[ibis-macro] Re: Follow-up on PDF discussion...
- From: "Mellitz, Richard" <richard.mellitz@xxxxxxxxx>
- To: "fangyi_rao@xxxxxxxxxxx" <fangyi_rao@xxxxxxxxxxx>, "Mirmak, Michael" <michael.mirmak@xxxxxxxxx>, "ibis-macro@xxxxxxxxxxxxx" <ibis-macro@xxxxxxxxxxxxx>
- Date: Fri, 13 Nov 2009 09:42:29 -0700
Before I start. There is NO data stream here! Get that notion out you head. We
start with an edge response however. This could be derived from a system
response.
A DATA pdf is the likelihood of data having a 1/0 or 0/1 transition. For purely
random data it's a 50/50 chance or uniform.
For example given purely random data the first edge has 50/50 chance of
occurring. It's either 0/1 or 0/1. The second edge that is 1 UI away also has a
50/50 change of occurring. Coding and specific data patterns change this i.e. a
DATA pdf. Now the superposition of edges with a 50/50 likelihood at the cursor
and an edge that has a 50/50 chance of happening at shifted UI yields a
likelihood of 25%. Now the edge that is delayed by 1 UI also has probability to
be off a little. We can call this jitter. The data edge shifting is really
convolution and the positional jitter is also convolution. So far we have a
data pdf used to convolve edges and jitter convolution to position edges. If
we continue this until we get a desired BER we have an exact non-extrapolated
solution. To make things more interesting we define a vertical axis probability
which is the likelihood of a signal really switching at the threshold. I.e.
vertical axis convolution. This whole process can be have co-channel
interference added as well. (crosstalk)
This PDF convolution ( edge, jitter, data, voltage) process can be referred to
as statistical simulation and what most IBIS-AMI has been using I call bit
stream convolution.
Ref:
http://www.designcon.com/media/2008/7-TP1--Richard_Mellitz.pdf
-----Original Message-----
From: fangyi_rao@xxxxxxxxxxx [mailto:fangyi_rao@xxxxxxxxxxx]
Sent: Thursday, November 12, 2009 1:00 PM
To: Mellitz, Richard; Mirmak, Michael; ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Re: Follow-up on PDF discussion...
Hi, Rich;
Thanks for your reply. Can you provide more clarification on what the PDF of
data edge represents?
1. It seems to be a function of both time and voltage. Correct?[Mellitz,
Richard] not really
2. Does it include effects of Tx jitter?[Mellitz, Richard] no
3. Does it include effects of ISI?[Mellitz, Richard] not really
4. Does the jitter PDF in your email represent Tx jitter or Rx sampling
jitter?[Mellitz, Richard] no
5. Do you need separate PDFs for logic 1 and logic 0?[Mellitz, Richard] unless
symmetry holds.
Thanks,
Fangyi
-----Original Message-----
From: Mellitz, Richard [mailto:richard.mellitz@xxxxxxxxx]
Sent: Thursday, November 12, 2009 6:20 AM
To: RAO,FANGYI (A-USA,ex1); Mirmak, Michael; ibis-macro@xxxxxxxxxxxxx
Subject: RE: [ibis-macro] Re: Follow-up on PDF discussion...
Hi Fangyi,
The PDF we are talking about is the PDF of data edges. That PDF convolves in
with a jitter PDF in time and voltage PDF in for receive sensitivity. Notice
there is no data stream here at all. We get an exact BER CDF in this manner.
I.e. no extrapolation.
... Rich
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx]
On Behalf Of fangyi_rao@xxxxxxxxxxx
Sent: Tuesday, November 10, 2009 2:48 PM
To: Mirmak, Michael; ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Re: Follow-up on PDF discussion...
Hi, Michael;
Isn't Tx_Jitter in AMI the same as Tx jitter PDF? I think that's all needed in
Casper's approach.
Thanks,
Fangyi
[Mellitz, Richard]
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx]
On Behalf Of Mirmak, Michael
Sent: Tuesday, November 10, 2009 11:12 AM
To: IBIS-ATM
Subject: [ibis-macro] Follow-up on PDF discussion...
FYI, in an earlier mail, I asked whether PDFs could be used directly under
IBIS-AMI. To clarify, the usage of PDF was as "probability distribution
function."
The reason for the question, plus some concerns about TX jitter treatment, are
outlined in a paper by Bryan Casper, Ganesh Balamurugan et al, entitled "Future
Microprocessor Interfaces: Analysis, Design and Optimization." It's available
through IEEE and other sources. The full citation follows.
IBIS-AMI support for this style of data input would be highly useful.
- MM
Casper, B.; Balamurugan, G.; Jaussi, J.E.; Kennedy, J.; Mansuri, M., "Future
Microprocessor Interfaces: Analysis, Design and Optimization," Custom
Integrated Circuits Conference, 2007. CICC '07. IEEE , vol., no., pp.479-486,
16-19 Sept. 2007
URL:
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=4405777&isnumber=4405667
ISBN: 978-1-4244-1623-3
Additional material in a similar vein is available here:
Balamurugan, G.; Casper, B.; Jaussi, J.E.; Mansuri, M.; O'Mahony, F.; Kennedy,
J., "Modeling and Analysis of High-Speed I/O Links," Advanced Packaging, IEEE
Transactions on , vol.32, no.2, pp.237-247, May 2009
URL:
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=4811945&isnumber=4982886
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector: //www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector: //www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
---------------------------------------------------------------------
IBIS Macro website : http://www.eda.org/pub/ibis/macromodel_wip/
IBIS Macro reflector: //www.freelists.org/list/ibis-macro
To unsubscribe send an email:
To: ibis-macro-request@xxxxxxxxxxxxx
Subject: unsubscribe
Other related posts:
