EC-19.5 The Design Cycle, and a First Build That Works

What this is and why it exists

The value of this topic is finishing something.

A reader who has completed one small build has a reference point for every abstraction that follows. They also know from experience that the first version does not work, and that this is normal rather than a verdict on them. That second lesson is worth more than the first and cannot be delivered by reading.

The engineering design course later formalises the same loop, with reviews, verification and integration. The board course later assumes you have already wired something up. This topic is what makes both of those land.

The vocabulary

• Design cycle — the repeating sequence of understanding, proposing, building, testing and revising.
• Iteration — one pass around that cycle.
• Prototype — a build made to learn from rather than to keep.
• Bring-up — powering a new build for the first time and checking it stage by stage.
• Bisection, in fault finding — measuring at the middle of a path to halve the region a fault can be in.
• Signal path — the route information takes from input to output.
• Scope creep — the growth of a project's aims while it is being built.
• Bill of materials — the list of every part a build needs, with quantities.
• Post-mortem — the short honest account written once a build is finished.

The mental model

Design is a loop rather than a line. Understand the problem, propose something, build it, test it, revise. Then go round again.

Going round the loop twice quickly beats going round once carefully. The reason is that the first pass tells you things no amount of planning would have. A component that is out of stock, a connector that does not fit, a reading that is ten times what you expected. Those discoveries are the point of the first pass rather than evidence that it failed.

The step people skip is the first one. Skipping it produces a device that works perfectly and solves the wrong problem, which is the most expensive kind of success. This is why the previous topic came before this one.

Choosing the project is the next decision and it is where most first builds are lost. A good first project has three or four components, one clear behaviour, and no part that has to be ordered from far away. A light that comes on when a room gets dark. A buzzer that sounds when a door opens. A display showing a temperature.

Ambition at this stage produces an unfinished box rather than a lesson. The aim is a completed loop, not an impressive object, and the loop can be run again on something larger once it is familiar.

Turning the block diagram into a circuit is the step that connects the abstract to the physical. Each block becomes a small group of components. Each labelled arrow becomes a wire carrying a voltage between them. Doing this once by hand, badly, teaches more than reading three finished designs.

Then build it, and build it in stages. Power one block, check it behaves, and only then add the next. This is the single habit that most separates a build that comes together from one that does not. Powering everything at once turns one findable fault into a puzzle with several possible causes at the same time.

When it does not work, and it will not, the method is bisection. Measure in the middle of the signal path. If the signal is correct there, the fault is downstream; if not, it is upstream. Repeat on the half that contains it. A path with eight stages is narrowed to one in three measurements.

The alternative, guessing at a likely culprit and replacing parts, is slower and teaches nothing. It also introduces new faults, because every removal is another chance to damage the board.

Finishing means writing it down. A short record of what the thing does, what it cost, what parts it used and what you would change is what closes the cycle. Check it against the one-page brief from the earlier topic, and be willing to write that a requirement was not met.

Naming the one thing you would do differently is the part that carries forward. Everything else about a first build is replaced by better work within a year; that one sentence is not.

What you should now be able to explain or do

• State the five steps of the design cycle and say why two fast passes beat one slow one.
• Choose a first project small enough to finish, and say what makes it small enough.
• Turn a block diagram into a circuit you can put on a breadboard.
• Bring a build up one block at a time rather than powering all of it at once.
• Find a fault by bisecting the signal path, and say why replacing parts is worse.
• Write a short honest record and check it against the brief you wrote first.

Check yourself