1. Crosstalk Analysis in UDSM technologies
Based on the work presented in “Crosstalk Noise Analysis in Ultra Deep Submicrometer Technologies” by :
, Kannan S. Tharmalingam and Magdy A.bayoumi
Proceedings of the IEEE Computer Society Annual Symposium on VLSI.
By :Arashk Noroozpour
Professor : Dr. S.M. Fakhraie

2. Interconnect Delay and Noise
The interconnect delay and noise have become the dominant factors in determining circuit performance [1]
Crosstalk noise, which is known as coupling noise, imposes three side effects on digital design:
It can affect timing, causing a delay failure.
It can increase the power consumption due to glitches.
It can cause functional failure [1]
Coupling capacitance between neighboring nets is a dominant component in today’s DSM design.
Noise from inductive coupling can also present problems for VLSI wires [2]
Scaling could affect both capacitive and inductive noises. [2]

3. Impact of Noise on Circuit Performance and Interconnect/Driver parameters
Place of crosscoupling
Signal direction between two parallel wires
The Operating Frequency
Spacing , wire length , wire width, resistances
Coupling length
Input rise-time
Driver strength
Peak Noise Amplitude (for victim)
Aggressor Delay
Noise Pulse Width

4. Transmission Line Model
Using three transmission lines allows us to vary L1, L2, L3 separately, and to be able to accurately notice the effects of each of the wire length and the cross-coupling length.
The figure on top shows the transmission line model, and the figure below shows three transmission lines with 2 conductors [1]

5. Simulations: Victim Amplitude vs. spacing
The noise relatively decreases faster for a small Ra [1]
The aggressor and victim widths have approximately no effect on the noise amplitude [1]
The effect of wire sizing diminishes if the coupling location is close to victim driver. But it is very effective when coupling location is close to victim receiver [1]
Spacing could be so effective for long coupling lengths.

6. Noise and Aggressor Delay
For the effect of sizing on the aggressor delay, Table 1 shows results.
The effect of aggressor and victim wire lengths with a constant coupling length between them, shown in the right-hand-side figure.
The peak noise decreases as the aggressor length increases as long as the coupling length is kept constant.
Victim length has no effect on the aggressor wire delay.

7. The Impact of the Coupling Length
The figure on top shows the increase of the victim peak amplitude with the increase of the coupling length. This is of course due to the coupling capacitance.
Aggressor delay is also affected by coupling length, especially for larger input rise-times.
Another important parameter that affects the noise especially in high-speed circuits is the input rise-time.

8. Noise Pulse Width Noise pulse width is affected with spacing.
Now we have three metrics for the noise:
peak noise amplitude
noise pulse width
peak noise occurring time
Some noise-avoidance techniques like increasing the spacing between two adjacent lines , and driver sizing or wire sizing seem to be reasonable.

9. Capacitive and Inductive Coupling
Capacitive coupling is a large problem for weakly driven nodes. [2]
Inductive coupling pushes the victim in the opposite direction from the capacitive coupling: a rising aggressor capacitively couples the victim up, but inductively couples the victim down.[2]
While capacitive coupling is mostly the “nearest neighbor” phenomenon, inductive coupling has a much larger range [2]
Inductive noise becomes a problem only when a large number of wires switch at the same time in bus-like situations [2]

10. Scaling effects on wire metrics
Noise coupling both capacitive and inductive, should be mostly unchanged under scaling as long as the wires scale in length.
For wires that do not scale in length, inductive noise can grow relative to the power supply, but more likely, these wires will be repeated.
Repeaters break up the current return paths effectively, making each repeated segment independent from the rest, and preventing inductive noise from growing over technologies.
Wires:
Wires that scale in length: the wire delay scales with technology, showing constant resistance and falling capacitance.
Wires that do not scale in length: show an increasing disparity with gate delays.
Designers rarely use global wires without repeaters. The increase in noise for long wires is another reason for using repeaters.

11. Conclusions
Interconnect/Driver parameters can affect crosstalk noise metrics. Noise metrics such as peak noise amplitude, aggressor delay and noise pulse width are the most important of all.
Both capacitive and inductive coupling, can present noise and thus threat functionality.
Inductive coupling pushes the victim in the opposite direction from the capacitive coupling
Wire scaling could have different effects on noise performance due to capacitive and inductive coupling.
Wires scale diversely due to technology scaling, local wires scale in length, while global wires do not.

12. References
1. Crosstalk Noise Analysis In Ultra Deep Submicrometer Technologies
by: Mohamed A.Elgamel, Kannan S. Tharmalingam and Magdy A. Bayoumi
proceedings of the IEEE Computer Society Annual Symposium on VLSI
2.The Future of Wires
by: RON HO, MEMBER, IEEE, KENETH W.MAI, STUDENT MEMBER, IEEE, and MARK A. HOROWITZ, FELLOW IEEE
proceedings of the IEEE
3.Digital Integrated Circuits: A Design Perspective
by: JAN M. RABAEY , ANATHA CHANDRAKASAN, BORIVOJE NIKOLIC