[SI-LIST] Re: Circle bus topology; Circular Firing Squad?

  • From: Scott McMorrow <scott@xxxxxxxxxxxxx>
  • To: silist <si-list@xxxxxxxxxxxxx>
  • Date: Thu, 26 Jul 2007 19:24:53 -0400

The loop topology busses I am aware of were used to drive the address 
lines on large memory systems.  Back in the day, memory address busses 
were often daisy chained from chip to chip.  If the memory driver was 
series terminated, the time for the bus to stabilize would be one round 
trip down the line and back.  If you got clever and looped the bus back 
on itself, the the driver is effectively driving two lines.  Four nice 
things occur.  First, a stronger driver could be used to drive the two 
lines in parallel.  Thus twice the power could be driven into the same 
total capacitive load without the need to increase the  trace width from 
50 ohm nominal.  Second, if a driver with series termination that was 
adjusted to be about 10-20% stronger than perfect termination was used, 
you could minimize the distributed RC time constant, by allowing a bit 
of overshoot and ring back.  Third, since the loop was closed, the bus 
would stabilize in one trip around the loop, which is 1/2 the time it 
takes for a daisy chained bus to stabilize.  Finally, if the memory was 
placed correctly, the routing could be quite easy.  If the driver 
already had the right output impedance, like some of the old TTL drivers 
we used to use, you didn't even need external series resistors, since a 
loop is perfectly balanced, and the traveling waves will meet each other 
right at the driver
However, if you analyze the structure, you'll also realize that you can 
do exactly the same thing with a driver driving two lines with two daisy 
chained lines, from two series resistors.  A low impedance driver and 
two series resistors are needed to dampen any imbalances between the two 
sides of the bus.  This was the much more robust solution.  Of course, 
today we used balanced matched Tee structures on memory busses, and 
dampen them with parallel series elements.  The problems are the same, 
but the data rates are a bit faster than they were in 1995.  
Unfortunately, DDR address drivers are generally not strong enough at 
the slow corner, to allow us to get to fancy in our topology splits and 
termination.

BTW, it was a real easy "quick fix" for a problem memory module 
(remember back when these were physically large things?).  The addition 
of one rework wire to create the loop would speed up an address bus 
significantly.   It wasn't testable, but who cared?  It fixed a 
production line problem due to a bad design.


Scott McMorrow
Teraspeed Consulting Group LLC
121 North River Drive
Narragansett, RI 02882
(401) 284-1827 Business
(401) 284-1840 Fax

http://www.teraspeed.com

Teraspeed® is the registered service mark of
Teraspeed Consulting Group LLC



olaney@xxxxxxxx wrote:
> No, the waves in each direction have the same phase at the source.  They
> would meet and *reinforce* each other at the electrical halfway point,
> continue around the loop in both directions, and meet again at the
> source.   Unless the source was back terminated, they would reflect from
> the low source impedance, this time in inverted phase from the original
> sense, and repeat the exercise until losses cause the waveform to die
> out.  This is just a wierd example of exactly what happens on a regular
> single ended trace with an open end, except that you are driving two
> traces of identical length in parallel that have their ends shorted
> together.  Also note that to back terminate the source, its impedance
> would have to be half that of the transmission path, because it sees both
> ends of the line in parallel.  Example: 50 ohm source to drive a 100 ohm
> looped trace.  Or, you could terminate the halfway roundtrip point the
> same way.  Or both, if you double the drive level (great for video but
> not good practice for logic signals).  Basically, ring layouts in copper
> foil are strange birds, useful for narrowband microwave purposes, but not
> of much use for wideband, time domain waveforms for all the reasons
> mentioned.
>
> Orin Laney
>
> On Thu, 26 Jul 2007 14:21:57 -0700 "Salkow, Steven"
> <steven.salkow@xxxxxxxx> writes:
>   
>> Mr Townsend et All made good points.
>>
>> DFM software would flag this as an error. 
>> Electrically, I think the idea is that the propagating standing 
>> magnetic
>> wave of each arc path would exactly cancel each other when the waves 
>> met
>> and no reflections would reflect back to the source. If one side of 
>> a
>> path was not exactly equal, the point at which the waves cancel 
>> would
>> not be at an physical midpoint rather at an electrical midpoint 
>> which
>> may fall on a driven node. The point made by others already is 
>> better
>> schemes exist already such as star distribution where end point
>> termination may be used to easily terminate a line with quite
>> predictable results. 
>> ss
>>
>> -----Original Message-----
>> From: si-list-bounce@xxxxxxxxxxxxx 
>> [mailto:si-list-bounce@xxxxxxxxxxxxx]
>> On Behalf Of Townsend, Fred
>> Sent: Thursday, July 26, 2007 9:05 AM
>> To: David.Carney@xxxxxxxxxx; si-list@xxxxxxxxxxxxx
>> Subject: [SI-LIST] Re: Circle bus topology; Circular Firing Squad?
>>
>> David:
>> I have to ask what would be gained from such a topology? Ron and 
>> Scott
>> both make some very good points. In the case of the Mux bus 
>> (1553)the
>> ring encompasses the airplane and the double ring structure gives
>> increased reliability. Microwave can make use of some structures 
>> like
>> directional couplers to help reduce reflections. If rings are good 
>> why
>> aren't we still using the token ring? Again the token ring was over 
>> an
>> area much bigger than a PCB. A ring in a PCB would have all of the
>> problems with no apparent gain. Think about your PCB router. Rings 
>> would
>> drive the router nuts?
>>
>> Fred Townsend
>>
>> -----Original Message-----
>> From: si-list-bounce@xxxxxxxxxxxxx 
>> [mailto:si-list-bounce@xxxxxxxxxxxxx]
>> On Behalf Of David Carney (Neenah)
>> Sent: Thursday, July 26, 2007 6:47 AM
>> To: si-list@xxxxxxxxxxxxx
>> Subject: [SI-LIST] Circle bus topology
>>
>> Has anyone ever experimented with a circle bus topology.  The basic
>> concept would be a bus with several devices attached.  They would be
>> routed in a daisy chain topology, and then the two ends of the daisy
>> chain would be connected together.  The PCB routing would look like 
>> a
>> circle or a loop for each net on the bus.  Pointers to references 
>> such
>> as papers or application notes would be greatly appreciated.  I'm
>> particularly interested in signal integrity and EMC implications of 
>> this
>> topology.
>> =20
>> Thanks.
>> =20
>> =20
>> ----------
>> David Carney
>> Senior Hardware Engineer
>> Plexus Corp.
>> Phone - 920-751-5646
>> =20
>> ------------------------------------------------------------------
>> To unsubscribe from si-list:
>> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject 
>> field
>>
>> or to administer your membership from a web page, go to:
>> //www.freelists.org/webpage/si-list
>>
>> For help:
>> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>
>>
>> List technical documents are available at:
>>                 http://www.si-list.net
>>
>> List archives are viewable at:    =20
>>                 //www.freelists.org/archives/si-list
>> or at our remote archives:
>>                 http://groups.yahoo.com/group/si-list/messages
>> Old (prior to June 6, 2001) list archives are viewable at:
>>                  http://www.qsl.net/wb6tpu
>>  =20
>>
>> ------------------------------------------------------------------
>> To unsubscribe from si-list:
>> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject 
>> field
>>
>> or to administer your membership from a web page, go to:
>> //www.freelists.org/webpage/si-list
>>
>> For help:
>> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>
>>
>> List technical documents are available at:
>>                 http://www.si-list.net
>>
>> List archives are viewable at:     
>>                 //www.freelists.org/archives/si-list
>> or at our remote archives:
>>                 http://groups.yahoo.com/group/si-list/messages
>> Old (prior to June 6, 2001) list archives are viewable at:
>>                  http://www.qsl.net/wb6tpu
>>   
>>
>> ------------------------------------------------------------------
>> To unsubscribe from si-list:
>> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject 
>> field
>>
>> or to administer your membership from a web page, go to:
>> //www.freelists.org/webpage/si-list
>>
>> For help:
>> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>
>>
>> List technical documents are available at:
>>                 http://www.si-list.net
>>
>> List archives are viewable at:     
>>                 //www.freelists.org/archives/si-list
>> or at our remote archives:
>>                 http://groups.yahoo.com/group/si-list/messages
>> Old (prior to June 6, 2001) list archives are viewable at:
>>                  http://www.qsl.net/wb6tpu
>>   
>>
>>
>>
>>     
>  
> ------------------------------------------------------------------
> To unsubscribe from si-list:
> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>
> or to administer your membership from a web page, go to:
> //www.freelists.org/webpage/si-list
>
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
>
> List technical documents are available at:
>                 http://www.si-list.net
>
> List archives are viewable at:     
>               //www.freelists.org/archives/si-list
> or at our remote archives:
>               http://groups.yahoo.com/group/si-list/messages
> Old (prior to June 6, 2001) list archives are viewable at:
>               http://www.qsl.net/wb6tpu
>   
>
>
>   


------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field


List technical documents are available at:
                http://www.si-list.net

List archives are viewable at:     
                //www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu
  

Other related posts: