Technology Description Benefits Drawbacks Development Cycle

Technology Description

An integrated circuit consists of very small structures which are very sensitive to mechanical damage and to exposure to chemicals (including water). They are so fragile, in fact, that handling by people or by the equipment used to manufacture electronic assemblies would destroy most integrated circuits - hence the use of packages to provide protection. The packages used by the electronics industry provide extremely good protection. It is typical that packaged integrated circuits can last for decades in very difficult environments.

As you can see in the picture, the packaged integrated circuit has a footprint that is much larger than that of the integrated circuit alone. It is not uncommon for a packaged integrated circuit to require as much as 4X the footprint as would be required for the die alone. This is particularly true for small die measuring 0.1" on a side. The larger the die, the less of a penalty paid there is. The package footprint for die measuring about 0.5" on a side (considered very large die) may be only about 2X larger than the die itself.

In the next picture, you can see than by placing multiple die into a single package - thus creating a multichip module (MCM) - the overall footprint of the resulting device has a much smaller footprint than would be required for individually packaged integrated circuits.

The reduction in footprint (size) reduction has been one of the primary reason that MCMs were developed. But, as is presented below there are several other good reasons for using MCMs, as well as several penalties for their use as well.

Benefits

• Size

  As was seen in the example above, the number one reason for using MCMs has been the footprint reduction that can be achieved, with a 4X reduction in size not unusual.

  The example given above used the most typical approach to mounting the integrated circuits inside the MCM - mounted flat against the interconnect. However, there are several other ways of attaching the chips.

  Many companies have demonstrated the capability of mounting die on edge or stacked one on top of the other to reduce the footprint. These techniques are collectively know as 3D packaging and have gained popularity in the last few years.

• Weight

  Along with the size reduction there is a considerable weight reduction as well, approximately linear with the size reduction. Weight can be a significant factor in hand held products, but is most important in military applications such as satellites where launch costs can reach up to $50,000 per pound.

• System Cost

  In cases where the electronic components of a product design will not fit on the available PWB area, additional PWBs must be used to hold the components. By using MCMs to reduce the required footprint to that available on a single PWB, the development and recurring costs of the additional PWB can be avoided.

  Whether the increased development and recurring costs of the MCM are offset by the reduction in additional PWB development and recurring costs usually depends on the quantity of products to be manufactured.

• Performance (speed)

  The speed at which electrical signals can be transmitted between two ICs is a linear function of the distance between them. On conventional PWB assemblies it is common for ICs to be separated by as much as 5-10 inches. Signal propagation typically takes place at about 1ns per foot, so for low frequency PWB designs the distance between ICs is not critical. However as clock frequencies increase above 500MHz, the distance (hence the propagation delay) can begin to have detrimental effects on the performance of the designs.

  Typical MCMs rarely exceed 2" on a side. So in an MCM, the distance between the ICs is very small and the effect of signal propagation delays is conderably reduced - allowing performance at up to 50% higher speeds than is possible using non-MCM approaches.

• Power

  The traces in an interconnect act as capacitors which ICs must charge in order to propagate electronic signals. It is not unusual for as much of 30% of the total power dissipation of an IC to be a result of the power required to charge the trace capacitance. Since capacitance is linear with length of the interconnect traces, the 1-2 inch traces in a PWB result in a 5X-10X reduction in that portion of the overall power dissipation of the IC - up to 30% depending on the specific design.

  It is also the case that interconnect trace resistances dissipate power, but the effect is oonsiderably less than the effect of line capacitance.

• Pre-tested Functions

  One of the more expensive aspects of developing an electronic assembly is the cost of creating the software to test the finished product. By incorporating an available MCM, (which is pre-tested by the MCM manufacturer) into the design, significant reductions in cost and schedule may be achieved.

• Package Cost

  For more complicated, high performance integrated circuits the IC package can be very complicated - costing more than the integrated circuit itself. Although an MCM package is also relatively expensive, replacing multiple IC packages with a single MCM package can result in an overall cost savings.

Drawbacks

• Development Cost/Schedule

  Developing an MCM is a fairly expensive, time consuming process. Development costs of several hundred thousand dollars and 6-9 months of calendar time are not unusual, whereas existing, individually packaged ICs are usually available off-the-shelf with short delivery times and carry no development cost at all.

• Production Cost

  An MCM is almost always more expensive than the cost of the individually packaged integrated cicuits. The low production volumes of MCMs result in comparatively high labor content and correspondingly high cost. This is true not only for the MCM assembly, but also for the interconnect fabrication and package manufacture. Also, because of the low volumes, manufacturing yields do not mature to reach the high levels enjoyed by individually packaged ICs. IC manufacturers often achieve yields in the 97%-99% range, whereas it is not uncommon for MCM manufacturers to see first pass test yields below 90%. Subsequent rework prevents the MCMs from being scrapped, but the rework cycle adds costs which are not found in the manufacture of integrated circuits.

  A production MCM may require from 1-3 hours of labor to assemble and test the MCM. At the relatively high overhead and labor rates of high tech companies, labor contributes several hundred dollars to the price of the MCM over individually packaged integrated circuits.

  In many cases, particularly for military applications where the volume is very low and the performance/environmental requirements are very high, MCM costs (above the price of the IC costs used in the MCMs) of several thousand dollars are not uncommon.

• Risk

  Developing a custom MCM is a fairly complex effort. There is a definite risk that the first pass of the design may not be successful, requiring a second pass to correct the deficiencies. Further, the custom interconnect and package, which are often sub-contracted to external vendors represent similar risks - as do delivery of the die and passive components which make up the design of the MCM. All of these risks can be managed by the application of a mature development process, but collectively represent a level of risk that is not present when a design consists only of individually packaged integrated circuits.

  Of course, the design of a schematic using individually packaged integrated circuits also has risk, but errors in this case typically cost less to correct.

  In many cases, MCM designers elect to use individually packaged integrated circuits to build an functional equivalent to the MCM - specifically because corrections to the schematic design are easier to make prior to spending the labor needed to design the MCM packaged version of the schematic. The MCM equivalent assembly is called a breadboard, and is much easier to assemble, test, or otherwise handle than is an MCM. Once the breadboard confirms that the schematic provides the intended functionality, the schematic is then used to commit to a final MCM design.

Development Cycle

Requirements Concept PDR Detailed Design CDR Prototypes Documentation

It cannot be emphasized enough how important it is to bring in a knowledgable MCM engineer at the first sign of need. There are compatibility issues between the MCM and the PWB on which it is to be assembled. If not handled correctly, these issues can result in major cost or schedule overruns. All to often, program engineers complete their design based on assumptions about the MCM only to find out that a re-design of the PWB or MCM is required. Here are some examples of what can happen:

• Thermal dissipation within the module requires additional heatsinking that is not in the PCB design
• Complexity of the MCM schematic prevents routing of the signal I/O to the pre-assigned locations, requiring a re-design of the MCM schematic
• Height of the MCM exceeds that of the other components on the PWB, requiring a non-standard PCB spacing or a re-design of the PWB to accomodate the MCM height
• Additional circuitry is required in the MCM to allow for adequate testing and troubleshooting of the the MCM - resulting in a re-design of the MCM footprint and of the PWB
• The selected component vendors do not offer the ICs in bare die format, so the MCM schematic must change, which in turns changes the design of the PWB
• The bare die or passive components are not compatible with the MCM factory process flow - such as process temperatures or acceptable feature sizes. New components must be selected, changing the MCM and PWB schematics

Requirements

Regardless of how much sense that makes, it is not unusual for requirements to be defined late in the development cycle, or simply to change in mid-stream. Most often, schedule or cost pressure is at fault - not enough time, or not enough money to include a rigid requirements definition phase, or not definition. Personnel often think they "know" what is needed, but when the details are worked it is not unusual to find that something was overlooked.

Concept

Regardless of the degree of preparation which was achieved, the Concept phase of the MCM development cycle is to come up with the specifics of the proposed MCM design:

• Dimensions
• Technology selection (package, interconnect, manufacturing & test processes)
• Component list
• Cost/schedule/resource allocation
• Performance analyses (electrical as well a mechanical)

PDR

Detailed Design

Generally, since the package must be sized to meet the interconnect design, the package design cannot be finalized until the interconnect design is complete. In practice, however, the customer requirements often define the package dimensions and the interconnect design must accomodate those requirements. It is not unusual for the MCM design team and the customer to negotiate changes to the requirements during this phase.

CDR

At this point in the development cycle, the MCM engineer is expected to have identified and resolved all issues with the design. It is expected that no issues will arise during the CDR, allowing the MCM engineer to proceed with release of the package/interconect for fabrication.

The team review of the complete design package serves to confirm such status, and if needed to identify issues/design flaws which have been missed that would require a return to the detailed design phase of the program.

Introducing design changes at this stage of the MCM development cycle would have significant impacts on schedule, but it would be worse yet to discover those needed changs after the incurring the expense of manufacturing flawed modules.

Prototypes

Depending on the company, the small lot goes by a variety of names - proof of design (POD), proof of manufacturing (POM), pipe cleanears, prototypes, scout lot, engineering evaluation units, to mention a few. Regardless of the name, the goal is to confirm that the MCM performs as expected, validating the design, documentation, test, and manufacturing flow.

Documentation