On a sunny day (Sat, 12 Dec 2015 15:55:45 -0500) it happened Phil Hobbs wrote in :
You may find this interesting:
As we were talking SRAM, look at the conclusion on page 6 point 21.
Counterintuitive but true. A smaller percentage of the array is composed of active elements (more wiring, etc.).
Yep. When we were doing the SOI PPC for Nintendo a radiation detector was added to the design to get past ITAR regulations. Before the chip made it to market the ITAR limits were raised so the radiation detector was dropped.
That certainly depends on the application.
Well, the way to answer such questions is within the area of Reliability Engineering. You build a mathematical model using reliability prediction method such as MIL-HDBL-217. There are softwares that make this easy to do now.
My guess would be the NE555 (if it is a legitimate part of good quality, and used within its limits). Building the equivalent circuit with discretes means a lot of solder joints -- and high-rel soldering is not easy.
The NE555 has behavioral peculiarities, which require careful design -- but that is not part of your question, right?
I see military and space-grade versions are available too. That might seal your choice (if their high costs aren't an issue).
Well, the way to answer such questions is within the area of Reliability Engineering. You build a mathematical model using reliability prediction method such as MIL-HDBL-217. There are softwares that make this easy to do now.
My guess would be the NE555 (if it is a legitimate part of good quality, and used within its limits). Building the equivalent circuit with discretes means a lot of solder joints -- and high-rel soldering is not easy.
The NE555 has behavioral peculiarities, which require careful design -- but that is not part of your question, right?
I see military and space-grade versions are available too. That might seal your choice (if their high costs aren't an issue).
Handbook 217 is a low-quality random number generator. It predicts (among other absurdities) that removing protection circuits increases reliability. I worked on one project that used it, long ago, and never ever again.
Cheers
Phil Hobbs
I'm interested in what would you recommend in its place, Phil?
Cheers
John S
I'm not what you'd call a reliability expert, since I've never designed a h igh volume product.
I have done a fair amount of firefighting for the server guys at IBM, but t hat was mostly inspection stuff, like the time some turkey installed washer s burr-side down on a whole bunch of Regatta server boards, which caused a few shorts and a big re-inspection headache.
In my somewhat limited experience, reliability is mostly controlled by a fe w bad actors: connectors, switches, electrolytic caps, blown-up inputs, sho rts on outputs or power supply inputs, and anything with too little design margin.
Cheers
Phil Hobbs
There's apparently companies peddling/buying old glass that's less radioactive for special applications too.
A radioactive oxide was once a common colouring for ornamental glassware/ornaments.
It was a luxuriant yellow, but I can't remember which oxide.
That's harder, since iirc the main radioactive species in soda glass is potassium-40, whose half life is over a billion years.
The alphas in lead are mostly due to neutron activation from other isotopes in the ore body, and the half lives are (iirc) a few decades.
Uranium.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
IIRC, one of those methods used an overly simple method of estimation (linear or exponential with whatever conditions, including time and temperature). So the addition of any component strictly decreased lifetime. When really, one must account for the statistics of external influences (such as EMI, surges) and environmental (temperature excursions, internal heating, time at full ratings?), for which said protection components really do do exactly what they say they do. (They're also easier to repair in event of catastrophic surges: hopefully, being applied sacrificially while the rest of the circuit survives.)
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
You are not even SEEING the real cosmic rays. What you are seeing is the
2nd or even 3rd or 4th level interaction from the original cosmic ray, which fortunately for us has interacted with an air molecule 20 miles above your head.The real cosmic rays are things like Oxygen or Neon muclei at GeV energy levels. These things flew out of supernova explosions and have been travelling for millions of years looking for something to hit. Those huge bursts you see are a shower of Gamma rays and Alpha particles that were produced when the original cosmic ray, or more likey one of the earlier shower particles/rays hit an air molecule or the roof of your building. The original rays will produce thousands of particles and rays on the first interaction. It is just like the stuff they do in high energy Physics experiments, but happening all the time right in our atmosphere.
And, of course, for sensitive electronics, there are all sorts of terrestrial sources of radiation, too.
Jon
Uranium glass is more of a green
-- \_(?)_
On a sunny day (Mon, 14 Dec 2015 18:01:33 -0600) it happened Jon Elson wrote in :
Yep, 100% correct. Fun is when you take a scintillator crystal, and go sit in the dark for a while, have your eyes adapted. You can see the flashes! Maybe one every 10 minutes or so here (average).
It was a long long time ago (~1982), but even then it struck me that adding up FITS (failures-in-time, i.e. failures per billion device-hours) was completely bogus. (This was on the space telephone project before I went to grad school.)
The other thing about that project is that they had designed in hot-spare redundancy for a whole lot of boards on the central-office side of things, but made no effort to standardize the circuitry for switching between main and spare, so there were probably a dozen completely different designs doing (nominally) exactly the same thing.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
That's a good thing Phil. It means that a design error in the failover card won't result in loss of redundancy for the entire system. Or it could have been a failure to properly manage engineers...
Well, they could have had two designs for redundancy. ;)
IIRC I did mine using data-selector logic, because it minimized the package count.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
I ignored reliability because my stuff never lasted long enough for reliability to be a problem. The only time I got involved was when the customer insisted on BITE (built in test equipment), a rather ominous and appropriate acronym. The circuitry BITE was suppose to monitor worked quite reliably. The failures came from the BITE interface equipment.
My list of suitable culprits include all of yours with the addition of anything that moves, anything the customer can adjust, and damage caused by various service facilities. Expecting two service centers to do the same thing, and get identical results, is my definition of insanity.
For RF, the lack of design margin appears mostly in the gain distribution of the RF chain. Inexperienced engineers tend to trust the datasheet specifications and use the nominal component specs. After a few disasters, they switch to using the minimum gain specifications. One memorable (audio) design I reviewed during a job interview, which included almost as many gain adjustment pots as there were gain stages. I failed the interview, because I didn't know that this was the interviewers design, which he was prepared to defend to the death. Only the paranoid survive, and this abomination made it into production just long enough for engineering to provide a replacement.
Some companies take the easy way out and provide sockets for their IC's and discrete devices on the assumption that it was ok for a device to fail if it was easy to replace. I have several pieces of test equipment from SSI, Wavetek, Cushman, and HP with sockets that I'm constantly "adjusting" to deal with intermittent connections.
I also did a fail-over transceiver, with redundant everything. The idea was to provide 100% uptime. Building such a beast with automatic fail-over invariably involved some single point of failure points in the design. As might be expected, these were the first to fail, bringing down both radios simultaneously. The best of such a design was the emergency backup generator system for the Bay Area ATC (air traffic control) system many years ago. In 1995, during a test of one of three generators, a failure in the generator under test, transferred the load to the 2nd generator, which failed in the same manner, which transferred the load to the next generator, etc, until the entire system collapse and the ATC center went off the air. Now, technical progress has now allowed us to provide some impressive electrical failures in software:
"The bigger and more complex they are, the harder they crash."
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
In a previous lifetime, I was responsible for the maintenance and troubleshooting of the controls for a 1976 vintage GE HVDC link (280 kV,
1320 A). The controls used redundancy, but weren't complicated enough to monitor their redundancy. They were also full of those little black leaky capacitors (CK5?, CK7?) and nickel plated iron connectors. There was a simulator you could connect the controls to for maintenance work, but it was even less reliable than the controls.Maintenance was hell. It would take a week (of our two week maintenance outage) just to hook the controls up to the simulator and get it working. We could never get the maintenance completed, and the redundancy masked just enough failures that we never knew if the controls were truly failure free or not. One year, we just pulled out every card, cleaned the connectors with Freon and Cramolin and jammed them back in. It appeared to work as well or better than our regular maintenance program. My successor in that job finally installed a multi-channel digital recorder and started looking at waveforms during operation and managed to exorcise most of the latent failures out of the system.
You do recall properly. This was discovered by Timothy May at Intel.
Oddly enough, he still posts to Usenet. A recent message from him: snipped-for-privacy@att.net
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.