Can IDE cable's end connector be left idle ( w-o affecting data transfer ) ?

[Ron Martell]

In fact cable select is the only reason for this. There is no other
one. If you jumper the devices to slave/master you can have them in
arbitrary order, number and position on these cables.

On more that one occasion, because of the layout of the motherboard
and the design of the computer case, I have connected drives using the
middle connector for the motherboard with a drive at each end. 40
wire cables with master/slave jumpers set appropriately, of course.
But zero problems with using the drives.

Ron Martell Duncan B.C. Canada
--
Microsoft MVP (1997 - 2006)
On-Line Help Computer Service
http://onlinehelp.bc.ca

"Anyone who thinks that they are too small to make a difference
has never been in bed with a mosquito."

 

[Folkert Rienstra]

Same for 40 pin 40 wire cables.


In fact cable select is the only reason for this. There is no other
one. If you jumper the devices to slave/master you can have them in
arbitrary order, number and position on these cables.

With cable select the device at the end is the master (during
boot-up) and the device in the middle the slave. Some devices will
fail to start or have problems if they are used as slave and no master
is present.

AFAIK

Which is zip.

this is more of a historic problem today.

Nope, most if not all harddrives still have jumpers for that situation.

 

[Folkert Rienstra]

Stop feeding the swiss troll.

You are correct that *on a single IDE channel*, the device arrangement
can be Master/Slave or Slave/Master or single Master or single Slave.
What is not allowed by the IDE controller, though, is Master/Master and
Slave/Slave on the same channel. But on different channels, you could
have ch.0 Master/ ch. 1 Master and ch. 0 Slave/ ch. 1 Slave and
ch. 0 Slave/Master/ch. 1 Master/Slave. The arrangement prohibited is
having the same jumpering for both devices on the same IDE channel,
i.e. on the same cable. That is imposed by the controller's need to be
able to tell the two devices apart which are on the same cable. I think
that's what you intended to express, but it wasn't clear.

And the implications go farther: A Slave HD can be the "boot drive"
just as well as well as a Master HD. All that is required is that the HD
be at the head of the HD boot order. The default settings (or settings in
BIOSes which don't enable adjustment of the HD boot order) put the
Master on ch. 0 at the head of the HD boot order. But this can be
changed by readjustment of the HD boot order. Indeed, it's quite
legitimate to have the Slave on ch. 1 (the secondary IDE channel) be
the "boot drive".

*TimDaniels*

 

[Arno Wagner]

Wrong, it reduces the 'signal echoes' which are actually
the sharp edges getting reflected off the impedance
discontinuity. Thats reduced with a drive on the end.

Irrelevant. The drive has a different impedance to no drive.

Very relevant. An unterminated drive connected to the
end actually makes things worse for the device in the
middle.

Arno

 

[Arno Wagner]

If the input impedance of the device matches that of the
cable, there will be no reflection. That is why high-freq comm
receiver designers try to match the input impedance to the
cable impedance. Are you saying that the input impedance
of ATA devices do not match the impedance of IDE ribbon
cable?

You are entirely correct on how signal termination works.

The input inpedance of an ATA device is a CMOS input plus ESD
protection. Some mega Ohm or more. The ATA ribbon cable has
an impedance of 200 Ohm, if I remember correctly. No termination
effect at all.

For the device in the middle that is how it should be.
Termination in the middle of a signal path is very bad. But
the end-device has no terminator it turns on. I suspect that
at some time that was planned, but at least the last time
I looked at the physical ATA bus characteristics, there was
no mention of it anywhere.

Arno

 

[Arno Wagner]

You are correct that *on a single IDE channel*, the device arrangement
can be Master/Slave or Slave/Master or single Master or single Slave.
What is not allowed by the IDE controller, though, is Master/Master and
Slave/Slave on the same channel. But on different channels, you could
have ch.0 Master/ ch. 1 Master and ch. 0 Slave/ ch. 1 Slave and
ch. 0 Slave/Master/ch. 1 Master/Slave. The arrangement prohibited is
having the same jumpering for both devices on the same IDE channel,
i.e. on the same cable. That is imposed by the controller's need to be
able to tell the two devices apart which are on the same cable. I think
that's what you intended to express, but it wasn't clear.

Sorry. Yes, that was what I wanted to say. Of course there is not
physical interaction between multiple IDE busses / channels (of which
you incidentially can have more than two or only one).

And the implications go farther: A Slave HD can be the "boot drive"
just as well as well as a Master HD. All that is required is that the HD
be at the head of the HD boot order. The default settings (or settings in
BIOSes which don't enable adjustment of the HD boot order) put the
Master on ch. 0 at the head of the HD boot order. But this can be
changed by readjustment of the HD boot order. Indeed, it's quite
legitimate to have the Slave on ch. 1 (the secondary IDE channel) be
the "boot drive".

Again true. It is a start-up issue for the device detection only. Not
for booting. The Master is supposed to be detected first. Then the
slave. After that both are identical, except that they have different
select signals. And yes, the BIOS is free to assign them any order
wanted.

Arno

 

[Arno Wagner]

On more that one occasion, because of the layout of the motherboard
and the design of the computer case, I have connected drives using the
middle connector for the motherboard with a drive at each end. 40
wire cables with master/slave jumpers set appropriately, of course.
But zero problems with using the drives.

Interesting.

Arno

 

[Eric Gisin]

The input inpedance of an ATA device is a CMOS input plus ESD
protection. Some mega Ohm or more. The ATA ribbon cable has
an impedance of 200 Ohm, if I remember correctly. No termination
effect at all.

You are a blithering idiot. ATA says there are resistors on all inputs.

For the device in the middle that is how it should be.
Termination in the middle of a signal path is very bad. But
the end-device has no terminator it turns on. I suspect that
at some time that was planned, but at least the last time
I looked at the physical ATA bus characteristics, there was
no mention of it anywhere.

Both devices are terminated. Read the ****ing specs, you moron.

 

[Arno Wagner]

When did you study electrical engineering (if at all)?

EE is only a hobby of mine.

For those frequencies, attached piece of cable does not behave like
nothing.

Trye. They make matters worse.

And YES, there is a termination as described by T13:
"4.2.2.2 Series termination required for Ultra DMA"
Otherwise, signal reflections will cause increased and substantial
interference.

Standard documents disagree with you:

[... disagreeng part snipped ...]

You can snip whatever you like,

What are you insinuating?

T13 document is still there:
http://www.t13.org/docs2002/d1410r3b.pdf

Ok, they did put in some form of termination since I last read the
standard. It is a _really_ screwed up form of termination, since each
device (i.e. also that in the middle) is terminated. It is also the
el-cheapo form, where you just have a resistor to ground or Vcc.

I honestly can't say what the effect of putting a device in the middle
only with this sort of termination would be. And it might actually make
maters worse in a 2 device configuration for the device at the end,
compared to no termination.

But I guess, that if you are experiencing problems with a single
device in the middle of the cable, then you should move it.
Personally I have 22 disks running UDMA5 that are connected in the
middle (no device at the end) and have not produced any CRC errors
in years of operation.

Arno

 

[Oscar]

Very relevant.
Nope.

An unterminated drive connected to the end actually
makes things worse for the device in the middle.

Nope. The impedance of that drive is
better reflections wise than no drive at all.

 

[Oscar]

You are entirely correct on how signal termination works.

The input inpedance of an ATA device is a CMOS
input plus ESD protection. Some mega Ohm or more.

Its more complicated than that with most
of the lines that arent pure receivers.

The ATA ribbon cable has an impedance of 200 Ohm,
if I remember correctly. No termination effect at all.

Wrong, the drive presents an impedance
that is different to no drive at all.

For the device in the middle that is how it should be.
Termination in the middle of a signal path is very bad.

Its more complicated than that too if termination is used.

But the end-device has no terminator it turns on.

It has an impedance anyway.

I suspect that at some time that was planned,

Not a shred of evidence of that. ATA has
always been an unterminated system.

but at least the last time I looked at the physical ATA bus
characteristics, there was no mention of it anywhere.

Because it has always been an unterminated system.

That does NOT mean that the drive impedance isnt relevant tho.

 

[Peter]

But

the end-device has no terminator it turns on. I suspect that
at some time that was planned, but at least the last time
I looked at the physical ATA bus characteristics, there was
no mention of it anywhere.

Arno

Look again:
http://www.t10.org/t13/docs2004/e04106r0.pdf
Page 21 and 22.
"Series termination" "Device termination" or "Recipient termination" is the
key term.
You may also look at Figure E.27.

 

[Peter]

First, IDE cables are not terminated. That means that for the drive
on

When did you study electrical engineering (if at all)?

EE is only a hobby of mine.

For those frequencies, attached piece of cable does not behave like
nothing.

Trye. They make matters worse.

And YES, there is a termination as described by T13:
"4.2.2.2 Series termination required for Ultra DMA"
Otherwise, signal reflections will cause increased and substantial
interference.

Standard documents disagree with you:
[... disagreeng part snipped ...]

You can snip whatever you like,

What are you insinuating?

What do you think?

Ok, they did put in some form of termination since I last read the
standard. It is a _really_ screwed up form of termination, since each
device (i.e. also that in the middle) is terminated. It is also the
el-cheapo form,

Here I agree. It is cheap. Much cheaper than SCSI termination.

where you just have a resistor to ground or Vcc.

That is something else...

I honestly can't say what the effect of putting a device in the middle
only with this sort of termination would be. And it might actually make
maters worse in a 2 device configuration for the device at the end,
compared to no termination.

compared to no termination (in the middle of the cable).

But I guess, that if you are experiencing problems with a single
device in the middle of the cable, then you should move it.
Personally I have 22 disks running UDMA5 that are connected in the
middle (no device at the end) and have not produced any CRC errors
in years of operation.

Only 22? Yeah, sure, they came installed like that.

Arno

Good luck to you.

 

[Folkert Rienstra]

And it shows.

For those frequencies, attached piece of cable does not behave like
nothing.

Trye. They make matters worse.

And YES, there is a termination as described by T13:
"4.2.2.2 Series termination required for Ultra DMA"
Otherwise, signal reflections will cause increased and substantial
interference.

Standard documents disagree with you:
[... disagreeng part snipped ...]

You can snip whatever you like,

What are you insinuating?

What do you think?

That's the problem, it doesn't. Not capable of it. It just babbles.

Which must have been half a decade ago.

It is the type that works without having to change the standard
too much and still allow old and new to work together.

It is also the el-cheapo form,

Here I agree. It is cheap.

Much cheaper than SCSI termination.

SCSI termination is cheap as well.
They just let us bleed through the nose for it.

That is something else...
Oh?

Like we and T13 are interested on what you have to say on the subject.

It has been working for years already, no reason for you to re-invent it
again, Babble.

compared to no termination (in the middle of the cable).

Yes Babble, that is what the spec says.

Only 22? Yeah, sure, they came installed like that.


Good luck to you.

Ostriches don't need luck. They just stick their heads in the sand.

 

[Folkert Rienstra]

Yes, that was what I wanted to say.

And that's what you said. Too stupid even to understand it's own ramblings.

Of course there is not physical interaction between multiple IDE busses /
channels (of which you incidentially can have more than two or only one).

Babble, babble, rant.

Again true.

It is a start-up issue for the device detection only.

Nope. Device initialization at power-up or reset only.

Not for booting.

The Master is supposed to be detected first. Then the slave.

Completely irrelevant.

After that both are identical, except that they have different
select signals.

No they don't.

 

[Folkert Rienstra]

Interesting.

In that it shows how you don't have the faintest clue.

 

[Jeff Richards]

Your comments are confused. There's no such thing, in SCSI or IDE, as a
terminated or unterminated device. There's a device, a terminator, or no
device. The terminator is a cheap way of pretending some sort of device is
connected, and it's important for SCSI mainly because of the cable lengths
involved. It doesn't really matter for IDE because the impact of an
unterminated connector on the signal quality (ringing, overshoot) is much
less with the shorter lengths involved, and the IDE interface can cope with
it. But there is still a difference in signal quality between an IDE cable
with a device connected and one without.

An 'Inactive' IDE device is not particularly high resistance, and certainly
has an impact on signal quality compared with an open connector.

 

[Arno Wagner]

Look again:
http://www.t10.org/t13/docs2004/e04106r0.pdf
Page 21 and 22.
"Series termination" "Device termination" or "Recipient termination" is the
key term.
You may also look at Figure E.27.

I looked. There is no bus termination, but apparently a device
termination was added. How this plays out if you have a UDMA
and a non-UDMA capable device on the same bus, I have no clue.

Arno

 

[Arno Wagner]

Your comments are confused. There's no such thing, in SCSI or IDE, as a
terminated or unterminated device. There's a device, a terminator, or no
device. The terminator is a cheap way of pretending some sort of device is
connected, and it's important for SCSI mainly because of the cable lengths
involved. It doesn't really matter for IDE because the impact of an
unterminated connector on the signal quality (ringing, overshoot) is much
less with the shorter lengths involved, and the IDE interface can cope with
it. But there is still a difference in signal quality between an IDE cable
with a device connected and one without.

An 'Inactive' IDE device is not particularly high resistance, and certainly
has an impact on signal quality compared with an open connector.

SCSI is complicated today. U2W cables are actively terminated. Older
standards allow terminated devices (i.e. devices with internal
termion) as well as external terminators.

As it turns out UDMA capable devices has some sort of half-assed
device terminaton, i.e. _every_ device on the bus has some pull-up
or pull-down resistirs. Quite inelegant IMO.

Arno

 

[Arno Wagner]

First, IDE cables are not terminated. That means that for the drive on
the middle connector it makes no difference whether there is a drive
at the end or not! If anybody says different, then they do not
understand the subject matter.

Thats your personal opinion.

No, that is electrical engineering. Read any text on digital signal
transmission of copper wires. And T13 does specify the inputs
in a way that no termination at all is going on.

When did you study electrical engineering (if at all)?

EE is only a hobby of mine.

For those frequencies, attached piece of cable does not behave like
nothing.

Trye. They make matters worse.

And YES, there is a termination as described by T13:
"4.2.2.2 Series termination required for Ultra DMA"
Otherwise, signal reflections will cause increased and substantial
interference.

Standard documents disagree with you:
[... disagreeng part snipped ...]

You can snip whatever you like,

What are you insinuating?

What do you think?

That I somehow tried to hide something? Would I have announced what
I snipped, if it was for anything else than brevity?

Here I agree. It is cheap. Much cheaper than SCSI termination.

That is something else...

No, that is the device termination as in the standard. At least
as in the version you cite. It also has a non too clear schematic,
where there are two resistors in the data lines, one pull-down
near the logic input and a series resistor between cable and the
logic input. The tables do only specify one of these resistors.
Looks kind of fishy to me.

compared to no termination (in the middle of the cable).
Yes.

Only 22? Yeah, sure, they came installed like that.

Actually they did not. They came in a large syrofoam HDD shipping
crate.

Arno