What circuit should be expected to be more reliable in the long term? Consider e.g. a monostable multivibrator made of an NE555 and its BJT counterpart. The latter is composed of more parts and more joints, but the former has very fragile metallization layers, prone to electromigration and finer bonding wires (?). Assume the electrical conditions are all OK (all signals within the absolute maximum ratings, no ESD etc.). What, in your opinion, would last longer and why?
Best regards, Piotr
On a sunny day (Thu, 10 Dec 2015 18:21:09 +0100) it happened Piotr Wyderski wrote in :
In my Trio oscilloscope (many many transistors) the 555 in the timebase is the only part that ever failed, apart from the input stage that I once blew up by touching a 6 kV booster diode anode in a TV. I replaced the 555 and recalibrated,. OTOH I have replaced many transistors and other components servicing and repairing very complicated stuff.
Most is a design question, under-rated or otherwise bad circuits. Always the same errors in specific equipment.
It is true however that you cannot always go by the definition of MTBF.
I think these days it is electrolytic capacitors that limit lifetime of electronic equipment. I'd say integrate as much as you can, for glue logic use FPGA etc, or maybe a PIC, a PIC has also 2 hardware comparators and can do a lot more than a 555, only limited by how long FLASH memory lasts. So are your pods and pads and TVs and anything that is not based on an ASIC.
We live in a throw away world. In the example I gave of a robot getting credit for work done, and it having an internet connection to buy stuff [1], it would collect more and more rubbish ^H^H^H^H^H^H stuff over time, and needs either garbage collection recycling and or a destructive war to get rid of it.
[1] better robot oil, better robot batteries, robot spray, robot anti-oxidant, robot perfume, makes you a better robot, and work harder so you can goto [1]Sorry
[2] Republicans love robots too, Much more than demonrats. Both ways the system needs no humming beans.Maybe Big Bot as leader? Different from Big Blue, but who knows?
The 555 is just an example of the circuit's complexity, as the comparison of a Joule thief and a Pentiumn Pro would have no value.
Yes, I forgot to mention that there will be no wet electrolytics in the circuit. Tantalum and ceramic caps are allowed. I also assume that the lesser parts the better, so I am not afraid of P-JFETs, depletion-mode MOSFETs etc.
20 years according to the manual, which isn't a particularly long time.But I am interested in building a very reliable circuit or two. What would a hi-rel guy do if all what he actually wants is the true hi-rel itself, not the associated medical/military certificates?
Best regards, Piotr
On a sunny day (Thu, 10 Dec 2015 19:03:41 +0100) it happened Piotr Wyderski wrote in :
I really do not know, maybe ask NASA, those vogagers and pioneer spacecraft are still working.
PS I have a little 8052AH based computer here that is still working since when was it about 1985 when did 8052AH come out? That makes 30 years:
Largely scrap parts used, all caps and resistors are from old TVs so are the small transistors, even that nicad is from an old Philips VCR IIRC.
30 years and still working. BASIC in ROM in that chip.Good enough? I tested it again last year. But it was not on all the time. But probably oldest electronics I still have.
I try to avoid Tantalum caps, as I have replaced MANY MANY of them in older equipment. It seems that these fail more in cases where equipment is stockpiled WITHOUT power for years, and then plugged in and the caps fail.
I have a lot of gear that I made in the 1970's and 1980's that are still in use, some of it continuously powered since that time. Keep things relatively cool, don't use IC sockets, make sure all connectors are gold on gold (none of those tin-plated Molex-style connectors) and watch out for long runs of cable that can pick up magnetic induction from lightning strikes. If you have to have cables running hundreds of feet, then put circuit protection on them.
Jon
Maybe my experience is different - I see relatively few tantalums fail on supply decoupling, more often its in signal circuits where even the smallest reverse voltage can cause trouble.
Beware of lytics. Run everything well below max v,i,p. Support the PCB well to avoid flexing/vibration. Tie all wiring down. Avoid fans. Use robust wi re to pcb connectors & external connectors. Minimise switch arcing. Avoid p ots. Use protection all round. Add limp modes & automated reboot if process or controlled.
NT
Better yet... avoid processor control >:-} ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
MTBF is generally inversely proportional to complexity, and failure rate is generaly proportional to complexity.
Reliability can be calculated in any environment, given accurate base data, but it is better demonstrated.
The data for calculation is drawn from a history of demonstration, as much as it is possible and can improve in accuracy with technical maturity of the materials and processes employed, by either increasing or reducing. A factor that penalizes 'immature' components is a regular feature of most calculation methods.
Only a component previously demonstrated to be reliable can be expected to repeat this performance in the future.
RL
And burn-in to get units past the early failures in the bathtub curve. If you have the budget, part testing to pick up functional but subpar parts, eg caps with higher than ideal ESR.
NT
It is a fact,that the greater the number of connections, the lower the reliability; even at six nines, the power law can kill the device reliability. That tends to argue against a discrete design.
Especially if you're talking about a SRAM built from CK722s. ;)
Cheers
Phil Hobbs
On a sunny day (Fri, 11 Dec 2015 23:47:05 -0800 (PST)) it happened Phil Hobbs wrote in :
So true, but on the other end when transistor size gets to be a few atoms, any co[s]mic ray may sort of change the math.
Modern multi-level FLASH is very vulnerable. Probably other chips too, like high density FPGAs.
Actually, cosmic ray bit flips were a bigger deal back in the day.
Even for a mere 640kB, they wasted a whole 11% of the entire memory store on parity!
You might think smaller transistors would lead to bigger errors (the ion trail from a cosmic ray, or alpha particle, could disrupt multiple transistors in a row), but it seems being a smaller target is the bigger effect.
SOI apparently gains many orders of magnitude further, which is what all? fine pitch hardware is made on now.
And nowadays, we have the disposable computing power (and speed) to put error correction into everything, so memory doesn't need to be perfect, it only needs to be good enough.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
If I recall properly, naturally-occurring radioisotopes in the IC packaging materials were a bigger part of the problem than cosmic rays, at least for devices operating near sea-level. Alpha- or beta-decay taking place a millimeter or two away from your RAM cells is bad news.
The biggest problem was crud in the SiO2 (oxide) diffusing into the underlying Silicon. ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
On a sunny day (Sat, 12 Dec 2015 10:09:32 -0600) it happened "Tim Williams" wrote in :
Yes, but not sure it is that simple. I had a discussion online some years ago with somebody from a manufacturer of FLASH, and they had huge problems increasing memory size because of radiation.. I have done some measurements myself, as I have this nice crystal and PMT gamma spectrometer, and really, it is hard to believe how powerful these co[s]mic rays can be.
Sometimes when I had the PMT voltage way way down it would just be hit by an out of scale event. Testing with both GM tube and the scintillator showed that those events sometimes happen in bursts. Unbelievable energy, and I am not even on the top of a mountain but as sea level. Anyways, there is a lot of info online if you google:
It is true that CRC checking and other forms of correction (more complicated) can catch _some_ errors, but most certainly not all. I have tested that too, a 16 bit CRC is easily defeated in a short time if enough errors happen.
I would not (and I think there are no) want consumer devices that fail every now and then, even more so where your life depends on it, like electronics in cars and what not, pacemakers, other stuff, or simply the light in your room...
Maybe the old magnetic core memories from Apollo days were less susceptible.
Anyways, nothing lasts for ever, the universe, what's the word, entropy... And I do not want a ticking noise in my mp3 collection...
And yes, I have backups of that in magnetic (harddisk), optical (DVD), and FLASH. So far so good.
I always called bullshit on the "cosmic ray" excuse vendors had for not replacing memory modules. It's funny how super super high energy cosmic rays would keep hitting and stopping one one memory module, yet scientists down in the bottom of some water filled cave with balloon sized photmoultiplier tubes have a hard time detecting them.
Even more interesting is with the really high density memory we have these days, the cosmic ray problem seems to have gone away.
Mostly alphas from the lead in solder. BITD the IC makers went the rounds of churches with lead roofs, offering them new roofs for old, like Abanazar with Aladdin's lamp.
An energetic particle causes a lot of ionization in the chip substrate. SOI construction prevents all that charge from getting into the electronics.
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
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.