12 April 2011
PCB layouts and 3D designs move closer together
For years, the worlds of electronic and mechanical computer aided design have evolved along very separate paths. It was never intended to happen this way.
For example, the Design Automation Conference (DAC) was never meant to be a conference purely about electronic design, the way it is now. The first DAC, held in 1964, had the nascent field of computer assisted electronic design as just one part of the programme and it was not seen as a major user of compute power at the time. Even ten years later, much of the design work performed by electronics engineers would be on nothing more sophisticated than rubylith strips on mylar film.
Federico Faggin, who led the design of Intel's groundbreaking 4004 microprocessor at the start of the 1970s, recalled years later how the engineers would find bits of transistor or routing stuck to their elbows, rather than where they were supposed to be. In 1964, the content was largely about 3D mechanical design, while Ivan Sutherland described his Sketchpad concept, one of the first attempts to build a graphical user interface for computers.
Slowly, the mechanical elements of CAD fell from the DAC programme. In fact, the content moved away from CAD altogether and towards design automation in its most general sense. It so happened that the largely abstract world of electronic circuit design, leading to the forerunners of the Spice analogue modelling language and various logic optimisation tools, was more suited to the non graphical computers of the time. And primitive graphical displays could cope more easily with the largely 2D world of PCB layout.
Mechanical CAD developers found other venues to discuss their technology. For years, there was little need for mechanical and electrical CAD to overlap. Manual processes worked well enough, especially when most of the circuitry could be constrained to fit a certain board size and when most components could be expected to project no more than millimetres from the surface. Tall electrolytic capacitors might cause concern in terms of height clearance, but this was relatively easy to accommodate in most systems.
Some companies saw a future in making the link between electrical and mechanical CAD but, because few systems had exotic mechanical demands, found limited enthusiasm among customers. When it picked up the remains of the Dazix business in the early 1990s, Intergraph saw an opportunity to bring together the two worlds under one broad product line. But, yet again, the idea did not take off. Some advanced users have repeatedly called for better integration between electrical and mechanical CAD.
NASA has been one of the driving forces for better cooperation, organising conferences to discuss how best to share data between different types of CAD tool. A growing number of users are now encountering problems similar to those faced by NASA. Consumer markets now provide a further impetus for merging mechanical and electronic CAD tools. Manufacturers want to pack more into a small space, while not creating hot spots that will cook sensitive components.
But smaller spaces leave less room for error and, if you start using the chassis for cooling as well as vibration protection, close cooperation between the two types of design team is essential. Any change may have a knock on effect on other components or subsystems, so the process of lobbing engineering change orders over the partitions between design departments becomes ineffective. Ideally, the layout tool would check component clearances against 3D data to ensure, for example, that tall electrolytic capacitors do not collide with the back of a display or battery compartment.
Heat transfer through the system is where the link between electronic and mechanical CAD has historically been strongest. Tools such as those developed by Flomerics – now part of Mentor Graphics – have been used to identify hot spots in a system design and to suggest workarounds. To work effectively, these tools need accurate models of component placement and structure in order to work out how hot air will flow around boards and heatsinks and follow conducted heat through the copper traces and around cutouts in the PCBs themselves.
Being able to transfer 3D models directly into the simulators will generally be much faster than reconstructing manually from schematics and layout files, although it may make sense during early prototyping to put together simple mockups that loosely resemble the planned PCB layout. There is a growing need for mechanical and electronic tools to interoperate and share data, but which standards they should use is less clear, despite the fact that some formats in use have been around for years.
These formats were developed originally for the aerospace community, which has struggled for years with the problems caused by electrical design not marrying up with the physical world. The Airbus A380 exemplified problems faced by engineers dealing with the interface: their software was not geared up to deal with the non standard bending requirements of weight saving aluminium electrical cabling versus the copper based harnesses that had been used until then. The key standards for these high end design environments come from US based organisation PDES.
The two main standards are STEP AP210 and AP214. Part of the problem is that STEP standards were defined with the idea that tools would support them natively. For the most part, vendors want to stick with their own file formats and translate data between them. Altium's software can export PCB data to the mainstream STEP format as well as generating its own 3D visuals to show how a physical board will look – and it tool will perform clearance checks based on 3D data.
Zuken, which worked on the AP210 and AP214 specifications together with Mentor and with the help of STEP specialist LKSoft, wound up using direct links to Dassault Systèmes' Catia V5 software from its CR-5000 PCB design package. Mentor and PTC decided to team up to adopt and promote a later iteration of the STEP work. Taking requirements from a series of workshops held in 2005, the ProSTEP iVIP group worked on the EDMD data exchange format.
This approach, based on the XML language used in internet software, adopted many elements of AP210 and AP214. This resulted in the IDX format, which has yet to prove popular with other vendors, who favour the older, but more widely used, IDF format. IDF is also supported by Pro/Engineer, among other mechanical CAD tools. The availability of free tools such as Farnell's Eagle – picked up through the acquisition of CADsoft – and a version of EasyPC distributed by RS Components under the DesignSpark name, may increase the use of 3D in design.
DesignSpark will export IDF files and the distributor has produced 3D models of many of the components it sells to allow their use in free mechanical CAD tools, such as Google's SketchUp. Tools exist for Eagle to export design files and generate 3D models in SketchUp. Alternatively, there is the option to use Eagle3D in the paid for Pro release of the PCB design software. However, the link between SketchUp and electronic CAD is not easy to achieve.
Google is concentrating mainly on the Collada format developed for interactive 3D modelling and animation, rather than engineering oriented formats; partly because the internal representations used by SketchUp are based around surface modelling. They do not readily describe solid objects, unlike tools such as SolidWorks. A further reaction to the demand for increasing density in electronic systems is to push more of the system design into multichip packages, resulting in 3D ICs. These 3D IC packages may include a variety of different chips, including logic, memory, analogue and MEMS sensors.
Companies such as Cadence do not reckon a big change is needed in design. A number of them have had IC package design tools on the market for some years and these are being extended to take care of the needs of a new generation of multichip modules. As with the link to mechanical CAD, the biggest problem lies in defining interfaces to other tools. Generally, the PCB layout engineer will know where the package's external pins or pads are located, but will have little idea of the design of the ICs inside. This could turn out to be a hindrance to efficient layout.
Designers working with large FPGAs have found that it is often worth being able to swap pins on the package to make it easier to route signals to other devices on the board. Very often, the effect of a pin swap on the internal logic is negligible, other than the effort involved in making it happen. Many 3D ICs are likely to use interposers and signal redistribution layers to carry signals to the individual chips. As with FPGAs, PCB layout engineers are likely to see a benefit from being able to move signals around on these intermediate chips so they can bring out key signals on better positioned pads.
This will call for a greater level of codesign between the 3D IC and the PCB. Similarly, it may make sense to move the ICs around inside the package, based on the overall layout of the PCB. Processors will generally run much hotter than the memory chips around them, so PCB designers may want to move them away from other hot spots or bring ICs together under one efficient heat sink – possibly entailing a major change to the orientation of chips inside the 3D stack. The core technology for this kind of work is in place, but not yet necessarily in place in commercial tools right now.
Some 40 years since the two branches of CAD went their separate ways, a bond is being formed between the largely 2D world of traditional electronic design and 3D real world. But the work to bring them together is not over yet.