All,
Answers below.
Walter
Walter Katz
Work 508.647-7633
Cell 720.417-3762
[Description: Description: Visit MathWorks.com]
From: Fangyi Rao <fangyi_rao@xxxxxxxxxxxx>
Sent: Tuesday, March 23, 2021 4:58 AM
To: Walter Katz <wkatz@xxxxxxxxxxxxx>; ibis-macro@xxxxxxxxxxxxx
Subject: RE: One more correction to Tuesday's presentation
Hi Walter,
Following are problems I identified in the slides.
#1. Regarding the flow on slide 11 (modified existing flow), the input to Tx2
Init is the output of Rx1 Init. If Tx2 is Init-only, in time-domain simulations
how does the EDA tool find the needed combined impulse of Tx2 and the
downstream channel (when Rx2 has GetWave) or combined impulse of Tx2, the
downstream channel and Rx2 (when Rx2 is Init-only)? By contrast, in BIRD210 the
input to Tx2 Init in the existing flow is still the impulse of the downstream
channel, as specified in 7.0 Spec.
WMK> The AMI_Init is required to apply the equalization of the model to all
aggressor columns. For any Init Only model, the EDA tool can add a Unit Impulse
Response “aggressor”. The EDA tool performs the GetWave functionality of this
model by convolving the impulse response output of this column with the
waveform intended for the missing GetWave. This applies to any Init Only model.
See slide 5.
#2. Regarding both flows on slides 11 and 12, the input to Rx2 Init is the
cumulative impulse starting from Tx1. If both Tx2 and Rx2 are Init-only, in
time-domain simulations how does the EDA tool find the needed combined impulse
of Tx2, the downstream channel and Rx2? By contrast, in BIRD210 one of the
input to Rx2 Init is the combined impulse of Tx2 and the downstream channel.
WMK> The AMI_Init is required to apply the equalization of the model to all
aggressor columns. For any Init Only model, the EDA tool can add a Unit Impulse
Response “aggressor”. The EDA tool performs the GetWave functionality of this
model by convolving the impulse response output of this column with the
waveform intended for the missing GetWave. This applies to any Init Only model.
See slide 5.
#3. Regarding the flow on slide 11 (modified existing flow), the input to Tx2
Init is changed from the downstream impulse, as specified in the 7.0 Spec, to
the upstream impulse. If an existing Tx2 model’s Init needs to use the
downstream impulse, the flow will break Tx2 Init.
WMK> There are no existing Tx2 models that use the downstream impulse. Redriver
model makers know that the flow is IBIS 7.0 is fundamentally flawed and have
requested that we apply the downstream channel to the output of Tx2 as we
stated in April 2014.
#4. Regarding the flow on slide 12 (the new flow), in impulse_matrix there is
no additional column for the model to return the impulse of its LTI part.
Without that how is deconvolution eliminated? By contrast, BIRD210 adds an
additional column in impulse_matrix for the impulse of the model’s LTI part.
WMK> The AMI_Init is required to apply the equalization of the model to all
aggressor columns. For any Init Only model, the EDA tool can add a Unit Impulse
Response “aggressor”. The EDA tool performs the GetWave functionality of this
model by convolving the impulse response output of this column with the
waveform intended for the missing GetWave. This applies to any Init Only model.
See slide 5.
Fangyi
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
<ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>> On
Behalf Of wkatz@xxxxxxxxxxxxx<mailto:wkatz@xxxxxxxxxxxxx>
Sent: Thursday, March 18, 2021 4:06 AM
To: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] One more correction to Tuesday's presentation
CAUTION: This message originates from an external sender.
All,
Thanks again for Curtis doing a careful editorial review of Tuesday’s
presentation. Slide 12 changed to reflect that the new flow parameter set to
true only changes the input impulse matrix of the Redriver Tx2, and that the
EDA tool will continue to convolve the downstream channel with output of Tx2 as
it does in the 7.1 flow when the new parameter is false or is not present in
the Tx2 .ami file.
Walter
Walter Katz
Work 508.647-7633
Cell 720.417-3762
[Description: Description: Visit MathWorks.com]