Re: Re[2]: GND Plane Bounce Question

Samie,

I agree with your assesment. Also, if you only care about the bounce
during the first few 100 ps's of the transient around the via, I have
the following numbers you might be interested. 2 - 3 inch away from the
via, the bounces are about 1.0 mV, 1.5 mV and 2.0 mV at 100 ps, 200 ps,
and 400 ps, respectively. 5 - 6 inch away from the via, they are 0.1 mV,
0.4 mV and 1.0 mV, respectively, and about 10 inch away, the bounce is less
than 0.1 mV at 400 ps. (all assuming 10 mA current).

I will also send you some information about our power/ground simulation tool
later (the tool is still under development and has not been released jet).

Regards,

Frank

> From uunet!ccm2.jf.intel.com!Samie_Samaan Tue Aug 22 09:16:17 1995
> Date: Tue, 22 Aug 95 08:43:00 PDT
> From: Samie Samaan <uunet!ccm2.jf.intel.com!Samie_Samaan>
> To: uunet!vhdl.org!ibis, uunet!fyuan%qdt.com%uunet,
> uunet!uunet!uunet!ccm2.jf.intel.com!Samie_Samaan
> Cc: uunet!uunet!qdt.com!jonp, uunet!uunet!qdt.com!fyuan
> Subject: Re[2]: GND Plane Bounce Question
> Content-Length: 7935
>
>
> Text item:
>
> Frank,
>
> Thank you very much for taking the time to address this problem. In
> fact I would like to know more about this tool, if you can send me
> some info on it. My address is with Quad, and also at the bottom. The
> numbers you give me are a big help. With a formula provided by someone
> else, I will have a good idea of what to expect.
>
>
> I would like to comment on this whole issue a bit, and address what
> prompted my question in the first place:-
>
> Some of the colleagues who answered my inquiry, correctly indicated
> that I should designate the source of the current and not just its
> sink. So I answered one of them by saying the following: Assume that I
> can magically inject the current into the GND plane thru the
> via...etc. The reason I say that is because I was really interested in
> local bounce, with the closest discontinuity to the GND plane being a
> few inches away. Such a case arises when you are dealing with, say,
> open collector buffers, where the collector is tied to a signal trace
> on the top layer of my hypothetical board, while the emitter pumps the
> GND plane thru the via. The source of the current has to tie back to
> the plane at some distant point, but that is not very relevant (in my
> opinion) during the first few 100 ps's of the transient around the
> via.
>
> The real reason I asked the above question has not been stated by me
> yet. Well, here it goes: If one is trying to measure Clk-out delay
> (Tco) of a buffer using a scope, and places the scope ground near my
> via, while placing the scope's active lead at the bit of interest,
> then, I believe, that if many bits are switching, that the waveform
> seen by the scope would be distorted by the GND plane bounce. Say you
> have a falling edge (while many bits toggle down), then since the GND
> bounces up, the scope would read a smaller voltage (its gnd lead is
> moving up too), and show the edge to be just as fast as when a single
> bit switches (or nearly so) even though the edge has --in reality--
> slowed down due to SSO. So one has to be more clever as to where to
> measure Tco is such situations.
>
> Do others agree with my assesment ??
>
> Sincerely,
>
> Samie Samaan
> Intel
> 2111 NE 25th Ave
> Hillsboro, OR 97124.
>
> ______________________________ Reply Separator _________________________________
> Subject: Re: GND Plane Bounce Question
> Author: fyuan%qdt.com%uunet@uunet.uu.net at SMTPGATE
> Date: 8/21/95 6:31 PM
>
>
> Hi, Samie,
>
> I did some quick and simple simulations for your GND plane bounce problem
> using the power/ground plane simulation tool we have at Quad Design.
> The PCB is same as yours, and the size is 10x10 inch. Assume the current
> is 10 mA (instead of 10 Amps in your message), and there are no de-caps or
> other loads between the power/ground plane, the local ground bound is about
> 8 ~ 10 mV, assuming the current source has a 50 ohm internal resistance.
> The ground noise propagates all over the plane, so the bounce away from the
> via is lower but not too much than that at the vis. Depending on location,
> the bounce is 2 ~ 6 mV. The location of the via also have some effects on
> bounce, but is usually small.
>
> Note that if you really have a 10 Amps current, the bounce would be 8 ~ 10
> volts. That will be a truly unusual or extrame situation.
>
> The above results assuming the power/ground planes are open (or no other
> terminations), This usually gives higher estimate of GND bounce because of
> the resonance of the plane. From my experice, the things affect the ground
> bounce mostly are your loads, especially decoupling capacitors. Proper
> placement of de-caps of 10~100 nf can significantly reduce the GND bounce
> in the above situation, by 5 ~ 10 times. Resistive loads have little
> effects on the GND bounce unless it is really small ( about 0.1 ohm for
> the above example).
>
> Our tool can do a much better job in extimating GND bounce if you have
> more information about the geometry, the ternimation or load conditions,
> and the source or device characteristics. Hope this is useful for you.
>
> Regards
>
> Frank Yuan
> Quad Design Technology
> (805) 988-8250
>
>
> > From uunet!ccm2.jf.intel.com!Samie_Samaan Mon Aug 21 13:11:20 1995
> > Date: Mon, 21 Aug 95 11:22:00 PDT
> > From: Samie Samaan <uunet!ccm2.jf.intel.com!Samie_Samaan>
> > To: uunet!vhdl.org!ibis
> > Subject: GND Plane Bounce Question
> > Content-Length: 1709
> >
> >
> > Text item: Text_1
> >
> >
> > All IBISians (SI experts by Default !!)
> >
> > I would like to pose the following SI question, and solicit help from
> > those who actually know the answer or have a very good idea of how to
> > estimate it by scaling from previous experience:
> >
> >
> > Assume one GND and one power plane in a PCB separated by say 10 mil,
> > with a dielectric constant of say 4. assume that the power plane has a
> > via opening where current is injected thru the via down to the GND
> > plane. Let's say that the current is 10 Amps. with a riserime of 1 ns.
> > Assume nominal via diameter. The drawing below shows the structure:
> >
> >
> > | Vcc
> > ---------------- | ------------------
> > | GND 10 mil, Er=4
> > -----------------|-------------------
> >
> >
> > What I would like to hear from anybody who actually has experienced
> > GND plane bounce, is:
> >
> > *howmuch do you expect the GND plane (not the via stem) to bounce up
> > (locally) given the above assumptions, and
> >
> > *howfar away from the via (radially) do you expect the bounce to drop
> > to 10% of the peak which occurs right where the via meets the plane ?
> >
> > I would apprciate a good guess (and Why: perhaps from previous
> > experience), or direction as to "who ?" you think could answer such a
> > question. I know that I might need to do a 3-D or radial transmission
> > line analysis of the system, but I'm really just looking for a quick
> > answer.
> >
> >
> > Thanks for taking the time to read or answer this question.
> >
> > Samie Samaan
> > Intel Corp.
> >
Received on Tue Aug 22 20:44:46 1995