SCSI termination question

[Patrick]

I'm a real novice when it comes to SCSI, so I'm hoping this is an easy question for someone to answer.

I'm setting up a friend's computer for DV editing. The motherboard is an Asus P2B-DS (with built in SCSI card) running Win98. The boot drive is IDE, and the two SCSI capture drives are identical 18 GB Seagate Barracudas (ST118273W). There is also (usually) a SCSI scanner connected on the end of the same SCSI cable at the back of the computer. I'm not sure what to do with the termination jumpers on pins 15 and 16 on each drive. Although the instructions at the Seagate site seem to indicate that the jumpers should be removed from the "middle" drive(s), both drives have apparently always had the termination jumpers installed while in this computer. I can't vouch for how well the drives or the system have worked up to this point, never having seen it run. All I know is that each drive took over an hour for me to format which seems pretty slow!

So I actually have two questions:

1) With everything (including scanner) connected, should the termination jumpers on both drives be removed?

2) When the scanner is NOT connected (as it is now because I don't have it here), does that make a difference whether or not the drive farthest from the "host" has its termination jumper removed?

Any replies appreciated...

[The PIT]

Re: SCSI termination question

Originally posted by Patrick


1) With everything (including scanner) connected, should the termination jumpers on both drives be removed?

The last drive in the chain should be terminated. The scanner will also need terminating.


2) When the scanner is NOT connected (as it is now because I don't have it here), does that make a difference whether or not the drive farthest from the "host" has its termination jumper removed?

The drive should should have it's terminator on.

Any replies appreciated...

[Patrick]

What took you so long...?

Thanks for the quick response!

Just so I understand completely, it sounds to me like I should remove the termination jumper from the drive closest to the SCSI "host" and leave the termination jumper on the drive that's farther away. And it also sounds like having the scanner connected or not makes no difference in regards to the termination jumpers on either drive. Have I got it right?

[VJ]

I suspect that the drives are internally connected and the scanner (obviously) externally. The config then looks like this :

scanner--host adapter--hdd1--hdd2

The both the scanner and hdd2 should be terminated, hdd1 should not be terminated. The ribboncable connecting the hostadapter, hdd1 and hdd2 should not have "dangling ends"; meaning that both ends of the cables should be used (there may be empty connectors in between, but no unused connectors past the last devices).

And now the killer : the scanner probably uses a narrow bus, while both drives use a wide bus. Therefore, when terminating the scanner, the upper 8 bits of the wide scsi-bus are not terminated. The host adapter should thus be set to either "automatic" termination, but as this can sometimes fail, it is possible to set the hostadapter to only terminate the high byte. So the setup looks actually like this (- means narrow bus, = means wide bus) :

scanner--hostadapter==hdd1==hdd2

where scanner and hdd2 are terminated (fully, but these don't offer a choice), and where the host adapter only terminates the high byte (either obtained automatically, or set manually).


If the hostadapter has a bios (which is not really requiered, as no scsi device is being used as boot device), this termination setting can be done in the bios (check manual, for Adaptec you should press ctrl-a when the message appears - just prior to detection of the devices). If the card has no bios, there most likely is a software tools that allows you to set this.

Hope this helps!


Jörg



PS: If the setup looks like this (with 2 external drives) :

hostadapter==hdd1==hdd2--scanner

Then the hostadapter and the scanner should be terminated. As most harddisks don't allow terminting only the high bus, a seperate high-bus terminator should be purchased (this is often incorporated in a wide/narrow converter, but not allways !).

[VJ]

I just noticed your reply:

Presence of the scanner is irrelevant for the termination of the drives.

It does matter for the termination of the host adapter, but if this one is set to "automatically", there should be no problem what so ever. If you need to set to only terminate the high byte, there could be trouble if the scanner is removed (lower byte is then no longer terminated)


Jörg

[Patrick]

Ummmm.....

Jörg, I need a simpler answer.

It appears from what you say that the presence of the scanner makes no difference. Do I then leave the termination jumper on the farthest drive and remove the termination jumper from the drive nearest the host? I'm a little confused as you can probably tell.

[VJ]

Euhm.... sorry....

1. Both the last devices on the chain need to be terminated, this means the harddisk that is last on the cable, and the scanner. Whether or not the scanner is present, does not make a difference for the jumpers of the harddrives.

2. The host adapter and the drives are connected with a ribboncable (a bit like an ide cable, but with perhaps more connectors). Both connectors at the end of the cable should be used (1 connected to the host adapter, and 1 to the harddisk that is terminated). The other harddisk can be hung on any connector inbetween.

So this is what it should look like on the inside of the computer :

h--u--u--u--dt

(u= unused connector, h=host adapter, dt=disk with termination, -= cable)

Any of the unused connectors can be used to connect the other harddisk (which must not be terminated).

This for instance is not allowed :

h--d--dt--u

The last connnector on the cable must be used (as in the previous example).

With these rules in mind, it should all work. You can connect/disconnect the scanner (if the system is not running), and the bus will remain correctly terminated.

In case of problems, then this is a possible cause :
The scanner probably uses an 8 bit bus, whereas the drives probably use a 16 bit bus (each individual wire inside the cable must be terminated). If this is the case, than the bus is not properly terminated. Normally the host adapter should be set to terminate its side of the bus automatically, which solves the problem (host adapter will terminate the 8 bits the stop at the host adapter, while letting the other 8 bit go to the scanner). Should the system pose problems, then it could be worth checking how the hostadapter is terminated (either in its bios, or in some supplied softwaretool).

Hope this clarifies things...

Jörg

[Greebe]

Terminate the only the scanner and also the furthest HD.

You should only terminate the furthest internal and furthest external device connected.

[VJ]

I forgot to mention this : the termination of the drive closest to the host adapter must be off (so you'll need to remove the jumper)

[VJ]

Some additional reading on this subject, might help to clarify things. There are some nice images (towards the bottom of the page) of correct scsi-configurations.

Adaptecs support site


Jörg

[Patrick]

We're getting there...

Jörg, this is the way the current setup is now:

h--dt--dt--u

Keep in mind that the "u" is the end of the cable that goes to the rear panel where the scanner plugs in.

I just ran HD Tach and the results weren't too bad. I'll change the setup to the following and run the test again:

h--d--dt--u

I may have to post my results tomorrow as it's 1:15 in the morning here! Thanks for your help!

[Patrick]

Much appreciated...

Thanks Greebe, and also to "The Pit".

[VJ]

This does change things....

So the last connector goes to the outside of the case, and has the scanner on it ? This makes the config look like this:

(with scanner) h--d--dt--s
and
(without scanner) h--d--dt--u

In both cases, the SCSI-specs are violated...
1. With the scanner connected, then you have a device on the bus beyond the terminated one (while this can be done / is sometimes allowed with external cabling, I'm unsure about the consequences when doing this with ribboncables).

2. Without the scanner connected, there is a "dangling end" on the cable...

3. I doubt your overall cablelength is less then 1.5 metres, which is a requirement for SE-UW (non-LVD). As the scanner will not be LVD, the presence of 1 device will result in the entire bus SE-mode (single ended), which seriously limits cablelength.

(see scsifaq )


I thought the scanner was connected to the external connector of the host adapter...

Could you post more specs :
Which host adapter are you using ?
What cabling does the scanner have ?


Jörg

[Patrick]

Ahh, who needs sleep anyway?

Ok, I couldn't resist staying up and removing the termination jumper from the drive nearest the host. I then ran HD Tach again. I have good news and I have bad news...

The good news is that the drives weren't any slower.

The bad news is that they weren't any faster.

The results for both drives were almost identical:

Random access.........9.6
Read burst.............26.3
Max read.........14984.0
Min read..........13359.0
Average read...14613.0

Not having any experience with SCSI drives, I don't know whether these results are good or not. I understand that these Seagate Barracuda drives are a few years old now. Does anyone know whether these results are acceptable for these drives?

Jörg, the built in SCSI card is an Adaptec AIC-7890 Ultra 2 & Ultra Fast/Wide. All I can tell you about the internal cabling is that it's quite short, about half a metre. I don't know what the cabling is for the scanner as I don't have it here.

I'm fading...I gotta get to bed now...

[VJ]

I cannot comment on the perfomance of those Seagates (haven't tested them, I have a Quantum Atlas 10K and an IBM Ultrastar 36LZX).

According to the manual I downloaded, the Asus P2B-DS has 3 SCSI-connectors : one 50 pin narrow, one 68 pin Ultra2 and one 68 pin UltraWide. Currently, you have all devices hanging on the 68 pin UW; correct ? (this is the connector located closed to the DIMM sockets)

The drives should be at maxium performance (and the bus at maximum stability) if you'd connect them as I stated previously :
h--u--u--dt
with a non terminated disk on any of the unused positions. You could try benchmarking the drives in such setup without the "dangling end" (if you happen to have a UW SCSI cable, or if the current cable can be used in such a way).

It then ought to be possible to connect the scanner to any of the other 2 SCSI-connectors (scanner terminated, host adapter should be capable of terminating only the upper byte). The only issue that would remain, is the cablelength : although the scanner and hdd's would be on different connectors, they'd still form a single bus and thus the overall cablelength would count (1.5 metres).
-
I myself had to buy a second adapter, just due to the cablelength issues :
Adaptec 2940UW has
--Quantum Atlas 10K (connected to 68 pin internal)
--IBM Ultrastar 36LZX
--Plextor 4220 (connected to 50 pin internal)
Adaptec 2906 had
--Agfa Snapscan 1236s (connected to 50 pin external)
--Iomega Zip 250

bus layout:
Plextor--u--u--2940UW--u--u--Quantum--IBM--SE/LVD terminator
2906--Agfa--Iomega
(My harddisks are LVD, they have no terminator. I cannot use their full potential, due to the UW-controller that is the bottleneck. But with 2 drives, the effect is limited)


Jörg