Hi Bob,
The alignment is described in 17-JAN-2017 and 24-MAY-2016 docs archived at
http://ibis.org/atm_wip/archive-date.html
On the impulse side, the Rx2 Init function takes care of the alignment between
the AFE impulse hAFE and the DFE impulse hDFE. Denote the input impulse to Rx2
Init as hin. As described in aforementioned docs, the Rx2 model must align hDFE
with the convolved impulse of hin*hAFE. Note that hDFE is sampled at the same
sample rate as hin and hAFE, allowing such alignment (that also means hDFE is
oversampled and has spikes at the tap positions).
On the waveform side, if the Rx2 has Getwave, then neither hAFE nor hDFE is
used. If the Rx2 model doesn’t have GetWave, then the EDA tool applies hin*hAFE
to the analog waveform coming down from upstream and hDFE to the digital
stimulus waveform. As described in aforementioned docs, because hin*hAFE and
hDFE are already aligned, the EDA tool needs to align the two waveforms. This
is equivalent to having the EDA tool recover the clock since the Rx2 model
doesn’t have Getwave.
Regards,
Fangyi
From: Bob Miller (APD) [mailto:bob.miller@xxxxxxxxxxxx]
Sent: Monday, May 08, 2017 2:51 PM
To: Walter Katz <wkatz@xxxxxxxxxx>
Cc: RAO,FANGYI (K-USA,ex1) <fangyi_rao@xxxxxxxxxxxx>; IBIS-ATM
<ibis-macro@xxxxxxxxxxxxx>
Subject: Re: [ibis-macro] Re: Fangyi and I had an AR to describe what bad
things can happen if IBIS 166.2 is approved in 7.0 and the Keysight BIRD with
additional IR is approved in IBIS 7.1 ...
I have what may be an unenlightened question about DFE in a bit-by-bit
simulation of Fangi's (tx2,rx2 are init_only) simulation.
If the rx2 returns a separate afe and dfe transfer-function impulse how does
the EDA tool correctly align the rx2 dfe impulse (essentially replacing the rx2
cdr function)? As far as I can tell, there is no IBIS-AMI defined time
relationship between the symbol alignment in the bit-by-bit stimulus at the rx2
input and the corresponding impulse into rx2 in the AMI_Init call. If the dfe
equalization is not applied in the correct alignment, less-than optimal results
ensue. The EDA tool can of course provide a cdr proxy for rx2, but how does it
know how the recovered clock aligns to the stored dfe impulse returned by rx2's
AMI_Init?
Or am I missing something?
[I have long felt that the time relationship between an impulse in AMI_Init and
a bit-by-bit pattern in AMI_Getwave should be defined to make these problems
simpler]
regards,
Bob
On Mon, May 8, 2017 at 7:09 AM, Walter Katz
<wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>> wrote:
Fangyi,
You have described a valid flow problem with one of the 81 cases that can occur
in IIS 6.1 models. And I agree that by adding additional IR outputs to Rx and
Tx AMI_Init (then there are 625 cases) these kinds of problem goes can go away
for if the model maker writes AMI_Init with the additional IR output.
The point I have been trying to make is that when we implement in the standard
the optional ability to add these additional IR outputs to AMI Init, in IBIS
7.x, then the flows will still need to support all 625 cases, including the 81
cases in IBIS 6.1.
* Will not the 7.x flows need to describe the flows for these 81 case?
* Will not the flows for these 81 cases need to be the exact flows I
describe in BIRD 166.3?
I claim the answer to both of these questions must be yes. And if this is the
case then nothing I have written in BIRD 166.3 for IBIS 7.0 becomes invalid for
these 81 cases.
I repeat the question that you have not answered.
What bad things happen if BIRD 166.3 is included in IBIS 7.0?
Walter
Walter Katz
wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>
Phone 303.449-2308<tel:(303)%20449-2308>
Mobile 303.335-6156<tel:(303)%20335-6156>
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>]
On Behalf Of fangyi_rao@xxxxxxxxxxxx<mailto:fangyi_rao@xxxxxxxxxxxx>
Sent: Monday, May 8, 2017 4:17 AM
To: wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>;
ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>
Subject: [ibis-macro] Re: Fangyi and I had an AR to describe what bad things
can happen if IBIS 166.2 is approved in 7.0 and the Keysight BIRD with
additional IR is approved in IBIS 7.1 ...
(Resend with typo fixed)
All,
In the attached doc I showed that the BIRD166 flow may unphysically amplify
redriver nonlinearity, internal noise and crosstalks when they reach downstream
DFE and that this amplification equals the upstream channel loss in dB. This
problem can’t be resolved by incremental changes in current flow. That’s why
additional columns are added to impulse response matrix in the Keysight BIRD.
BTW, there is no need to count the number of possible combinations. As more
aggressors and cascaded redrivers are added to the link, this number will keep
growing. The flow in the Keysight BIRD, described in a separate doc, is
designed in a scalable way so that one only needs to deal with 4 basic
combinations and all others are automatically covered.
Fangyi
From: ibis-macro-bounce@xxxxxxxxxxxxx<mailto:ibis-macro-bounce@xxxxxxxxxxxxx>
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz
Sent: Wednesday, May 03, 2017 11:40 AM
To: IBIS-ATM <ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx>>
Subject: [ibis-macro] Fangyi and I had an AR to describe what bad things can
happen if IBIS 166.2 is approved in 7.0 and the Keysight BIRD with additional
IR is approved in IBIS 7.1 ...
All,
Fangyi and I had an AR to describe what bad things can happen if IBIS 166.2 is
approved in 7.0 and the Keysight BIRD with additional IR is approved in IBIS
7.1.
I am enclosing a presentation which is a short review of the current IBIS
166.2, and adds a section demonstrating that no bad things will happen if 166.2
is included in IBIS 7.0 and the Keysight BIRD with additional IR is approved in
IBIS 7.1.
In summary:
* There are 81 possible flows based on IBIS 6.1 models.
* There are 625 possible flows in the Keysight BIRD.
* The 81 flows in IBIS 7.0 remain the same in IBIS 7.1.
Therefore the flows in IBIS 7.1 will not change any of the flows in IBIS 7.0.
Walter
Walter Katz
wkatz@xxxxxxxxxx<mailto:wkatz@xxxxxxxxxx>
Phone 303.449-2308<tel:(303)%20449-2308>
Mobile 303.335-6156<tel:(303)%20335-6156>
