Harddrives, IDE Cables and Jumpers

[hongman]

Hi all

I heard a rumour, and I wanted to check ot its integrity.

I got told that if you set an IDE device as master by the jumper, you HAVE to put it on the end connector on the IDE cable. And if you set a device as slave on the IDE Device, it HAS to be connected to the middle IDE Connector.

Is this true? I would think it makes sense especiallyin the case of CS, but I wasnt sure.

Thanks all

Hong

[ski]

This may help to answer your questions:
http://www.pcguide.com/ref/hdd/if/ide/confCS-c.html

[hongman]

Thanks Ski, that was a very good read.

But I still have the question... if you use a traditional IDE cable (not a y shaped one or sorts) and jumpered a drive as master, then stuck it on the middle connector, would it matter?

I'm sorry if that link answered this, but I read it and although it provded great info on CS, this question still remains.

Thank

Hong

[ski]

You're welcome, hongman.

I believe the following blurb from that article may answer your question:

"For signaling reasons, it's best to put a single drive at the end of a cable, not put it in the middle leaving a "stub" of wire hanging off the end of the channel."

[hongman]

So is that a 'YEs' to my question?

Sorry if I'm being stupid. I can see that you should put a single drive at the end, for reasons detailed below...but would it cause a config problem if I was to jumper:

1 drive as master and connect it to the middle
1 drive as slave and connect it to the end

of the IDE cable...

[Philgo]

hongman, they will normally work fine either way if jumpered as Master and slave. If they are CS then the end one will be treated as Master.

[hongman]

Thanks.

Ha, looks like my A+ tutor was wrong yet again... (ok and me but at least I can admit it!)

Thanks

Hong

[DrMDJ]

The "official" ata specs dealing with drives that will use UDMA modes 3 and up (aka ATA66 and up) and therefore require a 40-pin/80-wire cable specify that the master drive (in a master/slave configuration) be at the end of the cable and the slave be in the middle. The main reason for this is as ski quoted, so as to prevent signaling issues/interference that might be caused by having the cable end hanging (a stub) if the master was in the middle (in a single drive configuration). The overall idea here is kind of along the lines of scsi termination. There's also a less technical reason for the ata spec recommendation.

BUT, as with a lot of things there are specs and there are reality. In the case here the reality is that it would be the rare case (I've certainly never encountered it) where the the above spec really needs to be adhered to. Either in a single or dual drive setup the drives will work fine no matter where on the cable the master is and where on the cable the slave is.

PS. The specs also say that the maximum cable length can only be 18 inches. Here again this "rule" gets violated countless times with no ill effect.

[hongman]

thanks for confirmin that doc!

What a wealth of knowledge

[shiva_42]

Does anyone know why the specs even suggest that the cable length be limited to 18 inches?

If it's not an impedence issue, I can't imagine why it would matter. If it's a "loss of signal" equation, shouldn't using a round shielded cable offset this?

Just curious... my cables are round shielded cables that are approximately 24", each with two devices attached, no known problems whatsoever...

[SuperSparks]

I think the issue is one of signal timing Shiva - as the cables become longer so the data signals in the parallel cables become more and more out of sync. That's why high-speed serial interfaces have taken over from parallel over the years.

I've never heard of anyone having a problem with a 24" cable, but it's probably pushing things to go much longer than that.

[DrMDJ]

The 18-inch cable limit (which could actually end up being less because this related to the full electrical path length) started with the ATA-2 standard I believe (pre ultra ata). It is my understanding (I admit I have not looked at the spec in some time because I don't like getting the headaches such things can produce ) that given the speeds they were trying to obtain at that time, with the 40 wire cables being used (unsheilded and unterminated) there weren't enough (or they were afraid there wouldn't be enough) ground signals. As a result it was felt that if the length was allowed to go past 18 inches there would be both performance and data integrity issues.

With ultra-ata there was an increase in the number of ground wires in the cable. In addition the error detection/correction capabilities built in at both the hardware and non-hardware levels were changed/enhanced. So on one hand this took care of some of the justification for the original 18 inch limit. On the other hand though, I guess because of the increased speeds and to play it safe they decided to leave the 18 inch limit in the specs.

[DrMDJ]

By the way hong... This is another one of those cases where if your teacher says that a master must go at the end of and ide cable then he/she is absolutely 100% correct. At least until you are done with the class and have passed the test. After that then he/she can be considered dead wrong.

[hongman]

Hey doc, I know exactly what you mean!!

[Platypus]

The limiting factor with 40-wire cables was crosstalk - since there is no shielding earth wire between lines, too much length of wire running adjacent could couple a signal between data lines causing high error rates.

For 80-wire cables, which have an earth wire between signal lines to give some shielding against cross-coupling, the main limitation is signal rise and fall time.

A digital signal has a signal voltage that is either "low or "high" voltage to signal a logical 0 or 1. For a signal to be "clean" in a digital sense , the transition between these states can only remain in the indeterminate region (a voltage that is not recognised as either low or high) for a small percentage of the total cycle time. Otherwise the chance of data errors rises dramatically.

The circuitry driving the data lines can only supply a certain amount of current. Each line will inevitably have a certain amount of capacitance to ground, and the rise and fall time is governed by how long the available switching current takes to charge or discharge this capacitance. As the length of cable increases, the capacitance on each line will rise, so the signal gets less "true", as the signal starts to take too long to get from one state to the other, and again the likelihood of data errors rises.

18" is a "safe" measure, longer lengths are proven to still work satisfactorily, but the error correction may be working harder.