[SI-LIST] Re: delay vs. transmission line length
- From: Raymond.Leung@xxxxxxxxxxx
- To: si-list@xxxxxxxxxxxxx
- Date: Thu, 12 Jun 2003 16:18:06 +1000
Scott has correctly mentioned that the output edge rate
is faster than the input rate, of which I actually have
tried 0.5n input Tr yet the conclusion is the same. I
therefore try the simulation in a more realistic rise time,
Tr=0.2ns. I have the following figures:
Case A: Open T-line
measured output Tr=0.35ns (approx., each one is a bit different)
Characteristic length l =.35n/140p = 2.5 inches
l/3 = 0.83 inches
results: T-line length (inch)
0.5 1 2 3 5 8
delay 0.22n 0.237n 0.237n 0.238n 0.238n 0.219n
As you can see, this time I cannot assure my last conclusion.
Likely the mechanism is more complex than anticipated (say,
reflection details would affect the outcome).
I would say that this kind of simulations is just for fun
because it is very risky to let the T-line being unterminated in a
length more than 3 times of the characteristic length.
Regards,
Raymond
Former message:
The discussion has become more interesting that I need to do some
simulations by myself. To simplify the work I have assumed perfect
ground/vdd/ouput connection; i.e., no inductance between the internal
driver pads to the outside world. In case any other suggestions I will put
them to the next simulations.
Some technical details:
driver: simple inverter, no decoupling (because of perfect gnd/vdd)
p size: 200u/0.28u
n size: 80u/0.28u
additional output cap: 3pF
input rise time: 1ns
process: our company's 0.15u 1.8V process
supply: 1.8V, full swing output implied
T-line: 60 ohm 140ps/inch delay U element
Case A: open T-line
measured output Tr=0.72ns (approx., each one is a bit different)
Characteristic length l =.72n/140p = 5.14 inches
l/3 =1.71 inches
results: T-line length (inch)
1 2 5 10
delay 0.495n 0.504n 0.473n 0.43n
Case B: terminated with 60 ohm
measured output Tr = 0.43ns
Characteristic length l= .43n/140p = 3.07 inches
l/3 = 1.02 inches
results: T-line length (inch)
1 2 5 10
delay 0.504n 0.504n 0.504n 0.504n
The above results show that a well terminated T-line is resistive to
the driver (Tr faster, delay independent of the T-line length). While
for no termination case, the delay is shorter when the T-line length
is away from the lumped model length.
regards,
Raymond
"Scott McMorrow" <scott@xxxxxxxxxxxxx> on 12/06/2003 12:02:07
Please respond to scott@xxxxxxxxxxxxx
To: Raymond Leung/sdc@sdc
cc: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: delay vs. transmission line length
All,
Okay, I'm tired of hearing about Report generation software now, so
instead of discuss something that I truly know nothing about I'll
respond to this thread.
Hmmm, if an ideal transmission line is perfectly terminated in it's
characteristic impedance, the length of the transmission line does not
"matter" to the driver. As Art points out, the transmission line looks
resistive to the driver. What the original poster did not state was
whether the transmission line was terminated or not, and whether it was
a lossless or lossy transmission line. In addition, he did not state
whether the driver was simulated by itself, or with the associated
device package, and if simulated with the package, whether package
coupling was included.
Now, if the transmission line is open ended, I would expect that the
measured Tpd would be reduced as the transmission line becomes shorter
and shorter, since the reflection will add to the output voltage of the
driver, causing a seemingly faster transition through the threshold
w.r.t. the driver input stimulus. If the driver is driving an ideal
transmission line with perfect termination, I would expect absolutely no
change in Tpd at all, and, in fact, delay can then be measured into a
pure resistive termination. The only times I've ever seen the sort of
Tpd relationship that Yoni described are in two cases. First, where a
power delivery SSO problem is this issue, with some sort of power/ground
bounce causing delay changes. Or second, where there is a package and
the package has high reverse crosstalk, and this reflected reverse
crosstalk causes an apparent increase in Tpd when the device drives and
unterminated load. But, I am only guessing.
Seems to me that in order to answer the original question and not
speculate, one needs to know the complete story.
regards,
scott
--
Scott McMorrow
Teraspeed Consulting Group LLC
2926 SE Yamhill St.
Portland, OR 97214
(503) 239-5536
http://www.teraspeed.com
Come visit us see our new
Teraspeed IBIS Model Library
and IBIS Reference Page
Raymond.Leung@xxxxxxxxxxx wrote:
>
>Art,
>
>No one would object to your analysis about T-line, after the popularity
>of the "black magic handbook". However, the issue is talking about the
>loading effect to the driver that the delay within the driver will be different
>when the T-line length is changing. In view of this theory you can say
>there is no added "C load" in short T-line, yet the EFFECT of cap loading
>is apparent to the driver, as analised both by you and Jon.
>
>Raymond
>
>
>
>Art Porter wrote:
>
>Well, that's the definitive behavior of a transmission line. A properly
>terminated transmission line "looks" resistive (i.e. voltage and current are in
>phase at all frequencies). If it isn't properly terminated, it still "looks"
>resistive at the time of the incident wave. The driver doesn't "see" the
>reflection of whatever is at the end of the transmission line until a time
>equal to twice the delay of the transmission line. A "short" transmission line
>doesn't have any inherent added "C load." With a short transmission line, there
>is less time between the incident edge and the reflection from the C (or
>whatever else) at the end of the transmission line. If the length of the
>transmission line is short compared to the transition time of the edge, then
>it's difficult to distinguish the incident edge from the reflection. That is
>the source of the "rule of thumb" that you can model transmission lines as
>lumped elements if their length is <1/5 (or 1/3, or 1/10, depending on which "a
> uthority" you prefer) of the transition time.
>
>Art
>
>-----Original Message-----
>From: Raymond.Leung@xxxxxxxxxxx [mailto:Raymond.Leung@xxxxxxxxxxx]
>Sent: Tuesday, June 10, 2003 6:20 PM
>To: si-list@xxxxxxxxxxxxx
>Subject: [SI-LIST] Re: delay vs. transmission line length
>
>
>
>
>
>I think it is more or less like a resistive load seen by the driver
>when the T-line is long enough. As what Jon has described
>below, the C load of a short line would cause longer Tpd.
>
>Raymond
>
>
>
>
>"Jon Powell" <jonpowell@xxxxxxxxxxxx> on 11/06/2003 01:07:31
>
>Please respond to jonpowell@xxxxxxxxxxxx
>
>To: yonitz@xxxxxxxxxxxxx, "Si-List (E-mail)" <si-list@xxxxxxxxxxxxx>
>cc: (bcc: Raymond Leung/sdc)
>
>Subject: [SI-LIST] Re: delay vs. transmission line length
>
>In a lot of ways this relates to a previous question on how to measure
>time-of-flight and relate that properly to the CLK->Q data in a static
>timing program. Since the actual CLk->Q (Tpd or whatever) of the device is
>dependent on load, the datasheets will spec it into a specific load and then
>it becomes the job of the SI and Timing tools to figure out how to properly
>change that data for accurate total path timing. It used to be that most
>drivers were spec'd into 50pf loads because that happened to be the loads on
>the chip testers. Now days 50pf is so far away from the real load that that
>spec isn't good enough for many purposes.
>As other people have pointed out, if the load is closer to the driver than a
>round-trip, the C of the load is seen by the driver and can change the
>transition rate of the driver, which effectively changes Tpd. If the load is
>farther away than a round-trip it is effectively non-existent to the driver
>(at least starting from steady-state) and can have no effect on Tpd.
>
>-----Original Message-----
>From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]On
>Behalf Of Yoni Tzafrir
>Sent: Tuesday, June 10, 2003 5:45 AM
>To: Si-List (E-mail)
>Subject: [SI-LIST] delay vs. transmission line length
>
>
>hi,
>i run some simulations, for measuring Tpd from input of the of the =
>driver to the output of the driver.=20
>
> |\
> | \driver transmission line
> input------------| =
>\---------------------------------------------------LOAD
> | /out
> |/
>
>I have noticed that the longer my transmission line, the Tpd becomes =
>shorter.
>does it make sense? as i understand it a longer transmission line means =
>more capacitance (and resistance) so the Tpd should be longer, doesn't =
>it?
>
>Yonatan (Yoni) Tzafrir
>*Tel: (972) - 3 - 7552300 (T/L: 351)
>*Fax: (972) - 3 - 6177130
> *Mobile: (972) -54- 459469
>*E-mail: <mailto:yonitz@xxxxxxxxxxxxx>
>
------------------------------------------------------------------
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 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 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:
