The first prototype should not be striving to be final, it should just diminish design uncertainties as much as it can. Design it as a prototype.
The first board of Bart was overall in good shape, but it lacked a few prototype qualities. The power lines were not separable (worrying over power line impedance), and the Tag connect footprint was lacking the mounting holes. (worrying over board space)
The Tag connect has mounting holes that enable the connector to clip in. This footprint is actually for once in a while programming.
Well, a designer has a way of doing their things. Although it is a good trait overall, the first board is more of an experiment than a product.
In the set of uncertainties that a design has in its inception, there are ones that we are aware of their presence, and there are ones that only become evident when the design is physically realised.
When a design is meant to be a prototype, we capture all of the first set, and if we are lucky, a portion of the second set.
This, in my opinion, better than the other way around. Although it is nice to have a mean revision count around 2, in most worthwhile projects that number will exceed 2.
MA
A few days lost. Had to learn Fusion to design a holder. Still trying to find the where the current sinks.
- Name prototypes as prototypes. Make that explicit and people will look at it as one.
- Write down all the 'experiments' in the schematic and in the issue management system.
- Order prototype boards with a distinguishing color.
- A lot of probes, jumpers.