Question for SCSI folks

Gerry Houlder Gerry_Houlder at notes.seagate.com
Wed Nov 1 10:52:13 PST 1995


Faisal Haq's questions were:

1) What are min/max values for (a and c) out there, today? Is
   an assumption of 55ns minimum, for each of these correct? 

My answer: These delays are dependent upon the initiator protocol
   chip, waveform rise times, and cable turnaround delay. A typical
   rise time is 7 ns, a typical turnaround delay for 10 foot cable is
   (10 ft) * (1.7 ns/ft) * (2  [need to go up and back on the cable]) = 34 ns.
   7 + 34 = 41 ns even if the turnaround delay in the protocol chip is 0.
   Also remember that differential interface needs to add delay through
   the transmitter and receiver (another 40 ns typical) to the above numbers.

2) As a target, is it my responsiblity to ensure that (a+b+c+d)
   is always a fixed number? If so, what is that number? I've heard
   175ns (5.7MB/s), 200ns (5MB/s) among others. Is there a min,max range
   to follow? In any case, this means I would have to shorten (b and d) 
   to compensate for really slow hosts.

My answer: No, the target is only responsible to make delay periods b and
   d as short as possible. The target cannot "compensate" for a slow host.

3) Is there a minimum assertion/deassertion time for req, in
   addition to maintaining proper handshake protocol? I've heard
   25ns for assertion and 75ns for deassertion. Is this true?

My answer: The only minumums are determined by the propagation delays
   called out in the answer to question 1. Note that even a zero turnaround 
delay
   in the protocol chips at both ends will still result in 82 ns width with 10 
ft. cable
   (for S/E).

4) Is there ANY data hold time requirement for targets, past the leading 
   edge of ACK.

My answer: The data setup and hold times are specified in the standard. Look
   in the "Logical Characteristics" section, "information transfer phases" 
sub-section
   of SCSI-2.




More information about the T10 mailing list