Hi All, We overlooked the fact that there may be a case where Init_Return_Impulse may be True AND Getwave_Exist is also True and Backchannel may only be available for either Getwave or Init or both. This is now made possible by changing the Training Reserved Parameter to accept Integer values of 0,1 or 2 letting the user/EDA tool know which mode the model supports. Thanks to Vladimir who pointed this out to me. The latest version is attached. Thanks, Ambrish. From: ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Ambrish Varma Sent: Thursday, April 03, 2014 7:03 PM To: wkatz@xxxxxxxxxx; ibis-macro@xxxxxxxxxxxxx Subject: [ibis-macro] Re: Revised Backchannel BIRD Hi Walter, Thanks for your question. The communication between the Tx and Rx is sensitive to the direction of message flow. Here is how it works: Time Domain (Getwave based backchannel): Tx --> Rx communication The Tx tells the Rx a) how many taps there are and b) whether it has any scope to move up or down. This is done by the Tx by selecting 0, -1 or 1 for its values for each tap, where 0 is telling the Rx that it has room for movement up or down, -1 for reaching its lower limit and 1 for reaching its upper limit. Ex: "(BCI(taps (-1 0)(0 0)(1 0)))" AND "(BCI (taps (-1 -1) (0 0) (1 1)))" Rx --> Tx communication The Rx reads the string from the Tx and interprets the message and after analyzing the waveform, sends back a string with suggested units to increment or decrement. This is done by the Rx by specifying the number of units that it thinks the Tx taps should go or down by. Ex: "(BCI (taps (-1 -1) (0 0) (1 -2)))" : The Rx is suggesting to the Tx to decrement the pre tap by 1 unit and post tap by 2 units. If the Rx returns all 0s and the Training_Done parameter, then the training is complete. The Tx does its interpretation of the string coming back from the Rx and adjust its taps. In the example in the BIRD, we use a resolution of 1/32 for each tap - so the Tx decrements the pre tap by 1/32 and the post tap by 2/32. It is completely up to the Tx AMI model to decide how much they want to increment/decrement. Statistical Domain (Init based backchannel): Tx --> Rx communication The Tx tells the Rx a) how many taps there are and b) what range can the tap values lie in. if only 1 value is specified, then the Tx is telling the Rx that that tap cannot be changed. All the tap values are relative to the main tap value of 1. Ex: "(BCI (taps (-1 -0.25 0)(0 1)(1 -0.3 0.3)))" : The Tx is telling the Rx that the pretap value can lie between -0.25 and 0 and the post tap value can lie between +-0.3 Rx --> Tx communication The Rx simply decides what value it want the taps to be and conveys the suggested relative values to the Tx. These values may be in the range that the Tx specifies. Ex: "(BCI (taps (-1 -0.2)(0 1)(1 -0.1)))" The Tx may use these tap values to modify the impulse response which gets sent to the Rx for modification in the Init flow. The communication between the Tx and Rx regarding the parameter taps values is described in section 1.5 of the BIRD. Please let me know if this clears up the confusion. Thanks, Ambrish. From: Walter Katz [mailto:wkatz@xxxxxxxxxx] Sent: Thursday, April 03, 2014 3:39 PM To: Ambrish Varma; ibis-macro@xxxxxxxxxxxxx Subject: RE: [ibis-macro] Revised Backchannel BIRD Ambrish, I am really confused on how the parameter taps is used in the Statistical and Time domain flow. In some places it is a coefficient between -1 and 1, and some places it is an increment and decrement. And some places there seems to be an relationships of 1/32 between increments and coefficients. Can you clarify this please. Walter From: ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Ambrish Varma Sent: Wednesday, April 02, 2014 9:56 AM To: ibis-macro@xxxxxxxxxxxxx Subject: [ibis-macro] Revised Backchannel BIRD Hi All, The latest version of the Backchannel BIRD (draft 8) has been uploaded on the ATM website and is available here: http://www.vhdl.org/ibis/macromodel_wip/archive/20140401/ambrishvarma/Backchannel%20support%20BIRD%20147%20draft%208/Backchannel_BIRD147_Draft8.docx Or http://tinyurl.com/o8cqyqr This is the version that was discussed at the 04/01/2014 ATM and includes support for AMI_Init/Statistical Backchannel training. Please let me know if there are any questions. Thanks, Ambrish. Ambrish Varma. Cadence Design Systems.
Attachment:
Backchannel_BIRD147_Draft9.docx
Description: Backchannel_BIRD147_Draft9.docx
