A number of us are frantically sorting through our old work files for something that will help fill the gap of the lack of instruction and training being provided to today's engineering/Comp Sci students.

      We simply do not teach methodology. We teach pieces of it. The engineers showing up on the job are only capable of some of the tasks they should be able to perform 5 minutes out of school. We won't discuss their lack of computer skills. So what good are documents that provide application support for parts no longer produced? And does this violate an antiquated NDR?

• The parts in questions are no longer produced.(Bears repeating.)
• The copyright has long ago expired.
• The original company may have dropped the product line, the business unit, or be itself out of business.
• The parts so discussed can be freely considered to be homeless.

Note: and this is important, no one releases patent information or process information. We are speaking of the user-manual design-guide, datasheet, application-note level of information.

      First of all, the parts may not be in production, but they are still running around, creating very interesting support issues. If they don't make it anymore, how do we replace it when it dies of old age? Reverse engineering rears its little head. Trust me. They are (and always have been) very good at taking apart an existing design. You would think that little piece of Silicon is sacred ground – no it is not!

      Second, while the TECHNOLOGY itself may be obsolete, the design methodology is not. Make no mistake; we follow the same rules for today's designs. We think. What has happened is that, while trying to follow the old methodology, the experts have died off or been put to pasture (possibly kicking and screaming as they went). Without the mentors and the experienced designers running around, we forgot the recipe.

      By looking back at the history of the evolution of the high-tech industry, perhaps we can help today's student examine the processes that worked. How America got to be a leader and not a follower, as it pretty much is becoming today.

      Of course, we need a national mind-set, a goal that we need to reach besides re-learning the basics. The older folk had the challenge of putting a man on the moon. Well, now we want to put one on Mars!

Classic Case in Point

      The AMD 2900 family was so popular, it has become soft IP (RTL – Verilog) and you can buy it that way and drop it into your ASIC. The parts never "died". Mention the 2910 controller and most engineers (over say 35) know what it is. Depending on where they went to school, the younger ones may know it too. This is a part from the 1970s!

      The old handouts, seminar, datasheets and Application notes refer to a technology we are decades beyond. The industry hadn't leapt into ASICs at the time; we had no customed or even semi-customed ASICs. They came charging in behind the 2900 family.

      The family is clever, effective, and efficient. Designs are still being done using the basic part design, possibly with a few changes (deeper stack on the Am2910 for one; an altered instruction for another).

      By looking at the design methodology (as in designing with as well as the design of), we can pass on the knowledge of how things should be done. What questions do you ask? What trade-offs should be considered? What issues can arise?

      And, at the core, how did these clever little parts really work?  

AMC_Arrays ASIC design methodology - pre-design compiler (And the rules remain the same - we just have tools now)

AMD_2900 The Am2900 Family

See also Wikipedia

Bit-Saver's Archive hosts the full AMD Databook from the 1990s