[SI-LIST] Re: min/max trace length
- From: "DAVID CUTHBERT" <telegrapher9@xxxxxxxxx>
- To: bmgman <bmgman@xxxxxxxxxx>
- Date: Wed, 13 Aug 2008 16:05:30 -0600
The specified clock jitter seems like a wild card. If they would specify
clock jitter as RMS, and if it is truely Gaussian. I take a look at the
nature of the clock jitter (using a good o'scope) and derive my own RMS and
systematic jitter numbers. Then calculate the jitter excursion for say 12
months and add that into the timing budget.
Dave Cuthbert
On Wed, Aug 13, 2008 at 12:20 PM, bmgman <bmgman@xxxxxxxxxx> wrote:
> Bernd,
> There are times when a simulator is not strictly necessary, and a good
> spreadsheet timing calculation will do. Your clock speed suggests you
> may be in the spreadsheet domain.
>
> Build your spreadsheet with provision for every delay variable that you
> can think of. Some of the delays are not going to be well-specified,
> such as simultaneous switching push-out and crosstalk-induced jitter.
> Include them in your spreadsheet, and make a guesstimate for their
> values. Zero is probably not the right guess! As you gain knowledge,
> the guesstimates can be refined, giving you a better picture of your
> real margins.
>
> The setup calculation involves determining the path length from the
> latest data launch to the earliest capture clock. You find the latest
> common point (usually in the clock distribution) and time from there,
> using min delay in the clock path analysis and max delays in the data
> path analysis. Forget the data trace length and "margin" for a while.
>
> You should end up with the earliest clock arrival time being later than
> the latest data arrival time, or you are a priori broken. If not
> broken, the difference in time is the sum of the maximum trace delay
> plus the setup margin. Allocate as your system and engineering
> judgement allow.
>
> The hold calculation involves determining the earliest launch time and
> shortest data path versus the latest capture clock arrival in the
> launching cycle. Again, forget the data trace length and margin for a
> while. The data arrival time must be later than the clock arrival or the
> system is potentially broken. Should the data arrive before the clock,
> the difference is the required minimum trace delay for zero hold
> margin. Added length provides more hold margin.
>
> The difference between the maximum permissible and required minimum
> trace lengths is available to your board designer as a tolerance on the
> routing length. Note that you have been calculating in terms of time,
> and the trace delay is a function of Er. Be sure to talk to your board
> vendor to find out what his process variables are.
>
> Note that when you are calculating the fastest clock in the setup
> spreadsheet, you do not have to use the absolute Min delay from your
> data sheets. Why? Because the slowest data paths occur at minimum Vdd
> and max temperature, and the Min specs apply only at Max Vdd and min
> temperature. The clock (short) path can be calculated at Max delay
> minus a factor for the process variation. You'll have to research your
> vendors to find out what that may be. Similar reasoning applies to the
> calculation of the slow path in the hold spreadsheet.
>
> I've used spreadsheets with good results at system speeds considerably
> faster than you are considering. The big caveat is that SI
> considerations must be met independently of the timing calculation.
> Ringing, or pedestals, on any of your signals must either be eliminated
> by design, or incorporated into your timing spreadsheet. Net delay is
> not always simply (length * Tpd). For this, you need a simulator, or a
> well trained sense of intuition in the SI arena. Complex net topologies
> will blow most people's intuition out of the water, so SI simulation is
> highly recommended.
>
> Regards
>
> Mike
> - - - - -
> bernd schuster wrote:
>
> >Hi,
> >Unfortunately, I`ve no simulation tools... and therefore I read a lot of
> >datasheets and application notes from different manufacutres. Their
> >recommendation about different trace lengths are very often between 10mm
> and
> >20mm. Maybe that`s the best conditions on the one hand to be routable and
> on
> >the other hand....
> >
> >What about the skew and jitter calculation? I`m working with an ARM9
> >microcontroller with internal sdram controller and an external sdram. Is
> it
> >correct that the clock length have to be as long as the longest databus
> >length to get small skew
> >
> >
> >
> >>Of course, if you mean a synchronous interface, you may need a simulation
> >>
> >>
> >and timing calculation to determine the trace lengths.
> >The pcb layout software is able to calculate the normal length of a trace
> /
> >not the manhatten length - but it should be enough
> >
> >best regards
> >Bernd
> >
> >
> >2008/8/13 Benny Yan <zyan@xxxxxxxxxx>
> >
> >
> >
> >>Bernd,
> >> Setup and hold margin are the results from the layout length, output
> delay
> >>and input setup and hold time requirements.
> >>I think that you mean a source synchronous interface. In this case, the
> >>absolute trace lengths are not the main constraint.
> >>The setup margin 2ns and hold margin 0.8ns gives you a chance to tune the
> >>length matching between clock and data traces.
> >>Of course, if you mean a synchronous interface, you may need a simulation
> >>and timing calculation to determine the trace lengths.
> >>Or simply, you may use 1ns per 6inch to do a rough estimation, but don't
> >>forget to reserve enough margin.
> >>
> >>Benny Yan
> >>www.iometh.com
> >>
> >>-----Original Message-----
> >>From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]
> >>On Behalf Of bernd schuster
> >>Sent: Wednesday, August 13, 2008 5:13 PM
> >>To: si-list@xxxxxxxxxxxxx
> >>Subject: [SI-LIST] min/max trace length
> >>
> >>Hi,
> >>I often read minimum and maximum trace lengths in datasheets of sdram
> chips
> >>(for example). As I figured out the length depends on the setup margin
> and
> >>hold margin.
> >>
> >>If I have a setup margin of 2ns and hold margin of 0.8ns (e.g.) - which
> >>formulas will tell me the min. and max. trace length? The system is
> working
> >>at 100MHz with a fall time / rise tim of 2ns.
> >>
> >>How do I have to modify the formula when I want to calculate the timing
> for
> >>an external flash memory (without clock traces)?
> >>
> >>best regards
> >>Bernd
> >>
> >>
> >> ------------------------------------------------------------------
> >>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 technical documents are available at:
> >> http://www.si-list.net
> >>
> >>List archives are viewable at:
> >> //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:
> >//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:
> > //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:
> //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:
> //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:
//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:
//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:
