Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [10GBASE-CX4] Xmt Level




Steve,
	The thing about continually moving the Tx amplitude down because there
is at least a 20dB loss in the system therefor the change at the
receiver is very little adds up after a while. Taking you suggestion of
having the minimum peak amplitude be 750mVppd and compare that to say
moving it up to 1Vppd and instead of a 5mV difference at the receiver
you now have a 25mV difference at the receiver.  That could make or
break the reciver design.
	5mV difference at the receiver might not seem like a big deal but if
you have a 3-sigma offset for the receive comparator around 20-30mV then
5mV is 16-25% increase in margin!
	It is also unreasonable to expect to be able to cover every single
design possiblitiy in every process that exists.  I agree that you can
create a design in one of today's existing processes that will have a
very tough tijme reaching 800mV but 750mV is acceptable.  The thing is a
very typical design in a very typical process can meet 800mV with more
than 200mV of margin.

Howard Baumer


"Dreyer, Steve" wrote:
> 
> Shawn,
> 
> Two processes we use have issues at 800mV.  Maybe other processes
> can support it, but not all.  I think spec should be inclusive
> of all processes, and moving down by 50mV  to 750mV to be more
> inclusive seems reasonable and not too big a price to pay.
> 
> As to the multiple supply issue, I agree that many chips use multiple
> supplies because of the reasons you stated, but that is not true
> in all cases.   And many customers want single supply, low power
> operation.  It seems that that spec should be designed
> to allow this and not exclude it.
> 
> Steve
> 
> > -----Original Message-----
> > From: Rogers, Shawn [mailto:s-rogers@ti.com]
> > Sent: Friday, January 03, 2003 11:29 AM
> > To: Dreyer, Steve; stds-802-3-10GBCX4@ieee.org
> > Subject: RE: [10GBASE-CX4] Xmt Level
> >
> >
> > Steve, we do not see a problem with 800mVp-p differential at
> > 130nm using no
> > special tricks or transistor construction.   Our 130nm
> > process is pretty
> > generic so I would suggest that the issue is not the amplitude.
> >
> > I would prefer to leave the xmt amplitudes as is.
> >
> > BTW, most, if not all, 130nm processes require multiple supplies for
> > backward I/O compatibility, so the point of adjusting this
> > spec to allow a
> > single supply device is not valid.  I'd like to have a single
> > supply device
> > too, but the fact remains that I/O voltage levels always lag process
> > technology.  This dictates a world of where IC's have
> > multiple supplies.
> >
> > Regards,
> > Shawn
> >
> > ____________________________________________
> > Shawn Rogers,   PMP               s-rogers@ti.com
> > High Speed Serial Link Marketing
> > Texas Instruments
> > 12500 TI Boulevard / M/S 8732/ Dallas, Texas 75243
> > Office: 214.480.2678                        Cell: 214.549.4868
> > ______________________________________
> >
> >
> > -----Original Message-----
> > From: Dreyer, Steve [mailto:steve.dreyer@intel.com]
> > Sent: Friday, January 03, 2003 10:15 AM
> > To: stds-802-3-10GBCX4@ieee.org
> > Subject: RE: [10GBASE-CX4] Xmt Level
> >
> >
> > Howard,
> >
> > The process you are using is probably different than ours,
> > that is why we get different results.   But  we have seen
> > 800mV being difficult to impossible on on 2 different processes.
> > So, I would make the claim that it is a real issue that others
> > could experience as well.
> >
> > And lowering the amplitude by 50mV is not that big a deal for the
> > receiver.
> >
> > Using thick ox devices is not a very good solution because
> > it requires a higher supply, which means more power, and now the
> > IC requires dual power supplies.
> >
> > Our customers want a single 1.2V supply and low power.  I think if
> > lowering the level by 50mV can make this achievable  by everyone,
> > that is a big plus.
> >
> > Steve
> >
> >
> >
> > > -----Original Message-----
> > > From: Howard A. Baumer [mailto:hbaumer@broadcom.com]
> > > Sent: Friday, January 03, 2003 8:08 AM
> > > To: Dreyer, Steve
> > > Cc: stds-802-3-10GBCX4@ieee.org
> > > Subject: Re: [10GBASE-CX4] Xmt Level
> > >
> > >
> > > Steve,
> > >     There areexisting 130nm parts that put out > 800mVppd.
> > > Our simulations
> > > for our designs shows that this is possible.  For the 90nm process a
> > > designer can use the higher voltage fets to achieve even a 1 or 1.2V
> > > output.  Yes these fets are slower than the stnadard fets
> > of 90nm but
> > > they will be faster than the stnadard fets for 130nm.
> > >
> > >
> > > Howard
> > >
> > >
> > > "Dreyer, Steve" wrote:
> > > >
> > > > Howard,
> > > >
> > > > My comment to your comment to my comment below.
> > > >
> > > > Steve
> > > >
> > > > > "Dreyer, Steve" wrote:
> > > > > >
> > > > > > Howard,
> > > > > >
> > > > > > Took a quick look at the proposed draft and have the following
> > > > > > top level comments:
> > > > > >
> > > > > > 1. The min xmt amplitude of 800mV is probably not doable on a
> > > > > >    chip with 1.2V only supply when all worst case
> > conditions are
> > > > > >    considered, it probably needs to be dropped another 50mV
> > > > > >    or so.  I plan to have a quick presentation showing
> > > > > >    this next week.  For reference, the min XAUI level is
> > > > > >    a lot lower than 800mV using the far end method.
> > > > >
> > > > > <HAB>
> > > > >       The 800mVpp is for a differential signal so that is
> > > > > only a 400mV single
> > > > > ended swing at the driver output, this is attainable with a 1.2V
> > > > > supply.  Also the 800mVpp min is not the minimum the driver
> > > > > puts out, it
> > > > > is the minimum for the lower limit of the peak in the
> > > transmit output
> > > > > template you reference in #3 below.  This will have to be
> > > > > clearified in
> > > > > the text.
> > > > > <HAB>
> > > > >
> > > >
> > > > <SD>
> > > > Our simulations show that 800mV dif pp (or 400mV single
> > ended) can't
> > > > be done at 1.2V when worst case conditions are put in
> > > place.  We have
> > > > simulated with both a 130nm and 90nm process.
> > > >
> > > > Propose we reduce the min level a bit based on these simulation to
> > > > maybe 750mV (will have exact number next week).
> > > >
> > > > The 750mV is still larger than current XAUI min (if far
> > > > end spec is used), and reducing xmt level by 50mV reduces receive
> > > > sensitivity by 5mV, not a big impact on receive.
> > > >
> > > > I plan to show simulation  results next week at
> > Vancouver.  Maybe we
> > > > should wait until then to discuss further.
> > > >
> > > > Steve
> > >
> >