[SI-LIST] Re: Relation between slew rate and ISI

  • From: wolfgang.maichen@xxxxxxxxxxxx
  • To: "Preethi Ramaswamy" <preethi.gowtham@xxxxxxxxx>
  • Date: Tue, 6 May 2008 17:30:37 -0700

The source of your confusion is that you can get reflection at an 
impedance change, and you can get reflection at a parasitic element. In 
the former case you have constant, homogeneous line impedance Z1 before 
the change, and constant, homogeneous line impedance Z2 after the change, 
so you get a constant sized step back (rho=(Z2-Z1)/(Z2+Z1)), i.e. 
reflection amplitude form this particular discontinuity will not change 
over time, unless the incident signal changes again). On the other hand, a 
parasitic per definition is a localized deviation, i.e. spatial extent of 
the order of your signal rise time or less. In this case you see a 
"lumped, total effect", and the longer your signal rise time, the more 
this effect gets smeared out in time (smaller nut wider). You can picture 
a parasitic as a short section of transmission line with an impedance that 
is either higher (inductive) or lower (capacitive) than the line impedance 
before and after - and with a prop delay shorter than your signal rise 
time; in this picture you actually see TWO reflections (from the 
discontinuity at the beginning and the discontinuity at the end) lumped 
together, which explains the different behavior - if your signal rises 
very slowly, then both discontinuities see almost the same signal level 
(and same change) at every instant, and their responses mostly cancel - 
thus a long reflection (approx. as long as the rise time), but very 
shallow. On the other hand, a faster rise time means the signal level at 
the second discontinuity is lagging the first one quite significantly, and 
the composite reflection becomes larger - but on the other hand, the game 
is more or less over after the signal rise time (plus prop delay through 
the short section), so the reflection is shorter as well.
Yet another way to explain this is that the parasitc has a frequency 
dependent impedance; e.g. a parasitic capacitance C will have an impedance 
of Zc=1/(2*pi*C*f), thus since Zc varies with frequency, your reflection 
amplitude will vary with frequency. On the other hand, a homogenous, 
lossless  transmission line has an impedance that is independent of 
frequency, hence no influence of the rise time (remember, rise time is 
proportional to 1/frequency).

Hope that helps

Wolfgang







"Preethi Ramaswamy" <preethi.gowtham@xxxxxxxxx> 
05/06/2008 05:16 PM

To
"wolfgang.maichen@xxxxxxxxxxxx" <wolfgang.maichen@xxxxxxxxxxxx>
cc
si-list@xxxxxxxxxxxxx, si-list-bounce@xxxxxxxxxxxxx
Subject
Re: [SI-LIST] Relation between slew rate and ISI






Thanks Wolfang. That helps me concenptually justify what is happening in 
my simulation.
I found that reflections are the cause of fast case being worse than slow 
at the point where I'm tapping the signal.
 
I'm still finding it hard to wrap my head around the fact that faster the 
rise time, more pronounced the reflection is. How do you explain that 
formula ?
Reflection due to parasitics is caused due to discontinuity in the line 
impedance. So, everytime a signal hits a parasitic, based on the impedance 
of the parasitic, there is a positive or negative reflection and the 
amplitude of the reflection is proportional to the discontinuity in the 
impedance. Why should this amplitude be any different for different slew 
rates ?

 
Your input is much appreciated.

On 5/6/08, wolfgang.maichen@xxxxxxxxxxxx <wolfgang.maichen@xxxxxxxxxxxx> 
wrote: 

What your worst case will be depends on a variety of factors. 

If your main limitation is the drive signal rise time, then yes, slower 
rise time will degrade ISI as well as your vertical eye opening. Same goes 
if your path is dominated by skin effect and/or dielectric loss. 

On the other hand, if you have capacitive or inductive parasitics in the 
path, they will cause reflections, which will get more pronounced (larger 
amplitude but shorter duration) with faster signal rise times. As a rough 
rule of thumb, 

reflection = (approx) Tc/Tr x 100% 

where Tc is the time constant of the parasitic (Zo x C / 2 for capacitive, 
2 x Zo x L for inductive), and Tr is the signal rise time (assuming 
linearly rising edge from 0% to 100%). A slower rise time will smear out 
the reflection in time, so it's longer but not as high. If the reflection 
happens to come back at the middle of the data eye then it will reduce 
your vertical eye opening. On the other hand, even in this case rise time 
has only second order influence on data dependent jitter (ISI) - shorter 
rise times create larger reflections, but at the same time shorter rise 
times reduce the timing hit of a given size reflection, so in first order 
approximation that is a wash. 

Wolfgang 







"Preethi Ramaswamy" <preethi.gowtham@xxxxxxxxx> 
Sent by: si-list-bounce@xxxxxxxxxxxxx 
05/06/2008 03:52 PM 


To
si-list@xxxxxxxxxxxxx 
cc

Subject
[SI-LIST] Relation between slew rate and ISI









I'm looking for some information on the relation between slew rate and ISI
effect on high speed memory data signals.
From my SI simulation I'm observing that a signal looks much worse with a
higher slew rate than a lower slew rate. But the point I'm tapping is not 
at
the receiver but a point before the receiver. I don't expect the signal to
look as good as at the receiver but I was hoping that the trends match. At
the receiver itself, the fast corner signal looks better than the slow
corner signal. The bus is properly terminated.
Looking at the waveform, I see that in the fast corner case, whenever 
there
is a 1010 pattern, the signal is not reaching its intended Vhigh and Vlow
level. A similar thing happens at the slow corner but the signal swing is
much better. Hence, the eye diagram in the fast corner is almost closed
whereas the slow corner looks open.

While searching some artciles on this, I found the opposite stated. It 
said
that ISI effects are more pronounced in the slow corner case and Crosstalk
is more pronounced in the fast corner case.

Any experience or insight into this topic would be very helpful.


------------------------------------------------------------------
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 technical documents are available at:
               http://www.si-list.net

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
 






------------------------------------------------------------------
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 technical documents are available at:
                http://www.si-list.net

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: