This is coupled with the fact that, particularly with components in the personal computer market, there is increasingly a minimum clock speed. No longer can designers necessarily use a low-speed component, and so high-speed design techniques are becoming more and more common in relatively low-performance systems.

However, the key for successful, cost-effective design and quick time to market is to have the signal integrity and EMI tools tightly integrated into a constraints-driven design flow. For some engineers, this is fundamentally different to the way they have previously used such tools as add-ons to the existing design flow.

Consistency and concurrency are key to the design tools-integrating SI and EMI analysis into the design tools and adding fast simulation that gives an indication of whether the design will pass, fail, or is borderline, all in a matter of minutes rather than hours or even days.

The concurrency is vital as placement and routing can be an interactive process at best, especially with tighter space and cost constraints. To add another constraint of EMI and SI analysis at the end of the process can render such designs unrouteable. It's far better to be aware of the SI and EMI problems at the beginning, and design with those constraints in mind or in the tool. Such technology has to be tightly integrated as part of prototyping tools.

A constraints engine

Therefore, third-party tools tacked onto the edge of the design flow just won't work for modern designs which have to get to market in a reasonable time frame, avoiding multiple cycles through the EMC test center.

At the heart of the design flow has to be a constraints engine, which in the case of Zuken (Bristol, UK - (see Figure 1) who recently acquired the Incases EMI and SI analysis technology, is in the form of a spreadsheet called the Constraints Manager.

This Constraint Manager administers the constraints, giving an overview of the design at any point in the process. Additionally, the spreadsheet can modify the constraints from data in the design to refine them for the next stage of the design, and it opens up the possibility of synthesizing new constraints that may not have been clear earlier.

This can be done by linking the spreadsheet to a knowledge database that captures the expertise of the designer from previous designs, automatically recognizing elements or components in the new design and importing the previous constraints.

The Constraint Manager contains the required data for driving the floor planning, the routing, and the physical design. For example, in the physical design, the critical nets can be pre-routed and analyzed and then fixed before the rest of the PCB is routed.

But the designer also has to avoid the trap of over-constraining a design and rendering it impossible to build. This is where the human being is vital and why design can't be handled fully automatically.

Engineering expertise is essential in determining the best way of achieving the design objective.

As an example, one company tried to build a board with a microprocessor using the design rules supplied by the semiconductor manufacturer and ended up with a 16-layer board that was too expensive to make, but was guaranteed to work. However, the semiconductor manufacturer could provide a 12-layer board at a lower cost by using its engineering expertise to loosen some of the constraints and squeeze some others to reduce the complexity of the board and make it cost effective.

Tight integration

Design today isn't a linear flow, so having the world's most accurate point tool is of less value if it slows down the whole interactive cycle. What is more important is having the tight integration of tools to reduce that interactive cycle and at the same time having tools that offer the best ratio of accuracy to performance.

Simulation is a key element of any design flow and signal integrity-and EMI is no exception. In the past this has required highly accurate tools that took many hours to run, ruling out their use in the design flow unless absolutely vital. By using other simulation techniques, SI and EMI can be fully integrated into the design flow.

The enabling technology for this comes from developments at the University of Missouri at Rola (UMR) and a number of EDA companies, which included Incases, on a new set of algorithms. These are based on fast radiation screening rather than using techniques such as finite element analysis or method-of-moments, where the currents on all conducted surfaces are computed in detail and are then used to determine the resulting fields at any point in space. Accurate techniques, such as finite element analysis or method of moments, can be used to analyze radiation of specific structures like a clock distribution network of a fast bus in detail. The difficulty is, however, in determining which structures to analyze, and to interpret the results. Incases (and now Zuken) are using method-of-moment solvers in a tool called radiation workbench to provide this power to expert users.

By contrast, the UMR algorithm partitions the different radiation effects as the result of the differential mode and the common mode noise in the design.

The common mode voltage drop induces currents in structures such as heatsinks and cables, creating a radiating dipole. The algorithm takes care of all these parasitic dipoles and adds them together to create the overall radiation level.

This is a rather "rough and ready" approach, but gives the designer a red, amber, or green guide to the emissions-if the result is high, there's a problem; if it's on the limit, more analysis has to be done and if it's well inside the limit, there's no problem.

The advantage is that this analysis takes all of thirty minutes, not many hours, even for a complex board. Because the algorithm is based on partitioning, it also allows the designer to backtrack through the sources to find which particular dipole is causing the problem and so offer a potential solution.

Evaluation of the different algorithms shows that this screening methodology gives similar shape and magnitude results to those of accurate 3D field solvers if all the elements are modeled. But this type of screening is more vulnerable to changes in technology such as new materials and buried structures (for example, microvias and inductors) and the models have to be regularly updated to make sure that the test results are valid. To accomplish this, field solvers are used to model a new radiation effect for accurate results. And based on what is learned from this, new fast-screening algorithms can be developed.

Signal integrity analysis

Similarly, for the signal integrity analysis a combination of accuracy and performance is vital. Frequency domain analysis is fast, handling over 100 nets per second on a 500MHz Pentium III machine, but only accurate for linear component behavior. At the other end of the scale, time domain analysis will handle the transmission line analysis, including non-linear elements such as buffers, but at the cost of simulating at 1 net per second.

Sometimes the frequency domain simulation can even have its advantages. For drivers or receivers with clamping characteristics, the frequency domain will show the ringing normally clamped away by the diodes where the time domain simulation won't. Some designers like to see that ringing effect, as aggressive clamping will consume unnecessary energy and may increase the failure rate as the clamping diodes are dissipating the energy to ground and, with a high clock rate, can fail faster. With the knowledge of the ringing, the designer can change the design to avoid this problem.

But all these simulation tools can only provide value if they are tightly integrated into the design flow. Therefore the engineering environment not only has to provide powerful analysis, it also has to integrate best-in-class tools such as the auto router.

Following the acquisition of Incases by Zuken, Werner Rissiek became the Board Integrity Development Manager responsible for Zuken's high-end PCB design solutions. Prior to Incases, Rissiek worked on signal integrity and EMC issues at Siemens-Nixdorf.

To voice an opinion on this or any other article in Integrated System Design, please e-mail your comments to sdean@cmp.comd