Reliability
I've found a lot of computers coming in with faults that a
few years ago would have been unheard of.
Power supplies
The newer ATX PSU labels carry better and better ratings and
yet are physically smaller than ever. They fail regularly. The
older bigger one's with less power printed on their labels seem
to go on for ever.
Inside the new PSUs the components are getting smaller and
smaller for the same ratings.
Is this the reason they're failing?
If a capacitor gives up its sometimes impossible to find a
replacement physically small enough to fit in the case.
Is it that margins are getting squeezed?
Are manufacturers asking too much of the component suppliers?
Not just component problems though. I still occasionally find
the odd PSU with resistors inadequately rated for their job. Wattage
is OK but in the front end of the PSU where there are voltages
in excess of 320, some designers still use resistors with totally
inadequate voltage ratings. Of course these fail prematurely.
In fact, it's not just mains variation, but to some extent
whatever domestic appliance being operated by your next-door neighbour
may influence the peak voltages in the front end of your power
supply. Whenever the spin on his washing machine drum changes
direction a thump of power is induced into the local mains supply.
The waveform of the voltage entering your computer is no longer
a pure sine wave as intended by the generating company, it may
be a lop-sided affair with large switching transients. The smaller
of these may be dealt with by tiny devices across the power supply
input but sometimes not. Sometimes the poor thing will end up
charred beyond recognition and sometimes cracked into pieces.
Once this protector has given up, any repetition of the event
will pass through the diode bridge and into the smoothing capacitor
where chance may play its part and end-of-life follow quite quickly.
The real reason for a reducing PSU MTBF (mean time before
failure.. or a way of averaging out over lots and lots of similar
things) is a combination of causes...
Yes.. physical size is being reduced, pressing component manufacturer's
to cut margins on their specifications.
Prices are being squeezed in line with increased competition.
This means pennies saved on components as margins are cut back.
BUT... in the UK the mains supply isn't what it used to be.
Pressure from Government to cut prices has cut costs. Cutting
costs has been made by reducing staff. This has resulted in less
preventative maintenance and hence more random power cuts, more
mains fluctuations and much more seriously... reducing the number
of local power conditioners (sub-stations.. or those humming masses
of grey pipes or a little brick building with a lightning bolt
on its door sitting in the middle of some housing estates) with
the result that your mains voltage isn't as tightly controlled.
Basically the mains voltage has been allowed to rise. Why? Because
less tolerance in the voltage means less expense.
The power station doesn't send out 230 volts. If it did there
certainly wouldn't be 230 volts at your mains socket. Because
of voltage drop along the distribution cables more voltage has
to be added to deal with the losses. Least loss is at higher voltages
so the primary transmission may be over 100,000 volts. As this
is too high for most peoples electric doorbells and suchlike it
must be transformed into something more manageable and more useful.
Progressively the voltage is transformed downwards until it's
at the right level for domestic users. Of course the actual voltage
at the local mains socket is still determined by the local demand
on current and as this may vary from next to nothing to huge amounts
of current when that football match on TV gets to half time and
thousands of kettles are switched on, some means of governing
the voltage must exist. A technique using feedback of the amount
of current being consumed is therefore made at the local sub-station.
In turn this equipment will feed back its demand on the equipment
further back in the chain and so on. Reduce the number of such
equipments and you'll cut costs but of course with the end result
that mains variation will increase.
Most mains supplies are well above the nominal rating of your
computer power supply! Look at the labelling. It will probably
say "230 volts". What is your supply voltage? Probably
around 245 or more. Hence reduced reliability from the stressed
components. A few years ago a PSU may have been rated at up to
265 volts. I don't think this applies any longer to most. Before
the Government allowed the voltage to be plus 10%/minus 6% it
was plus/minus 6%. Did they think about the consequences of their
action (said to allow harmonisation with the rest of Europe)?
I think not.
What about electrical equipment manufactured for use on the
continent (I mean Europe of course, or for Europhiles...the rest
of Europe)?
In France and Germany, equipment such as TV sets will be marked
220 volts. Also light bulbs and computers and suchlike are designed
to run at 220 volts. Plug one of these European leads into a UK
mains socket at 2am and there will be a significant mismatch.
In practical terms.. the life of a 220 volt item will be markedly
less when operated in the UK. How much? It's hard to say. It depends
on the margins built into the particular design. Plug into UK
mains a device intended for US mains, nominally 115 volts at 60
cycles, and the lifetime may be measured in milliseconds or up
to a minute if you're lucky. It's all a matter of degree. One
of my customers bought a "Millenium Clock" in Florida.
This was a large circular thing covered with LEDs and had a count-down
to 2000. When he plugged it into the mains on his return it barely
counted one second before illuminating hundreds of LEDs in one
last gasp then extinguishing and expiring in a puff of smoke.
Not very satisfactory you might say. But what about the situation
in England say in 1930? Such was the totally haphazard growth
of mains, from its concept in late Victorian times to the mid-30s
that a person might be faced with a mains socket supplying AC
or DC. If it was AC it may be anything from say 100 volts in Brompton
and Kensington to 250 volts in Durham. DC supplies were just as
varied being from around 100 volts at Herstmonceux to a toe-tingling
500 volts in parts of Eccles. AC wasn't always 50 cycles either.
It varied across the country from 25 cycles to 100 cycles, presumably
depending on the history of the particular power station and its
initial use... as sometimes lower frequencies were considered
best for some jobs and higher for others. Even a simple task of
purchasing a light bulb must have been fraught with difficulty,
no wonder gas lighting was so popular for so long.
Wireless set designers were forced to offer their products
in such a way as to enable a specific user to tailor the voltage
settings and cope with the two types of supply from his mains.
Hence we see the term "AC/DC". These sets did not use
a transformer (useless for DC mains) but relied on a large ballast
resistor carrying lots of tapping points to iron out the difference
between the lowest and the highest voltage to be met in practice.
AC mains sets had a more economic (cooler) means of catering for
voltage differences. This was a set of tapping points on the mains
transformer primary, usually via a rotatable 2-pin plug. Even
as late at the 60s were AC/DC sets for 200 to 250 volts still
being made.
Not only computer PSUs but also failing more regularly are
the small PSU's for things like scanners and printers. These are
usually marked 230 volts and their small physical size, coupled
with their location, where heat dissipation is well nigh impossible,
means that they will surely fail if their designers do not provide
an adequate margin. Unfortunately "margin" in this sense
invariably equates to heat and heat invariably equates to failure.
If you don't believe me. What's the modification being recommended
by lots of people that service satellite receivers? Take off the
lid and add a small electric fan to cool the chips.
A side effect of increased mains voltage is reduced life for
your light bulbs. When a lamp is designed there's a trade-off
between brightness and life. The brighter a lamp for a given voltage
and wattage the shorter it's life. That's been a fact known since
lamps were first invented!
Today the mains is 230 volts plus or minus 10%. Except for
rare occasions when load is very high, that means PLUS. In times
of lightly loaded mains the voltage is going to be at least 253
volts and this may be even more when tolerancing is poor. A modern
light bulb is marked "240 volts" so it's lifetime that
was calculated at 1,000 hours will be a lot less than this. Continental
electrical goods are rated at 220 volts so beware when plugging,
say French equipment, into UK mains! If it has a lamp in it, it'll
be very bright but not for long. Strangely, many of those new
bayonet-fitted long-life fluorescents are marked 220-240 volts.
Longer life in Europe and shorter life surely in the UK? At least
they'll be brighter here.
Back to computer PSUs... I've had a few PSU's that appear
to be perfect but the computer stubbornly refuses to work. This
is because at initial switch-on, in a fraction of a second the
PSU output is checked by the motherboard and if found wanting
will be turned off. Usually the 5 volt rail say.. will drop a
fraction of a volt under load and this is enough to deem it bad.
Make the "power good" line active and the PSU will come
on quite happily with voltages looking normal, but this test is
not as stringent as the computer test. The usual cause of the
problem is poor smoothing capacitors running at way above their
designer's ratings. If the capacitors are low in value or if their
impedance rises the PSU output integrity will be compromised and
it won't work.
Hard drives
I must have replaced four hard drives in a couple of weeks.
These weren't very old either. A few months ago I fitted a new
drive under guarantee. Not surprising why it had failed though.
The owner had set defrag to operate every day as a matter of routine.
The poor hard drive had just worn itself out!
The latest problems...
Couldn't read because the circuit board had failed (Fujitsu..
I contacted them about a new circuit board, as the data was needed
urgently, but they just couldn't have cared less). I made sure
the new drive wasn't a Fujitsu.
Made a noise like a marble in a tin bucket when spinning up.
A Seagate.
Just couldn't be recognised except when cable select set up
instead of master or slave. A Connor.
Couldn't be set as master but could be read as a slave. Another
Connor.
See elsewhere in my scribblings about the GREAT Fujitsu scandal!
Motherboards
Not very often do I get a faulty motherboard. A Pentium Pro
board wouldn't accept more tha 16Mbyte of memory. A socket 7 board
from a Tiny computer lost its graphics facilities.
CDROMs
Lots and lots of these have been replaced over the last few
years. They seem to last no more than 2 to 3 years now. I have
a suspicion it may be their lasers getting tired. Question...
is the laser switched on continuously? Is it switched on only
when a disk is present? It it switched on when a read is requested?
Maybe the ones that fail most frequently do so because the laser
is on too much? The same holds good for domestic CD players. I'm
forever fitting new lasers to machines not much more than 2 or
3 years old.
With computer CDROMs.. it's not worth repairing these because
the cost of a new model (with a much better spec) is far cheaper
than the likely cost of repair.
Cooling Fans
These things, used in PSUs and atop the processor chip, often
start to get noisy with advancing years but quieten down as they
warm up. Don't be fooled though because when thet quieten they
sometimes spin too slowly and when it's a non-Intel chip needing
maximum cooling the fan won't keep the temperature down and the
chip r u n s s l o w .
Because of consumer grumbles much quieter fans are now around.
Time will tell whether these are any good.
Return to the computer faults page