve
about equally well as the "ground" plane, but you need to consider the
noise rejection margins both ways for asymmetrical logic families like
TTL. CMOS doesn't care. It's really necessary to distribute
low-impedance Tantalums around the board and have 0.1uF and 0.001uF
MLC caps at each device, since the two sizes have different frequency
responses in the GigaHertz range.
Thermal reliefs at the power plane vias make hand soldering and rework
far easier. They are the dashed rings at some pins on the Internal
If the impedance is critical such as in a microwave digital radio or
oscilloscope front end you may have to discuss the board's fabrication
with the vendor since they can't simultaneously control the thickness
of all internal layers and the overall board thickness. Something
needs to squish, like the prepreg between 3 and 4 if the outer layers
are controlled. It's a good idea to learn all the standard and
extra-cost limits of their processes.
I haven't heard of any other experienced circuit board designers who
thoroughly understood complex circuits. Management wasn't real happy
paying their lab manager to perform that apparently menial task, but I
proved I could deliver challenging boards that worked well the first
They were small rectangles of copper on the board, capacitively
coupled to the ground plane. The trim was to bump them up part way
between power and ground with a hand-selected voltage divider, to
control the amount of charge they coupled into the cap. The S/H sample
cap was as small as acceptable to better follow rapidly changing input
This is what they are, if you're unfamiliar with measurement circuits:
"If the input value was permitted to change during this comparison
process, the resulting conversion would be inaccurate and possibly
completely unrelated to the true input value."
The project engineer was a beard, ponytail, sandals and VW bus genius
Ph.D who had designed magical instruments at Keithley. He had me build
a current meter to characterize dielectric absorption currents which
could measure to a picoAmp in a milliSecond, which he said was state
of the art. Its output was a scope display of current vs time. I had
to heavily rigidize and shield it to eliminate mechanical vibration
from the ultrasonic building alarm and us talking. Even the less
sensitive (more stable) instruments in the production machine could
detect a person's electrical field at 10-20 feet if unshielded. I used
that to make it wake up and greet anyone who approached it after
hours, mostly the programmers. For some reason it ignored me sitting
still next to it, perhaps it accepted me as its master.
The molecular polar moment that gives a high dielectric constant
causes some delayed current flow as it physically relaxes after the
voltage changes. That's why electrolytics are poor for audio circuits.
Dielectrics with low absorption result in capacitors that are
physically large for their value, with unwanted parasitic effects.
They aren't the simple, perfect components that EEs learn them as in
wrote: >>For a 6 layer board you put the power planes on layers 2 and 5 and
I thought CAD software slowed me down because I could see the whole
board at 2x or 4x magnification on the light table, but only a small
window on the screen, and I could use both hands to apply the tape
strips. The snap-to grid in CAD was a little easier to stay on and
much easier to change than the printed grid sheet on the light table.
You plan out the path of each connection separately, the same way
you'd plan a trip on a map. The difference is that you run each new
trace alongside the previous ones so you don't quickly clog the
remaining free space, and move vertically on one side and horizontally
on the other. On a 386 computer a good designer could hand-route
almost as fast as the autorouter, like man vs machine chess. The
program advanced incrementally, tried every dead end, and couldn't
step away for a global overview.
I complained of eye strain to get my boss to buy a $2000 20"
long-persistance (no flicker) CAD monitor, which helped a lot because
I could see both ends of the trace I was hand-routing to minimize the
number of inductive corners and unwanted coupling to other signals. I
still had to avoid red which the eye focuses at a different distance
from other colors.
74AS hi-to-lo transitions are fast enough to excite GigaHertz
resonances. Another department brought me a board to troubleshoot with
a problem that had eluded them. I found that they had used low-speed
analog design rules to make a star-pattern ground plane and where 74AS
signals crossed from one lobe of it to another the ground bounce
during a hi-to-lo bus transition was over 3V, because the signal's
return reflection in the plane couldn't make it out to the
single-point plane junction and back along the other side in a
Being bested by a lab tech doesn't do a Ph.D's self-image any good. I
had to present the evidence very clearly on a scope display, which
usually took longer than finding the problem, and be as diplomatic as
possible, letting them see the evidence and make their own judgement.
It was surprising how much trouble some of them had with mental math,
like converting the time between pulses to frequency. Quick, what's
the frequency for a 60 nanoSecond period?
Many discrete components' properties look terrible when you sweep them
with a microwave network analyzer. This shows the deviation of a chip
cap in an SMT package that's larger than necessary for its value.
When the company was bought the competitors sued that it was "unfair"
that the biggest company in the ATE field had acquired the small one
they acknowledged to be the best. The cost of fighting that suit
Engineers don't learn those strange ways unless they encounter them in
practice. I did a job for a well-respected mechanical engineering prof
who devised an overly complex way to assemble a robot chassis because
he had never heard of Pemnuts.
I wasn't speaking incrementally, rather about the difference between
doing it by hand then, and doing it now with sophisticated software.
(Yeah, I know, you had to take the long, hard route. Condolences.)
I toyed with the freebie downloads of autorouters for the low-level
stuff, just for fun. A decade later, both software and hardware were
multi-generations better, faster, and smarter.
Remember the yelonblu screens way back when? Ayieeeeeeeeeeeee!
Nice monitors are great. While you played with CAD, I was buying
better than average (but nowhere your grade) monitors for graphic
design work. A friend had an old Multilith 1250 press and I'd print
out 8.5x14" polyester plates (4-color seps) with my HP LJ5p laserjet.
That was low-tech fun. Once I got a better monitor, I could see the
problems beforehand and avoid printing them. Resolution is God.
Hah! That's a helluva bounce, a bit more than a couple pf could
handle, wot? (But how do you damp the ground plane?)
Oh, joy! Diplomatic Dog & Pony Shows suck.
"I don't need to know that. Ask my secretary."
Suckage. More often than not, lawsuits destroy everything they touch.
Indeed. Too many of those uberProfs never get into the trenches at
all, missing 2/3 of knowledge in their field by default. The result
is that I seldom trust anyone with anything over a Master's. Too
often, it _is_ Piled Higher and Deeper.
Give me the luxuries of life.
I can live without the necessities.
Heh! I hired an East Indian once who hired on as hourly 'work' staff at
the fireworks plant, but who professed to be a degreed mechanical
engineer by trade.
It became clear early on that he had zero manual skills, and no knowledge
of tools or physical methods. He was simply astounded and enthralled by
a simple scissor-style PVC pipe cutter, when it was demonstrated to him.
I once asked him, as inoffensively as I could, how it could be the case
that he was both degreed in the trade, and absent the manual skills. He
said, "Oh, engineers nebber get deyer hands durty! Dat is for de
laborers! We design... Dey make de products!"
Never met an engineer worth his salt who couldn't turn a wrench!
On Fri, 26 Sep 2014 05:32:23 -0500, "Lloyd E. Sponenburgh"
Actually it is quite common in parts of Asia. Largely because in the
developing countries people who attain a collage degree usually come
from wealthy families. Do you spend all that money to have the kid out
mucking about with the hired help :-)
I've met and worked with a lot of them and they didn't know what to
make of me; multilingual, very well educated and entirely willing to
take on delicate manual lab work. I did the miniscule soldering for a
Chinese female engineer who couldn't.
The Industrial Revolution and the modern world arose mainly from the
discoveries of well-educated Englishmen of the Royal Society who were
willing to experiment with their own hands.
plus a few foreign experimentalists like Lavoisier.
I've debated this with an engineer from Bangladesh who couldn't
understand why Europe had so suddenly surpassed Indian culture, and
didn't like my answers
Right. Engineers Hammer is a 3 1/2 pound double headed hammer that they
use in the work. No claws - double flat heads.
Engineers move around the world making and putting up stuff from tunnels
to rails to bridges to buildings......airplanes...missiles...to large
and small machines.
I know - I grew up with one in the house and one who daughters lived
together waiting for Dad to come home. Both grown. Did that all
of their and their Mom's life. He was a civil engineer and moved
mountains. He worked all over south America.
My Dad, I and my next brother were all engineers and we moved world
wide doing work as needed.
On Friday, September 26, 2014 12:32:23 PM UTC+2, Lloyd E. Sponenburgh wrote
Don't get me started. I have a good friend who has a high position in aviat
ion safety engineering. He has such an astoundingly poor understanding of b
asic physics, mechanics, and electricity that I don't understand how he eve
n manages his household, let alone his job . Many examples, but one day we
were talking about car brakes, and I mentioned that the rear discs were typ
ically smaller than the front ones. He had no clue as to why that should be
. Still, he makes way more money than me. My practical knowledge seems to h
ave thwarted my joining the 1%, but I have friends in the 0.1% who say that
when the shit really does hit the fan, then it's going to be good old Rob
who gets them through it. I like that.
Don't get me started. I have a good friend who has a high position in
aviation safety engineering. He has such an astoundingly poor
understanding of basic physics, mechanics, and electricity that I
don't understand how he even manages his household, let alone his job
. Many examples, but one day we were talking about car brakes, and I
mentioned that the rear discs were typically smaller than the front
ones. He had no clue as to why that should be. Still, he makes way
more money than me. My practical knowledge seems to have thwarted my
joining the 1%, but I have friends in the 0.1% who say that when the
shit really does hit the fan, then it's going to be good old Rob who
gets them through it. I like that.
In my experience the most capable engineers resist being moved into
management, so the less qualified ones become the bosses.
Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here.
All logos and trade names are the property of their respective owners.