Everyone knows that flexible product designs make the best sense in today's rapidly changing markets, where requirements can change from day to day. So with all this focus on flexibility and adaptation, why is everyone not developing upgradeable systems as a matter of course?
The answer is that old habits and an out-of-date understanding of the costs and benefits of incorporating flexibility stand in the way. So, in this article, let us see what we can do to explode the 'Six myths of microcontroller choice'.
Myth # 1: Embedded ROM costs less than embedded Flash memory
This used to be true, because a true ROM manufacturing process is simpler than a Flash process. But several factors now work against ROM and in favour of Flash memory.
Dense processes and large wafers mean there can be more than 10 000 devices on a wafer, and 250 000 units in a 25-wafer manufacturing lot. Minimum order quantities are 5000 to 10 000 units, resulting in poor manufacturing utilisation due to equipment loading and high scrap rates when partial wafers are discarded.
Flash memory manufacturing processes also make it easy to add complex analog functions to microcontrollers, such as power-on reset, brown-out/low voltage-detect and oscillators with more clock source types. These analog peripherals help the microcontroller adapt to varying power supply conditions and sensor inputs as the environment changes.
Myth # 2: ASIC designs are the lowest-cost solution
Application specific integrated circuit (ASIC) technology has been used to increase integration and potentially cut costs for digital logic designs. But it is unclear if the strategy always works, since some designs can become over-integrated to the point of requiring highly sophisticated packages and cooling technology for heat dissipation.
The design, validation and manufacturing costs of advanced ASICs are now putting them out of the reach of all but the highest volume applications, a trend that is at odds with today's shortening product life cycles. Add some precision analog peripherals to the ASIC and the complexity of development and testing increases dramatically.
In short, ASICs take too long to design, cost too much and are too inflexible to meet the majority of today's product development needs.
Myth # 3: Smaller lithography lowers cost
Moore's Law teaches us that devices get faster and more cost-effective as manufacturing processes use denser lithography. But as we enter the world of deep sub-micron processes, the number of masks used is rising, increasing costs and lowering the throughput of process equipment.
Since the smaller lithography usually increases the overall wafer cost, the die must become much smaller to offset the higher wafer cost. But certain functions, such as bonding pads, may be constrained by the mechanical limitations of packaging technology and so are unable to scale down in line with process improvements. Analog circuits may also not scale well, so designs with a high proportion of functions that do not scale may actually cost more on denser processes.
Myth # 4: 16- and 32-bit MCUs replace 8-bit MCUs
Many 4-bit MCUs have been replaced by 8-bit versions, so many people expect 16-bit MCUs will inevitably replace 8-bit parts. But this is not so. Some applications just do not need to use longer datatypes.
For example, the ASCII code used to handle text characters fits into eight bits and is best handled by an 8-bit processor. 16- and 32-bit processors have to do special operations to handle 8-bit data in their 16- or 32-bit natural data types. Such processors also need larger program memories and more complex manufacturing processes to handle their wide memory buses.
16- and 32-bit processors also often use sophisticated architectures to increase their speed, at the expense of determinism. Many times an 8-bit MCU can be added alongside a larger processor to handle high-frequency interrupts from realtime sub-systems. This approach can offload the larger processor, increasing its performance, simplifying its software and reducing its power consumption.
Myth # 5: All MCUs have the same development environment
Many designers believe their most difficult task is to choose the lowest-cost MCU and that their system development time will be the same, regardless of the chosen vendor.
As system complexity has increased, the demands on the development tool chain have exploded. Many MCU vendors have not made the investment in development tools necessary to enable system designers to diagnose tough problems like realtime interdependencies. So designers should evaluate the entire development tool chain before selecting an MCU.
Development tools must provide as much visibility into the system as possible without being intrusive. Basic tools must address hardware/software development, integration and final testing. In-circuit emulators should allow full visibility into the processor and peripherals so all interactions can be observed. An integrated development environment binding all phases of the software development helps minimise errors and implement source-code control for product quality.
While most high-end processors offer extremely capable development tool chains, most 8- and 16-bit MCU environments have not been updated in several years. As system complexity increases the emphasis on development tool capability becomes even greater than the MCU itself.
Designers should also be wary of MCU vendors who give away tools to attract customers to their MCU product family. With no revenue coming in from development tools, can the vendor afford to keep the tool chain up-to-date?
Myth # 6: It is best to have a single processor supplier
In the 1990s the idea of vendor reduction programs became popular. The idea was that buying all your processors from one supplier should make things easier on the development and logistics organisations. Porting software between 8-, 16- and 32-versions of one architecture would allow easy scalability with minimal development.
It is an attractive theory but less good in practice. Most vendors use different development groups to design each processor family, and so little time is spent on making it easy to migrate between them. To check this for yourself, examine the development tool chains for each processor family and see how much commonality there is. The various families may also compete for manufacturing resources at the potential vendor, putting your supplies at risk if another of the vendor's processor families suddenly experiences massive demand.
Conclusions
There is a host of myths that surround the choice of an MCU. Choosing a vendor for 8- and 16-bit MCUs should be based on more than just the product attributes. Vendors dedicated to the market, actively developing new products, updating technology, adding new capacity and increasing tool-chain capability should be considered leading candidates over vendors simply treating the product as a cash source with little or no new development.
For more information contact Avnet Kopp, 011 809 6100, Memec SA, 011 897 8600, or Tempe Technologies, 011 452 0530.
| Email: | willem.hijbeek@tempetech.co.za |
| www: | |
| Articles: | More information and articles about Tempe Technologies |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version