I think older people neturally want to teach. That's a tribal, selfish-gene survival thing.
When something in a new circuit doesn't work, people are sometimes inclined to replace a "bad part." Most often, the part wasn't bad and all that's accomplished is to rip pads off the board.
So my lesson of the day, just delivered to a young engineer, is "Don't solder; measure and think."
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
I worked for a company making array processors at the Cray performance level. One of the test technicians was seeing a problem quite often and blaming the FIFO chip because he could fix it by replacing that part... sometimes taking two or three replacements though.
Turns out the designers used a gate (or two) to generate a timing delay for a write pulse. Sometimes it didn't generate a wide enough pulse to actually make the chip work. The problem wasn't that the FIFOs were bad or that the delay gates were bad, this was just bad design.
Another time we found that a lot of buffer chips were messing with the integrity of the clock. Seems some brands had asymmetrical delays on the rising and falling edges. This was early days of the logic family in use and not everything was considered before releasing the design I guess. We had to replace a lot of those chips though. It wasn't a design issue on our boards per se. Still, no one realized the problem until the test floor stopped replacing the chips wholesale and said, "Hmmm... this isn't right".
-- Rick
Some people almost brag about "buckets full of burned out transistors"; such people are almost certainly fools...
If I do say so myself, my record of burned transistors would probably fit in both hands cupped together, maybe just one hand.
I recall reading that the development of the F-1 rocket engine combustion assembly (see: Saturn 5) involved many more-or-less random attempts at getting the injector plate and/or combustion chamber perforations correct (specifically relating to trans-sonic fluid-gas flow stability layers between the incoming cryogenic liquids washing over the combustion chamber, before entering the center of the chamber proper where they mix and explode).
But then, to be perfectly honest: they were fools, too. And they knew it, which is why they did it that way.
And in fairness to them, we're not much better off today. Fluid dynamics is notoriously, ludicrously hard to solve for.
Though SpaceX recently reported achieving GPU optimized CFD,
As is, there is NO way to solve for boundary conditions for a desired CFD operating envelope, just as there is no solution to solve for a circuit topology and values based on desired waveforms.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com "John Larkin" wrote in message news:uijcmal8bi6tfrkostpa5qs8seq4ik3jhh@4ax.com... > > > I think older people neturally want to teach. That's a tribal, > selfish-gene survival thing. > > When something in a new circuit doesn't work, people are sometimes > inclined to replace a "bad part." Most often, the part wasn't bad and > all that's accomplished is to rip pads off the board. > > So my lesson of the day, just delivered to a young engineer, is "Don't > solder; measure and think." > > -- > > John Larkin Highland Technology, Inc > picosecond timing precision measurement > > jlarkin att highlandtechnology dott com > http://www.highlandtechnology.com >
Hey, we did this on purpose:
-- John Larkin Highland Technology, Inc picosecond timing laser drivers and controllers jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Yeah sure, but sometimes you need to swap stuff to see that more thinking is needed.
For we who are less smart, more data may be needed.
George H.
Occasionally, the manufacturers are the fools; I tested an IRF740 and found, either RthJC is half the max spec (seems likely!), or the bastards can tolerate a TJ over 200C. They also typically drop the DC SOA curve, but the oldest datasheets used to specify it. Strange, because apparently the SOA is way larger than they'll admit to.
As you found, few modern parts even dare to provide DC SOA; those that don't, remove it because it's ugly. And when they do provide it, it's ugly anyway.
(Were "linear rated" FETs even available when you were doing those tests? I suppose not, as that would've saved a lot of R&D time and cost.)
A modern part, of mostly comparable ratings (Fairchild something or other,
6A 600V) died exactly where the datasheet said, not a watt more. Impressive.Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
As my old friend and mentor Henry used to say, "Measure twice, cut once."
Cheers
-- Syd
They almost certainly can, probably for over a thousand hours.
Cheers
-- Syd
I've used a proprietary "genetic optimiser" for designing filters. Give it a coupe of "goodness" functions (e.g. a desired waveform and some preferred way to determine how bad a given approximation is - least squares, maximum error, peak group delay, whatever you can write an expression for, and maybe reciprocal of chip area for a second goodness function) and some constraints, and it tweaked values and ran SPICE till it found much better values than I ever could have, from a solution space that would have been unfeasible to search exhaustively or randomly. I was very impressed. These things also do a good job of choosing transistor sizes in high-speed logic gates. It convinced me how efficient evolution-type processes can be at solving some problems that most human designers could not solve directly. On the other hand, I think that it would not be good at solving anything that takes so long to simulate that tens of thousands of runs would be impractical.
A related story: Years ago I was visiting a chemical plant, and they showed me their new, computer based tank gauging system. It was ZX80 days, and the gear used double sided PCBs crammed with up to 50 or so DIPs of various sizes. They explained how all their instrument techs had been trained on diagnosis and repair of the system, but later on, in the privacy of his office, the maintenance chief told me that if a board was suspected bad, the techs unsoldered and changed every chip on the board. It was a strongly militant union site so I guess it was easier to just let things be.
Sure, but replacing a chip that doesn't seem to work seldom fixes a design problem, and doesn't provide any more data.
In our current case, I advised to not replace the chip, and it turns out there was nothing wrong with the chip. The output was wrong because the input was wrong.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Yeah, It can be hard helping young people. You have this lifetime of experience, you want to just pour it in to their brain. But if they are anything like me, they have to make their own mistakes to learn.
I was over at the UB physics dept on Friday... planning some of my own work. But when I got there the grad student had a dewar full of helium and no signal from the FTIR spectrometer! (That demanded instant attention, helium is wicked expensive these days.) We spent some time confirming that the detector was working... and more explaining. Detector was a photoconductive device R ~10 Meg, but it avalanches at ~0.5 Volts, so you need low currents. He was trying to measure the resistance with a DMM. After we found the problem, (the spectrometer computer had lost it's mind and had a mirror pointing in the wrong direction.) I asked him to analyze his detector circuit. Voltage source feeds "bias" resistor to photoconductor, signal being the change in voltage across the PC. I figured maybe 15 minutes, but he didn't know how to get started. He got the answer after some coaching.
It's not easy to get people to think... or at least think in the "right" direction.
George H.
In otherwords, they were simply part replacers. I know the type, just change a whole board and don't bother to even diagnose it, later to find it wasn't that board. Some go as far as leaving the new board in and putting the old board in the new box, there by taking a chance of putting some intermitting or some previous hacked job repair.
Techs like that are very costly to business. It does pay at times to actually have real techs onboard.
There comes a time however when the cost of labor to replace the actual bad components is not cost effective and a new board is put in, at least it makes it a learning experience just to diagnose the old one for accurate faults. It could help the maker of the PCB for the next revision. Jamie
Well, replacing all the chips maximizes union labor.
I've been in places, a big turbine manufacturer, where I couldn't turn a trimpot on my own gear. Only the union guy was allowed to touch a screwdriver. I had to tell him how much to turn it.
That company is now gone, or course.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
There are fundamental differences between a repair on a system which was working, and getting a new design to work.
If the system was working then something has failed ( even if it was a design flaw which caused it to fail ), so fixing it must involve at least a replacement, hence the soldering iron approach, preferably still after some thinking and tracing to isolate what exactly has failed.
But in a new design, its much more likely to be a design problem, so the thinking comes first to find out what part of the design is wrong. Reaching for the soldering iron is the last thing to do. Apart of course from replacing the components which failed because the design was wrong.
-- Regards, Adrian Jansen adrianjansen at internode dot on dot net Note reply address is invalid, convert address above to machine form.
Yeah, I call the first trouble shooting. (Seems like most problem solving (I do) is trouble shooting.)
I don't have a good word for "de-bugging" a new design. Of course that's where you learn stuff. "Oh, this can't work without a cap here, or a resistor over there."
George H.
I had the same situation once. The union said the company tech had to do the DMM and scope probing and adjusting under my directions. I said I can't work that way. They said I had to. I began putting my equipment away. They asked what I was doing. I told them I was leaving to catch the next flight back home. It wasn't a bluff.
They changed their minds.
Many years ago when I was designing circuits, I had a tech who would build all circuits on the smallest piece of perf beard possible. So if a part had to be added a new board was required.
So I learned to put in all the parts that might be needed. The circuit would usually work. Then I would disconnect the parts that might not be needed and see how well it worked.
Dan
Usually they'll compromise: you can do the testing, as long as there's a union member standing next to you, doing nothing.
I worked for one guy, in a job that involved going out on ships. He paid off the union, and they left us alone.
-- John Larkin Highland Technology, Inc picosecond timing laser drivers and controllers jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
When you cut one end of a wire it will too short at the other end.
-- Reinhardt
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.