Comments on Ultra SCSI

Tak Asami (Asami asami at
Mon Feb 14 18:12:48 PST 1994

Steve Finch says:
> Is Ultra SCSI for differential only?  I think it will have to be, as 
> single ended has significant problems.  

I must agree with what he says.
And I'll even ask an open question.  Are we interested in doing Ultra
SCSI at all?

It would be difficult to precisely control the driver slew rate, and 
you probably wouldn't like the price of the product if we did.
On top of that, in our experience, the driver with 5nsec rise/fall time
runs into waveform integrity problems in at least one of your customer's 
systems.  Guaranteed.

Whenever someone used significant amount of stubs, or sufficnent impedance
discontinuities, like several cable connectors, back planes and ribbon
cable combinations, etc. (which cover just about every system out there)
generates enough reflections that reaches 100+MHz in frequency.
This is why we worked hard to slow down our drivers to run very, very slowly.
But now it won't work if you speed up the cycle time (it already looks 
like a triangular wave at 100nsec cycle in some cases).

With advent of shorter cycle time (50nsec or less), the reflection from 
which you suffer may be generated from several cycles before, unless
cable length and stub rules are significantly tightned up.

The thing is, for fast SCSI, the cable length is already "recommended"
to be 3m or less.  If you make it any shorter, is it worth the trouble
it will take?  Who's going to buy this stuff?

John Llohmeyer puts it at "80+%" of the market is single ended.
I put more like "95%" is single ended.  With that being the case, is
it worth the trouble of working Ultra SCSI at all?

We've put a lot of work to get the fast SCSI working up to now.
I can't believe someone wants to put additional obstacle to overcome so
we'll get to do it over again.

I am trying to ask a basic question what is driving us to do it.
Let's not do something simply because it's interesting; figure what
problem we are trying to solve.

By the way, I believe the meeting will be on March 2, right?
You guys are trying to confuse me......

Tak Asami
Western Digital Corporation
OEM Storage Product
(Office) 714 932-7621, (Fax) 714 932-6496
Internet: asami at

More information about the T10 mailing list