On Bob's request in the ATM teleconference today, I am forwarding the email
thread
on the DC_Offset BIRD draft that took place during this past week. Keep in
mind that
this email thread was discussed in the meeting today, and the questions seemed
to
have been settled without objections.
Thanks,
Arpad
==================================================================
From: Muranyi, Arpad
Sent: Tuesday, May 14, 2019 1:26 PM
To: 'Fangyi Rao' <fangyi_rao@xxxxxxxxxxxx>; Walter Katz <wkatz@xxxxxxxxxx>;
ambrishv@xxxxxxxxxxx; 'Randy Wolff (rrwolff)' <rrwolff@xxxxxxxxxx>; 'Justin
Butterfield (jdbutterfiel)' <jdbutterfiel@xxxxxxxxxx>
Subject: RE: new draft for DC_Offset BIRD197
I have a comment on:
"If we are to state in the standard that DC_for_Statistical is also the
threshold"
Thresholds are usually not a single value, as there is usually some hysteresis
at the input.
For that reason, I don't think we can use a single value parameter
DC_for_Statistical for
threshold purposes...
Thanks,
Arpad
=====================================================================
From: Fangyi Rao [mailto:fangyi_rao@xxxxxxxxxxxx]
Sent: Monday, May 13, 2019 6:04 PM
To: Walter Katz <wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>>;
ambrishv@xxxxxxxxxxx<mailto:ambrishv@xxxxxxxxxxx>; 'Randy Wolff (rrwolff)'
<rrwolff@xxxxxxxxxx<mailto:rrwolff@xxxxxxxxxx>>; 'Justin Butterfield
(jdbutterfiel)' <jdbutterfiel@xxxxxxxxxx<mailto:jdbutterfiel@xxxxxxxxxx>>
Cc: Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx<mailto:Arpad_Muranyi@xxxxxxxxxx>>
Subject: RE: new draft for DC_Offset BIRD197
Hi Walter,
By "simulated signal" I meant the Rx latch eye computed by statistical
simulation.
Fangyi
From: Walter Katz <wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>>
Sent: Monday, May 13, 2019 5:53 AM
To: Fangyi Rao <fangyi_rao@xxxxxxxxxxxx<mailto:fangyi_rao@xxxxxxxxxxxx>>;
ambrishv@xxxxxxxxxxx<mailto:ambrishv@xxxxxxxxxxx>; 'Randy Wolff (rrwolff)'
<rrwolff@xxxxxxxxxx<mailto:rrwolff@xxxxxxxxxx>>; 'Justin Butterfield
(jdbutterfiel)' <jdbutterfiel@xxxxxxxxxx<mailto:jdbutterfiel@xxxxxxxxxx>>
Cc: 'Muranyi, Arpad' <Arpad_Muranyi@xxxxxxxxxx<mailto:Arpad_Muranyi@xxxxxxxxxx>>
Subject: RE: new draft for DC_Offset BIRD197
[EXTERNAL]
Fangyi,
My comments below
Walter
From: Fangyi Rao <fangyi_rao@xxxxxxxxxxxx<mailto:fangyi_rao@xxxxxxxxxxxx>>
Sent: Monday, May 13, 2019 4:09 AM
To: Walter Katz <wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>>;
ambrishv@xxxxxxxxxxx<mailto:ambrishv@xxxxxxxxxxx>; 'Randy Wolff (rrwolff)'
<rrwolff@xxxxxxxxxx<mailto:rrwolff@xxxxxxxxxx>>; 'Justin Butterfield
(jdbutterfiel)' <jdbutterfiel@xxxxxxxxxx<mailto:jdbutterfiel@xxxxxxxxxx>>
Cc: 'Muranyi, Arpad' <Arpad_Muranyi@xxxxxxxxxx<mailto:Arpad_Muranyi@xxxxxxxxxx>>
Subject: RE: new draft for DC_Offset BIRD197
Thanks for the input, Walter.
Given that DC component is affected by numbers of bits of logic 1 and logic 0
and by the difference in duty cycle between 1 and 0, in the attached revision I
rephrased the definition of parameter DC_for_Statistical from "DC component at
the Rx latch" to "mid-point of the steady state high and low voltages at the Rx
latch" to make it more precise. I also changed the value in the
DC_for_Statistical example from 0.5 to 0.001.
As for Walter's suggestion, I'd like to discuss the following related questions
1. Do we need to address the threshold issue when we are defining the
simulated signal? I recalled that Ambrish preferred not to in a previous
discussion
WMK> By "simulated signal" I assume you meant the Rx GetWave output. I thought
you concluded that if there was an offset (e.g. DC_for_Statistical=.001) , the
Rx GetWave would include it in its output. I agreed with this conclusion.
1. The mid-point voltage at the Rx latch may not be the same as the
threshold that the Rx decides to use to sample the signal at the latch
MWK> You proposed DC_for_Statistical, because the output of Rx AMI_Init is an
Impulse Response and without a parameter DC_for_Statistical the only number the
EDA tool has to sample the signal is at the midpoint.
1. If we are to state in the standard that DC_for_Statistical is also the
threshold, then is this threshold used only in statistical simulation or in
both time-domain and statistical simulations? Currently the Usage rule for this
parameter is for statistical simulation only.
WMK> As I responded to 1.0, the DLL has a method in GetWave to shift the signal
by DC_for_Statistical. DC_for_Statistical would be double counted if the EDA
tool applie it in GetWave.
Regards,
Fangyi
From: Walter Katz <wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>>
Sent: Saturday, May 11, 2019 1:16 PM
To: Fangyi Rao <fangyi_rao@xxxxxxxxxxxx<mailto:fangyi_rao@xxxxxxxxxxxx>>;
ambrishv@xxxxxxxxxxx<mailto:ambrishv@xxxxxxxxxxx>; 'Randy Wolff (rrwolff)'
<rrwolff@xxxxxxxxxx<mailto:rrwolff@xxxxxxxxxx>>; 'Justin Butterfield
(jdbutterfiel)' <jdbutterfiel@xxxxxxxxxx<mailto:jdbutterfiel@xxxxxxxxxx>>
Cc: 'Muranyi, Arpad' <Arpad_Muranyi@xxxxxxxxxx<mailto:Arpad_Muranyi@xxxxxxxxxx>>
Subject: RE: new draft for DC_Offset BIRD197
[EXTERNAL]
Fangyi,
Excellent. You might want to add that DC_for_Statistical is the threshold
voltage at the Latch. To have confidence that the level is high at the latch
the waveform must be greater than DC_for_Statisticalplus +
Rx_Receiver_Sensitivity and to assure that the level is low at the latch, the
waveform must be less than DC_for_Statisticalplus - Rx_Receiver_Sensitivity.
Walter
From: Fangyi Rao <fangyi_rao@xxxxxxxxxxxx<mailto:fangyi_rao@xxxxxxxxxxxx>>
Sent: Saturday, May 11, 2019 3:22 PM
To: ambrishv@xxxxxxxxxxx<mailto:ambrishv@xxxxxxxxxxx>; Walter Katz
<wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>>; Randy Wolff (rrwolff)
<rrwolff@xxxxxxxxxx<mailto:rrwolff@xxxxxxxxxx>>; Justin Butterfield
(jdbutterfiel) (jdbutterfiel@xxxxxxxxxx<mailto:jdbutterfiel@xxxxxxxxxx>)
<jdbutterfiel@xxxxxxxxxx<mailto:jdbutterfiel@xxxxxxxxxx>>
Cc: Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx<mailto:Arpad_Muranyi@xxxxxxxxxx>>
Subject: new draft for DC_Offset BIRD197
Hi All,
In the last ATM meeting we observed that
- The Rx GetWave output waveform can have non-zero DC component
regardless whether the standard specifies this waveform to be differential or
not due to non-ideal effects in Rx device
- It is not necessary for the standard to specify that the EDA tool
adds a DC offset to the Rx GetWave output. Such DC offset can be included in
the Rx GetWave output by the model. The standard is more concise if we just let
Rx GetWave include the DC offset in the output waveform
- To support statistical simulations, the Rx model needs to provide
the EDA tool the DC component at the Rx latch
Based on that, we agreed to make the following modification to the BIRD
- Change DC_Offset's Usage back to In. The physical Rx input waveform
is the input waveform to Rx GetWave plus DC_Offset.
- Rx GetWave output waveform is the physical waveform at the Rx latch
- Introduce a new reserved parameter named DC_for_Statistical for the
Rx model to support statistical simulations. The parameter is optional, and its
Usage is Out. Default is 0V. It's only used by the EDA tool in statistical
simulations
I incorporated these changes in the attached draft. Please review and provide
feedback. I plan to send it to the reflector before next ATM for discussion.
Thanks and regards,
Fangyi
