Re: 911 Fatal Error on TransWarp GS?

[mjmahon@aol.com (Michael J. Mahon)]

Rubywand wrote:

>Frank Townsend writes ...
>>
>> Rubywand" <rubywand@swbell.net> wrote ...
>>>
> ....
>>>   If the 0911 errors vanish, it is fairly likely that the problem is
>system
>>> noise resulting from the load represented by the TransWarp GS in
>combination
>>> with skimpy Power Supply leads and skimpy power traces going to the Slots.
> ....
>>
>> I thought Michael Mahon's nicely reasoned essay discussing the resistance of
>> the IIgs power supply leads pretty well debunked this hypothesis.
>> 
>> See
>> 
>>
>http://groups.google.com/groups?q=%22power+supply%22+mahon&hl=en&lr=&ie=UTF-
>> 8&oe=UTF-8&selm=20020111033043.25525.00001341%40mb-fc.aol.com&rnum=2
>> 
>
>     Thanks for the link. Missed that message from Michael.
>
>     Agree. He makes a reasonable case. It does depend on some ideal
>conditions. 
>
>     In practice, the lines to the motherboard are not ideal 18 gauge lines,
>the power supply is not an ideal battery, and the voltages in the lines are
not
>pure DC.

Actually, there isn't much to an 18-gauge wire except resistance, inductance,
and capacitance, and all those are addressed in my reply.  Assuming only
that most of the strands are not broken, the analysis stands without change.
Note also that a break in half the strands is bridged by the fact that strands
are not insulated from one another, so that the incremental increase in
resistance is very small (though in extreme cases, local heating could be
a problem).

Nothing in my analysis depends on how "ideal" the power supply is--only
that it meet the voltage, current, and filtering requirements of the Apple II.

Except for ripple, which should be in the millivolt range, the power supply
voltages are, indeed, DC.  My analysis assumes that the poser supply is
properly decoupled for high frequencies by the main board, and for low
frequencies by the power supply filter capacitors.  If this is not true, then
the problem is not in the wires, but in the decoupling components.

>     Most fairly filled-out IIgs systems probably draw 3-4 Amps on the +5V
>line. At this level, measurements showed a noticeable DC drop across the
lead--
>think it was around .3V-.4V. When powering the IIgs, the instantaneous current
>demand due to clocked switching of gates should be higher and so should the
V's
>appearing across the line (especially at high frequencies).

And, because high frequencies are decoupled on the board, these transient
currents will be supplied by the decoupling capacitors.  My analysis dealt
with a low-frequency delta-I of 2 amps, which is quite generous.

>     So, the user should expect that gates will have lowered noise immunity
and
>that more noise will be present. This should lead to system crashes. 

Again, not if the main board decoupling is working (for high frequencies) and
the power supply filter capactors are working (for low frequencies).  (And,
of course, the power supply regulation is working for very low frequencies.)

If any of these things are not working, they are the problem, not the
power supply wires or PC board conductors.

>     When programming in Applesoft or in the monitor the user is doing a lot
of
>ADB accesses (key presses). On a noisy IIgs, the ADB sub-system seems a fairly
>likely place for a programmer to experience a glitch leading to a crash.
>
>     Over time, the noise spikes appear to degrade Power Supply performance.
>Perhaps the feedback leads to circuit conditions which damage some components.
>For sure, many IIgs power supplies seem to fail for no obvious reason.

The reasons for power supply failure are pretty much the same for all
switching power supplies:  line transients, temperature, and time.

A single large line transient can take out an input diode, and consequently
a fuse and/or and input capacitor.  Repeated smaller line transients can
do the same over time.

High temperatures accelerate aging of all electronic components, but
expecially electrochemical components, like electrolytic capacitors--
most often resulting in lowered capacitance and/or higher equivalent
series resistance (ESR).  Both effects contribute to lowered filtering
capability and lower power supply margins against specifications.

Finally, time increases the cumulative probability of both random
failures and wear-out or stress failures.  Again, it is the electrolytic
capacitors that are the most probable victims.

>     Real world support for doing the suggested Power Supply mods is pretty
>good. Our IIgs is a Woz machine with the ROM-01 upgrade. It has the Power
>Supply and motherboard Slot wiring mods. It's fairly loaded-- 4MB GS-Ram III,
>64k ZipGS+, RamFAST SCSI, Focus drive, Stereo Card, Mockingboard, ... . There
>is a System Saver IIgs to supply cooling, etc..
>
>     Crashes are very rare-- maybe once every ten sessions or so; and, the
>Power Supply (which has never had a cooling fan attached) is, except for
>having
>the leads swapped out,  the one that came with the computer back in 1986.
>
>     Plus, one or two users plagued with crashes and noise who have done the
>Power Supply mod report good results.

As I pointed out, making such a modification will slightly improve margins in
a marginal situation.  Given the continuous worsening of most power supply
symptoms, I would be very surprised if the marginally lowered resistance
of the power supply leads added much useful life to a gradually failing supply.

It is difficult to perform controlled experiments.  For example, replacing
the wires on a power supply plug, or replacing the plug in the process, may
result in cleaning the contacts or even slightly enlarging them, inadvertently
repairing a marginal power plug/socket connection.  (Replacement also
repairs any cold solder joints (one hopes ;-), but I have not seen this
problem on Apple boards and power supply connectors.)

To find out what is actually going on, use an accurate voltmeter and an
oscilloscope to check the voltages (DC and transient) at each end of
each power supply wire, being particularly careful about grounding.

I think you will find that the same or better results could be obtained
by bridging a 1000mf-25v capacitor from +5v to ground, either at the
power supply end or at the main board.  (This will "stiffen" the power
supply filtration back into specifications.)

Although one could also add or replace high frequency decoupling
on the main board (with more 0.1mf caps), it is unusual for them to
fail (except by mechanically breaking off the board), so replacement
is seldom required.

-michael

Check out 8-bit Apple sound that will amaze you on my
Home page:  http://members.aol.com/MJMahon/