The Adventure of the Uncertain Failures

It is amazing that simple physics was the answer but IBM with all their engineering prowness couldn't figure it out. I wonder why IBM does so little engineering and manufacturing now.

I am currently dealing with nearly the exact problem. It is a permanent magnet which passes by the reed switch and causes an electrical connection thru the switch. Had approx. 30% that couldn't get switch positioned correctly on 1st or even 2nd try sometimes. They used to (70's) have a strong magnet in lab they ran them all across and have since lost it. I guess we'll be buying another magnet.

To answer C.Rama Murthy, this occurred in the summer of 1966. Reed switches were used by IBM for many purposes, some of which were: (1) a magnetic moved near one detected position to activate a circuit as you described, (2) as relays by putting the reed switch in a coil and mounting it on a circuit board, (3) many more of which I was not aware of.
There were articules in electronic hobby magazines at the time (such as "Popular Electronics") where a reed switch would be hidden inside a locked cabinet in multiple points in a series circuit, then when multiple magnets were placed in the right position the cabinet would open by activating a solenoid.

Very interesting and educative. How many years ago did this happen? Would you kindly tell us the purpose for which these reed switches were used? About more than two and a half decades ago, I bought an old computer keyboard(in junk market) which had the key shaft arranged to move an annular (ring) magnet towards a reedswitch to detect key press.I wonder if these keys were made by IBM besides the mainframes and many other things that they were famous for.

This is a prime example of why teaching basic problem solving is so important. One can learn all types of academic information, but if they can't apply it to solve problems, including making improvements to existing processes / products, then they can't be effective at their jobs. I also agree with Mike Z. - why shouldn't engineers be rewarded for process improvements / cost reductions like this, when money peddlers be paid big bonuses, even though their work is just a house of cards that can easily LOOSE money?

Doing what you are paid to do is "the job", not an extra mile thing. Years ago my job was to patent the Bondhus Balldriver. I did, and got a thank you & 50 bucks to look over the patent papers. I was pleased to have improved the product.

I think everyone is missing the whole point of the story; First off, the original engineer of the reed switches didn't do his homework. He should have known that the reed switches would be magnatized by the coil. I beleive I first learned about magnatizem in like 2nd or 3rd grade and it was reveiwed many many times both in school and latter in electronics training and on the job; And its something that is ever present when working with metals and electric currents; With things like saturated coils and transformers, to magnatized picture tubes that need deguaused, etc. And the other glaring problem is that over the years everyone just ASSuME'd the reed swithes were bad; Rather than finding the real problem. Reed switched are realy simple devices; Having 50% bad is a joke. It should have been obvious from the start something else was up; Such as the testing procedures, etc.
I beleive this is one of the most major problems with our school systems, people don't realy learn problem solving skills, they just memorize things for the test and then forget it all afterwards.

"...This was a very expensive loss, but over the years no one had figured out the cause".
How can managemant be so numb as to allow such a loss for years??

I think that Philip Oakley missed the "post test" part of the demagnetization. Removing the residual magnetism will make all the reed switches "the same" and ready for the bias that they will receive in their ultimate installation. If not demagnetized, 50% of the switches will be installed "the right way" and 50% "the wrong way". The "wrong way" units will need more cycles or a higher current to activate at first.
Of course, a good, old-fashioned bulk tape eraser swept across a batch of reed switches would probably be enough to degauss the whole lot.

"Steve Greenfield commented: Demagnetizing the reed switch is the wrong thing to do."
From a test perspective and achieving consitency it is a reasonable thing to do. However your point is well made that in situ the switch will become magnetised. This probably affects real circuit performance, apart from that consistency issue.
There is a second common oocurence of this fault, which is when zero crossing switches are used with transformers. These again have a hysteresis bias and potentially a very large inrush current. About 10 years ago I purchased a piece of US equipment for use in UK, which was provided with a basic (no spare capacity) transformer for voltage switching. It blew all the fuses!
Getting a larger transformer solved the problem - don't let it go into saturation.
The 50Hz/60Hz change didn't help either.
Philip

In response to Steve Greenfield: Reed switches are normally soldered into a circuit and stay there permanently. The degaussed switches will build up a magnetic bias in response to the electromagnetic pulses supplied by their respective coils and should work fine for the remainder of their service life. The only concern then would be that the coil supplies a strong enough field for the first few actuations to operate an unbiased reed.

With all due respect to tom s: Demagnetizing the reed switch is the wrong thing to do. They won't get demagnetized in circuit.
The method of testing he came up with first puts the reed switch in the state of magnetization it will be after it has been in circuit and energized many times. So what is the purpose in undoing this?

Final missing step was to demagnetize the switches post test with an alternating and diminishing field.

Hi, guys the answer is: IBM told me that if I was a full time employ would have received 1% of the savings for the next 10 years (which would have been about $10000/yr - about equal to my first full time salary at JPL). But since I was NOT a full time salary I received nothing except an excellent reference letter which may have helped me get the job at JPL.

And this fascinating case is exactly why engineering as a profession has a bleak outlook in this country. In response to the lead comment "I hope IBM gave Prem a big bonus" I'd be willing to make a sizeable bet that he got the good will of his managers but a token payment (if even that). By contrast on Wall Street the people that set up deals feel entitled, and indeed get paid, to share in the profits they make for the company. Someone smart enough to solve tricky engineering problems is smart enough to set up deals profitably shuffling pieces of paper around.

Not to sound like a greedy, materialistic SOB, but I hope IBM was greatful enough to share some of the cost savings that Mr. Sobel was able to generate. People are asking all the time how to get students interested in engineering, financial rewards would at least make them look.

Excellent article. I hope IBM gave Prem a big bonus and used his work as an example for others. I'd also like to thank whoever is responsible for properly capitalizing the titles of the articles. English grammar seems on its way to becoming a lost art on the internet. I deeply appreciate your efforts to teach by example.

This is the kind of engineering that we need to keep in this country! Can you teach it? Perhaps not, but the environment of always trying to figure out how things actually work is key and that is influenced by the culture. I still believe the American culture has this ingredient and many other places on earth do not. We are not better, just different.