Re: The OTHER problem with Netgear WGT624 (and probably others)

On Wed, 16 Aug 2006 09:54:30 -0700 Jeff Liebermann <jeffl@xxxxxxxxxxxxxxxxxxxxxx> wrote:
| phil-news-nospam@xxxxxxxx hath wroth:
|
|>When weighing in a wide range of expenditures, probability plays a very
|>important role. Given 100 purchases where the failure rate over the usage
|>period is 3% then you can expect approximately 3 failures. Plan for it.
|>But you don't have to plan for 50 failures in this case.
|
| None of the low end wireless hardware (Linksys, Dlink, Netgear, etc)
| specify an MTBF estimate and therefore has no track record or numbers
| with which to calculate a failure rate. Cisco 1300 series units are
| at 132,000 hrs (15 yrs). 400,000 hrs (45.5 yrs) for the PoE adapter.
| Sonicwall specifies the TZ150 at 79,000 hrs (9 yrs), and the TZ170 at
| 69,000 hrs (7.9 yrs).

Have those estimates proven accurate?


| I can make a fairly bad guess, based local circuit board heating, at
| the MTBF of some of the low end wireless hardware. It's not going to
| be very good. Ignorning infant mortality it's MTBF is generally not a
| concern unless it's less than the expected lifetime of the product. My
| guess(tm) is that low end hardware MTBF is at or below the expected
| lifetime.

Then what is the "expected lifetime"?


|>Businesses that are not flush with cash have to figure this in so they
|>can minimize the expenditure on as many things as possible.
|
| Sure. Using what numbers for this calculation? Most businesses don't
| have the need, time, or expertise to estimate failure rates. They
| dive into consumer reports, read a few reviews, talk to consultants,
| assume that spending more money will get them a better product, and
| then buy exactly what their competitors are using because they know it
| works. Immitation is the safest way to buy.

It's not done very accurately. There is an assumption that a certain
percentage of anything will fail, or otherwise need to be replaced, at
a certain rate, usually around 5% per year. That is factored in to
the overall cost, not of any one individual thing, but of everything
overall. It can turn out that some items will fail early, and that
such failure could have been predicted, had there been the resources
to investigate it. But for that to be viable, _every_ purchase would
have to be investigated, and that is not practical to do. So that
leaves the practice of buying everything at commodity pricing, except
for what one already knows to have a higher or earlier failure rate,
and leave a percentage of financing available to cover a normal rate
of failure overall.


|>Where glitches don't matter, but having some power during an outage does,
|>then I suppose that is an OK choice for that power size. I would not use
|>such a thing for mission critical functions.
|
| Sure. Nothing is mission critical until it fails with a shop full of
| customers. Then, it magically becomes a crisis. I sometimes put a
| big fat capacitor across the 12VDC power input of various routers to
| minimize power related glitches. In the distant past, before cheap
| UPS's, I used to do the same thing inside PC's that were overly
| sensitive to power line glitches.

Or you could use a battery (on the 12 VDC stuff, not the PC).


|>| Yep. SmartUPS series. No switching time, but there's a price. They
|>| suck power in standby. I recently measured an APC1400RH, which burned
|>| about 40 watts (70VA) doing nothing. At $0.15/kw-hr, that's about
|>| $50/year in electricity.
|>
|>If you need glitch-free reliability, wouldn't that be worth it?
|
| Sure, if it were disclosed by the manufactory. It seems that standby
| power consumption isn't commonly specified. I bought a "Kill-a-Watt"
| cheapo power meter for the purpose. Try selling a UPS to a commercial
| customer and then tell them that it will cost them $50/year to own.
| The switching time verus opertating cost discussion is always
| entertaining.

I bet it is an entertaining discussion.


|>The larger microwave ovens in hospitals are 900 MHz.
|
| Not at the 3 local hospitals I'm familiar with. The heart monitors
| and crash carts all run on 900MHz wireless. 900MHz "industrial" were
| banned long ago. However, the real reason is that there is no
| international 915-928MHz frequency band. Therefore, the autoclave and
| oven manufacturers would need to make a different model for US and
| Europe. Too big a PITA so they settled on 2.4GHz, which also works
| somewhat better.

So what frequency do the "industrial" large cavity microwave ovens use?
They can't use 2.45 GHz in the larger sizes.


|>That was probably
|>a number that got them so concerned. A shame they totally misunderstood
|>technology and you were unable to convince them of what was right.
|
| I'm not sure what they were thinking. The logic of the refusal was
| never adequately explained. My guess is that they took a look at the
| big 900MHz dish or yagi in the broshures and decided that it would be
| too ugly or something. Never mind the antenna farm and cell site on
| the roof.

There are some nicer looking panel and rod antennas now days.


|>That's not unlike the benefit of doubling the data speed when it increases
|>the packet failure (requiring retransmission) rate by say 15%. It's still
|>a win. OTOH, if latency is an issue, it might not.
|
| Increasing the speed also reduces the range. Not much of a win if you
| can't communicate. Latency is affected by retransmissions, which are
| always bad. However, with adequate signal strength, the latency is
| hardly noticeable. If you're thinking of speed-o-light, at 10Km the
| increased latency is only 66 usec.

Increasing speed loses noise immunity and that is related to loss of
range, since there is less signal at greater range to begin with to
overcome the noise.


|>Turns out he just assumed that "never go down" was a simple
|>configuration choice.
|
| I can see it now.... Web based configuration page with a settable
| reliability figure. If I need some revenue, I just set the router to
| fail every few months. Nice feature.

Interesting concept. But it's probably already patented.

--
|---------------------------------------/----------------------------------|
| Phil Howard KA9WGN (ka9wgn.ham.org) / Do not send to the address below |
| first name lower case at ipal.net / spamtrap-2006-08-16-2038@xxxxxxxx |
|------------------------------------/-------------------------------------|
.

Relevant Pages

  • Re: FBOFW Letters for May
    ... failure rate is too high, and if someone else argues that of course it's ... arguing about the definition of abstinence vis-a-vis sexual activity. ... is NO control over usage, and therefore no control over failure rates. ...
    (rec.arts.comics.strips)
  • Re: Report claims 16.4% of all Xbox 360 consoles will fail
    ... found that the failure rate of an Xbox 360 is not as high as ... I hate it when Call of Duty 4 multiplayer games crash ... These dropped host, shit spawns and scopes not being accurate are ...
    (alt.games.video.xbox)
  • Re: Letter to US Sen. Byron Dorgan re unpaid overtime
    ... >> will decrease the failure rate? ... large enterprise system failure, a true and arguable claim that has ... But I sense in this ng a hatred of meaning itself. ... moment at which we were able to reflect, independent of a programming ...
    (comp.programming)
  • Re: [angband] Percentage for spells ?
    ... For a ranger, the identify spell has minimum level 23 ... The adjustment to failure chance is ... increased to match the minimum failure rate. ...
    (rec.games.roguelike.angband)
  • Re: Fun with Lead-Free soldering
    ... ESD control. ... A guy I know in Wellington switched to antistatic bags & handling, and saw his PCB failure rate halve. ... We were pretty good with antistatic, but when we moved to SMT and the volume went up, our failure rate stayed constant, and became a problem. ... We used a Samtec bottom-entry connector on the control PCB, which had the PCB pins folded over the outside of the housing. ...
    (sci.electronics.design)
Loading