1) Design according to the needs of the project. 2) You design based on assumptions 3) Mitigate risk with planning Does your design have a long lifetime in the field? Are you going to be producing your design for a long time? Will you have to support or maintain your design for more than a year? Will you be refactoring the design in a year? Does your design have extreme cost sensitivity? Does it have size, power consumption, or other needs that make a particular part more attractive? Does it require the use of the latest cutting-edge technology? Once you've determined what the requirements are, then you write down all your assumptions, and make simple plans if that assumption fails. If the assumption is very unlikely to fail, spend little time working on the mitigation plan. The following assumptions are a good set to start with, as long as you've got a good parts qualification, design, and testing cycle: I assume any bugs that would require a hardware design change will be discovered during part qualification and prototyping with the dev kits. I assume the parts I use in the hardware design after qualification will not have bugs that require me to redesign the hardware. I assume that if there's a software fix required before release, the manufacturer will supply the fix and it'll take less than a week to integrate and fix. I assume that we will perform thorough enough testing that any existing bugs in the silicon will be found before release, or do not affect us due to the design and use of our product. When asking the question, "How do I best mitigate an assumption failure?" the best answers are (in order): Modify the design process such that the failure is less likely to occur. Modify the process so that it is detected as early as possible. Modify the design and the design process so the impact is very little. Determine alternatives as the design progresses so changes can be made quickly. There are other options as well. Some companies in cost-is-no-object time-critical projects actually have two seperate redundant teams working on the same project. Often I'll design the PCB to accept several different designs, primarily for second source issues with parts that aren't generic. PIC is nice in that (so far) they've been plentiful, rarely EOL a good part, and many different parts are pin compatible, so if X isn't available I can use Y, though it may cost more. In the case of this ethernet chip, there is no second source, and given that it's a rather big departure from the typical parallel bus ethernet, and Microchip has no proven experience with ethernet, I would wait at least a year or two after this one is officially in production, or (better) wait for them to release the next version of the chip (or integrate it into a PIC). This would give enough time to find all bugs that affect most designs, and most bugs that affect many designs. But that would be for an 'important' project. For a one off I'd not hesitate to use it primarily for simplicity and ease of use reasons - they are more important than any minor bugs that have yet to be discovered. -Adam On 10/4/06, stef mientki wrote: > I wonder how you guys DESIGN circuits ? > > Having a lot of trouble making a design, using the ENC28J60, > working around all the bugs in this chip, > I thought a few weeks ago I had a very good design. > It ran for weeks, without any noticeable trouble. > > Until to day, a new errata appeared in my mailbox, > (about a bug in the checksum calculations) > and now I can redesign my software again :-( > > I think this chip is almost a year old, and still new bugs arrive, > when will it stop? > Or should we only use components older than say 2 years ? > Is this common for uChip products ? > What other experiences are there with new uChip products. > > Ok, I can appreciate that uChip gives the possibility to subscribe to > errata, > but I always hope I never get mail from them ;-) > > cheers, > Stef > > > > > -- > http://www.piclist.com PIC/SX FAQ & list archive > View/change your membership options at > http://mailman.mit.edu/mailman/listinfo/piclist > -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist