[SI-LIST] Re: DDR2 eye-diagram masks creation
- From: steve weir <weirsi@xxxxxxxxxx>
- To: Hermann Ruckerbauer <Hermann.Ruckerbauer@xxxxxxxxxxxxx>
- Date: Tue, 09 Oct 2012 08:49:20 -0700
Hermann, we may look at this a bit differently because I concern myself
more with the physical channel. It is my position that the most
accurate characterization of a channel is always from the point of view
of the receiver timing reference: DQS for DDRx data, and CLK for A/C.
What that does to a Tvac measurement is reduce the measured Tvac valid
interval based on jitter between the data signals and the strobe. I
contend that if a reliable channel is what we are after that this
approach is the most reliable.
Best Regards,
Steve.
On 10/9/2012 7:57 AM, Hermann Ruckerbauer wrote:
> Hello Steve,
>
> we are totally in line with this one and I never said anything different
> on the dataeye.
> This is the reason why I still use eye diagrams for memory, even this
> there is no mask defined.
> Yes, as you mention you can read out if tDS and tDH is met and e.g. if
> the tDIPW is causing any problem. It is just not possible to generate a
> diamond shaped mask with tDS/tDH and have all parameters verified by
> this mask.
>
> I never thought about generating multiple independent masks and place
> them in the same eye... They need to be aligned independently inside the
> eye because they will will need to be measured from different times
> (t_VAC from AC crossing, tDIPW from Vref Crossing, tDS/tDH from DQS
> reference).
> But this is still not correct: Also the eye would need to be generated
> differently for this different masks. For t_VAC the generated eye needs
> to be timing referenced to the AC level crossing, for tDIPW to the Vref
> crossing. So one need to generate different eyes for each mask.
>
> I have never seen this approach and I still think it is simpler to
> generate the eye and do a timing budget based on this one. In parallel
> the automated simulator/scope functions to do an edge by edge compliance
> test can be executed.
> The second one I'm usually adding when doing large sweeps with a lot of
> variations. These kind of test are simpler to automatize to get some
> PASS/FAIL information than looking to hundreds of eyes an calculating
> timing budgets
>
> Hermann
>
>
>
> Our next Events:
> ================
>
> "Open the Black Box of Memory"
> Seminar on 08/09. November 2012
>
> Check our website or contact us for details
>
> EKH - EyeKnowHow
> Hermann Ruckerbauer
> www.EyeKnowHow.de
> Hermann.Ruckerbauer@xxxxxxxxxxxxx
> Veilchenstrasse 1
> 94554 Moos
> Tel.: +49 (0)9938 / 902 083
> Mobile: +49 (0)176 / 787 787 77
> Fax: +49 (0)3212 / 121 9008
>
> schrieb steve weir:
>> Hermann, the need for multiple qualification masks does not alter the
>> either the reality or the utility of the eye. If for example the eye
>> is narrower than tDPIW one immediately knows that they have a
>> problem. TSimilarly, Tsu, and Th masks will determine if those
>> parameters have been met. In all cases the geometry gives a clear
>> view of margins, positive or negative and clues as to what the major
>> impairments are.
>>
>> Best Regards,
>>
>>
>> Steve.
>> On 10/9/2012 3:59 AM, Hermann Ruckerbauer wrote:
>>> Hello Steve,
>>>
>>> I did not say, that an eye can not be drawn. I think the question was to
>>> generate a Mask and for memory up to DDR3 this is not possible (guess
>>> DDR4 changes this one):
>>>
>>> One can not take the setup/hold times and generate an eye mask with the
>>> AC/DC levels.
>>> It is not allowed to apply tDS(min) and tDH(min) at the same time,
>>> because this would violate tDIPW (data input pulse width).
>>> Or if you take tDS(min) and draw t_VAC for special speed grades the
>>> resulting "eye mask" is much larger as just tDS + tDH.
>>> So there are several timings that need to fulfilled at the same time and
>>> it is not possible to combine this in a single mask.
>>> In the video on my website the case with t_VAC is shown.
>>>
>>> As you mentioned one need to take care HOW the eye is generated. In the
>>> Source Sync system it is required to take the timing reference (so for
>>> DQ the DQS rising and falling), place this in the center and draw the
>>> corresponding DQ edges (before and after) around. This needs to be done
>>> for each DQS edge. I know that scopes can do this in the meantime .. I'm
>>> not sure how many simulators are able to evaluate an eye this way ..
>>>
>>> Even I know it is not correct: My standard Eye evaluation is an AC/DC
>>> based eye opening.
>>> Based on this I can calculate Timing budgets (accounting for DQS
>>> jitter).
>>>
>>> Maybe this is some discrepancy in wording: An eye Mask defines an area
>>> the signal is not allowed to cross. The AC/DC open Data window
>>> evaluation is NOT an eye mask even it looks somehow like this.
>>>
>>> Hermann
>>>
>>>
>>>
>>> Our next Events:
>>> ================
>>>
>>> "Open the Black Box of Memory"
>>> Seminar on 08/09. November 2012
>>>
>>> Check our website or contact us for details
>>>
>>> EKH - EyeKnowHow
>>> Hermann Ruckerbauer
>>> www.EyeKnowHow.de
>>> Hermann.Ruckerbauer@xxxxxxxxxxxxx
>>> Veilchenstrasse 1
>>> 94554 Moos
>>> Tel.: +49 (0)9938 / 902 083
>>> Mobile: +49 (0)176 / 787 787 77
>>> Fax: +49 (0)3212 / 121 9008
>>>
>>> schrieb steve weir:
>>>> Hermann, I disagree a bit here. Source synchronous busses have timing
>>>> eyes just like channels with embedded clocks. Eye diagram captures for
>>>> source synchronous busses should trigger on the timing strobe. Levels
>>>> need to satisfy the AC signal thresholds. Measuring the eye is a
>>>> little
>>>> bit of a PITA because timing is against both clock edges. One way
>>>> around that is to perform separate captures on each clock edge.
>>>>
>>>> Steve
>>>> On 10/9/2012 12:40 AM, Hermann Ruckerbauer wrote:
>>>>> Hello,
>>>>>
>>>>> there is no real eye mask for DDR2/DDR3. On can generate some AC/DC
>>>>> based eye opening, but this is not correct as DRAMs do have a source
>>>>> snyc edge by edge specification instead of an eye mask.
>>>>> Some time ago I generated a movie/presentation that you can
>>>>> download on
>>>>> my website.
>>>>>
>>>>> Go to www.EyeKnowHow.de
>>>>> under Download/Movies you will find a DDR3_SetupHold_tutorial at the
>>>>> bottom. This was done for DDR3, but is basically valid for DDR2 as
>>>>> well..
>>>>>
>>>>> For anybody how is interested there in more detail: just visit our
>>>>> Seminar (see below) ;-)
>>>>>
>>>>>
>>>>> Hermann
>>>>>
>>>>> Our next Events:
>>>>> ================
>>>>>
>>>>> "Open the Black Box of Memory"
>>>>> Seminar on 08/09. November 2012
>>>>>
>>>>> Check our website or contact us for details
>>>>>
>>>>> EKH - EyeKnowHow
>>>>> Hermann Ruckerbauer
>>>>> www.EyeKnowHow.de
>>>>> Hermann.Ruckerbauer@xxxxxxxxxxxxx
>>>>> Veilchenstrasse 1
>>>>> 94554 Moos
>>>>> Tel.: +49 (0)9938 / 902 083
>>>>> Mobile: +49 (0)176 / 787 787 77
>>>>> Fax: +49 (0)3212 / 121 9008
>>>>>
>>>>> schrieb danko miocinovic:
>>>>>> Dear Experts,
>>>>>> I am doing some verification of LPDDR2 memories and my question is
>>>>>> about creating eye-diagrams. I got some material from Agilent but
>>>>>> it gives not enough details.
>>>>>> How do you create eye-diagram mask, what are the boundaries? If we
>>>>>> imagine the central romboid of a mask, I am interested if there is
>>>>>> some detailed tutorial about setting those boundaries.
>>>>>> If I have e.g. 533 memory and I know exactly what the frequency
>>>>>> is, I can calculate the UI x-size and easily create it.
>>>>>> What are boundaries for Y coordinates in this central part
>>>>>> (grayed-out area), how to create them? Do I put some minimal or
>>>>>> maximal values there (maximum values sound as the worst case)?
>>>>>>
>>>>>> I hope I was clear enough about what I would like to know, if not
>>>>>> I could provide more info.
>>>>>>
>>>>>> Thank you.
>>>>>>
>>>>>>
>>>>>> Kind regards,
>>>>>> Danko Miocinovic
>>>>>> ------------------------------------------------------------------
>>>>>> To unsubscribe from si-list:
>>>>>> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>>>>>>
>>>>>> or to administer your membership from a web page, go to:
>>>>>> //www.freelists.org/webpage/si-list
>>>>>>
>>>>>> For help:
>>>>>> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>>>>>
>>>>>>
>>>>>> List forum is accessible at:
>>>>>> http://tech.groups.yahoo.com/group/si-list
>>>>>>
>>>>>> List archives are viewable at:
>>>>>> //www.freelists.org/archives/si-list
>>>>>> Old (prior to June 6, 2001) list archives are viewable at:
>>>>>> http://www.qsl.net/wb6tpu
>>>>>>
>>>>> ------------------------------------------------------------------
>>>>> To unsubscribe from si-list:
>>>>> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>>>>>
>>>>> or to administer your membership from a web page, go to:
>>>>> //www.freelists.org/webpage/si-list
>>>>>
>>>>> For help:
>>>>> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>>>>>
>>>>>
>>>>> List forum is accessible at:
>>>>> http://tech.groups.yahoo.com/group/si-list
>>>>>
>>>>> List archives are viewable at:
>>>>> //www.freelists.org/archives/si-list
>>>>> Old (prior to June 6, 2001) list archives are viewable at:
>>>>> http://www.qsl.net/wb6tpu
>>>>>
>>>>>
>>
>
--
Steve Weir
IPBLOX, LLC
150 N. Center St. #211
Reno, NV 89501
www.ipblox.com
(775) 299-4236 Business
(866) 675-4630 Toll-free
(707) 780-1951 Fax
All contents Copyright (c)2012 IPBLOX, LLC. All Rights Reserved.
This e-mail may contain confidential material.
If you are not the intended recipient, please destroy all records
and notify the sender.
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
