 | |
Solid capacitors vs electrolytic...anyone with an opinion either way?
Hardware Chips Processor, cache, memory chips, etc. (comp.sys.ibm.pc.hardware.chips)
 |
| | Thread Tools | Search this Thread | Display Modes |
#31
 
| ||||
| ||||
|
Re: Solid capacitors vs electrolytic...anyone with an opinion either way? -
04-10-2007
, 09:47 AM
|
me (AT) privacy (DOT) net> wrote: +--------------- | Rob Warnock wrote: | >Since almost *all* electronic device failures occur during power | >transitions | | Evidence, please. +--------------- Over 40 years of experience in the industry, |
Has it been your experience that flashing LEDs fail more
quickly than alway-on LEDs? That resistors, diodes, etc.
that aren't at the edges of the rating evvelope fail much
more quickly when subjected to power on/off cycles?
Quote:
|
designing and manufacturing computers & related devices, mainly, but also using them. |
"all electronic devices."
Quote:
|
Note, however, that I was *not* referring to devices with a high component of mechanical motion such as fans or disk drives [as another branch of this thread seemed to get fixated on]. |
devices", but I am inclined to assume that you meant pure
electronic rather than electromechanical. Clearly they have
differing failure mechanisms.
Quote:
|
I was referring primarily to printed circuit boards and the non-moving parts normally mounted on them, such as chips, capacitors, resistors, etc. |
(many resistors in opamp circuits dissipate far less than
that) will last about as long as one that cycles from 1/16
watt to 0 watts and back again. And when the latter does
fail it will be at a random time, not at the exact moment
the power transition happens.
#32
| |||
| |||
|
|
Rob Warnock writes: | >Since almost *all* electronic device failures occur during power | >transitions. for a large number/style of systems in a wide range of conditions, frequent power cycling generally tends to result in less total time before failure. |
Quote:
|
There are exceptions, of course, at the extremes of operating regimes. If you only turn your system on once a year for one hour per session, it might very well last longer than my machines which run continuously. But for myself, personally, I consider that operating regime to be nearly useless. YMMV. +--------------- | Also, I have seen a number of machine failures during power-on | conditions, from stuck fans to dead RAM and dead disks. +--------------- Uh... I didn't disagree with that. Whether frequently or infrequently power-cycled, the most likely moment for a failure is during a power-on event [and the second-most-likely is a power-off event]. But also see my other reply, where I mentioned that I wasn't especially addressing devices with a high component of mechanical motion or stress. Those have much more complex tradeoffs between frequency of power-cycling, on-time per session, total on-time, and MTBF. [Bearings dry out while running; seals dry out while *not* running; "stiction" is aggravated by long "off" times; heads can crash from overly-*short* "off" times (power flicks); etc.] +--------------- | >A power cycle is stressful in two ways: voltage/current surges, | >and temperature cycling. The latter is the more serious in the | >long term but sets up the conditions for failure due to the | >former. | | OTOH, power-on hours also take their toll on the electronics. | For semiconductors, electromigration will eventually make the | device fail; +--------------- True, but for devices manufactured since the phemomenon of electromigration was understood and (somewhat) mitigated against, thermal cycling is probably a more serious stressor. The worst case -- when both are active -- is a 95% "on" time with frequent "off" cycles just long enough for the system to cool off. +--------------- | for electrolytic capacitors, the liquid tends to evaporate faster | with higher temperature, and they are hotter when the machine is | powered on. +--------------- Also somewhat true, though any machine which is operated at such a high temperature that this is a dominant effect is in serious danger from failing for many *other* reasons!! ;-} Under normal "room-temperature" conditions with circuits designed not to overly- stress the capacitors' A.C. current-handling rating [which is IME a more-serious source of high temps in electrolytics than ambient temp per se], thermal cycling is more likely to be the cause of failure, e.g., by the cracking of seals or solder joints. Though I will agree with you completely about those old *huge* electrolytics that used to be used in large-computer linear power supplies [back before switchers!], e.g., the 100,000 uF 16 V caps that were used in the DEC PDP-10 supplies. I used to help run a PDP-10 that was in a small closed room with barely adequate air conditioning. If the air conditioning failed -- which it sometimes did in the hot Atlanta summers -- and the PDP-10 got hot enough for the internal thermal sensors to shut the machine down, then, just like clockwork, about two weeks (yes, weeks!) after the air conditioning outage one of those monster caps would suddenly explode, spraying electrolyte all over the place [and resulting in hours of DEC Field Circus time to get it fixed]. We were convinced the failures were due to the caps having dried out during the HVAC outage, causing the A.C. impedance to go up, which caused the caps to continually overheat later, which accelerated the aging in a runaway cycle until... *BOOM*! But modern desktop system mainboards don't contain those sorts of "wet cell" electrolytics any more. -Rob ----- Rob Warnock <rpw3 (AT) rpw3 (DOT) org 627 26th Avenue <URL:http://rpw3.org/ San Mateo, CA 94403 (650)572-2607 |
#33
| |||
| |||
|
|
A 1/4 watt resistor that disssipates a constant 1/16 watt (many resistors in opamp circuits dissipate far less than that) will last about as long as one that cycles from 1/16 watt to 0 watts and back again. And when the latter does fail it will be at a random time, not at the exact moment the power transition happens. |
without expecting to first comply with them yourself.
Furthermore, what evidence do you have that resistor
failures cause most device failures? They are easy to
find, but finding a failed resistor doesn't mean it caused
the failure. Resistor burnout is just as likely to have
resulted from some other initiating cause.
-- Robert
#34
| |||
| |||
|
|
Furthermore, what evidence do you have that resistor failures cause most device failures? |
If you want to make an argument about the failure of
electronic systems instead of "all electronic devices"
please call them electronic systems, not electronic
devices. Note thet the original "all electronic devices"
claim was in the context of a discussion of capacitors,
which are devices, not systems.
Many electronic devices comprise an electronic system.
If the subset of devices that tend to fail at the moment
of power cycling fail far more often than the subset
that does not, then the system itself will reflect the
failure characteristics of the fail-on-power-cycle subset.
That seems like a very real possibility and a reasonable
argument. Rob Warnock's claim was (exact qoute) "almost
*all* electronic device failures occur during power
transitions" is quite simply not true.
#35
| |||
| |||
|
|
Robert Redelmeier wrote: Furthermore, what evidence do you have that resistor failures cause most device failures? A resistor is an electronic device. |
I'll indulge you, briefly:
First, a resistor is not an "electronic device" it is an electrical
part. electronics is a term reserved for semiconductors (perhaps
including tubes) where the unique behaviour of electron [holes]
(as opposed to current) is controlled
Second, a device is not a single simple part. Devices are
more complicated than parts but less complex than systems.
Quote:
|
Rob Warnock's claim was (exact qoute) "almost *all* electronic device failures occur during power transitions" is quite simply not true. |
a power transition as well, especially in the predominant
CMOS technology.
-- Robert
Quote:
| |
#36
| |||||
| |||||
|
|
In comp.sys.ibm.pc.hardware.chips me (AT) privacy (DOT) net wrote in part: Robert Redelmeier wrote: Furthermore, what evidence do you have that resistor failures cause most device failures? A resistor is an electronic device. I see, you wish to argue semantics and terminology. I'll indulge you, briefly: First, a resistor is not an "electronic device" it is an electrical part. electronics is a term reserved for semiconductors (perhaps including tubes) where the unique behaviour of electron [holes] (as opposed to current) is controlled |
graduated.
Quote:
|
Second, a device is not a single simple part. Devices are more complicated than parts but less complex than systems. |
I know you have.
Where is it you are located?
Quote:
|
Rob Warnock's claim was (exact qoute) "almost *all* electronic device failures occur during power transitions" is quite simply not true. |
Quote:
|
Oh, it what way? Please note, the clock may be considered a power transition as well, especially in the predominant CMOS technology. |
Quote:
|
-- Robert |
#37
| |||
| |||
|
|
Where is it you are located? |
from relay logic to semiconductors for critical safety systems.
Welcome to my world.
-- Robert
#38
| |||
| |||
|
|
Strangely, in my experience machines that are generally powered on all the time seem to be more likely to fail when power cycled. |
the university than the biannual tests of emergency power supply to the
building: because this meant turning off all power, not only did he have to
shutdown and powerdown the whole network, he could count on average about
three PSUs (from several hundrer) dying in the process.
Jan
|
« Previous Thread
|
Next Thread »
| Thread Tools | Search this Thread |
| Display Modes | |
Linear Mode
| |
Powered by vBulletin Version 3.5.3
Copyright ©2000 - 2008, Jelsoft Enterprises Ltd.
Copyright ©2000 - 2008, Jelsoft Enterprises Ltd.