Ambrish, Thanks for your comments. I was thinking along the same lines as you were: if the user is not aware of what jitter terms are included inside the DLL and he or she assigns a value to one of the jitter parameters, the simulator may produce excessive jitter. So the IBIS committee needs to be clear about how we define each jitter term and its association to a physical source. And the model developer needs to tell the user what jitter terms are already included in the DLL. I'll draw up a block diagram and share it with the list. Maybe we can agree on a common jitter model as we have agreed on a common flow. Greg Edlund Senior Engineer Signal Integrity and System Timing IBM Systems & Technology Group 3605 Hwy. 52 N Bldg 050-3 Rochester, MN 55901 From: Ambrish Varma <ambrishv@xxxxxxxxxxx> To: Gregory R Edlund/Rochester/IBM@IBMUS Cc: "ibis-macro@xxxxxxxxxxxxx" <ibis-macro@xxxxxxxxxxxxx>, "ibis-macro-bounce@xxxxxxxxxxxxx" <ibis-macro-bounce@xxxxxxxxxxxxx> Date: 08/24/2011 09:45 AM Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Thanks for your response Greg, My question was right after the ATM call – and there was a discussion of how the Rx takes care of the jitter, what jitter does the model maker add and how the EDA tool takes care of the rest – so the question might be a little confusing – sorry about that. I was following up on Fangyi’s remarks earlier – “Rx_Rj, Rx_Sj and Rx_DCD are Rx model parameters. They are defined as impairments external to receiver and serve the purpose of jitter budget analyses”. When I was talking about budgets, I was referring to the jitter that the EDA tool will inject to stress the Rx. This usually comes from the specifications. So I was wondering if there is any chance of double counting of the jitters if both the EDA tool and the model account for the same jitter. You suggestion of the block diagram makes a lot of sense and would help in clearing the cloud over jitter analysis for AMI simulations. Thanks, Ambrish. Ambrish Varma | Member of Consulting Staff P: 978.262.6431 www.cadence.com From: Gregory R Edlund [mailto:gedlund@xxxxxxxxxx] Sent: Wednesday, August 24, 2011 9:00 AM To: Ambrish Varma Cc: ibis-macro@xxxxxxxxxxxxx; ibis-macro-bounce@xxxxxxxxxxxxx Subject: Re: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Ambrish, When you say "jitter budget," are you talking about something the simulator keeps track of and reports to the user at the end of a simulation? When I use the term "jitter budget," I usually mean something outside of the simulator entirely, i.e. a benchmark for comparison that I use to convince myself I believe the simulator. The comments of Ambrish and others point out the need to have a clear understanding of the Physical Sources of non-channel jitter, both inside and outside the PHY. I would like to propose a block diagram, similar to what Walter has done already in "Time Domain GetWave Flow." If we hook up a TX circuit to an oscilloscope, we should see the same jitter components that we see in simulation. I volunteer to make the diagram with input of others on the list. Whatever solution the committee arrives at for this BIRD needs to allow the simulator to mimic the hardware. Greg Edlund Senior Engineer Signal Integrity and System Timing IBM Systems & Technology Group 3605 Hwy. 52 N Bldg 050-3 Rochester, MN 55901 Inactive hide details for Ambrish Varma ---08/23/2011 03:34:59 PM---Just wanted some understanding on how jitter budgets work: Ambrish Varma ---08/23/2011 03:34:59 PM---Just wanted some understanding on how jitter budgets work: Are the jitter inside the tx and rx a par From: Ambrish Varma <ambrishv@xxxxxxxxxxx> To: "ibis-macro@xxxxxxxxxxxxx" <ibis-macro@xxxxxxxxxxxxx> Date: 08/23/2011 03:34 PM Subject: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Sent by: ibis-macro-bounce@xxxxxxxxxxxxx Just wanted some understanding on how jitter budgets work: Are the jitter inside the tx and rx a part of the budget – or are they separate from the budget? If they are a part of the entire budget for the simulation and the ami file tells the EDA tool to add some Rx_Rj (for example) to the Time domain simulation without the knowledge of the user (assuming the user does not read the .ami file), wouldn’t the user be not counting that jitter in the budget? Thoughts? Thanks, -Ambrish. Ambrish Varma | Member of Consulting Staff P: 978.262.6431 www.cadence.com From: ibis-macro-bounce@xxxxxxxxxxxxx [ mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of fangyi_rao@xxxxxxxxxxx Sent: Tuesday, August 23, 2011 2:48 PM To: wkatz@xxxxxxxxxx; kumaran@xxxxxxxxxxxx; vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor If they are external to Rx, user/EDA tool inject them into simulation according to standard as simulation input instead of model parameters. If they are internal to Rx, model makers don’t define them. So in both cases, we don’t need them as Rx model parameters. Fangyi From: Walter Katz [mailto:wkatz@xxxxxxxxxx] Sent: Tuesday, August 23, 2011 11:42 AM To: RAO,FANGYI (A-USA,ex1); kumaran@xxxxxxxxxxxx; vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Fangyi, If the model maker did that, then he simply does not define Rx_Rj, Rx_Sj and Rx_DCD. Walter From: fangyi_rao@xxxxxxxxxxx [mailto:fangyi_rao@xxxxxxxxxxx] Sent: Tuesday, August 23, 2011 2:39 PM To: wkatz@xxxxxxxxxx; kumaran@xxxxxxxxxxxx; vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Walter; What I meant is when the Rx_Rj, Rx_Sj and Rx_DCD impairments are internal to Tx, model maker can combine them into Rx_Clock_Recovery_Rj, Rx_Clock_Recovery_Sj and Rx_Clock_Recovery_DCD for statistical simulation, and include them in clock_times in time-domain simulations. Fangyi From: Walter Katz [mailto:wkatz@xxxxxxxxxx] Sent: Tuesday, August 23, 2011 11:31 AM To: 'Kumaran Krishnasamy'; RAO,FANGYI (A-USA,ex1); vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Kumaran, Fangyi, Comments below (WMK>). Walter From: Kumaran Krishnasamy [mailto:kumaran@xxxxxxxxxxxx] Sent: Tuesday, August 23, 2011 1:39 PM To: fangyi_rao@xxxxxxxxxxx; wkatz@xxxxxxxxxx; vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Hi, To add to Fangyi’s first comment, users determine the amounts of jitter from the standards (for example PCIe, SFI) that describe the max amounts to be used to stress the Rx. WMK> You are correct. In this case the User/EDA tool will specify Rx_Rj, Rx_Sj and Rx_DCD from the particular standard. Regards, Kumaran From: ibis-macro-bounce@xxxxxxxxxxxxx [ mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of fangyi_rao@xxxxxxxxxxx Sent: Tuesday, August 23, 2011 10:29 AM To: wkatz@xxxxxxxxxx; vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Walter; Thanks for the answers. Please see my further comment/question in brown. How would a model user know how to set Rx_Rj, Rx_Sj and Rx_DCD? The BIRD states that they are external to the receiver, so it’s expected that model users determine and inject (through simulator) such jitters into simulations. In some cases, for example a forwarded clock the User/EDA tool can analyze the forwarded clock signal, in which case the EDA tool would be directed to use these jitter impairments. However, if the reference clock is generated inside the Rx chip, then the model maker does know these jitter impairments, and can set them inside the AMI model. In that case they are not external to Rx. Can’t they be combined into Rx_Clock_Recovery_Rj, Rx_Clock_Recovery_Sj and Rx_Clock_Recovery_DCD? (Sj could be tricky but can be handled) WMK> Are only Rx_Clock_Recovery_Rj, Rx_Clock_Recovery_Sj and Rx_Clock_Recovery_DCD Are only used in statistical simulation (unless the perverse case when Rx AMI_GetWave does not return clock_times). Rx_Rj, Rx_Sj and Rx_DCD are used in both statistical and time domain time domain simulations, so cannot be combined with Rx_Clock_Recovery_Rj, Rx_Clock_Recovery_Sj and Rx_Clock_Recovery_DCD. Fangyi From: fangyi_rao@xxxxxxxxxxx [mailto:fangyi_rao@xxxxxxxxxxx] Sent: Monday, August 22, 2011 9:02 PM To: wkatz@xxxxxxxxxx; vladimir_dmitriev-zdorov@xxxxxxxxxx; Arpad_Muranyi@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Hi, Walter; I wonder why Rx_Rj, Rx_Sj and Rx_DCD are Rx model parameters. They are defined as impairments external to receiver and serve the purpose of jitter budget analyses. So they are simulation inputs but not intrinsic Rx parameters. They should be set by model users instead of model developers. EDA tools can include them in simulations by allowing users to specify their values from the simulator. Regards, Fangyi From: ibis-macro-bounce@xxxxxxxxxxxxx [ mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz Sent: Wednesday, August 03, 2011 4:18 PM To: 'Dmitriev-Zdorov, Vladimir'; 'Muranyi, Arpad'; 'IBIS-ATM' Subject: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Vladimir, We are planning to discuss the Jitter BIRD Tuesday. I think we can resolve all of the questions at that time. Walter From: Dmitriev-Zdorov, Vladimir [ mailto:vladimir_dmitriev-zdorov@xxxxxxxxxx] Sent: Wednesday, August 03, 2011 3:35 PM To: wkatz@xxxxxxxxxx; Muranyi, Arpad; 'IBIS-ATM' Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Walter, After reading your comments, I’m more satisfied with the document. Otherwise it’s hard to understand some things. Still, I have a few remaining suggestions, shown in a few additional comments. Vladimir From: ibis-macro-bounce@xxxxxxxxxxxxx [ mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz Sent: Tuesday, August 02, 2011 5:00 PM To: Muranyi, Arpad; 'IBIS-ATM' Subject: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Arpad, I am including my response to Vladimir’s comments in the enclosed document. I stand by my earlier statement that I believe the current specification defines each parameter mathematically precisely and indicates how the parameters are to be used in either the statistical or time domain simulations. I would not object to adding additional words would be helpful to explain the precise mathematical definitions with “C” code or other standard statistical function definitions. Walter From: Muranyi, Arpad [mailto:Arpad_Muranyi@xxxxxxxxxx] Sent: Tuesday, August 02, 2011 5:55 PM To: Walter Katz; 'IBIS-ATM' Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Walter, Did you look at the comments Vladimir inserted in the attached Word document? I think addressing those would be a good start in this discussion. But there are some specifics even in the body of the email from Vladimir: “The Bird does not seem to be logically complete (does not cover all cases, such as Time domain and Statistical flow), or does not provide all necessary parameters affecting the results (such as either giving frequency content of Tx Gaussian jitter, or making it defined as white uncorrelated jitter, etc.) ” Thanks, Arpad ======================================================== From: Walter Katz [mailto:wkatz@xxxxxxxxxx] Sent: Tuesday, August 02, 2011 4:45 PM To: Muranyi, Arpad; 'IBIS-ATM' Subject: RE: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Arpad, Vladimir, There was great effort to define each of the parameters in strict mathematical terms. Please indicate specific parameters that the BIRD does not either define mathematically precisely, or indicate how the parameters are to be used in either the statistical or time domain simulations. Walter From: ibis-macro-bounce@xxxxxxxxxxxxx [ mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad Sent: Tuesday, August 02, 2011 4:43 PM To: IBIS-ATM Subject: [ibis-macro] Re: Comments on BIRD 123.2.2 from Mentor Walter, Here is the email (with attachment) I mentioned in the ATM meeting today. There was no response to it since I posted it. I think it would be good to get some discussion going on these questions/comments. Thanks, Arpad From: ibis-macro-bounce@xxxxxxxxxxxxx [ mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Muranyi, Arpad Sent: Friday, July 01, 2011 11:53 AM To: IBIS-ATM Subject: [ibis-macro] Comments on BIRD 123.2.2 from Mentor Walter and All, I am forwarding Vladimir’s comments on BIRD 123.2.2 to this list for discussion. Please look at his comments which are in the attached Word document. Since Vladimir is also on this email list, please direct your answers, questions and comments to him. (I don’t want to be the middle man in such conversations…). The general problem is ambiguity of given definitions when it comes to implementation in EDA tool. The Bird does not seem to be logically complete (does not cover all cases, such as Time domain and Statistical flow), or does not provide all necessary parameters affecting the results (such as either giving frequency content of Tx Gaussian jitter, or making it defined as white uncorrelated jitter, etc.) Thanks, Arpad ============================================================