Re: 911 Fatal Error on TransWarp GS?

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

Rubywand replied:

>"Michael J. Mahon" writes ...
>> 
> ....
>
>> 
>> What was the significant voltage drop?  How was it measured?  Did it
>> include the connector?  Did it include any main board traces?  How
>> much voltage drop was induced by inserting the current meter?
>> 
>> Have you looked at the +5 with a scope?  If the noise is present at
>> the power supply terminals, then fatter leads won't help.  If power is
>> clean at the power supply terminals and dirty at the main board
>> connector, then, and only then, bigger leads could help.  (Again,
>> proper gounding of the scope is essential.)
>> 
>
>     It's been a while since doing the measurements; so, I don't recall the
>details. The check was done outside of the IIgs into a purely resistive load.
>Removing the current meter would have been my usual practice for doing final
>voltage drop checks on low R leads. Pretty sure the drop at 3 Amps was around
>.3V or so. Don't recall putting the scope on power check points in the IIgs. 

This would correspond to a resistance of 0.1 ohm--which is 14 times the
resistance of a foot of #18 wire.  That is not attributable to the wire unless
it is _seriously_ broken, or has one or more defective connections.

>> Have you tried simply replacing the wires with known-good wires of the
>> _same_ guage?  Ohm's Law is not just a good idea--it works!  ;-)
>>
>
>     Replacing the main PS lines is sufficient bother so that, when you do
>it,
>you make sure to use heavier wires.

Only if your are predisposed to think that the designed wires are
too small.  I'm still not seeing the case for that.

"That it works" ignores the fact that the solder joints were re-soldered,
the connector re-soldered and re-plugged, etc.  And, of course, it's
possible (but unlikely) that the wire(s) themselves were broken in
several places...

>> > For instance, plugging in stuff like accelerator boards often leads
>> > to a rash of "mysterious" system crashes, increased noise in audio
>> > outputs, etc..
>> 
>> As would be expected on any system in which the power supply had
>> become marginal relative to the load.
>
>     If you mean that you have problems when the PS is significantly damaged,
>sure.

No, I mean where it is barely meeting, or just missing its specs, but still
working, mostly.  Most probably because of aging electrolytics, but
possibly also because of chronic operation beyond its rated load.

>     In most cases, though, we are dealing with good power supplies. Adding
>an accelerator will seldom push the load beyond the native capacity of the
IIgs
>power supply. If it does-- like you have the other Slots filled with power
>gobbler boards-- then, the PS should simply shut down and remove all doubt.

This is not clear.  Most Apple power supplies do not have a strong
overload detection capability.  They just try to regulate, and, at some
point where they are far out of spec, they shut down.  Most continue
to operate even into a load which, if sustained, will overheat them.

>     Most IIgs "power supply problems" are either obviously linked to a
>defective PS or they are really power distribution problems-- i.e. the lines
>from the PS are too skimpy, motherboard power traces are too skimpy,
>bypassing
>fails to handle noise produced by accelerators switching at frequencies 2-5
>times the normal IIgs max, etc.. 

I'm with you on the defective power supply possibility, but would like
more data on how far the Apple IIgs deviates from accepted practice on
power traces.  And bypassing actually becomes more effective with
higher frequencies, if it is properly designed (very close to chip power
leads, wide <low inductance> traces, etc.).

In the case of the Apple ][, the +5v supply to the slots was rated at
500ma for _all_ cards together, so a "good citizen" card would not
consume more than 500ma/8 or 67.5ma.  The early Apple-designed
cards actually went out of their way to power down ROMs, for example,
when the card was not selected.

Many later-designed cards fell into the practice of consuming _all_
of the 500ma budget, throwing themselves on the mercy of 1) most
users having few plug-in cards, and 2) the tolerance of the Apple,
including its power supply, to overload.

When the //e came out, increased integration levels actually
reduced the main board power requirements, so that tolerance
of out-of-spec card drain was reinforced (though the Transwarp
and a few others pushed the limits _hard_).

I don't have the IIgs hardware spec here, but I'll bet that the
card current drain is still speced to a pretty low value--well below
what many cards actually draw.  This is a creeping design issue,
since there was no "certification" program for cards, and therefore
no enforcement of specifications by Apple.  The _de facto_ rule
was, "if it works in my system, it's OK."  This is a recipe for
trouble down the line.

Perhaps this is what you mean when you say that the Apple
IIgs bus power wiring is not sufficiently heavy duty.  It may have
been a good idea to double the power supply capacity and
seriously beef up the slot wiring for the IIgs, but it is hard to
fault the designer for meeting the spec, not exceeding in by
a sizeable factor.

>> >     The voltage through a line is always, to some extent, modulated
>> > by changes in the load. Besides the PS leads, the IIgs has many
>> > power lines (traces) which act as resistances across which voltage is
>> > modulated.
>> 
>> Yes, but these traces are factored into the design of the IIgs, and should
>> operate properly if not improperly loaded.
>
>     Except for the Slot card plane design (which is known to be defective),
>the other motherboard lines may be fine.
>
> .... 
>
>> 
>> Slot power decoupling is specified to be on the plug-in cards, not on
>> the main board.  You would not  want the decoupling capacitors across
>> a connector from the load--in this case, the logic on the card.
>
>     There is just so much you can achieve with on-card decoupling. The Slot
>plane +5 line is definitely inadequate for many setups. For some boards, the
>+12 line is probably inadequate, too.

The purpose of bypassing on a card is to eliminate noise from
the card from getting back onto the main board.  If all cards do
this, there is no problem.  If they don't, it's faulty card design,
not faulty main board design.

With the exception of cards that supply +12 power to an external
component, like the Disk ][ Controller, loading on +12 should be
minimal.  The Apple ][ spec is 250ma of +12 for _all_ cards.  (This
is why the disk controller never turns on more than one disk at a 
time, and why +12 is loaded only when a disk is operating.)

>> > e.g. replace the chief power supply leads-- then, 
>> > motherboard bypassing may be roughly adequate. Jumpering the
>> > main power lines on the Slot plane-- e.g. from the PS connector
>> >t o Slots 3 and 7-- further reduces noise generation.
>> 
>> Again, this indicates that either the cards are not adequately bypassed, or
>> the transient loads are outside of spec, or both.  I suspect that some
>> accelerators (particularly the TransWarp) are
>> considerably outside the slot power spec, and
>> direct-wiring the power may be compensating for this design problem.
>> 
>
>     Most likely. This still leaves the user with the problem of dealing with
a
>serious Apple design error. People were plugging power gobbler cards into
>Apple
>II Slots years before the IIgs came along. Anyone who's ever designed a
>computer (that works) should have known the Slot power traces would be
>inadequate. 
>
>     Similarly, the power supply should come with heavier leads for +5, +12,
>and Ground. 

That may have been nice, but it certainly would have gone well beyond the
"agreement" that Apple had with card designers.

-michael

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