Strange SCSI behaviour - maybe missing terminator?!?

[Zac67]

UW-SCSI is very picky about termination quality, cable quality, device spacing(!) and cable length(!!). With >4 devices connected to the bus it runs a maximum of only 1.5 m.

Single Ended was abandoned beyond Ultra SCSI because of this. All later standards used Low Voltage Differential signaling.


The bus ends MUST be terminated actively (SE or LVD/SE) and the devices next to the ends should be supplying termpower. My guess is device spacing (any connectors added after manufacturing?), overall cable length or cable quality. Another possibility is a SCSI-2 device influencing the Ultra-SCSI transmissions. I've also seen bad wide-to-narrow adapters.

If you can't or don't want to change the cabling you can probably slow down CSPPC SCSI to Fast-Wide (20 MB/s max) in general to get rid of the problems. Another approach is to move the slower, non-Ultra devices to a separate bus (=controller).

[Zac67]

The SCSI 2 will slow your bus down to 10MB/s or less.

No. The bus will arbitrate the best speed with each device separately. Slow devices may use up a lot of bus time while they're in use, slowing down faster devices, but they do not slow down the bus in general.

The only exception to this rule is when connecting an SE device (up to Ultra SCSI) to an otherwise LVD bus (U2W+), forcing the whole bus to SE mode electrically. But there's no LVD controller for the Amiga, so this doesn't count. (There's even an exception to this exception rule when using a SCSI bridge like e.g. the Adaptec 2940U2W does.)

SCSI 1: 5MB/s. 8-bit data w/ asynchronous transmission

Not entirely correct - SCSI-1 has asynchronous (~3.5 MB/s max) and synchronous mode (5 MB/s max).

U160/320 = MegaBits/s (not MegaBYTES/s)

Nonsense - it's 160 / 320 megaBYTES/s.
Our old LTO-3 drive peaks at ~120 MB/s. 320 Mb/s would only be 40 MB/s.

Another note on termination:
Passive terminators are only supported up to Fast (Wide) SCSI. Ultra (Wide) SCSI requires active termination and U2W, U160, U320 require LVD(/SE) terminators.

[Zac67]

Well, maybe you should get rid of that feline hulk...

[Zac67]

This one's fine, too: http://www.ebay.co.uk/itm/AMPHENOL-G5925733-HD68-SCSI-LVD-SE-TERMINATOR-/280955499040?pt=UK_Computing_Drive_Cables_Adapters&hash=item416a40b220

only .99 GBP

[Zac67]

... or the Plextor CD drive that will be next to the end in the chain. The Quantum just plugs in and the Plextor plugs in with an adaptor.

Do NOT put an old narrow SCSI device at the end of the chain. Its terminator (resistor packs are useless anyway) would only terminate the lower 8 bits of the 16 bit bus, leaving the upper half unterminated. By far the easiest way is to put a wide device on the end, otherwise you need an adapter with a 'half terminator'.

Termpower is the power supply for the terminators, they won't work without. At least one device needs to supply termpower. I usually use the two at the ends powering their own terminators or the cable terminators right next to them.

[Zac67]

Are you thinking about the Quantum HD or the Plextor CD?

Any narrow device at the end of a wide bus is a problem.

[Zac67]

Active single ended terminators is what you need. There are no LVD (U2W+) controllers for the Amiga.

As long as you don't try to use the termination on the CD drive it shouldn't matter where it's attached.

[Zac67]

Forget about termination on a narrow (50-pin) device, it won't work. Terminate the bus with a wide device (68-pin) or a cable terminator. Using a cable terminator, you need to make sure that any on-device terminators are turned off.

[Zac67]

All terminators require termpower, regardless of type. At least one device needs to supply TP.

[Zac67]

Dunno about Plextor, but there's usually a TE jumper (for Terminator Enable) - remove. PE is Parity Enable (you'll want that on), TP Term Power (don't need it, remove). The ID must be unique, of course.

[Zac67]

The pic on the right is an SCA adapter. SCA is an interface intended for connecting into a backplane, there's no termination or termpower provided (there are some SCA adapters with terminators but this isn't one).

Termpower is often provided by several devices but I've had - rare - occasions when that was a problem somehow (which I can't explain I'm afraid). Best practice is to provide termpower by the host adapter and if that isn't possible from the devices at the end of the bus, near the terminators. Terminators with an LED have the advantage of showing you whether termpower's present.

SCSI IDs: the RDB 'last drive' option just stops scanning further drives for RDB automount purposes, so it's perfectly reasonable to put an optical drive on higher IDs. I'm running my boot drive on ID1, so I can add a temporary drive at ID0 without having to fiddle with HDToolBox. Opticals (if present) are on 4 & 5, a tape on 3. 6 I reserve for a 2nd host adapter (PC side) but I haven't 'SCSI-netted' for ages.

[Zac67]

When a narrow device is connected somwhere in the middle you can simply use an ordinary adapter.
Only if you connect that device to an end and need it to provide termination (no cable terminator) then you need a half-terminating adapter to take care of the high byte.

[Zac67]

And you can't mix SE and LVD on the same bus.

Yes, you can - it may or may not be smart though. A SCSI bus with a single SE device connected to it forces the bus to SE electrically: all return lines are grounded. Unless any of your LVD devices has a problem with SE in general (some - rare - are LVD only, not LVD/SE), everything is fine at UW speed (unless cable length for that mode is exceeded). Since the CSPPC is SE to start with, there's little to gain by forcing SE somewhere else.

On some PC HBAs there were even SCSI bridges to get an SE and an LVD domain in a single bus (e.g. AHA2940U2W).