[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: IIgs locking up after 15-20mins, TWGS likely overheating.

On 8/31/2010 12:47 AM, ict@ccess wrote:

Sorry Michael!  I am jinxing your voodoo and raising you a fact.


There's always some fact that causes people to believe in voodoo.  ;-)


You can fool some of the people some of the time with voodoo and you
can fool some of the people all of the time with facts but you can't
fool all of the people all of the time with (somebody else fill in
this word) --- bulls**t -- ice cream --- candy ---


In this case, the "voodoo" is the notion that the bigger wires will
increase the voltage or current delivered to the main board to a
significant degree--and that isn't so.  There are other effects,
though, as noted below...


Not at all.  My first reading of the reasons for installing bigger
wires were, bigger wires have less resistance, which reduces heat and
noise, which in turn allows for less voltage drop which means that the
power supply is sufficient to supply most slot cards which in turn
creates a more stable system.


Some of this is true, at an infinitesimal level, but none of it
should have any effect on a system which is not *very near* crashing.

The "reduces heat" claim is so small as to be unmeasureable--in fact,
to the degree that less resistance between a load and a bypass capacitor
reduces noise, it actually increases heat slightly by supplying peak
current during a fast switching event that would otherwise be limited.

It is important to discriminate between significant effects and those
which are insignificant.  If your system is so near its margins that
it responds to increasing the size or position of the power feed, then
there are much more effective (and easier) ways of increasing its
margins.

Rob, you are the target market of the Monster Cable folks.  ;-)  It
may be fun to attribute good juju to some physical aspect of wire,
but it is not good science.  An hour spent in a *different* way to
improve system stability would be much more effective.


I'd be willing to bet that your crashes would also stop if you
connected a high-value capacitor between +5v and ground in the
vicinity of slot 7.


Don't know.  If I would have read an explanation of how that prevented
crashes instead of reading how the increased wire size reduced crashes
and made the system more stable, I might have tried that instead.


Be careful what you read.  ;-)

Most system modifications are made by electronic novices--because
there are so many of them--and their "experiments" are often done in
an unscientific way.  For example, many do not "undo" the single change
they attribute their results to in order to verify that it alone was
responsible for the change in behavior they observed.

Just unplugging and replugging the power supply connector can be
responsible for making the difference they observe!


Big wire doesn't do a lot to increase DC voltage, but it does
decrease the resistance and inductance between the main board and
the big electrolytics in the power supply.  As a result, it is
causing the high-frequency noise to be attenuated at the bus,
improving noise margins.


Finally, we agree on something  :)


But read on--it's ability to reduce noise is quite marginal compared
to directly addressing the problem...


A better way to bypass the main board is to put capacitors right
on the main board, as near as possible to where rapidly changing
loads occur--the Zip GS and RAM card are good candidates.  (This
is why well-designed cards have sufficient bypassing on the card.)

It sounds like your system is running on the "ragged edge" of
reliability, since a solid system shouldn't crash in weeks!  This
makes it very sensitive to very minor changes in the noise, voltage,
and temperature environment.  This is exactly the environment in
which changing almost anything can make a difference.


Define solid system.  My system is solid if I replace the 8 MB serius
card with the 1.5 MB Apple RAM card.  Which tends to lead me to
believe there is nothing wrong with the ZipGS card, or maybe the
serius RAM card just doesn't like Zip.  And also don't forget, I went
from 10-12 crashes down to 1 with increased wire size.  Which lends me
to believe that the motherboard still can't quite handle 2 or more
large cards.


You may be right.  It sounds like the RAM card is the critical element
here.  RAM cards can also be big sources of transient loads, making
good bypassing critical for proper functioning.

Usually, DRAM cards have ample bypassing (lots of 0.1uF caps) to reduce
the effect of their fast current peaks, but I haven't seen a Sirius card
like yours.  Maybe it needs more bypassing.  Maybe some of its bypasses
are missing or broken (I once had a ][+ fail because the 0.1uF bypass
near the processor had one lead broken from fatigue).

The ZipGS should not interact negatively with the RAM card from a timing
or electrical perspective, so if the RAM is reliable, the ZipGS will not
alter that.  The only interaction would be additional power drain, but
your evidence suggests that your system supplies that power reliably.


So if a system is solid with minimum cards, does that make it running
on the "ragged edge" if too many cards draw too much power and makes
it unstable?  Everything has its breaking point.


It is certainly possible to overload a power supply by using too many
high power drain cards.  The correct solution in that case is to use
a more capable power supply--and provide additional cooling to deal
with the generated heat.

Each to his own, but I would never rely on a computer that is not
operating well within its margins, as evidenced by stable operation
over a long period of time and over all normal environmental conditions.
But then I'm not about finding "breaking points" (except for margin
determination), but about *using* the computer as a tool.


But a digital computer is designed so that you can change all sorts
of things by +/-10% and *nothing changes* in the functioning of the
machine.  That's the whole point of a digital machine--that small
"analog" perturbations are irrelevant to its operation, since it
has sufficient timing and noise margins to not respond to them.

Life is much better when you "back it off a little" and improve system
reliability by a huge factor.


Is it not easier to improve or make the slot cards more efficient with
less draw on the system than trying to improve the system itself?
Modifying the motherboard is probably more than most of us care to
do.  The system has already proven to be solid with no cards plugged
in.  But such is the imperfect world.


Of course it would be better to use more efficient cards--if they
exist.  That's why I prefer the ZipGS to the TWGS.  But if you must
live with what you have, simple bypassing modifications or using a
more capable power supply are a fine solution.


I haven't soldered a wire to slot 5 ground yet though because, just
call me plain lazy, but one crash a day is acceptable to me.  :)


And if just doing that makes your system more stable, then doing
something else (like additional bypassing or a better power supply)
would make it rock solid--which I assume you'd like even better.

Again, my point is that when a system is operating near the limit of
its margins, its reliability is *extremely* sensitive to almost any
change in conditions.  This is a *bad thing*.  If you fix the underlying
cause of the near-failing condition, the system will be operating well
within its margins, and it will be rock solid.  This is the goal for
any digital system.

When a system is sensitive to almost any change in conditions, it is
a fertile source of superstitions, since changing almost anything will
have some effect on its operation, for better or worse, causing a
casual observer to think that a critical "fact" has been discovered.

Maybe you remember how people fiddled with indoor TV antennas in the
early days of poor tuners and weak signals.  Someone would raise their
arm and the picture would improve and everyone would say, "Stop there!
That's it!".  Of course, that *wasn't* it--*it* was getting a better
antenna or a stronger signal!  ;-)


Wow.  To me, that's like saying my tires are fine because I only have
one blowout a day.  ;-)  You shouldn't have *any* crashes, except
possibly those due to software bugs.


I actually have tires that have been patched 3 times and still going
strong.  :)  But the blow outs are no fault of the vehicle or the
contents of the vehicle.  They are more a cause of the load that is
being placed on the vehicle.  (How's that for an analogy  :)


I wasn't reacting to any *cause* of the blowout, only that you
seemed satisfied with a system that was unstable enough to crash
on a daily basis.  That's about 1000 times less reliable than I'd
be willing to live with.  Even with backups, I'd hate to risk my
hard disk with a system crash.


Even 1 mVolt goes a long way to make the IIGS more stable and the
power supply strong enough for most purposes.  I have even installed a
Version 01 motherboard into a PC tower and adapted a 200 watt power
supply to work with the IIGS motherboard with no better results.  I
still crash about once per day.

my 2 cents worth (solder and #12 gage wire extra)


Millivolt variations in supply voltage are completely irrelevant to
a stable system.


Again with the stable system  :)
you are not trying hard enough to break it  :)
half the fun is testing the limits


I get the smileys, but of course the essence of a digital system
is stability.  We couldn't have computers unless we could design for
relative independence from environmental variation.


You have at least one serious hardware instability
in yours, and I'd bet that its the ZipGS.


Could still be possible, but ZipGS is way more stable with a 1.5 MB
RAM card than the 8 MB Serius RAM card.  I can leave the computer on
for a week without crashing.  But I really want to use my 8 MB RAM
card.  Testing with the ZipGS turned off is not really an option which
makes GSOS excruciatingly slow.


Again, from what you've said, I'm becoming more suspicious of the RAM
card...


Many overclockers just keep increasing the frequency until it crashes,
then back off one notch.  A year later, margins have decreased to the
point where it is now crashing at the lower frequency.  Degrading
capacitors and power supplies, or increased loads (more cards) are
the usual mechanism for decreasing margins.  Better to fix them than
to "learn to live with them."


I believe fixing them requires new technology with lower power draw
demands.  Like, my serius RAM card would probably have lower power
consumption if I had the 1x 1MB chips instead of the 8x 1MB and thus
be more stable.  And the CFFA drives have almost no write errors
compared to the Focus Drive, but not just because of the lack of
mechanical parts, but also partly due to the less power draw.  Fixing
the motherboard is not really an option.


It would be nice to have lower-power options, and I recommend them
when they exist, but there's no point in waiting for what may never
come to the Apple II when there are good ways of dealing with what
does exist.

Good bypassing and good power supplies are easy solutions that are
readily accessible.


If you look inside most power supplies, there's an adjustment for the
+5v level.  Measure the +5v at the main board and adjust the pot for
5v (or even 5.1v).


Done that too.  But I am still not going to let the "benefits of
thicker wires" go.


The benefits you describe are not worth the trouble of installing them.
They are, literally, a drop in a bucket.  Better to fill the bucket.


I have done most of what has been suggested with the exception of the
capacitor to ground mod but to date, the wire mod has been the most
successful.


If the problem is excessive digital noise, then the wires are the
*only* thing you've tried.  The capacitor(s) will be much more
effective (and easier and less visible).


None of this really matters as the Bernie emulator on a beige G3 has
become my favorite computer to program on.  No problems of crashing
there.  But like most, I like to return to the original to play the
odd game.


I completely agree.  Emulation has its advantages, but the real thing
is still great!


Still only giving my 2 cents worth but if there is a reply to this
post, then I will have to increase to 5 cents due to inflation :)


;-)

I have two reasons for replying:  1) to help prevent the spread of
superstition about the Apple II, and 2) to help you make your system
more reliable.

-michael

NadaNet 3.1 for Apple II parallel computing!
Home page:  http://home.comcast.net/~mjmahon/

"The wastebasket is our most important design
tool--and it's seriously underused."