EIDE Cable Specs

[GreasyItalianAdmin]

What features should one look for in the zillion or so EIDE cables/mfgs.
out there to ensure fast/reliable data transfer for DMA 66+ hard drives?
Thanks for your opinions.

 

[NunYa Bidness]

What features should one look for in the zillion or so EIDE cables/mfgs.
out there to ensure fast/reliable data transfer for DMA 66+ hard drives?
Thanks for your opinions.

DMA 66 is in the UDMA family of EIDE transfer modes.

It will function with a standard 40 pin cable or even better with
the enhanced 80 pin UDMA cable.

The master goes on the end connector. The blue connector goes into
the MOBO, and the center connector is for the slave drive.

Then, the only remaining requirement is that you plug it into a UDMA
EIDE interface connector on the motherboard or add-in card.

Features would be if they actually conformed to the color coding
spec as it relates to the connectors. Like blue for the interface
end, gray for the slave, and black on the other end for the master.

Other features these days are that they have them bundled into
"round" cables now. A bit more pricey, but yield less airflow
restriction and a bit less electrical noise.

 

[John Gilmer]

Other features these days are that they have them bundled into
"round" cables now. A bit more pricey, but yield less airflow
restriction and a bit less electrical noise.

I have always had mixed "emotions" over those flat cables.

I first say them used in the old DEC computers. My first impression is
that is seemed SLOPPY! But when I think about it and realize that every
other wire is a virtual ground (usually) I have come around to thinking that
I would be hard pressed to come up with a "round" cable that I could
guarantee would work as well or better.

 

[krw]

I have always had mixed "emotions" over those flat cables.

I first say them used in the old DEC computers. My first impression is
that is seemed SLOPPY! But when I think about it and realize that every
other wire is a virtual ground (usually) I have come around to thinking that
I would be hard pressed to come up with a "round" cable that I could
guarantee would work as well or better.

I did a lot of work with the IDC cables many years back. As long as
every-other pin is ground, they're quite good, electrically. IDE
cables don't have enough grounds though and are rather poor (but good
'nuff). The 80-conductor cables are better, but still not enough
ground pins to make a great transmission environment. Although again,
good enough for the application.

OTOH, I've been using round IDE/ATA cables because they make routing
the cables so much easier. I haven't studied their properties, though
they too work "good enough".

 

[[email protected]]

What features should one look for in the zillion or so EIDE cables/mfgs.
out there to ensure fast/reliable data transfer for DMA 66+ hard drives?
Thanks for your opinions.

I have some 6' braided cables (4 ½' longer than the spec) and they
work fine on DMA66 drives. This is really a pretty robust interface if
you have good cables.

 

[GreasyItalianAdmin]

What features should one look for in the zillion or so EIDE
cables/mfgs. out there to ensure fast/reliable data transfer for DMA
66+ hard drives? Thanks for your opinions.

Thanks to all the good responses.
Quickly reading the replies, maybe I was not clear. What I was wondering
specifically is what type of construction is best to eliminate errors?
Some of the companies advertise various shielding materials and being
ignorant in this area, I was wondering what construction materials are
best and how do I separate the cheaply made stuff from the good stuff?
The regular flat ribbon cables in the machine presently (40 wire, I
think) are all twisted up and I think some of the sharp bends were
giving me multibit errors-I tried to restraighten them and the multibit
errors went away. Also I have read from some articles on the net that
the round cables do not conform to the ATA standards? Also suprised to
hear from one poster that he used such a long cable, everything I've
read says no more than 18 inches.

 

[NunYa Bidness]

The master/slave plugs are there so that you can use cable
select settings on the drive(s). If you don't use cable
select, you can stick either drive on either connector -
which sometimes makes cable routing easier.

Interesting. I thought it was an issue surrounding the fact that
the "controller" actually resides on the hard drive chosen as master,
and its location of the "bus" the cable offers.

All these years that we have been actually selecting the drive
assignment, when we could have been placing the selector on csel the
whole time.

Query though: When one makes both drives set to cable select, the
cable position determines the master?

Is that how I read it? Thanks.

 

[NunYa Bidness]

I have always had mixed "emotions" over those flat cables.

I first say them used in the old DEC computers. My first impression is
that is seemed SLOPPY! But when I think about it and realize that every
other wire is a virtual ground (usually) I have come around to thinking that
I would be hard pressed to come up with a "round" cable that I could
guarantee would work as well or better.

IIRC those rounded versions are ribbons that have been tucked into a
tube. I do not know how a 40 or 80 strand round configuration
conductor (shirt of twisted pair configurations) could work at the
data rates given for the differentially de-coupled "noise free" ribbon
cables.

 

[NunYa Bidness]

I did a lot of work with the IDC cables many years back. As long as
every-other pin is ground, they're quite good, electrically.

That rules ALL 40 pin cables out.

IDE
cables don't have enough grounds though and are rather poor (but good
'nuff).

They only have one at the 40 pin level.

The 80-conductor cables are better, but still not enough
ground pins to make a great transmission environment.

Huh? Try every other conductor.

Although again,
good enough for the application.

Most assuredly, since it DOES "have enough grounds".

OTOH, I've been using round IDE/ATA cables because they make routing
the cables so much easier. I haven't studied their properties, though
they too work "good enough".

A simple test is to run a hard drive throughput benchmark with a 40,
80, and round cable configuration, log the results, and examine the
log.

 

[NunYa Bidness]

I have some 6' braided cables (4 ½' longer than the spec) and they
work fine on DMA66 drives. This is really a pretty robust interface if
you have good cables.

Braided? Wouldn't that be "twisted pairs"? Speaker wires got
braided, but ribbon cables get twisted pair configurations.

The new step these days are differential pairing.

 

[NunYa Bidness]

Thanks to all the good responses.
Quickly reading the replies, maybe I was not clear. What I was wondering
specifically is what type of construction is best to eliminate errors?
Some of the companies advertise various shielding materials and being
ignorant in this area, I was wondering what construction materials are
best and how do I separate the cheaply made stuff from the good stuff?
The regular flat ribbon cables in the machine presently (40 wire, I
think) are all twisted up and I think some of the sharp bends were
giving me multibit errors-I tried to restraighten them and the multibit
errors went away. Also I have read from some articles on the net that
the round cables do not conform to the ATA standards? Also suprised to
hear from one poster that he used such a long cable, everything I've
read says no more than 18 inches.

I noticed that the 6 foot spec he mentioned I have not seen
elsewhere. Maybe for the old IDE spec, but UDMA certainly has very
short limits.

IF your motherboard supports UDMA 66 then you should replace all of
your 40 pin cabling with 80 pin UDMA cables. They will both eliminate
the errors, and allow the auto sensed speed capacity to get jacked up.

Your problems will disappear!

 

[NunYa Bidness]

I did that and yet my garden still floods every Winter...

Yet you still make me desire your companionship for... ever!

I love girls that are smarter than I am. That is also why I have
been lonely for decades too!

 

[[email protected]]

Braided? Wouldn't that be "twisted pairs"? Speaker wires got
braided, but ribbon cables get twisted pair configurations.

Yup they do act like twisted pairs. These were originally used on IBM
3480 tape drives but they had the same connector so I tried them.
Originally I used them on a CD drive but later I put an external hard
drive out there. This is on my MP3 player machine out in the pool bar,
the system unit is in the garage. Cables go through a hole in the
wall.

 

[[email protected]]

I suspect that he meant a braided sheath.

Nope, grounds are braided into the signal cables. They also use bigger
wire than a usual IDE

 

[krw]

That rules ALL 40 pin cables out.

Not if there are 20 signals and 20 grounds. Note that I said IDC
cables. I did *not* say IDE cables.

They only have one at the 40 pin level.

Not according to my information. Pins 2, 19, 22, 24, 26, 30, and 40
are ground. Note that there is a ground next to the strobes (23 =
/IOW, 25 = /IOR, 31 = IRQ).

Huh? Try every other conductor.

I said "pins", not conductors.

Most assuredly, since it DOES "have enough grounds".

No, it really doesn't for a good transmission environment. More ground
*PINS* are needed. It's "good enough" for IDE, since there are few
signals that are really important (data and address lines are allowed
to settle before being qualified by a strobe).

A simple test is to run a hard drive throughput benchmark with a 40,
80, and round cable configuration, log the results, and examine the
log.

I prefer to look at signals with a scope to see how good the interface
is. A TDR on the cable is also useful.

 

[NunYa Bidness]

Not if there are 20 signals and 20 grounds. Note that I said IDC
cables. I did *not* say IDE cables.

Even those would be dependent on how the PCB layout person laid out
the board. He may not have used it that way. it wasn't required in
any sense. Differential pairs weren't in wide practice (in the PC
realm anyway) until recently with EIDE. It has been around a while in
SCSI, and known about in parallel conductor data pass design for a
long time. They have allowed for higher speed signaling without
reflections or crosstalk over longer passes. That's cool. Speed is
good. :-]

You are right though, configured the way you bespeak.

Maybe we can have those new vertically oriented (magnetically) hard
drives at 2.5" form factor or such, and have an array of them, all hot
swappable on SATA passes, and have a huge parallel set of SATA
conductors, all segregated from each other that way.

A person could wear a belt that has a fault tolerant, one terabyte
RAID file server on it, field repairable, connected to a battery, and
backpack laptop. Everyone could wear a live video recorder!

 

[NunYa Bidness]

I said "pins", not conductors.

Yes, but I said conductors, as we are talking about successive
parallel conductors on a ribbon, before the header attachment. That
is where the noise cancellation takes place.

 

[keith]

Even those would be dependent on how the PCB layout person laid out
the board.

One doesn't just lett the PCB layout person do anything he wants! There
are guidelines, groundrules, and quick-kicks to the butt!

He may not have used it that way. it wasn't required in
any sense.

It was on any work *I* did, whether I specified the transmission
interfaces or designed the entire system! The layout guy never did
anything outside the spec. Good grief! How do you think things are done?

Differential pairs weren't in wide practice (in the PC
realm anyway) until recently with EIDE.

The PC is hardly the end-all. ...and EIDE *still* doesn't use differntial
pairs. BTW, who was talking about differential pairs?

It has been around a while in
SCSI, and known about in parallel conductor data pass design for a long
time. They have allowed for higher speed signaling without reflections
or crosstalk over longer passes. That's cool. Speed is good. :-]

Sure, but much SCSI wasn't differential either. Better than IDE, but not
all differential.

You are right though, configured the way you bespeak.

I know. ;-) I did several interfaces with 50-pin IDC cables with
10 connectors spaced at .75". The quality of the transmission environment
really surprised me, but was a little low (80ish ohms, rather than about
110 for long cables, IIRC).

Maybe we can have those new vertically oriented (magnetically) hard
drives at 2.5" form factor or such, and have an array of them, all hot
swappable on SATA passes, and have a huge parallel set of SATA
conductors, all segregated from each other that way.

Why?

BTW, I'm less than impressed by SATA. There is little gain. If only
they'd packaged the power in with the signal cable...

A person could wear a belt that has a fault tolerant, one terabyte
RAID file server on it, field repairable, connected to a battery, and
backpack laptop. Everyone could wear a live video recorder!

Why? I'm happy with a couple of sub gigabyte USB sticks on my keychain.
....then again, I'm not much into video (radio on a stick, OTOH... ;-).

 

[keith]

Yes, but I said conductors, as we are talking about successive
parallel conductors on a ribbon, before the header attachment. That
is where the noise cancellation takes place.

Wrong. Without PINS to carry those ground currents, most of the benefit
of the ground conductors is lost. Inductance, Luke.

 

[NunYa Bidness]

The PC is hardly the end-all. ...and EIDE *still* doesn't use differntial
pairs. BTW, who was talking about differential pairs?

UDMA cables are differentially paired lines.