Problem with U320 drive, high byte termination required?

Hello folks, sorry this is a bit long-winded,

I have a Fujitsu U320 36GB 80-pin drive which I must hook up to a 50-pin
SCSI host, naturally this is through a 50-80pin adapter. I do this often
with other LVD drives, and there has never been a problem. This drive, a
MAP3367NC, now connects fine to the 68 pin bus (this time via a 68-80 pin
adapter) on a 2940UW card, however it runs at half the speed it did before
i.e. 17MB/sec. I know it was 33MB/sec when I was testing the drive a couple
of weeks ago. Maybe this is a clue to what is wrong? Although, the speed of
the drives is unimportant as long as they can get to about 5MB/sec.

When I try to connect it to a known working 2940U with a working cable and
50-80 pin adapter, the drive spins up but is not ever seen by the host
controller, whether its the 2940U or the 50 pin host that I wish to attach
it to. If I replace it with a similar U160 IBM or Seagate unit, there's no
problem. The other drives work fine.

I've taken off the Fujitsu's circuit board and inspected it with a
magnifying glass, and re-seated the motor connector and cleaned all external
contacts. There's no difference though, not on a 50 pin bus at any rate.

Earlier last night, I couldn't get the Fujitsu to connect at all to my
2940UW's 68-pin bus. Now since I cleaned the contacts, it works on the
68-pin bus. At half speed, like I said. But any 50-way bus isn't showing the
drive.

Nothing in the drive's manual seems to indicate that this Fujitsu would not
be backwards compatible. I tested for error codes and problems with a
Fujitsu SCSI utility but everything was fine.

I'm no expert on different terminators but the 68 pin bus that works with
the drive on the 2940UW is terminated by a fancy circuit that has the words
LVD/SE moulded onto it. When I run the drive on a narrow bus, I am using
(what I believe to be passive) centronics 50 terminators.

Other LVD drives work fine using those centronics terminators. Is there any
reason that this U320 drive would not work at all? I thought high-byte
termination was only required when connecting narrow devices to wide SCSI
cable / bus?

Perhaps I need an active or LVD/SE terminator even though the terminator
will be attached to a 50pin centronics connector anyway?

Hopefully, the drive is simply faulty and I can indeed use U320 drives of
any brand, especially Fujitsu, on 50-pin buses.

Thanks kindly for any advice you can offer.

Dave.

Dave Scaramanga wrote:
>
> [...]
> When I try to connect it to a known working 2940U with a working cable and
> 50-80 pin adapter, the drive spins up but is not ever seen by the host
> controller, whether its the 2940U or the 50 pin host that I wish to attach
> it to. If I replace it with a similar U160 IBM or Seagate unit, there's no
> problem. The other drives work fine.

This sounds very much like a "high-byte termination" problem.

> I've taken off the Fujitsu's circuit board and inspected it with a
> magnifying glass, and re-seated the motor connector and cleaned all external
> contacts. There's no difference though, not on a 50 pin bus at any rate.
>
> Earlier last night, I couldn't get the Fujitsu to connect at all to my
> 2940UW's 68-pin bus. Now since I cleaned the contacts, it works on the
> 68-pin bus. At half speed, like I said. But any 50-way bus isn't showing the
> drive.

How did you measure the speed?

> Nothing in the drive's manual seems to indicate that this Fujitsu would not
> be backwards compatible. I tested for error codes and problems with a
> Fujitsu SCSI utility but everything was fine.

It should be backward compatible. The fastest transfer mode for the SE
physical layer is "Fast20" (40MBytes/s on a wide bus), so the 33MBytes/s
should be OK.

> I'm no expert on different terminators but the 68 pin bus that works with
> the drive on the 2940UW is terminated by a fancy circuit that has the words
> LVD/SE moulded onto it.

That is fine even though the 2940UW have a SE-only interface. This is a
"multimode" terminator that can detect SE and LVD. It automatically will
switch to the matching circuit.

> When I run the drive on a narrow bus, I am using
> (what I believe to be passive) centronics 50 terminators.

The Fast20 transfer mode requires good cables and active terminators on
SE-busses. If it really is a passive terminator it is not suitable for
the Fast20 transfer mode. Maybe the drivers on your host have detected
parity errors and have re-negotiated to the Fast10 mode with the half
speed.

> Other LVD drives work fine using those centronics terminators. Is there any
> reason that this U320 drive would not work at all? I thought high-byte
> termination was only required when connecting narrow devices to wide SCSI
> cable / bus?

The other way round. If you connect a narrow device to a wide bus there
are no open lines and no problem (assumed that the wide bus is correctly
terminated).

If you connect a wide device to a narrow bus, the high byte of the
device-interface is floating. Some devices don't work at all in this
configuration. A termination for the high byte must be provided by the
wide/narrow-adapter or the device itself. Because 80pin SCA drives have
no terminators onboard, the adapter must do it for them.

> Perhaps I need an active or LVD/SE terminator even though the terminator
> will be attached to a 50pin centronics connector anyway?

Yes, at least for the Fast20 transfer mode.

> Hopefully, the drive is simply faulty and I can indeed use U320 drives of
> any brand, especially Fujitsu, on 50-pin buses.

Try the correct termination before declaring the disk as faulty. SCSI
don't allow floating signal lines, a wide interface requires wide
termination. Some devices may tolerate the floating high byte, but don't
expect that this will work. An exception are devices with a
"Narrow"-Jumper that can be forced in narrow mode.

To avoid the special "high-byte terminated adapter" think about this
solution:
Use a wide cable and connect the U320 disk with a mechanical-only
80/68-adapter (LVD-capable). Now use a mechanical-only 68/50-adapter to
connect the wide cable to the narrow SE-host (a SE-only adapter is
sufficient at this place). The wide terminator at the end of the cable
pulls all unused lines to defined levels (common for all disks if you
connect more than one!). The high-byte is unterminated at the other
side, this is not perfectly legal but it should work because there will
never be a data transfer on that lines (the narrow host can't do it).


Micha

"Michael Baeuerle" <michael.baeuerle@gmx.net> wrote in message
newstije6-f41.ln1@micha.freeshell.org...
> Dave Scaramanga wrote:
>>
>> [...]
>> When I try to connect it to a known working 2940U with a working cable
>> and
>> 50-80 pin adapter, the drive spins up but is not ever seen by the host
>> controller, whether its the 2940U or the 50 pin host that I wish to
>> attach
>> it to. If I replace it with a similar U160 IBM or Seagate unit, there's
>> no
>> problem. The other drives work fine.
>
> This sounds very much like a "high-byte termination" problem.

Firstly, thank you for taking the time to help, I really appreciate this.
The devices that I need this drive to work with are Akai S5000 / S3000
series digital audio samplers with 50-pin SCSI host controllers, I assume
they are SCSI-1. I only use my PC's SCSI cards to perform maintenance upon
the drives I fit to samplers.

>
>> I've taken off the Fujitsu's circuit board and inspected it with a
>> magnifying glass, and re-seated the motor connector and cleaned all
>> external
>> contacts. There's no difference though, not on a 50 pin bus at any rate.
>>
>> Earlier last night, I couldn't get the Fujitsu to connect at all to my
>> 2940UW's 68-pin bus. Now since I cleaned the contacts, it works on the
>> 68-pin bus. At half speed, like I said. But any 50-way bus isn't showing
>> the
>> drive.
>
> How did you measure the speed?

With a freeware program called HDTUNE. It benchmarks the speed and displays
seek times. I distinctly remember that the transfer rate was 33MB/sec, (only
tested on the 2940UW) from beginning to end. This was 2 weeks ago when I
first acquired the drive. A couple of days ago, I really needed to use it,
so I placed it into a 50-pin SCSI enclosure, like I have done with many
drives before. The Akai didn't recognise it, and when I attached it to my
PC's 50-pin Adaptec 2940U, that didn't see it either. I took it out of the
enclosure, connected it to my 2nd SCSI card - the Adaptec 2940UW, and 4
times out of 5 even that card wouldn't detect it.

I took off the circuit board and checked everything and cleaned all
contacts. now the 2940UW sees the drive every time. But not the 2940U with
its 50-way connectors. Nor the Akais.

>
>> Nothing in the drive's manual seems to indicate that this Fujitsu would
>> not
>> be backwards compatible. I tested for error codes and problems with a
>> Fujitsu SCSI utility but everything was fine.
>
> It should be backward compatible. The fastest transfer mode for the SE
> physical layer is "Fast20" (40MBytes/s on a wide bus), so the 33MBytes/s
> should be OK.
>
>> I'm no expert on different terminators but the 68 pin bus that works with
>> the drive on the 2940UW is terminated by a fancy circuit that has the
>> words
>> LVD/SE moulded onto it.
>
> That is fine even though the 2940UW have a SE-only interface. This is a
> "multimode" terminator that can detect SE and LVD. It automatically will
> switch to the matching circuit.
>
>> When I run the drive on a narrow bus, I am using
>> (what I believe to be passive) centronics 50 terminators.
>
> The Fast20 transfer mode requires good cables and active terminators on
> SE-busses. If it really is a passive terminator it is not suitable for
> the Fast20 transfer mode. Maybe the drivers on your host have detected
> parity errors and have re-negotiated to the Fast10 mode with the half
> speed.

I considered that the cables might be at fault. So I tried different 50-way
cables, different 50-80 pin adapters and different (passive?) terminators.
Whatever combination I used, no 50-pin interface would even detect the
presence of a drive.

I have a few other 68-way cables (with similar terminators) and I am getting
the same results with them as before, the drive is recognised, read speed is
17MB/sec.


>> Other LVD drives work fine using those centronics terminators. Is there
>> any
>> reason that this U320 drive would not work at all? I thought high-byte
>> termination was only required when connecting narrow devices to wide SCSI
>> cable / bus?
>
> The other way round. If you connect a narrow device to a wide bus there
> are no open lines and no problem (assumed that the wide bus is correctly
> terminated).
>
> If you connect a wide device to a narrow bus, the high byte of the
> device-interface is floating. Some devices don't work at all in this
> configuration. A termination for the high byte must be provided by the
> wide/narrow-adapter or the device itself. Because 80pin SCA drives have
> no terminators onboard, the adapter must do it for them.
>

Right, that makes sense. So it's the high byte of the *drive* that is
floating about, unterminated - and not the high byte of the *host
controller* that causes the problems as I originally thought?

>> Perhaps I need an active or LVD/SE terminator even though the terminator
>> will be attached to a 50pin centronics connector anyway?
>
> Yes, at least for the Fast20 transfer mode.
>

So I may just have been lucky up until now - I have three 80-pin LVD drives
attached to those old 50-pin Akais, with standard passive terminators - they
have performed without error for years. They even run in series with various
other SCSI devices, notably Jaz drive and DVD-RAM, and they have been rock
solid until now.

>> Hopefully, the drive is simply faulty and I can indeed use U320 drives of
>> any brand, especially Fujitsu, on 50-pin buses.
>
> Try the correct termination before declaring the disk as faulty. SCSI
> don't allow floating signal lines, a wide interface requires wide
> termination. Some devices may tolerate the floating high byte, but don't
> expect that this will work. An exception are devices with a
> "Narrow"-Jumper that can be forced in narrow mode.
>

Hmm, I don't think I've come across one of those. Occasionally there is a
'force SE' jumper. No good for me?

> To avoid the special "high-byte terminated adapter" think about this
> solution:
> Use a wide cable and connect the U320 disk with a mechanical-only
> 80/68-adapter (LVD-capable). Now use a mechanical-only 68/50-adapter to
> connect the wide cable to the narrow SE-host (a SE-only adapter is
> sufficient at this place). The wide terminator at the end of the cable
> pulls all unused lines to defined levels (common for all disks if you
> connect more than one!). The high-byte is unterminated at the other
> side, this is not perfectly legal but it should work because there will
> never be a data transfer on that lines (the narrow host can't do it).
>

Right, I see, that's not going to be easy since the drives must sit in
external enclosures which are all 50-pin. Unless I could obtain a 50-68 or
50-80 adapter with high byte termination built in.

Or, use 68pin instead of 80 pin drives and set the drive's own terminator
on - would that work (assuming I do that only at the end of the SCSI chain)?

>
> Micha

Thank you Micha, I sincerely appreciate that - even though it's bad news for
me....

Dave Scaramanga wrote:
> Michael Baeuerle wrote:
> > Dave Scaramanga wrote:
> > >
> > > [...]
> > This sounds very much like a "high-byte termination" problem.
>
> Firstly, thank you for taking the time to help, I really appreciate this.
> The devices that I need this drive to work with are Akai S5000 / S3000
> series digital audio samplers with 50-pin SCSI host controllers, I assume
> they are SCSI-1. I only use my PC's SCSI cards to perform maintenance upon
> the drives I fit to samplers.

Even if they are not SCSI1, it is likely that they use a slow/uncritical
transfer mode because it is sufficient for audio. The narrow Adaptec
seems to behave very similar than the Akais (don't see the drive at
all). This really look like a problem with the wide interface of the
disk (high-byte termination).

> [...]
> I have a few other 68-way cables (with similar terminators) and I am getting
> the same results with them as before, the drive is recognised, read speed is
> 17MB/sec.

Hmm ... and you are sure that you havn't limited the transfer mode to
Wide-Fast10 or Narrow-Fast20 in the setup of the 2940UW by accident?

> > > Other LVD drives work fine using those centronics terminators. Is there
> > > any
> > > reason that this U320 drive would not work at all? I thought high-byte
> > > termination was only required when connecting narrow devices to wide SCSI
> > > cable / bus?
> >
> > The other way round. If you connect a narrow device to a wide bus there
> > are no open lines and no problem (assumed that the wide bus is correctly
> > terminated).
> >
> > If you connect a wide device to a narrow bus, the high byte of the
> > device-interface is floating. Some devices don't work at all in this
> > configuration. A termination for the high byte must be provided by the
> > wide/narrow-adapter or the device itself. Because 80pin SCA drives have
> > no terminators onboard, the adapter must do it for them.
>
> Right, that makes sense. So it's the high byte of the *drive* that is
> floating about, unterminated - and not the high byte of the *host
> controller* that causes the problems as I originally thought?

Exactly. The 2940U has no high byte at all and the 2940UW can terminate
it with its onboard terminator if you connect a narrow bus.

Let me try to explain why floating lines are a problem:
Parallel SCSI use no dedicated address lines, instead the device address
(SCSI ID) is transferred over the data lines. But the address is not
binary encoded, every data line simply represent one address. This is
the reason why the number of devices on a narrow bus is limited to 8 but
a wide bus can work with 16 devices.
1) Arbitration
If a device want to access the bus it must do a so called arbitration
(required to avoid collisions). To arbitrate a device drive the BSY
signal and its own ID (the corresponding data line) active and wait a
defined time. Then it looks at the other data lines: If there is a data
line active that corresponds to a SCSI-ID with higher priority the
device has lost the arbitration and the SCSI protocol define that it
must release the bus now (the bus termination pulls undriven lines to
inactive state). At this time floating data lines can look active, like
devices that lost arbitration but don't release the bus and such a
protocol error may force the winning device to abort its transfer.
2) Selection
After a device has won arbitration it must select the communication
partner. Therefore it drives the BSY and SEL signals active and place
its own ID and the target ID together with valid parity on the data bus.
The SCSI protocol define that a target must not respond to a selection
if there are more than 2 data lines active or the parity is invalid. But
this is the case if floating data lines look active at this time.

In your case Problem 2) may occur at the device scan. The narrow host do
a correct selection but the wide device don't respond because it sees a
protocol error. After the selection timeout the host will give up and
continue with the scan.

> > > Perhaps I need an active or LVD/SE terminator even though the terminator
> > > will be attached to a 50pin centronics connector anyway?
> >
> > Yes, at least for the Fast20 transfer mode.
>
> So I may just have been lucky up until now - I have three 80-pin LVD drives
> attached to those old 50-pin Akais, with standard passive terminators - they
> have performed without error for years. They even run in series with various
> other SCSI devices, notably Jaz drive and DVD-RAM, and they have been rock
> solid until now.

This is no problem as long as a slower transfer mode is used. If the
Akai use Fast5 or asynchronous transfers, you can use 6m cable with
passive terminators.

> > > Hopefully, the drive is simply faulty and I can indeed use U320 drives of
> > > any brand, especially Fujitsu, on 50-pin buses.
> >
> > Try the correct termination before declaring the disk as faulty. SCSI
> > don't allow floating signal lines, a wide interface requires wide
> > termination. Some devices may tolerate the floating high byte, but don't
> > expect that this will work. An exception are devices with a
> > "Narrow"-Jumper that can be forced in narrow mode.
>
> Hmm, I don't think I've come across one of those. Occasionally there is a
> 'force SE' jumper. No good for me?

This has nothing to do with the bus width. The "Force SE" jumper simply
grounds the DIFFSENS line and prevent the disk from switching to LVD
mode. If you connect the Adaptec or the Akai, they will ground the
DIFFSENS line, the result is the same. The jumper is only needed if the
DIFFSENS signal is faulty wired.

> > To avoid the special "high-byte terminated adapter" think about this
> > solution:
> > Use a wide cable and connect the U320 disk with a mechanical-only
> > 80/68-adapter (LVD-capable). Now use a mechanical-only 68/50-adapter to
> > connect the wide cable to the narrow SE-host (a SE-only adapter is
> > sufficient at this place). The wide terminator at the end of the cable
> > pulls all unused lines to defined levels (common for all disks if you
> > connect more than one!). The high-byte is unterminated at the other
> > side, this is not perfectly legal but it should work because there will
> > never be a data transfer on that lines (the narrow host can't do it).
>
> Right, I see, that's not going to be easy since the drives must sit in
> external enclosures which are all 50-pin. Unless I could obtain a 50-68 or
> 50-80 adapter with high byte termination built in.

Yes.

> Or, use 68pin instead of 80 pin drives and set the drive's own terminator
> on - would that work (assuming I do that only at the end of the SCSI chain)?

This will work. As you state, because on disks you can normally only
enable all terminators (high+low) together the disk must be at the end
of the bus and no additional terminator must be present on the external
cases second connector.


Micha
--
http://micha.freeshell.org/