RE: Broken TED for visual6502.org

You hit on a primary mechanism; the higher the heat, the higher the
resistance, the higher the heat...

A s far as initial cause it could also have been "punched",  either a
voltge/current spike or an slight imperfection starts a hot spot and
breakdown in insulation.

Early failures (I.E. 20 years ago...lol) could have been caused by a punch
due to bad handling 1-2 months earlier.

NMOS and HMOS and rule changes were usually done by sizing in generally,
as in reduction ratio at photomask, we stated it as the L or length of the
(average) gate.  I think there would be other chip-wide changes for a
shrink or N to H such as the backbias generator geometry.  With that said
I am going to shoot an email to the same chip guy that answered last time
and see what he remembers.


-----Original Message-----
From: owner-cbm-hackers@musoftware.de
[mailto:owner-cbm-hackers@musoftware.de] On Behalf Of Segher Boessenkool
Sent: Friday, July 20, 2012 3:16 AM
To: cbm-hackers@musoftware.de
Subject: Re: Broken TED for visual6502.org

>>> Could be... But what caused the driver for A7 to die in that fashion
>>> though? I should have tested all the signals before mailing in the
>>> CPU to see if any of them were dead.
>>
>> A short circuit on the pin: if something (external) drives the
>> address line high while the CPU pulls it low, a nice big large
>> current goes through that big FET where the metal turned black.  It's
>> not built for such large currents, it's too high resistance for that,
>> you get heat, stuff melts and whatnot.
>
> The thing is, there have been external memory expansions for the
> C16 that must have done exactly that with the _CAS line from TED
> without killing it outright. The default is that the RAM in the C16 is
> mirrored 4 times and there is no way on the expansion port to disable
> it. The only way to add another 16 KB on an external board is to force
> _CAS for the internal RAM high at the proper address space and
> generate your own _CAS for the RAM on the board.

I dunno.  Maybe they limit the current to some "safe" value.

The other thing is, as soon as the output FET is damaged it will have
higher resistance, accelerating the process.  Same thing as happens with
electromigration on metal tracks: small failure leads to big failure.  It
could well be that this failure started through something else (your bad
passivation theory, for example); but it looks to me like it ended with a
nice big current frying the thing.

>> Sure, I'd love to see those as well, but the 6510 has bigger
>> differences to both endpoints we have pics for so far (6502 and
>> 8501).  And it's historically more interesting.
>
> I expect it to be quite similiar to the 8501, just in NMOS since I
> doubt they did a full relayout of the CPU from NMOS to HMOS.

We _know_ that the CPU core was redone somewhere between 6502 and 8501.
I find it more likely that it was done for the 8500 than that it was
already done for the 6510; esp. if you look at the actual layout on the
8501, all the transistors are resized, so if it was already redone for the
6510 it would have had to be redone twice.

We also know that all of the pad/drivers stuff is redone for HMOS: it has
to be much bigger relative to the core logic; you can see this on e.g. the
SID and VIC-II chips.

I don't expect the 8501 to differ much from the 8500 at all.  So if I had
to choose, I'd much rather see the 6510 than the 8500.  But we don't have
to choose, we'll see all of-em eventually :-)

Oh, and 7501...  :-)


Segher


       Message was sent through the cbm-hackers mailing list

       Message was sent through the cbm-hackers mailing list