7 SCSI devices on Ultra SCSI channel

[Steve]

I have an LSI 20860 Ultra (20MB/s) SCSI card. It has no BIOS, but is adopted as a bootable secondary channel by the LSI U160 also in my rig. Connected to the 20860 are seven drives,

three internally:
ID 6 = Toshiba SD-M1711
ID 5 = Teac CD-532S
ID 4 = Yamaha CRW2100S

and four externally in a caddy:
IDs 3,2,1 = Plextor 40x CD-RWs
ID 0 = Plextor 40x CD-ROM

When I boot the system and Ctrl+C into the U160's BIOS, I see all the drives listed on the 20860's channel except the Teac. And when I allow the system to continue booting, the scan of the 20860's device's IDs times out (pauses ten seconds) on ID 0, the Plextor CD-ROM. How do I troubleshoot and fix these?

Other questions that occured to me:

1) For termination, I'm using a simple plug that appears to be brainless. The only drive with termination enabled is the Yamaha internally. Should I have one of the external drives do it too, or seek an active terminator instead of what I'm using?

2) Does the limit of seven devices on the channel include the card, or exclude it?

 

[Mark R]

You have terminated your bus incorrectly.

You must have termination at both physical ends of the chain linking all the drives, and nowhere else. Where you have internal and external - then you can think of the card itself as being in the middle. Therefore you must disable termination on the card.

So if you system is connected as follows:

Yamaha->Toshiba->Teac->LSI<-Plextor<-Plextor<-Plextor-<PlextorCD-ROM.

You would terminate the yamaha and the plxetor CD-ROM because they are on the ends of the cable.

The limit of 7 excludes the card itself.

 

[Peter]

You need to have termination on both ends of the cable chain ONLY - preferrably on the drive on the far end of the internal ribbon cable, as well as on the unused SCSI plug at the end of the external device chain. Do NOT enable termination on any drives in external casing. Both terminators must be of the "active" kind. If your external terminator block doesn't explicitly say "Active" on it, don't use it.

Furthermore, with more than four devices on U20 SCSI, your cable length limit is 1.5 meters, and you need to have the devices spread onto this length as evenly as possible. 20 cm minimum distance between devices needs to be followed too.

U20 is highly fragile; you might have to back down to Fast10. I had to, and I only have five devices on (including the controller). You can do that from the configuration menu, "Device settings". Set the 53C860 item to "10 MT/s" to force the entire SCSI chain back to a maximum speed of 10. Or use a separate SCSI controller for the external branch. (Didn't we find a bargain on a dual-channel UW40/U20 card recently?)

 

[Steve]

Peter - I've switched to a shorter internal cable, but it's still not following the 20cm guideline (haven't cut one of my long cables yet). Now the Yamaha is unlisted rather than the Teac. The Plextor CD-ROM still times out when scanning its ID on boot. Against your advice (and my better judgment) I enabled termination for the Plextor CD-RW at the external end; of course that did nothing, but I was just humoring the responses to my post at other forums (the ones you didn't post at). In the meantime I've ordered a 3-foot external cable (to replace this six-footer) and an active external terminator (to replace this passive one) from Hypermicro. Wish they had a storefront, they're less than an hour away.

MarkR - this is what my chain looked and still looks like:

Internal end:
4)Yamaha CD-RW (TRM)
|
5)Teac CD-ROM
|
6)Toshiba DVD-ROM
|
card:
7)LSI 20860
|
external
0)Plextor CD-ROM
|
1-2-3)Plextor CD-RWs (#3 at end - TRM on it changes nothing)

Only before I had the Toshiba up top and the Yamaha at the bottom, because the cable I was using was keyed to connect the drives that way. The Yamaha was (and still is) at the end internally, with termination on.

 

[Steve]

Peter - I can't change any settings for the 20860, they're all white rather than yellow.

 

[Peter]

Hmmm ... I thought the 20860 did have a configuration EEPROM. Non-adjustable settings indicate it doesn't. Well then you have no choice but build an UltraSCSI compliant setup, or go with separate channels for external and internal.

 

[Steve]

Sorry to get back so late, work has kept me busy. I have done some troubleshooting which has produced these observations:

- All devices do indeed have a unique ID number, and parity enabled.

- The only device with termination enabled is whichever one I've put at the end of the chain inside my PC. This was the Toshiba, and is now the Yamaha, which remains unlisted in the SCSI BIOS.

- None of the drives in the external caddy have termination enabled, as that had a passive terminator and now has an active one. I've also changed the six-foot cable for a three-footer.

- If I replace the Plextor CD-ROM with another Teac I have, no ID times out. It seems only the Plextor CD-ROMs have this problem (I saw this before when trying them in my PC), but it goes away if I remove the LSI U160 card. Of course, then there's no BIOS to check (which isn't all that bad, I suppose). I've tried with four Plextor CD-ROMs, two with firmware 1.05 (they can go no higher), one with 1.14, and one UltraPlex Wide, with a 68-to-50 pin adapter, at firmware 1.05 too.

I would stick with the Teacs, only their eject buttons don't open their trays. Right-click and "Eject" works for one of them though But they're "only" 32x anyway.

Peter, sounds like I need to try for a properly spec'd setup (or just remove the LSI card, which I may do). I have one or two internal cables I can modify to the 20cm guideline (which should fix the Yamaha), but the big problem will be finding or making a good cable for the caddy, what with its 50-pin Centronics connectors.

I seem to recall reading that only Plextor CD drives did real SCSI (as in commands, etc.) so that would explain why their CD-ROMS are so finicky. If the burners can cope just fine though, why isn't this bug in the readers fixed yet?

 

[Peter]

hmmm ... I (faintly) recall that one of my SCSI opticals regularly fails to show up in SCSI BIOS, but that doesn't affect its operation at all.

The TEAC drives, like most others, have an "inhibit eject button" jumper (so they can be used with Macintosh computers). Check that.

 

[TRUMPHENT]

You are attempting to use a single channel scsi card as dual channel. The LSI20860 is a single channel card according to the manufacturer's website.
LSI20860

In other words, you can either use the external connector "OR" the internal connector. You cannot use both simultaneously.

 

[Steve]

How am I doing that? The devices are listed in the U160's BIOS in one list, one column, under one channel, device 0 thru 7 which is specified for Ultra SCSI.

 

[Peter]

Nevermind, you're doing everything right. The card passes the same SCSI channel to both connectors, and there's nothing wrong with having the host adapter in the middle of the affair - because it detects that and auto-disables its own termination. Remember that on SCSI, all participants are equal - and it's just as OK for a host adapter to sit in the middle of the cabling as for any other device.

 

[Steve]

I've been able to come back to pursuing my issue in this thread. Here's what I've tried and found:

-My long, seven-connector cable using connectors 1, 4, and 7 (beating the 20cm rule and nearly evenly spaced); same behavior occurs, even when the external caddy isn't turned on. Decided not to try 1, 3, and 5 and cut the excess, I'll bet it won't change it.

(note that I normally use a 70cm internal cable (30+20+20), and a three foot external cable, on my 20860)

-Popped in an Adaptec 2940U2W; same thing happens on it too. I therefore decided it was something in the drives' firmware, since none of their burners exhibit this problem.

-Contacted Plextor support via email and explained the situation. Below are each of their responses (or at least, the highlights):

#1
"If the system locks at the same SCSI ID each time you boot then the drive itself is probably defective. You have already tried everything we would have suggested."

#2
"A thought, on the Adaptec 2940U2W is the termination setting in its bios set to disabled, being that you have mixed terminations, a hardware termination and an active termination? I will forward this information on to see what else we might be able to come up with."

#3
"I think I remember a similar issue with another customer a couple of years back. It required a custom firmware to reduce the time from 13-15 seconds to a faster time. The standard firmware will probably time out with his card.

I suggest you use the Adaptec card and use the embedded BIOS utility (Crtl-A). You should go into the SCSI device configuration and disable the "Include in BIOS scan" option for the ID of the ROM. You would still be able to use one of the RW drives as a boot device if desired.

Also, Ultra SCS with that number of devices should not exceed 1.5 meters combined cable length for internal and external if they are on the same bus. If there are any anomalies or intermittant device operation, including this timeout issue, you should check for active termination on the last devices and verify the host adapter is not terminated. You need to include the internal ribbon cable, the external cable, and the length of the cable inside the tower in this measurement."

I was hoping just to sell the Adaptec, but I can see how disabling that drive's ID scan would help. In any case I'm thinking about the LSIU40SE. Looking through its manual:

http://www.lsilogic.com/files/docs/tech...torage_stand_prod/SCSIboards/U40SE.pdf

Page 3-11 indicates that I should be able to turn off that drive's ID for scan, which is fine by me.

I guess I'm just wondering if this is the right path to take? Considering my M847LU only has three PCI slots, I wonder if I should leave my 20860 in for the rest of the drives or get new cables for them to put on the U40SE, if that'll even work?

I will try and ask for the custom firmware first though

 

[Peter]

Adaptec or LSI won't help. All the ID disabling does is reduce the scan time when you're booting.

Going with a configurable card like the LSI U40SE, and either limiting the bus speed to 10 MB/s or using one of its channel for internals and the other for externals, both at full 20 MB/s, would be the way to go. You might even leave the 20860 in the box as well, for a total of three SCSI channels

 

[Steve]

I'm keeping the U40SE in mind but first I'm going to try Plextor's custom firmware for their drive. Here is the email:

"Hi [sm8000],
I've attached the firmware. It is beta based on v1.14. It is supposed to check the drive for media status. If it sensed a disc before a bus reset it will check again. If it checked no media before a bus reset it will ignore the scan. It should drop the time to about 10-11 seconds. If this does not work for you, there is no other option as the drive was EOL a couple of years ago and no firmware development will be done.

Regards,

[Applications Engineer]

 

[nweaver]

I thought the SCSI bus was 7 devices total, and the card was on one of those devices.


Have you tried each drive seperatly, to make sure the drive is operable outside the large chain? I would try that first. Then, assuming all drives work single, I would start adding drives to the chain, 2 at a time, maybe one at a time.

 

[Steve]

Ultra SCSI (20MB/s) is eight including the host, per channel.


I've tried those suggestions a long time ago, probably in a much earlier thread. At this point I'm positive Plextor's CD-ROM firmware is the culprit.

 

[Steve]

Guess what? The beta firmware does it! Now when I boot the drive's ID is scanned for a fraction of a second rather than a full ten. Just how I like it (it's a bit longer since it's looking for a bootable disc, so Plextor says). Anyhow I think everything is working well now. Thanks everyone for your input!