From: Sven Karlsson <svenka_at_it.kth.se>
Date: Tue, 22 Jan 2002 13:58:13 +0100

Hi Björn,

>I still have no idea how one manages to _destroy_ harddisks by simply
>writing bad things to the ATA interface. But clearly it's possible. :-(

What might have happened is that you accidentally were driving some
bi-directional IDE signals, e.g. the data bus, at the same time as the hard
drive is doing that. Depending on the strength and design of the drivers on
the hard drive, you could then potentially damage the buffers on the hard
drive, the buffers (the '245's) in the jukebox, or both. IIRC the only
bi-directional IDE signals are the data bus signals.

I'll try to explain how the buffers can be destroyed: Let's assume that the
IDE hard drive is driving a signal high, i.e., in this case 5V, while the
jukebox is driving it to low, i.e., close to ground. What happens in this
case is that the hard drive's buffer will output a very large electrical
current trying to raise a the voltage on the wire. In this process a couple
of transistors along the path of the current can be destroyed causing the
buffer to stop working which of course means that you will not be able to
communicate with the hard drive at all.

I have destroyed a few buffers, through software, during my years as a hobby
electronics designer. The trick to avoid that is usually to insert a few
wait states here and there in which the bi-directional signals just floats
and are not driven at all. This is sometimes very hard to achieve.

Best regards
 Sven
Received on 2002-01-22