[SI-LIST] Re: Ground nodes in spice

Fabrizio,

There are several basic assumptions in this model:

1) Signaling is pure balanced differential with no common mode.  If this 
is the case, there will be absolutely no net current flow in the ground 
blades of the connector.

2) The ground inductance is significantly lower than the signal 
inductance, so that the ground differential voltage from input to output 
due to signal switching is minor.  This is actually not a bad assumption 
as long as the signals are nearly pure differential mode.

3) The model has been reduced to a grounded differential circuit from 
the original partial inductance circuit, maintaining the correct delay.

This model is useful for understanding the typcial performance of the 
connector in your system simulation, but will not be accurate enough to 
characterize the interconnect for worst case.  In the worst case, there 
are common mode components which will flow through the ground blades of 
the connector.  These common mode components are caused by routing skew, 
driver skew, and adjacent signal crosstalk (trace-to-trace, via-to-via, 
and pin-to-pin in the connector).

One useful technique would be to perform some crosstalk and eye patter 
simulation studies of the fully coupled connector models in isolation 
with 50 +/-10% driver sources and 50 ohm+/-10% termination and various 
combinations of signal skew.  With only the connector and synthetic 
sources  you can perform some comprehensive crosstalk and return path 
studies, and determine the worstcase noise and eye closure across the 
connector with multiple switching patterns.  These studies can be 
performed in the time domain, or they can also be performed in the 
frequency domain.  The results can be used as the additional noise 
guardband required for your system to operate when simulated with only 
one differential pair and perfect grounding.  This will get you close 
enough.

Another solution would be to use frequency domain simualtion techinques, 
utilizing one of the many different methods.  Your VHDM-HSD multi-line 
spice model can be converted to touchstone S-parameters and simulated in 
a complete system simulation environment in either the frequency domain, 
or the time domain using a convolution simulator. (Agilent ADS, Ansoft 
Designer, CST, Nspice, Apsim ..... etc.)

If you are not pushing the envelope, you can use the above 
pre-characterization scheme to guardband your design.  But, if you are 
pushing the envelope in noise performance, more indepth modeling might 
be indicated, including modeling of the backplane and PCB PTH via 
structures.

best regards,

scott


-- 
Scott McMorrow
Teraspeed Consulting Group LLC
2926 SE Yamhill St.
Portland, OR 97214
(503) 239-5536
http://www.teraspeed.com



Fabrizio Zanella wrote:

> Thanks for your valuable comments, Bob and Scott.  I'm using Hspice, 
> the connector is the HSD6, and I'm using one of the differential pair 
> two line models, 'hsd6ab.cir'.  This model is made up of 23 cascaded 
> pi-sections, and contains five nodes in the following arrangement
>
> .SUBCKT HSD6AB
>
> +   100      101              102
>
> +           2401             2402
>
>  
>
> I've found that using the 6x6 style connector models makes the 
> simulation times much too long, especially when running for 200ns or 
> more for eye pattern simulations.  So with this 2 line model there's 
> only one ground which is common for every ground pin, on both ends of 
> the connector. 
>
> So I'd like to refer to the accuracy issue you brought up regarding 
> tying both ends of a connector model to the same ground node - using 
> this model is this not an issue?  With this connector model, how does 
> the ground pass from the daughtercard to the backplane if I were to 
> use separate reference nodes - for example, node 10 for the card, node 
> 20 for the backplane, etc?
>
> Is it advisable to connect these reference nodes to each other through 
> an LR circuit, and if so what would be reasonable values to use?
>
> Thanks, Fabrizio
>
>  
>
> -----Original Message-----
> From: Scott McMorrow [mailto:scott@xxxxxxxxxxxxx]
> Sent: Thursday, February 27, 2003 12:58 PM
> To: rhaller@xxxxxxxxxx
> Cc: Fabrizio Zanella; si-list@xxxxxxxxxxxxx
> Subject: Re: [SI-LIST] Re: Ground nodes in spice
>
>  
>
> Fabrizio,
>
> I would agree with everything that Bob has suggested, and emphatically 
> reinforce item 3.  Every  backplane connector model that I have had 
> experience with has been modeled with partial mutual inductance for 
> signal and power pins.  If both sides of the connector are tied to the 
> same ground node, (whether node 0 or some other node), then mutual 
> inductances are shorted out, changing the characteristics of the 
> model.  For some connectors this change is huge and can amount to 
> large differences in delay, impedance and coupling.
>
> Many board-level transmission line SI simulators utilize node 0 for 
> simulation across all boards and connectors.  This is, of course, 
> incorrect and will lead to significantly incorrect results.  There is 
> nothing that can be done but to move to another nodal simulator like 
> Spice.  When you do this, and separate the grounds from each section 
> of the design, then you do have to adhere to Bob's item #2.  The 
> reference for displaying your signals must be the local reference on 
> the card, chip, or thing that you are measuring.  Typically, all 
> waveform displays will reference all signals to node 0.  In order to 
> utilize the remote reference you will either have to use some math 
> with your waveform viewer and view the difference of the signal to the 
> remote reference, or use a level shifting circuit element in the net 
> list (my preference.).
>
> Finally, there are some connector models that are poorly constructed, 
> having a large dynamic range between the largest and smallest 
> inductive elements.  This wrecks havoc in simulation of connectors 
> with large mutual inductance matrices and will cause convergence 
> issues or significant slow down in Spice performance (10 to 100:1) 
> when compared with models from competitors.  A particular connector I 
> am thinking about was specifically designed for 2 to 3 Gig backplanes. 
>  When the grounds on either side of the connector are isolated (as 
> they should be) then huge convergence problems occur, since this 
> requires that a large number of extremely small partial mutual 
> inductance elements be simulated.  Tying both sides of the connector 
> to the same ground will eliminate the issue, but with a severe loss in 
> accuracy, and some still greatly increased simulation times.
>
>
> I hope this helps.
>
> regards,
>
> scott
>
>
> Robert Haller wrote:
>
>Fabrizio,
>
>    I have a couple of points you should be  aware of.
>
>1. You do not HAVE to tie all of the ground nodes set to zero, but if 
>
>you don't
>
>be prepared to accurately model your interconnect, connectors, decoupling
>
>and signal to return ratio's.
>
>2. When you are displaying an Eye pattern on a remote board not tied to 
>
>zero,
>
>use the remote reference for your signals reference.
>
>3. If you are using a connector model that contains partial inductances 
>
>you will violate
>
>the assumptions of the field solver used to create that model if you  
>
>ground (tie to zero)
>
>both sides of the connector.
>
> 
>
>Regards,
>
>Bob
>
> 
>
> 
>
>Robert J. Haller (rhaller@xxxxxxxxxx <mailto:rhaller@xxxxxxxxxx>)
>
>Principal Consultant
>
>Signal Integrity Software Inc.
>
>6 Clock Tower Place, Suite 250
>
>Maynard, MA 01754
>
>Phone: (978) 461-0449, ext 15
>
> 
>
>Fabrizio Zanella wrote:
>
> 
>
>  
>
>>I have a question regarding backplane simulations of serial 8b/10b
>>
>>signals, running 2.5Gbs and above.  Does it matter whether the ground
>>
>>nodes of the daughtercards and backplane are all set to 'zero' or to a
>>
>>floating node for eye pattern results?
>>
>>Thanks,
>>
>> 
>>
>>Fabrizio Zanella
>>
>>Principal Hardware Design Engineer
>>
>>Broadbus Technologies
>>
>>fzanella@xxxxxxxxxxxx <mailto:fzanella@xxxxxxxxxxxx>
>>
>> 
>>
>> 
>>
>>    
>>
> 
>
> 
>
>------------------------------------------------------------------
>
>To unsubscribe from si-list:
>
>si-list-request@xxxxxxxxxxxxx <mailto:si-list-request@xxxxxxxxxxxxx> with 
>'unsubscribe' in the Subject field
>
> 
>
>or to administer your membership from a web page, go to:
>
>http://www.freelists.org/webpage/si-list
>
> 
>
>For help:
>
>si-list-request@xxxxxxxxxxxxx <mailto:si-list-request@xxxxxxxxxxxxx> with 
>'help' in the Subject field
>
> 
>
>List archives are viewable at:     
>
>         http://www.freelists.org/archives/si-list
>
>or at our remote archives:
>
>         http://groups.yahoo.com/group/si-list/messages 
>
>Old (prior to June 6, 2001) list archives are viewable at:
>
>         http://www.qsl.net/wb6tpu
>
>  
>
> 
>
>  
>
>
>
>-- 
>
>Scott McMorrow
>
>Teraspeed Consulting Group LLC
>
>2926 SE Yamhill St.
>
>Portland, OR 97214
>
>(503) 239-5536
>
>http://www.teraspeed.com
>
> 
>
>  
>
> ------------------------------------------------------------------------
> This email message and any files transmitted with it contain 
> confidential information intended only for the person(s) to whom this 
> email message is addressed. If you have received this email message in 
> error, please notify the sender immediately by telephone or email and 
> destroy the original message without making a copy. Thank you.




------------------------------------------------------------------
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:
http://www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List archives are viewable at:     
                http://www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages 
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu

Other related posts: