On Wednesday 12 January 2005 12:04, Jon LaBadie wrote:
>On Wed, Jan 12, 2005 at 11:18:00AM -0500, Eric Siegerman wrote:
>> On Tuesday 11 January 2005 16:40, Jon LaBadie wrote:
>> >Also, I think that if both types of devices exist on the same
>> > bus, the lower performance one determines the performance of the
>> > entire bus.
>>
>> In theory, this is *not* the case.  ...
>>
>> On Tue, Jan 11, 2005 at 10:48:17PM -0500, Gene Heskett wrote:
>>
>> As to the whole chain being restricted to the speed of the slowest
>> device, I've heard that, but it doesn't, on the face of it, make a
>> lot of sense to me
>
>Gene, Eric,
>
>Thanks for the correction.
>I'm glad I remembered to say I wasn't a scsi expert.
>My comments provide the proof.
>
>> <off-topic>
>>
>> > To many folks forget that a a scsi bus is indeed
>> > an rf transmission line, subject to the usual rules about vswr.
>> >
>> >From the context it's pretty clear what "vswr" means, but what
>>
>> does it stand for?
>> </off-topic>
>
>Wondered too, as I'm posting anyway, I'll guess: v??? standing wave
> reflections

VSWR=Voltage Standing Wave Reflections.  Back in rf days, because the 
signal is repetitive, very high voltages could develop at certain 
points in a transmission line driving an antenna that wasn't an ideal 
load, so they came to be known as standing waves.  And it doesn't 
take a lot of miss-match to burn a 6 & 1/8" 75 ohm rigid copper coax 
with 30kw peak power going into it at channel 19 (about 507mhz), 
making you call in a tower crew and replace the burnt section and 
clean the teflon carbon out of the other 1000 feet of it.  Lets just 
say thats expen$ive...  But now lets get down to the much profaned 
scsi buss, where these same physical laws apply even if they don't 
make junk and smoke out of expensive copper and teflon parts.

Understand we are dealing with a bus capable of responding to a 10 
nanosecond or less signal for starters, often lots less.

Some ascii art if you'll all bear with me & use a monospaced font.
The vertical scale:
3.0 Volts a solid logic 1
2.4 Volts unk upper bound, usually a 1 99% of the time
1.8 Volts unk logic state
1.2 Volts unk logic state
0.6 Volts unk lower bound, usually a 0 99% of the time
0.0 Volts a solid logic zero

Now assume a worst case condition of no termination just to make it 
simple.

Input pulse from typical wired-or driver, active pulldown.

t0 t+5ns t+10ns t+15ns  t+20ns t+25ns t+30ns
1-------|  |----------------------------
? |  |
? |  |
? |  |
? |  |
0 ----------------|

Same pulse 18" down the cable after the pulse has reached the end of 
the 24" cable and bounced because its not terminated.
           |-----| <- this could be clipped
           |     |   |-----|
           |     |   | |     |-----
           |     |   | |     |etc
------------------------|  |-----|     |   | |     |etc
   |  |     |   | |     |etc
   |  |     |   | |     |etc
   |  |     |   | |-----|etc
   |  |     |-----| 
   |---------------|

As you can see, after the main data pulse, there are several forays 
into unknown territory as the rest of the circuit slowly absorbs the 
unwanted echo.  Also, in the real world, there would be some wibbles 
in the initial logic zero time I drew as a flat line, but most of 
these are absorbed by the very solidly turned on driver transistor at 
the time.  These largly disappear if the terminations are correct, 
but only active terms really match the cable that well.

This loss of the upper boundary noise margin in particular is one of 
the reasons I'd love to see a special, 5.8 volt supply made available 
just for supplying term power to scsi busses, the extra .8 volts to 
make up for the losses because the average card designer gets his 
choice of a low voltage drop schotkey diode for the isolator 
overridden by some friggin bean counter who doesn't understand that 
his arbitrary replacement of a 50 cent diode with a 10 cent si diode, 
and its .7 volt loss, has just cost the product about 90% of its 
logic 1 noise margin by reducing the upoper resting voltage of the 
buss to the 2.6 volt area.  And we wind up advertising for virgins to 
sacrifice to make the darned thing work.  Hell, even a Ge diode would 
be a better choice if they think they cannot afford that 50 cent 
schotkey.

Hopefully this poor ascii art might clarify the situation a bit.  If 
I'm curious, I turn on my scope and look, but TBT, my 100 mhz dual 
trace scope is actually a bit slow to really display stuff like this.
Stuff like this really needs a 500mhz scope to display it anywhere 
near in real time, but the chips on a scsi buss are in fact that 
fast!

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
99.31% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attorneys please note, additions to this message
by Gene Heskett are:
Copyright 2005 by Maurice Eugene Heskett, all rights reserved.