[IBIS-Users] RE: Understanding IV-VT mismatch errors / warnings and their fix

From: <birendra.rana_at_.....>
Date: Wed Mar 04 2009 - 01:12:58 PST
Hi Bob,

Thanks for the response and for spending your time to review the model.

I had given some thought to the idea of scaling the VT transition high
end point to make it equal to IV end point just to remove the
mismatches. But what prevented me from doing so is the concern of ending
up having IBIS transitions not matching Spectre model transitions. The
Spectre netlist simulations will show the min waveform reaching 0.31V
whereas IBIS will settle at 0.33V for 50ohm load connected to gnd.

Best regards,

Birendra Rana

-----Original Message-----
From: Bob Ross [mailto:bob@teraspeed.com]
Sent: Wednesday, March 04, 2009 1:02 AM
To: Birendra Rana (WT01 - Wipro NewLogic)
Cc: bob@teraspeed.com
Subject: IBIS Reflector Waveform Mismatch Eroror/Warning Questions

Hi Birendra

Yi IBIS reflector messages had some addtional HTML? attached at the end,
causing the total message to exceed the reflector size limits.
So the messages were not sent out and thye bounced to me.

You can try to send them out again in Text mode.
------
However, I can respond also briefly and directly to you based on a quick
review of your IBIS model.
Some notes are in your text
-----------
<........ snip..........>

This despite taking care of following:

1. VT table captures the full PAD rise / fall transition 2. IV and VT
tables are generated with same conditions/setup 3. IV clamp tables do
not double count clamp currents

BR --- The above steps and the tuning of Spectre below might have
helped, but the Warnings and Errors are still serious.

Spectre was used for generating the model. Playing around with some of
the spice options (primarily GMIN) did help in reducing the differences.
The "best" model I could get with changing GMIN across corners at time
of generation has the following parser message:

<........ snip..........>

I have the following questions:

1. What are the factors that lead to IV-VT mismatches? In my case what
would be the possible cause (despite of taking care of points 1 to 3
listed above)?

BR ---- Some models might require several cycles for the V-T data to
stabailze.  However, you tried several reasonable things later without
success, so I do not know know the source of the problem or if there is
a good stable solution.

2. Would like to know of what is an acceptable percentage? If I
understand correctly errors are reported for deviation > 10%. Can
mismatch warnings be waived off in case removing them is difficult?

BR ---- The acceptable precentage is less than 2%.  You should not have
any errors or warnings.  If your warning messages are slightly outside
the 2% range, you could just re-scale the V-T waveforms to match the
predicted I-V endpoints.  It is not acceptable to leave the errors or
warnings as is.

3. Transient analysis of the buffer seems to indicate that it takes some
time for the PAD to settle to a regular operation. The first output
transition does not reflect the correct behaviour of the PAD. So I got
the VT tables generated with 3rd rise and fall PAD transition. The IV
tables which are generated with DC conditions actually reflect the
initial buffer state, hence the mismatches remains. Tried generating IV
data by doing a very slow PWL sweep of PAD after ramping up the
supplies. However, the IV characteristic thus generated is highly
nonmonotonic and deviates from the expected. As a result mismatches are
even more. Can someone suggest a way to resolve this problem?

BR -------- At this point you tried several things ... Slow sweep for DC
extraction, third cycle for V-T extraction, etc. I would just use one of
the mismatched waveforms and re-scale them.  The t=0 STARTING Point that
matches the I-V extraction is an indicator of a solid starting point and
corresponding ending point of the opposite edge.  For example:  Typical

   Rising with 50 ohms to Gnd   0 to .31  rescale the rising waveform
   to about 0.33 to elliminate the warning.

   Similar adjustments apply for all the other waveforms that create
   Warning or error messages.

BR ------- The I-V data is non-monotonic, but in the clamping regions
outside of where the the V-T waveforms operate.  I would clean up the
non-monotonic data as a separate issue, but they are not a factor here.

I am copying the IBIS model below for reference.

Best regards,
Birendra Rana

-------------------------
Bob Ross
Teraspeed Consulting Group LLC     Teraspeed Labs
121 North River Drive              13610 SW Harness Lane
Narragansett, RI 02882             Beaverton, OR 97008
401-284-1827                       503-430-1065
http://www.teraspeed.com           503-246-8048 Direct
bob@teraspeed.com

Teraspeed is a registered service mark of Teraspeed Consulting Group LLC

------------------------------------------------------------------------
--------
From: Birendra Rana (WT01 - Wipro NewLogic)
Sent: Tuesday, March 03, 2009 8:33 PM
To: ibis-users@server.eda.org; ibis@server.eda.org
Subject: Understanding IV-VT mismatch errors / warnings and their fix

Hello experts,

I need help in understanding the possible cause of the IV-VT mismatch
errors and warnings reported by ibischk4 parser for a particular case.

The IBIS model is of an IO buffer which is configurable in two different
slew rates. I dont see any ibischk4 errors/warnings for fast slew mode.
However in slow mode, I get the following output from the parser:

WARNING - Model PAD_IO: The [Rising Waveform]
      with [R_fixture]=50 Ohms and [V_fixture]=0V
      has TYP column DC endpoints of  0.00V and  0.31v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 0.00V and  0.33V),
      a difference of  0.00% and  7.59%, respectively.
ERROR - Model PAD_IO: The [Falling Waveform]
      with [R_fixture]=50 Ohms and [V_fixture]=1.8V
      has TYP column DC endpoints of  1.53V and  1.80v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.56V and  1.80V),
      a difference of 11.50% and  0.00%, respectively.
WARNING - Model PAD_IO: The [Rising Waveform]
      with [R_fixture]=50 Ohms and [V_fixture_min]=0V
      has MIN column DC endpoints of  0.00V and  0.17v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 0.00V and  0.17V),
      a difference of  0.00% and  3.06%, respectively.
ERROR - Model PAD_IO: The [Falling Waveform]
      with [R_fixture]=50 Ohms and [V_fixture_min]=1.65V
      has MIN column DC endpoints of  1.49V and  1.65v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.55V and  1.65V),
      a difference of 60.53% and  0.00%, respectively.
WARNING - Model PAD_IO: The [Rising Waveform]
      with [R_fixture]=50 Ohms and [V_fixture_max]=0V
      has MAX column DC endpoints of  0.00V and  0.48v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 0.00V and  0.52V),
      a difference of  0.00% and  7.18%, respectively.
ERROR - Model PAD_IO: The [Falling Waveform]
      with [R_fixture]=50 Ohms and [V_fixture_max]=1.95V
      has MAX column DC endpoints of  1.56V and  1.95v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.48V and  1.95V),
      a difference of 17.06% and  0.00%, respectively.

This despite taking care of following:

1. VT table captures the full PAD rise / fall transition
2. IV and VT tables are generated with same conditions/setup
3. IV clamp tables do not double count clamp currents

Spectre was used for generating the model. Playing around with some of
the spice options (primarily GMIN) did help in reducing the differences.
The "best" model I could get with changing GMIN across corners during
the time of generation has the following parser message:

WARNING - Model PAD_IO: The [Rising Waveform]
      with [R_fixture]=50 Ohms and [V_fixture]=1.8V
      has TYP column DC endpoints of  1.52V and  1.80v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.53V and  1.80V),
      a difference of  4.80% and  0.00%, respectively.
WARNING - Model PAD_IO: The [Falling Waveform]
      with [R_fixture]=50 Ohms and [V_fixture]=1.8V
      has TYP column DC endpoints of  1.52V and  1.80v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.53V and  1.80V),
      a difference of  4.80% and  0.00%, respectively.
ERROR - Model PAD_IO: The [Rising Waveform]
      with [R_fixture]=50 Ohms and [V_fixture_min]=1.65V
      has MIN column DC endpoints of  1.48V and  1.65v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.55V and  1.65V),
      a difference of 63.49% and  0.00%, respectively.
ERROR - Model PAD_IO: The [Falling Waveform]
      with [R_fixture]=50 Ohms and [V_fixture_min]=1.65V
      has MIN column DC endpoints of  1.48V and  1.65v, but
      an equivalent load applied to the model's I-V tables yields
      different voltages ( 1.55V and  1.65V),
      a difference of 63.39% and  0.00%, respectively.

I have the following questions:

1. What are the factors that lead to IV-VT mismatches? In my case what
would be the possible cause (despite of taking care of points 1 to 3
listed above)?
2. Would like to know of what is an acceptable percentage? If I
understand correctly errors are reported for deviation > 10%. Can
mismatch warnings be waived off in case removing them is difficult?
3. Transient analysis of the buffer seems to indicate that it takes some
time for the PAD to settle to a regular operation. The first output
transition does not reflect the correct behaviour of the PAD. So I got
the VT tables generated with 3rd rise and fall PAD transition. The IV
tables which are generated with DC conditions actually reflect the
initial buffer state, hence the mismatches remains. Tried generating IV
data by doing a very slow PWL sweep of PAD after ramping up the
supplies. However, the IV characteristic thus generated is highly
nonmonotonic and deviates from the expected. As a result mismatches are
even more. Can someone suggest a way to resolve this problem?

I am copying the IBIS model below for reference.

Best regards,

Birendra Rana

Please do not print this email unless it is absolutely necessary. 

The information contained in this electronic message and any attachments to this message are intended for the exclusive use of the addressee(s) and may contain proprietary, confidential or privileged information. If you are not the intended recipient, you should not disseminate, distribute or copy this e-mail. Please notify the sender immediately and destroy all copies of this message and any attachments. 

WARNING: Computer viruses can be transmitted via email. The recipient should check this email and any attachments for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. 

www.wipro.com

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.


--------------------------------------------------------------------
|For help or to subscribe/unsubscribe, e-mail majordomo@eda-stds.org
|with the appropriate command message(s) in the body:
|
|  help
|  subscribe   ibis       <optional e-mail address, if different>
|  subscribe   ibis-users <optional e-mail address, if different>
|  unsubscribe ibis       <optional e-mail address, if different>
|  unsubscribe ibis-users <optional e-mail address, if different>
|
|or e-mail a request to ibis-request@eda-stds.org.
|
|IBIS reflector archives exist under:
|
|  http://www.eda-stds.org/pub/ibis/email_archive/ Recent
|  http://www.eda-stds.org/pub/ibis/users_archive/ Recent
|  http://www.eda-stds.org/pub/ibis/email/         E-mail since 1993
Received on Wed Mar 4 01:14:00 2009

This archive was generated by hypermail 2.1.8 : Wed Mar 04 2009 - 01:15:05 PST