8-bit IDE

From: Richard Erlacher <edick_at_idcomm.com>
Date: Mon Apr 17 09:52:30 2000

----- Original Message -----
From: <allisonp_at_world.std.com>
To: <classiccmp_at_classiccmp.org>
Sent: Monday, April 17, 2000 7:12 AM
Subject: Re: 8-bit IDE


> Took my existing 16bit interfaced IDE and wrote a pattern in sector 0,
> head 0, cylinder 0 or 0->ffh repeatedly till I filled the sector. Then I
> disabled the word translation logic, tried setting the bit in a read op
> and I got 0,2,4,6.... didn't work.
> > > Well wishful thinking had me check it out using several 85-130mb
> > > (quantum, Seagate, maxtor, WD) and none seem to do that. After all
> > > having that would make the interface a no brainer and save a simple
> > > for splitting read and writes. However, it was wishful thinking.
> > >
> > What???
> I tried to interface using only 8bit data and using the 8-bit IO bit and
> it didn't work. It was wishful thinking I could use it.
This has to be set up as a feature in the register setup when the interface
is initialized. The unconditioned interface defaults to 16-bit mode.
> > I personally like the CMOS much better since it drives harder, and since
> > pulls and pushes with the same impedance, unlike TTL which sinks 16 and
> > sources 1.6 mA. I've tried replacing all the bus interface buffers on
> > old S-100 cards with AC logic. In some cases I used HC or AHCT
> I've seen latchup on busses that ring negative using HC and HCT parts.
> I'd like the better drive but they would randomly flame on me due to the
> bus ringing. Obviously loading the bus with terminators would solve this
> but it's still something that worried me and made for a less robust card
> for handling and ESD.
Latchup most frequently occurs when the device is able to draw power through
its inputs before it can draw power through its power connections. That's
been a recognized problem with CMOS for over 20 years and the correct way to
avoid it is through proper design of system timing and power distribution
and filtering. Often a resistor in series with the offending signal is
required to delay the flow of current into the CMOS device's input. It is
quite common in environments where the technology is mixed, i.e. TTL and
CMOS hardware in the same circuit. I fix these problems by leaving out the
TTL, LSTTL, STTL, etc and sticking with CMOS. CMOS is faster, easier on the
power supply, and noise-immune.
> Allison
Received on Mon Apr 17 2000 - 09:52:30 BST

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:32:41 BST