| Author |
Message |
techie_alison
Guest
|
 Posted:
Mon Oct 31, 2005 11:45 am Post subject:
SCSI 1 Command Timings |
|
|
Hello,
Would anyone have any links which detail the timings and command structures
of the SCSI 1 interface? I'm searching high and low on Google at the
moment. A SCSI host adapter is being developed.
What's thrown a spanner in the works is that some bright spark has suggested
responses being in the region of 90nanoseconds. I mean, SCSI was about in
the early 80's and this 90nS figure to respond seems a bit extreme.
I'm developing the interface with an embedded 10MIPS microcontroller so
100ns is about the best I can do.
What I'm ideally looking for is detailed information of loading commands
onto the bus and handshaking the control lines.
Thanks,
Aly, Cambridge UK |
|
| Back to top |
|
 |
techie_alison
Guest
|
| Posted:
Mon Oct 31, 2005 3:57 pm Post subject:
Re: SCSI 1 Command Timings |
|
|
"PeterD" <peter2@hipson.net> wrote in message
news:ho8cm1pho8h1nnqfe4ct9rvrfi4l754htr@4ax.com...
| Quote: | On Mon, 31 Oct 2005 11:45:06 +0000 (UTC), "techie_alison"
techie_alison@yahoo.com> wrote:
What's thrown a spanner in the works is that some bright spark has
suggested
responses being in the region of 90nanoseconds. I mean, SCSI was about
in
the early 80's and this 90nS figure to respond seems a bit extreme.
Where did you see this?
|
Someone said it in the cam engineering department. He's due to be flogged
later this evening for crimes to techies. |
|
| Back to top |
|
|
techie_alison
Guest
|
| Posted:
Mon Oct 31, 2005 4:35 pm Post subject:
Re: SCSI 1 Command Timings |
|
|
"PeterD" <peter2@hipson.net> wrote in message
news:ho8cm1pho8h1nnqfe4ct9rvrfi4l754htr@4ax.com...
| Quote: | On Mon, 31 Oct 2005 11:45:06 +0000 (UTC), "techie_alison"
techie_alison@yahoo.com> wrote:
What's thrown a spanner in the works is that some bright spark has
suggested
responses being in the region of 90nanoseconds. I mean, SCSI was about
in
the early 80's and this 90nS figure to respond seems a bit extreme.
Where did you see this?
|
Do you have the information I need? |
|
| Back to top |
|
|
techie_alison
Guest
|
| Posted:
Mon Oct 31, 2005 7:04 pm Post subject:
Re: SCSI 1 Command Timings |
|
|
"Rob Turk" <_wipe_me_r.turk@chello.nl> wrote in message
news:kWs9f.219$zc1.36@amstwist00...
| Quote: | "techie_alison" <techie_alison@yahoo.com> wrote in message
news:dk5h4h$h6q$1@nwrdmz03.dmz.ncs.ea.ibs-infra.bt.com...
"PeterD" <peter2@hipson.net> wrote in message
news:ho8cm1pho8h1nnqfe4ct9rvrfi4l754htr@4ax.com...
On Mon, 31 Oct 2005 11:45:06 +0000 (UTC), "techie_alison"
techie_alison@yahoo.com> wrote:
What's thrown a spanner in the works is that some bright spark has
suggested
responses being in the region of 90nanoseconds. I mean, SCSI was
about
in
the early 80's and this 90nS figure to respond seems a bit extreme.
Where did you see this?
Do you have the information I need?
Your best bet is to go to www.t10.org and browse the SCSI-2 draft specs
that
are in place. From memory I don't think there's any such requirement of
90us
for responses. When the host adapter selects a target, it has to respond
within 250ms (yes, a quarter of a second..), after that the target takes
control of bus phases and transfers.
If your HBA supports synchronous transfers then a negotiation will happen
and you can tell the target how fast you can go. If you stick to the basic
asynchronous protocol then the REQ and ACK cycles are interlocked and you
can go as slow as you want. Some targets may not like going extremely slow
but your 100us cycles should be more than adequate.
Rob
|
Hi Rob,
Thanks for that :-)
Have found some further info this afternoon about all of this. That 250ms
figure sounds a lot more palatable, I won't know properly until the
prototype is in place and can then play about with timings in firmware. I'm
down to 100nS instruction execution on the microcontroller now so have built
in a little more space, which will hopefully pay off as a higher transfer
speed. Going down to 25nS instructions would have required complete new
programming equipment for the high end MCUs.
What I'm actually developing is an Atari ACSI <> IDE interface. The ACSI
standard being a significantly cut down version of SCSI with about 10
commands overall. We're also moving 16bit IDE <> 8bit ACSI without parity
to add just a little more spice. Hence the worry about timing constraints
when we have other things going on too. To indicate and jump to a new
interrupt will take about 500-800uS unless we're monitoring continually then
it's about 100-200uS.
Admittedly I'm more interested in getting this to work to prove it can be
done. 5Mhz being SCSI-1's maximum transfer rate in the best of conditions /
burst mode was beginning to make me wonder how much s-h-i-t-e this 90uS
figure was, since 5Mhz is 200uS. In pure logic it's entirely achievable so
I was starting to worry.
Thanks,
Aly |
|
| Back to top |
|
|
PeterD
Guest
|
| Posted:
Mon Oct 31, 2005 7:58 pm Post subject:
Re: SCSI 1 Command Timings |
|
|
On Mon, 31 Oct 2005 11:45:06 +0000 (UTC), "techie_alison"
<techie_alison@yahoo.com> wrote:
| Quote: | What's thrown a spanner in the works is that some bright spark has suggested
responses being in the region of 90nanoseconds. I mean, SCSI was about in
the early 80's and this 90nS figure to respond seems a bit extreme.
|
Where did you see this? |
|
| Back to top |
|
|
Rob Turk
Guest
|
| Posted:
Mon Oct 31, 2005 11:57 pm Post subject:
Re: SCSI 1 Command Timings |
|
|
"techie_alison" <techie_alison@yahoo.com> wrote in message
news:dk5h4h$h6q$1@nwrdmz03.dmz.ncs.ea.ibs-infra.bt.com...
| Quote: |
"PeterD" <peter2@hipson.net> wrote in message
news:ho8cm1pho8h1nnqfe4ct9rvrfi4l754htr@4ax.com...
On Mon, 31 Oct 2005 11:45:06 +0000 (UTC), "techie_alison"
techie_alison@yahoo.com> wrote:
What's thrown a spanner in the works is that some bright spark has
suggested
responses being in the region of 90nanoseconds. I mean, SCSI was about
in
the early 80's and this 90nS figure to respond seems a bit extreme.
Where did you see this?
Do you have the information I need?
|
Your best bet is to go to www.t10.org and browse the SCSI-2 draft specs that
are in place. From memory I don't think there's any such requirement of 90us
for responses. When the host adapter selects a target, it has to respond
within 250ms (yes, a quarter of a second..), after that the target takes
control of bus phases and transfers.
If your HBA supports synchronous transfers then a negotiation will happen
and you can tell the target how fast you can go. If you stick to the basic
asynchronous protocol then the REQ and ACK cycles are interlocked and you
can go as slow as you want. Some targets may not like going extremely slow
but your 100us cycles should be more than adequate.
Rob |
|
| Back to top |
|
|
Rob Turk
Guest
|
| Posted:
Tue Nov 01, 2005 1:42 am Post subject:
Re: SCSI 1 Command Timings |
|
|
"techie_alison" <techie_alison@yahoo.com> wrote in message
news:dk5prl$4ln$1@nwrdmz02.dmz.ncs.ea.ibs-infra.bt.com...
| Quote: |
Hi Rob,
Thanks for that :-)
Have found some further info this afternoon about all of this. That 250ms
figure sounds a lot more palatable, I won't know properly until the
prototype is in place and can then play about with timings in firmware.
I'm
down to 100nS instruction execution on the microcontroller now so have
built
in a little more space, which will hopefully pay off as a higher transfer
speed. Going down to 25nS instructions would have required complete new
programming equipment for the high end MCUs.
What I'm actually developing is an Atari ACSI <> IDE interface. The ACSI
standard being a significantly cut down version of SCSI with about 10
commands overall. We're also moving 16bit IDE <> 8bit ACSI without parity
to add just a little more spice. Hence the worry about timing constraints
when we have other things going on too. To indicate and jump to a new
interrupt will take about 500-800uS unless we're monitoring continually
then
it's about 100-200uS.
Admittedly I'm more interested in getting this to work to prove it can be
done. 5Mhz being SCSI-1's maximum transfer rate in the best of conditions
/
burst mode was beginning to make me wonder how much s-h-i-t-e this 90uS
figure was, since 5Mhz is 200uS. In pure logic it's entirely achievable
so
I was starting to worry.
Thanks,
Aly
|
I think you will have no problems at all then. Since you will be the Target
from a SCSI perspective, you control every aspect of timing and data flow.
All you have to do is respond to a SCSI Selection within 250ms and the bus
is all yours.
To put things in perspective, I built a CP/M system on Eurocard boards in
the early 90's and I designed a SCSI 'controller' for it from simple TTL
chips. The entire SCSI protocol was managed by the 6MHz Z80 CPU in assembly
and ran just fine ;-)
Rob |
|
| Back to top |
|
|
Folkert Rienstra
Guest
|
| Posted:
Tue Nov 01, 2005 5:04 am Post subject:
Re: SCSI 1 Command Timings |
|
|
"techie_alison" <techie_alison@yahoo.com> wrote in message news:dk5h4h$h6q$1@nwrdmz03.dmz.ncs.ea.ibs-infra.bt.com
| Quote: | "PeterD" <peter2@hipson.net> wrote in message news:ho8cm1pho8h1nnqfe4ct9rvrfi4l754htr@4ax.com...
On Mon, 31 Oct 2005 11:45:06 +0000 (UTC), "techie_alison"
techie_alison@yahoo.com> wrote:
What's thrown a spanner in the works is that some bright spark has suggested
responses being in the region of 90nanoseconds. I mean, SCSI was about in
the early 80's and this 90nS figure to respond seems a bit extreme.
Where did you see this?
Do you have the information I need?
|
That would be interesting ;-) |
|
| Back to top |
|
|
Folkert Rienstra
Guest
|
| Posted:
Tue Nov 01, 2005 6:13 am Post subject:
Re: SCSI 1 Command Timings |
|
|
"techie_alison" <techie_alison@yahoo.com> wrote in message news:dk5042$5u9$1@nwrdmz03.dmz.ncs.ea.ibs-infra.bt.com
| Quote: | Hello,
Would anyone have any links which detail the timings and command structures
of the SCSI 1 interface? I'm searching high and low on Google at the
moment. A SCSI host adapter is being developed.
|
http://www.t10.org/ftp/t10/drafts/s1/s1-r17b.txt
| Quote: |
What's thrown a spanner in the works is that some bright spark has suggested
responses being in the region of 90nanoseconds.
I mean, SCSI was about in the early 80's and this 90nS figure to respond seems a bit extreme.
|
Time to respond, don't think so, minimum time to sample lines, yes.
| Quote: |
I'm developing the interface with an embedded 10MIPS microcontroller so
100ns is about the best I can do.
What I'm ideally looking for is detailed information of loading commands
onto the bus and handshaking the control lines.
Thanks,
Aly, Cambridge UK
|
http://www.t10.org/ftp/t10/drafts/s1/s1-r17b.txt |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in