This is only a preview of the February 1997 issue of Silicon Chip. You can view 25 of the 96 pages in the full issue, including the advertisments. For full access, purchase the issue for $10.00 or subscribe for access to the latest issues. Items relevant to "Computer Controlled Dual Power Supply; Pt.2":
Items relevant to "Control Panel For Multiple Smoke Alarms; Pt.2":
Articles in this series:
Articles in this series:
Articles in this series:
Purchase a printed copy of this issue for $10.00. |
*!#$*&<at>*
COMPUTERS
Sorting out what’s really to blame
If there’s one thing I’ve learnt about computers
over the years, it’s not to jump to conclusions
when something doesn’t work properly. When
things go wrong, it’s all too easy to blame the
obvious, without getting to grips with what’s
really at fault.
By GREG SWAIN
How often have you heard that a par
ticular operating system is unstable?
Or that it doesn’t work on such and
such a computer? Or that a particular
item of hardware has a “bug” and
should be avoided? Or that something
just doesn’t work when your own ex
perience indicates otherwise?
When it comes to computers, there
are enough real problems to sort out
without having to also sift through a
mine of misinformation, straight-out
even been told that PCs are no good
in this role because “they just don’t
work” and because “they have font
problems”.
Well, you could have fooled me.
There we’ve been all those years, suc
cessfully producing a magazine using
PCs that don’t really work – at least
according to the hearsay of several
self-appointed experts.
What rubbish! We’ve used PCs in
the desktop publishing role for over
“A computer is really a box full of
gremlins, just waiting to wreak all
sorts of havoc at the user’s expense”.
bad mouthing and old wives’ tales.
I’d like a dollar for every time that
someone has rolled their eyes to the
ceiling when told that SILICON CHIP
is produced using PCs, for example.
I mean, everyone knows that Macs
are the all the go when it comes to
desktop publishing, don’t they? I’ve
4 Silicon Chip
six years now with very few problems
but try telling that to some people.
Not that it’s ever really worth the
bother – a few pointed questions in
variably reveal that such people know
very little about PCs, and are simply
basing their opinions on “common
knowledge”.
It’s all the stuff of myths and legends
but if it’s “common knowledge”, then
it must be true. I’m not seeking to belit
tle Macs here, by the way. I’m simply
making the point that the PC is a valid
alternative for desktop publishing, de
spite what many ill-informed people
will try to tell you.
The most common misconceptions
by far arise out of hardware and soft
ware upgrades. The reasons are not
too hard to find. Hardware upgrades,
in particular, are often not straight
forward for a variety of reasons. After
several recent experiences of my own,
I’m convinced that a computer is
really a box full of gremlins just wait
ing to wreak all sorts of havoc at the
poor user’s expense.
OK, so I’m exaggerating somewhat
but if you’ve ever attempted to add
hardware to a PC, you’ll know what
I mean. Even Windows 95’s much
vaunted Plug and Play (PnP) system
has problems in some circumstances.
Let me give you a couple of examples
of what can happen when even rela
tively simple upgrades are attempted.
The not-so-crook RAM
Recently, we decided to upgrade
the RAM in a couple of our office
machines from 32Mb to 64Mb. These
two machines used identical mother
boards and in each case, the existing
RAM consisted of two 16Mb SIMMs.
As a result, we decided to purchase
two new 32Mb SIMMs for the first
machine and transfer its existing 16Mb
SIMMs to the second machine.
Installing the new 32Mb SIMMs was
straightforward enough but when we
We solved the problem by leaving
the new SIMMs in the third machine
and sharing its original four 16Mb
70ns SIMMs between the first two
machines. So all three machines ended
up with 64Mb of RAM – it’s just that
the two new 32Mb SIMMs ended up
in an unexpected location.
But it’s easy to see how misun
derstandings can arise in this sort of
situation. We could have easily been
fooled into returning perfectly good
RAM to the supplier, demanding that
it be replaced. And of course, the re
placement RAM would have caused
exactly the same problems.
upgrades. Win95 cannot automati
cally assign interrupts (IRQs) to nonPnP expansion cards (now referred
to as “legacy” cards) and can easily
get itself into a knot if left to its own
devices.
To explain, a standard PC has 16
interrupts (0-15) available but most of
these are taken by the system, leaving
about six free for expansion cards (de
pending on the configuration). Each
expansion card must be assigned a
unique IRQ; if two cards have the same
IRQ, there will be a conflict and the
system won’t work properly.
The “minimalist” approach usually
works well when install
ing Win95,
particularly if you have a mixture of
legacy and PnP cards. This involves
removing all non-essential cards, such
as sound cards and network cards,
before installing the software. Once
the system is up and running properly,
you can add the expansion cards back
in, one at a time.
By the way, it’s best to add the legacy
cards first, as the PnP cards are auto
matically assigned the leftover IRQs.
In addition, the system assigns IRQs
to PnP cards in ISA slots before those
in PCI slots.
Note also that you should reserve
the appropriate IRQs for the legacy
cards in the system BIOS, where this
facility exists (ie, if the motherboard
has a PnP BIOS). This will usually be
found under a “Plug and Play Config
uration” (or similar) menu. For exam
ple, if you have a legacy card that’s set
to IRQ10, then you must assign IRQ10
for use by an ISA card.
The above step is quite important. If
you don’t do it, Win95 may automat
ically assign an IRQ that’s already in
use to a PnP card.
Operating system myths
The suspect motherboard
There are also plenty of myths
floating around regarding operating
systems. One that I’ve heard from a
couple of people is that Windows 95
needs a Pentium processor and won’t
run on a 486. Wrong! What they really
mean is that they couldn’t get it to
work on their particular 486 for some
reason or other.
To prove the point, we recently
installed the Win95 upgrade pack on
an old 50MHz 486 machine. It ran
without a hitch and I’ve even heard
of people running Win95 on a 386.
Hardware conflicts are often the
root cause of aborted operating system
It’s not just hardware conflicts that
can be a problem. Hardware bugs can
also cause problems and lead to unfair
criticism of an operating system or
even individual programs.
Consider my own experiences with
Windows 95 which is installed on my
main office machine. This is one of
the machines described above that
didn’t like the 60ns 32Mb SIMMs and
the problem I am about to describe
is directly related to the memory
upgrade.
As mentioned earlier, this machine
was originally configured with 32Mb
of RAM (2 x 16Mb SIMMs). It is set up
This new motherboard cured an unstable Windows 95 installation. The bus
speed of the motherboard in the original machine had apparently been pushed
beyond its design limitations – or, at least, that’s one theory.
switched on, the machine refused to
boot. Sometimes it would just hang
after completing the BIOS checks. At
other times, it would start to boot the
operating system and then halt, with a
screen full of obscure error messages.
Our first thought was that we must
have dislodged a cable when installing
the new RAM but a quick check re
vealed that all was as it should be. Our
next snap diagnosis was crook RAM
and this was seemingly confirmed
when it also failed to work in the sec
ond machine. Yet when we substituted
the old RAM, both machines booted
without problems.
That was it, of course – one of those
new 32Mb SIMMs just had to be faulty!
But was it? Perhaps the problem was
in the two machines which, after all,
were virtually identical. The only way
to find out was to try the new SIMMs
in a third machine with a completely
different motherboard. When we did,
it booted straight up and performed
flawlessly.
So much for our snap diagnosis of
faulty RAM! Instead, it appears that the
motherboards in the first two machines
weren’t happy with the 60ns RAM
on the new SIMMs (the older SIMMs
used 70ns RAM). And there were no
settings in the system BIOS that could
compensate for this.
February 1997 5
You can easily check for
hardware conflicts in
Win95 by double-clicking
the “System” icon in
“Control Panel”, then
selecting the “Device
Manager” tab. Doubleclicking on a specific
device then gives you
the “Resources” tab,
which lets you view the
resources allocated.
as a dual-boot Windows 3.11/Win95
system but I was never entirely happy
with the system under Windows 95
as it occasionally crashed when the
going got heavy.
My first inclination was to blame
which ever program I happened to
have running at the time but eventu
ally I began to suspect Windows 95
itself. However, the pattern was too
intermittent to really get to grips with
the problem.
The breakthrough came when
6 Silicon Chip
the extra 32Mb of RAM was added.
Windows 3.11 continued to work
normally but not so Windows 95. It
now frequently crashed, generat
ing
“Unexpected Exception” errors in the
process. And when it crashed, it would
often refuse to boot again unless the
additional RAM was removed.
This meant that the problem was
probably hardware related. Initially,
I simply tried replacing the original
RAM with the second two 16Mb
SIMMs but this made no difference.
I also tried swapping memory banks
and substituting the memory from the
machine’s twin without result.
By now, the finger of suspicion was
pointing fairly and squarely at the
motherboard. After all, it had previ
ously failed to work with the 32Mb
SIMMs, so it was definitely suspect
when it came to handling memory.
And there was another thing. The
machine ran a 133MHz Pentium
processor but the manual that came
with the motherboard only mentioned
75MHz and 90MHz processors. An
8-way DIP switch is used to set the
bus and processor clock speeds but the
manual gave no details of the settings.
Fortunately, the settings were
screen-printed on the moth
erboard
in a vacant area adjacent to the micro
processor. But what was interesting
was that the machine’s twin, which
was purchased four months later,
also showed an additional setting for
a 100MHz processor. And both ma
chines were set to this configuration.
That aside, the motherboard was
obviously originally designed to op
erate at a bus speed of either 60MHz
when used with a 90MHz processor, or
50MHz when used with a 75MHz (or
100MHz) processor. That’s because the
processor always runs at some ratio of
the bus frequency, the available ratios
in this case being 1.5x and 2x (ie, 1.5
x 50MHz = 75MHz, 1.5 x 60 = 90MHz,
and 2 x 50MHz = 100MHz).
So what was going on? A 133MHz
processor requires a 66MHz bus so
the board must have been tweaked to
run at this higher speed – probably by
the simple expedient of substituting
a different crystal in the clock circuit
during manufacture. Fairly obviously,
this was an existing design that had
been hastily adapted to cater for the
faster processor but it appeared that
it wasn’t up to the task – at least not
with Windows 95.
There was nothing for it but to
change the board.
It ain’t that easy
Now if you thought that changing
the motherboard in a Win95 system
was a straightforward task, think again.
The mechanical installation is easy
enough but getting everything up and
running again is a different matter.
Windows 3.11 was OK (it’s too dumb
to recognise the swap) but Windows
95 is too clever by halves, tying itself
in all sorts of knots when it discovered
SILICON CHIP SOFTWARE
Now available: the complete index to all
SILICON CHIP articles since the first issue
in November 1987. The Floppy Index
comes with a handy file viewer that lets
you look at the index line by line or page
by page for quick browsing, or you can
use the search function. All commands
are listed on the screen, so you’ll always
know what to do next.
Notes & Errata also now available:
this file lets you quickly check out the
Notes & Errata (if any) for all articles published in SILICON CHIP. Not an index
but a complete copy of all Notes & Errata text (diagrams not included). The file
viewer is included in the price, so that you can quickly locate the item of interest.
The Floppy Index and Notes & Errata files are supplied in ASCII format on a
3.5-inch or 5.25-inch floppy disc to suit PC-compatible computers. Note: the File
Viewer requires MSDOS 3.3 or above.
ORDER FORM
PRICE
❏
Floppy Index (incl. file viewer): $A7
❏
Notes & Errata (incl. file viewer): $A7
❏
Alphanumeric LCD Demo Board Software (May 1993): $A7
❏
Stepper Motor Controller Software (January 1994): $A7
❏
Gamesbvm.bas /obj /exe (Nicad Battery Monitor, June 1994): $A7
❏
Diskinfo.exe (Identifies IDE Hard Disc Parameters, August 1995): $A7
❏
Computer Controlled Power Supply Software (Jan/Feb. 1997): $A7
❏
Spacewri.exe & Spacewri.bas (for Spacewriter, May 1997): $A7
❏
I/O Card (July 1997) + Stepper Motor Software (1997 series): $A7
POSTAGE & PACKING: Aust. & NZ add $A3 per order; elsewhere $A5
Disc size required: ❏ 3.5-inch disc
❏ 5.25-inch disc
TOTAL $A
Enclosed is my cheque/money order for $A__________ or please debit my
Bankcard ❏ Visa Card ❏ MasterCard
❏
Card No.
Signature_______________________________ Card expiry date______/______
Name ___________________________________________________________
PLEASE PRINT
Street ___________________________________________________________
Suburb/town ________________________________ Postcode______________
Send your order to: SILICON CHIP, PO Box 139, Collaroy, NSW 2097; or fax your
order to (02) 9979 6503; or ring (02) 9979 5644 and quote your credit card number
(Bankcard, Visa Card or MasterCard).
✂
the new motherboard.
This problem apparently stems from
the fact that the parameters of the new
motherboard and its BIOS don’t match
the existing registry settings. In the
end, the only way around the problem
was to clean off the existing installa
tion and its associated applications
and reinstall all software.
If you’re ever in this situation, by
the way, don’t be tempted to simply
reinstall the operating system over
the top of the existing installation. It’s
best to clean everything off and go for
a fresh installation.
The Win95 reinstallation was not
without a small glitch, however. I’d
removed the sound card but left in
a SCSI card and a PnP network card.
Everything went fine until the first
boot. Windows 95 made it past the
logon dialogs but then announced that
it was searching for new hardware.
There followed a brief period of hard
disc activity, after which it just “hung”.
I tried switching the machine on
and off several times but always with
the same result. Eventually, I pulled
the network card and tried again. And
that was it – the system now booted
correctly and I was able to reintroduce
the sound and network cards.
Windows 95 now recognised the
network card, installed the correct
driver for it and automatically as
signed an available IRQ. Now why
didn’t it do that in the first place?
Did the new motherboard do the
trick? Well, based on my limited obser
vations so far, the answer is yes. I now
have a stable Windows 95 installation
but just think how easy it would have
been to jump to conclusions and badmouth the operating system.
Strangely enough, the other ma
chine with the identical motherboard
operates perfectly with its 64Mb of
RAM but then it’s running Windows
NT. So is Windows 95 fussier than NT
about the hardware company it keeps?
Or was it just a matter of manufac
turing tolerances between the two
boards? Or was the problem caused by
some subtle hardware conflict which is
now resolved and nothing to do with
the motherboard at all?
Finally, could I have cured the prob
lem by changing the DIP switches so
that the motherboard ran at a slower
bus speed (yes, this would have throt
tled back the processor)?
We’ll probably never know the an
SC
swers to those questions.
February 1997 7
|