Re: [SI-LIST] : Non-standard bus termination (2nd posting)

Ron Miller (rmiller@Brocade.COM)
Thu, 23 Sep 1999 09:23:52 -0700

--------------C749299B7AF85A87AC356D16
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Steve

You may be right for the case where much of the data values are fixed at one
state. I have
assumed that this is a data bus with significant activity, pseudo-random and
that the DC
termination value is not extremely important except that it should not draw too
much DC
current as with a termination to ground.

The purpose as I understood it was to provide some type of RF termination to
remove reflections
while reducing the number of components from two(an R and a C). In many cases a
DC termination
voltage works well with a simgle R on each leg. The single cap just makes the
DC value float some
around the average DC value of the whole data bus.

As far as the power dissipated and the terminating voltage go, you can't get
something for nothing.

Ron Miller

sweir wrote:

> Ron,
>
> I have to agree with D.C. that this sounds like a really bad idea. Let's
> assume that you have an arbitrarily large capacitor. Then, the voltage on the
> capacitor side of the termination resistors will be essentially constant for
> any signal period we care to evaluate, but may wander from GND to Vcc over
> time. Unless the bus has a state-space average value which happens to be a
> good terminating value, we get the worst of all worlds:
>
> 1. The power consumption of a parallel termination to either Vcc or GND for
> the fully charged and discharged states.
> 2. Non-uniform termination current, and therefore variable settling time at
> the load, which depends on the states of adjacent lines, ( read crosstalk
> distortion ).
>
> In the best case, the state-space average is a convenient terminating value,
> in which case, this just emulates an active termination. Considering that
> active termination can be had fairly easily, why rely on the average signal
> content to set the voltage?
>
> Regards,
>
>
> Steve.
> At 02:48 PM 9/22/99 -0700, you wrote:
>
>> Hi DC
>>
>> Make it a big cap.
>>
>> The cap is used to keep from DC loading the line.
>>
>> The resistor provides the line AC termination.
>>
>> Ron Miller
>>
>> "D. C. Sessions" wrote:
>>
>> > Ron Miller wrote:
>> > >
>> > > Use tie all the common sides of the resistors together, and use one
>> > capacitor from the
>> > > common point to ground.
>> >
>> > Not exactly a Real Good Idea unless you have lots of timing margin.
>> > The problem is that this creates a lot of jitter (edge times depend
>> > on prior state of all the lines.)
>> >
>> > > Kassem Abdallah wrote:
>> > >
>> > > > > Has anybody out of there any suggestion about the following problem?
>> > (I think it
>> > > > > is quite a common one).
>> > > > >
>> > > > > I?ve got a board with a non-standard bus of 34 traces. The bus drives
>> > a group of
>> > > > > memories.
>> > > > > The lenght of the bus is necessarily higher than the critical
>> > lenght (lcr =
>> > > > > tr/2tpd): a termination could be useful. The problem: there are so
>> > many traces
>> > > > > that one has to use too many components (34 resistors and 34
>> > capacitors for AC
>> > > > > parallel termination).
>> > > > >
>> > > > > What shall I do?
>> > > > > Are there any special components (array) to terminate this type of
>> > bus?
>> > > > > Shall I terminate the bus or not? (In past experience not always a
>> > termination
>> > > > > was mandatory)
>> > > >
>> > > > I am not the expert on SI Issues, but there are several resistor_packs
>> > > > for termination available from a variety of vendors. However, You have
>> > > > to be very carefull, because these packs sometimes cause noise and make
>> >
>> > > > things worst. It depends on how the resistors network is packaged for
>> > > > current/GND leads.( I suggest to look carfully at the schematics and do
>> >
>> > > > some noise analysis).
>> > > >
>> > > >
>> > > > Also, I "think" a 34-bit bus is not that WIDE, and if it is critical
>> > > > data path, you might be better off terminating with single loads( may
>> > > > be you can stick them on the backside of your board ?? It is a
>> > > > compromise).
>> > > >
>> > > > ********************
>> > > > |Kassem M. Abdallah|
>> > > > |Digital Designer |
>> > > > |EMC Corp. |
>> > > > |Hopkinton MA 01748|
>> > > > ********************
>> > > >
>> > > > **** To unsubscribe from si-list: send e-mail to
>> > majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
>> > si-list, for more
>> > > > help, put HELP. si-list archives are accessible at
>> > http://www.qsl.net/wb6tpu/si-list ****
>> > >
>> > > --
>> > > Ronald B. Miller _\\|//_ Signal Integrity Engineer
>> > > (408)487-8017 (' 0-0 ') fax(408)487-8017
>> > > ==========0000-(_)0000===========
>> > > Brocade Communications Systems, 1901 Guadalupe Parkway, San Jose, CA
>> > 95131
>> > > rmiller@brocade.com, rbmiller@sjm.infi.net
>> > >
>> > >
>> >
>> > --
>> > D. C. Sessions
>> > dc.sessions@vlsi.com
>> >
>> > **** To unsubscribe from si-list: send e-mail to
>> > majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
>> > si-list, for more help, put HELP. si-list archives are accessible at
>> > http://www.qsl.net/wb6tpu/si-list ****
>>
>> --
>> Ronald B. Miller _\\|//_ Signal Integrity Engineer
>> (408)487-8017 (' 0-0 ') fax(408)487-8017
>> ==========0000-(_)0000===========
>> Brocade Communications Systems, 1901 Guadalupe Parkway, San Jose, CA 95131
>> rmiller@brocade.com, rbmiller@sjm.infi.net
>>
>>
>

--
Ronald B. Miller  _\\|//_  Signal Integrity Engineer
(408)487-8017    (' 0-0 ') fax(408)487-8017
     ==========0000-(_)0000===========
Brocade Communications Systems, 1901 Guadalupe Parkway, San Jose, CA  95131
rmiller@brocade.com,  rbmiller@sjm.infi.net

--------------C749299B7AF85A87AC356D16 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit

<!doctype html public "-//w3c//dtd html 4.0 transitional//en"> Steve

You may be right for the case where much of the data values are fixed at one state.  I have
assumed that this is a data bus with significant activity, pseudo-random and that the DC
termination value is not extremely important except that it should not draw too much DC
current as with a termination to ground.

The purpose as I understood it was to provide some type of RF termination to remove reflections
while reducing the number of components from two(an R and a C).  In many cases a DC termination
voltage works well with a simgle R on each leg.   The single cap just makes the DC value float some
around the average DC value of the whole data bus.

As far as the power dissipated and the terminating voltage go, you can't get something for nothing.

Ron Miller

sweir wrote:

 Ron,

I have to agree with D.C. that this sounds like a really bad idea.  Let's assume that you have an arbitrarily large capacitor.  Then, the voltage on the capacitor side of the termination resistors will be essentially constant for any signal period we care to evaluate, but may wander from GND to Vcc over time.  Unless the bus has a state-space average value which happens to be a good terminating value, we get the worst of all worlds:

1. The power consumption of a parallel termination to either Vcc or GND for the fully charged and discharged states.
2. Non-uniform termination current, and therefore variable settling time at the load, which depends on the states of adjacent lines, ( read crosstalk distortion ).

In the best case, the state-space average is a convenient terminating value, in which case, this just emulates an active termination.  Considering that active termination can be had fairly easily, why rely on the average signal content to set the voltage?

Regards,
 

Steve.
At 02:48 PM 9/22/99 -0700, you wrote:

Hi DC

Make it a big cap.

The cap is used to keep from DC  loading the line.

The resistor provides the line AC termination.

Ron Miller

"D. C. Sessions" wrote:

Ron Miller wrote:
>
> Use tie all the common sides of the resistors together, and use one capacitor from the
> common point to ground.

Not exactly a Real Good Idea unless you have lots of timing margin.
The problem is that this creates a lot of jitter (edge times depend
on prior state of all the lines.)

> Kassem Abdallah wrote:
>
> > > Has anybody out of there any suggestion about the following problem? (I think it
> > > is quite a common one).
> > >
> > > I?ve got a board with a non-standard bus of 34 traces. The bus drives a group of
> > > memories.
> > > The  lenght  of  the  bus  is necessarily higher than the critical lenght (lcr =
> > > tr/2tpd):  a  termination could be useful. The problem: there are so many traces
> > > that  one  has to use too many components (34 resistors and 34 capacitors for AC
> > > parallel termination).
> > >
> > > What shall I do?
> > > Are there any special components (array) to terminate this type of bus?
> > > Shall  I  terminate the bus or not? (In past experience not always a termination
> > > was mandatory)
> >
> > I am not the expert on SI Issues, but there are several resistor_packs
> > for termination available from a variety of vendors. However, You have
> > to be very carefull, because these packs sometimes cause noise and make
> > things worst. It depends on how the resistors network is packaged for
> > current/GND leads.( I suggest to look carfully at the schematics and do
> > some noise analysis).
> >
> >
> >  Also, I "think" a 34-bit bus is not that WIDE, and if it is critical
> >  data path, you might be better off terminating with single loads( may
> >  be you can stick them on the backside of your board ?? It is a
> >  compromise).
> >
> > ********************
> > |Kassem M. Abdallah|
> > |Digital Designer  |
> > |EMC Corp.         |
> > |Hopkinton MA 01748|
> > ********************
> >
> > **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more
> > help, put HELP.  si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
>
> --
> Ronald B. Miller  _\\|//_  Signal Integrity Engineer
> (408)487-8017    (' 0-0 ') fax(408)487-8017
>      ==========0000-(_)0000===========
> Brocade Communications Systems, 1901 Guadalupe Parkway, San Jose, CA  95131
> rmiller@brocade.com,  rbmiller@sjm.infi.net
>
>

--
D. C. Sessions
dc.sessions@vlsi.com

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP.  si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****

-- 
Ronald B. Miller  _\\|//_  Signal Integrity Engineer
(408)487-8017    (' 0-0 ') fax(408)487-8017                 
     ==========0000-(_)0000=========== 
Brocade Communications Systems, 1901 Guadalupe Parkway, San Jose, CA  95131
rmiller@brocade.com,  rbmiller@sjm.infi.net
 
-- 
Ronald B. Miller  _\\|//_  Signal Integrity Engineer
(408)487-8017    (' 0-0 ') fax(408)487-8017                 
     ==========0000-(_)0000=========== 
Brocade Communications Systems, 1901 Guadalupe Parkway, San Jose, CA  95131
rmiller@brocade.com,  rbmiller@sjm.infi.net
  --------------C749299B7AF85A87AC356D16-- **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****