Re: [SI-LIST] : ATA source termination

About this list Date view Thread view Subject view Author view

From: Charles R. Patton (crpatton@pe.net)
Date: Mon May 01 2000 - 08:42:12 PDT


Prior to my retirement from Western Digital (WD), I spent some time
working on the ATA interface and I’d like to shed some anecdotal light
on the problem of ATA termination.

Historically, the IDE (now ATA) interface started out as an 8 bit bus
with interspersed ground wires, but early in its life those ground wires
were converted to the upper 8 bits to yield a 16 bit interface. That
loss of the interspersed ground wires is what makes the IDE cable
modeling so uncontrolled. A few points are:

1) The nominal impedance of 40 wire ribbon cable is around 80 ohms. But
this presupposes that the adjacent wires to the signal wire are
grounded. In point of fact they are driven signal lines. The
consequence of this is that the apparent impedance varies depending on
the signals in adjacent lines, i.e., the impedance is data pattern
sensitive, and can be much higher by a factor of 2 to 3 times when the
bus as a whole changes in the same direction. For this reason, a
minimum coupled line model should contain at least 5 lines to even begin
to get reasonable results. And even then, the model is never quite
right because it doesn’t model the lay of the cable against the
chassis. Sometimes the cable is dressed adjacent to the chassis while
other times it hangs randomly in the air – just look at the IDE cables
in your PC -- therefore the capacitance to ground is a random number in
practice. A modeling nightmare.

2) That variability together with the wide range of drivers being used
is, in part, why the suggested termination resistance is difficult to
pin down. We concluded that the then “in vogue” drivers were too strong
and should be weaker in order to increase their dynamic resistance for
better matching and lower slew rates to improve the crosscoupling
problem. Some of this philosophy made it into later ATA specs, but
legacy problems always muddied the works.

3) In early testing we found that the worst problem was a single line,
physically in the center of the group (D13 for instance) staying low,
while the rest of the bus transitioned high and low. This caused
maximum crosstalk, which leads to the next point.

4) The nominal electrical length of the cable is short and so one might
be led to believe that the reflections are over in a few nanoseconds,
but in point of fact, we consistently observed a 30 to 60 MHz ring
frequency in the signal lines excited by that crosstalk. This was
attributed to the end-to-end cable inductance resonating with the end
termination capacitance of some 10’s of pF. This ringing would often be
higher than the low threshold of TTL (remember this was a legacy
application.) This was always the BIG problem to solve.

5) As the interface specs were proposed by Quantum for ATA-33 and
ATA-66, this ringing was often right in the middle of the strobe window
which led WD to propose the 80 wire ATA cable solution. The extra 40
wires are grounded together with the already grounded pins in the
standard IDE interface in the connectors themselves to form the more
standard signal model of signal surrounded by grounds. Now the
impedance is more controlled and the crosscoupling is reduced below TTL
thresholds. We worked with several vendors to develop the needed
special connectors and cable and now this cable is in the ATA specs.

Charles R. Patton, Owner
Synergy Co.
“Creative, Cost Effective EMI, ESD and Analog Design Solutions”
21490 Camino Arriba
Murrieta, CA 92562
Phone: 909-698-9657
Fax: 909-698-0224
Email: charles.r.patton@ieee.org

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Nov 22 2000 - 10:50:09 PST