Any known issues with Dell computers and ISA bus functionality?

  • Thread starter Thread starter PC Guy
  • Start date Start date
P

PC Guy

I've got a 16-bit ISA card (part of a specialty piece of data
acquisition equipment) that I know works in several different
mainstream motherboards (Soyo, Tyan, Gigabyte, etc).

It doesn't function correctly in a computer that I've been asked to
test it on:

Dell optiplex gx150 manufactured April 2002.

This is a desktop tower computer with one 16-bit ISA bus slot that is
connected to the motherboard via a side-mounted edgecard connector (as
if the ISA slot was designed as an optional add-on component to the
motherboard).

While debugging this board with a scope and some machine code written
and executed from within the DOS debug program, it seems that the
motherboard is not correctly honoring the IOCS-16 bus signal being
generated by the ISA card.

I think it's generating two 8-bit writes in response to a 16-bit I/O
write. The IOCS-16 signal seems to be working correctly.

Are there any known issues with Dell computers like this and ISA bus
functionality?
 
PC Guy said:
I've got a 16-bit ISA card (part of a specialty piece of data
acquisition equipment) that I know works in several different
mainstream motherboards (Soyo, Tyan, Gigabyte, etc).

It doesn't function correctly in a computer that I've been asked to
test it on:

Dell optiplex gx150 manufactured April 2002.

This is a desktop tower computer with one 16-bit ISA bus slot that is
connected to the motherboard via a side-mounted edgecard connector (as
if the ISA slot was designed as an optional add-on component to the
motherboard).

While debugging this board with a scope and some machine code written
and executed from within the DOS debug program, it seems that the
motherboard is not correctly honoring the IOCS-16 bus signal being
generated by the ISA card.

I think it's generating two 8-bit writes in response to a 16-bit I/O
write. The IOCS-16 signal seems to be working correctly.

Are there any known issues with Dell computers like this and ISA bus
functionality?
Click to expand...

Have you checked for any relevant options in the system bios?
Also, does the ISA slot share IRQs or other resources with
another slot or onboard device?
 
Have you checked for any relevant options in the system bios?
Click to expand...

What I've encountered in the past with motherboards made from, say,
1996 to maybe 2000 is that BIOS settings that referred to "16-bit wait
states" played a role with the some ISA cards I've dealt with in the
past. This was also referred to as "16 bit I/O recovery time" with
choices like 0, 1, 2, or 3. I haven't seen any such bios settings on
motherboards with ISA slots made since 2000 or 2001.

With regard to the Dell Optiplex I'm working with right now, there is
no such BIOS setting that I can find. I have come across some
internet references that Dell might have a "compatibility" mode option
in the BIOS which I'll look for tommorrow.
Also, does the ISA slot share IRQs or other resources with
another slot or onboard device?
Click to expand...

I would expect that the IRQ's available on the ISA slot are shared (or
not?) in a manner that is typical with conventional motherboards with
the normal compliement of ISA/PCI/AGP slots.

However, the ISA board in question is a pure I/O device only - it does
not use IRQ's or DMA's. All reads and writes are 16 bit. When I boot
the computer into Windows XP, the hardware manager device I/O list
does not show any devices using the I/O address range that the board
is using.

The ISA card is using relatively modern programmable logic chips for
all operational aspects (including bus decode) so perhaps the IOCS-16
signal that it is generating is "too fast" or too short for the Dell
to see?
 
PC Guy said:
What I've encountered in the past with motherboards made from, say,
1996 to maybe 2000 is that BIOS settings that referred to "16-bit wait
states" played a role with the some ISA cards I've dealt with in the
past. This was also referred to as "16 bit I/O recovery time" with
choices like 0, 1, 2, or 3. I haven't seen any such bios settings on
motherboards with ISA slots made since 2000 or 2001.

With regard to the Dell Optiplex I'm working with right now, there is
no such BIOS setting that I can find. I have come across some
internet references that Dell might have a "compatibility" mode option
in the BIOS which I'll look for tommorrow.
Click to expand...

This is a longshot, but also check for a jumper on the motherboard.
I would expect that the IRQ's available on the ISA slot are shared (or
not?) in a manner that is typical with conventional motherboards with
the normal compliement of ISA/PCI/AGP slots.

However, the ISA board in question is a pure I/O device only - it does
not use IRQ's or DMA's. All reads and writes are 16 bit. When I boot
the computer into Windows XP, the hardware manager device I/O list
does not show any devices using the I/O address range that the board
is using.

The ISA card is using relatively modern programmable logic chips for
all operational aspects (including bus decode) so perhaps the IOCS-16
signal that it is generating is "too fast" or too short for the Dell
to see?
Click to expand...

Well, the fact that XP doesn't list a given I/O range as being
in use, doesn't necesarily mean it's unused. I've run into
problems installing legacy cards at various addresses (e.g.
Adaptec scsi cards with a default of 140h, audio cards at
350 and 250h, even joysticks at 200h, and ran into conflicts
even though Device Manager claimed these addresses
weren't in use.

Just out of curiosity, what range does your card use?
 
Well, the fact that XP doesn't list a given I/O range as being
in use, doesn't necesarily mean it's unused.
Click to expand...

I understand, but we are talking about a machine built in 2002, with
integrated video and integrated everything-else (no other boards
plugged into any other slots) so I would expect that all system
devices to be pnp compliant and to be present in the device manager
list.
Just out of curiosity, what range does your card use?
Click to expand...

A small handful of addresses (maybe 18) at various locations between
520h and 596h.
 
PC said:
(e-mail address removed) wrote:
I would expect that all system
devices to be pnp compliant and to be present in the device manager
list.
Click to expand...

Ah, but this is a Dell you're talking about, I wouldn't expect anything in
it to necessarily be compliant in the way you may understand it.
I tried to install an OEM CD burner in a fairly modern Dell recently. Having
changed every possible jumper, setting and cable I could think of, with no
success, I gave up, Put the burner in a generic PC and it worked first time.
But I bet if I bought a burner from Dell it would magically work. :-)

SteveH
 
PC Guy said:
I would expect that the IRQ's available on the ISA slot are shared (or
not?) in a manner that is typical with conventional motherboards with
the normal compliement of ISA/PCI/AGP slots.
Click to expand...

That's one place where ISA and PCI are different -- no shared IRQs allowed
on ISA.

However, the ISA board in question is a pure I/O device only - it does
not use IRQ's or DMA's. All reads and writes are 16 bit. When I boot
the computer into Windows XP, the hardware manager device I/O list
does not show any devices using the I/O address range that the board
is using.
Click to expand...

ISA modem cards, for example, usually had jumperable IRQ settings.

Another possible quirk is that it will work in one slot, but not another. I
recall having problems on at least one board with the slot closest to the
AGP slot, and have solved otherwise indecipherable problems in the past by
merely swapping ISA slots.
 
John said:
Another possible quirk is that it will work in one slot, but not
another.
Click to expand...

This dell only has a single ISA slot, mounted on it's own small strip
of PC board with a bit of circuitry, connected to the main motherboard
via a side-mounted high-density, low profile edgecard connector.

The bit of circuitry on the board looks to include a PCI-ISA bridge
chip.
I recall having problems on at least one board with the slot
closest to the AGP slot, and have solved otherwise indecipherable
problems in the past by merely swapping ISA slots.
Click to expand...

I've had problems in the past with the functionality of some PCI
boards plugged into the PCI slot closest to the AGP slot. I've never
had any similar ISA-board functionality issues like that, and have
never seen a motherboard with ISA slots immediately adjacent to an AGP
slot without there being PCI slots between them.
 
John Weiss said:
That's one place where ISA and PCI are different -- no shared IRQs allowed
on ISA.
Click to expand...

That's not quite correct. Multiple ISA devices cannot share
an IRQ, however on boards that have both ISA and PCI
slots it's common for the ISA to share with a PCI slot. E.g.
my Abit BF6 has 5 PCI/1 ISA, and the ISA shares with
one of the PCI slots.
 
Further to my original post, here is more information regarding this
ISA problem on the Dell computer.

This was a Dell Optiplex GX150 manufactured in April 2002. It has an
Intel Pentium-3, 1Ghz clock speed, 256 mb of system memory. The video
adapter and all other interface devices are integrated onto the
motherboard. The motherboard has several available expansion slots
(AGP, PCI, and 1 ISA) all of which were available (no pre-installed
card installed in any of them). The ISA slot was built into it's own
separate PC board that was connected to the main motherboard with a
high-density side-mounted connector. On this separate PC board was a
National Semiconductor PCI-to-ISA bridge (PC87200) which translates
I/O and data signals and commands from the motherboard to the ISA
slot. This National Instruments chip seems to be a first-generation
version PCI-to-ISA bridge introduced in early 1999 that allowed PC and
motherboard makers to add ISA slots to their systems and still be
classified as PC99 compliant.

Curiously, National Instruments makes no information available on it's
web site about this chip (no spec sheets, no white papers, etc). The
only reference to this chip is in this March 1999 press release:

NATIONAL SEMICONDUCTOR ANNOUNCES PCI-TO-ISA BRIDGE FOR NEW
PC99-COMPLIANT SYSTEMS

PC87200 Enables PC Manufacturers to Add ISA Slots to Their
PCI Systems

http://www.national.com/news/item/0,1735,361,00.html

My previous experience with ISA instability or non-functionality were
usually solved on older motherboards by changing the 16-bit I/O
wait-state setting in the system BIOS, but there was no such setting
on this Dell Optiplex. The Dell had bios revision 09, and it was
updated to the apparent last version made available by Dell (version
11) which made no difference.

ISA bus signals were investigated using an oscilloscope while
low-level data commands (I/O reads and writes) were performed to
addresses being decoded by a 16-bit data acquisition ISA card. While
the addresses were being decoded correctly, all transactions were
being performed in 8-bit mode instead of 16-bit mode.

When performing 16-bit I/O reads or writes on the ISA bus to even (not
odd) addresses, the bus controller can either perform this as a single
transaction, or it can perform 2 sequential 8-bit transactions. ISA
bus specifications dictate that the target ISA card controls this
transaction (16-bit or 8-bit) by asserting the IOCS-16 line if the
card is indeed capable of performing 16-bit transactions. The ISA
card was asserting the IOCS-16 line correctly (this line is normally
in the high state and is active-low when asserted, for the duration of
the address decode which was observed to be 500 nanoseconds).
However, the Dell Optiplex, and presumably specifically the National
Semiconductor PC87200 chip, does not seem to honor the state of the
IOCS-16 line and instead performs 16-bit I/O reads and writes as two
sequential 8-bit operations.

Perhaps this was a defect for this specific Dell computer, but there
are plenty of other situations where these mass-marketed computers
have shown to cut corners physically, electrically, and
compatability-wise with generally recognized standards and components.
 
card is indeed capable of performing 16-bit transactions. The ISA
card was asserting the IOCS-16 line correctly (this line is normally
in the high state and is active-low when asserted, for the duration of
the address decode which was observed to be 500 nanoseconds).
Click to expand...

Double check that time. If accurate there's your problem. Even at
4.77MHz a complete bus cycle would take fractionally under 210 ns.
I'll admit my memories of the AT bus are a little hazy but it
doesn't sound right.
 
ISA bus signals were investigated using an oscilloscope while
low-level data commands (I/O reads and writes) were performed to
addresses being decoded by a 16-bit data acquisition ISA card. While
the addresses were being decoded correctly, all transactions were
being performed in 8-bit mode instead of 16-bit mode.

When performing 16-bit I/O reads or writes on the ISA bus to even (not
odd) addresses, the bus controller can either perform this as a single
transaction, or it can perform 2 sequential 8-bit transactions. ISA
bus specifications dictate that the target ISA card controls this
transaction (16-bit or 8-bit) by asserting the IOCS-16 line if the
card is indeed capable of performing 16-bit transactions. The ISA
card was asserting the IOCS-16 line correctly (this line is normally
in the high state and is active-low when asserted, for the duration of
the address decode which was observed to be 500 nanoseconds).
Click to expand...

FWIW, here is a scan of the section of the original IBM AT Tech Ref
manual that deals with a Prototype Adapter:
http://www.users.on.net/~fzabkar/PC-AT/Proto.zip

- Franc Zabkar
 
Back
Top