Signal versus power integrity in high-speed embedded design -

Signal versus power integrity in high-speed embedded design

Signal integrity (SI) and power integrity (PI) are two distinct but related realms of analysis concerned with proper operation of digital circuits.  In signal integrity, the main concern is making sure that transmitted 1s look like 1s at the receiver (and same for the 0s).  In power integrity, the main concern is making sure that the drivers and receivers are provided with adequate current to send and receive 1s and 0s.  So, power integrity could be considered a subset of signal integrity.  Really, they are both analyses concerned with the proper analog operation of digital circuits.

The Necessities for Analysis
If computing resources were infinite, these different types of analysis might not exist.  Entire circuits would be analyzed at once, and problems in one part of the circuit would be identified and eliminated.  But other than being bound by the reality of what can practically be simulated, the advantage of having different realms of analysis is that specific problems can be addressed in groups without falling under the category of “anything that could possibly go wrong.” 

In signal integrity, for instance, the focus is the link from transmitter to receiver.  Models can be created for just the transmitter and receiver, and everything in between. This makes simulating signal integrity fairly straightforward.  Power integrity, on the other hand, can be a little more difficult to simulate, since the “boundaries” are a little less defined, and really have some dependence on items in the signal integrity realm.

In signal integrity, the goal is to eliminate problems with signal quality, crosstalk, and timing.  The same types of models are needed for all these types of analysis.  These include models for the drivers and receivers, the chip package, and the board interconnect, which consists of traces and also vias, discretes, and/or connectors.  The driver and receiver models include information about the buffer impedance, edge rate, and voltage swing.  Often, IBIS or SPICE models are used as the buffer models.  These models are used with the interconnect models to run simulations to determine what the signal will look like at the receiver.  

The interconnect will consist mainly of circuit board traces, which behave like transmission lines.  Such transmission lines, or T-lines, have characteristic impedances, delays, and loss.  Their characteristics determine how the connected drivers and receivers interact with one another.  The electromagnetic properties of the interconnect must be solved using some type of field solver, which characterizes them in terms of a circuit element or S-parameter model that can be used with a signal integrity simulator.  Most traces can be modeled as a uniform 2-dimensional cross-section.  That cross-section is sufficient to calculate the characteristic impedance of the trace.  It is that impedance that will affect the shape of the waveform at the receiver on a signal line.  The most basic of signal integrity analysis involves setting up a board stackup, including appropriate dielectric layer thicknesses, and finding the right trace width to achieve a certain target impedance for the traces.

Traces are relatively easy to model compared to vias.  Proper via modeling becomes important when doing signal integrity analysis on faster signals.  Signals in the multi-GHz range often require a model from a 3D field solver to be appropriately characterized.  Fortunately, these signals tend to be differential, which keeps their effects relatively localized.  Fast, single-ended signals that pass through vias interact very strongly with the power distribution network (PDN).  The return currents from these vias pass through nearby stitching vias, stitching capacitors, and/or plane pairs, the same components that make up a PDN and need to be modeled for power integrity analysis.

Figure 1.  Energy propagation on a trace cross-section, a signal via, and a PDN.

In power integrity analysis, higher-frequency energy is distributed through transmission planes.  This immediately makes the analysis more complex than basic signal integrity, since energy is moving in x and y directions, as opposed to just one direction down the transmission line. 

At DC, modeling is relatively simple in that the series resistance of traces, plane shapes, and vias needs to be calculated.  But for high frequencies, analyzing the impedance between power and ground at various locations on the PDN requires complex calculations.  The impedance will vary based on board location – where the capacitors are placed, how they are mounted, and what type and value of capacitor they are. 

High-frequency behavior, such as mounted inductance and plane spreading inductance, need to be included in the modeling in order to generate accurate results for decoupling analysis.  There is a simple version of decoupling analysis usually referred to as lumped analysis, where the impedance of the PDN is calculated as if it were one node.  This is usually a good, quick first-pass type analysis to ensure that there are enough capacitors and that they are the right values.  Then, running a distributed decoupling analysis ensures that all the impedance needs of the PDN are met at various locations on the board.

Signal Integrity Simulations
Signal integrity simulations focus on three main issues of high-speed signaling: signal quality, crosstalk, and timing.  For signal quality, the goal is to get signals with nice clean edges, no excessive overshoot, nor ringback.  Usually these problems can be solved by adding some type of termination to match the impedances of the driver to the transmission lines.  For multi-drop busses, matched impedances aren’t always possible, so a combination of termination and length changes on the topology are necessary to control reflections such that they do not adversely affect signal quality and timing.

Figure 2.  Using signal integrity analysis and design space exploration to eliminate signal quality and crosstalk issues.

These same simulations can be run to determine the flight times of the signals as they travel through the board.  Board timing is an important part of the system timing, and gets affected by the lengths of the lines, their propagation speed as they travel through the board, and the shape of the waveform at the receiver.  Since the shape of the waveform determines when the received signal crosses the logic threshold, it is essential to the timing.  These simulations usually drive changes in the length constraints put on the traces.

Another signal integrity simulation that is usually run is crosstalk.  This involves multiple transmission lines coupled to one another.  As traces get packed into dense board designs, knowing how much energy they are coupling onto one another is essential to eliminating errors due to crosstalk.  These simulations will drive minimum spacing requirements between the traces.

To read more of this external content,  go to “Power Integrity Simulations.  

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.