Requirements, Architecture Options & Concept Review

Why gate position matters

The first requirement comes from physics, so start there. Bill Pentz’s measurements of small-shop systems give the numbers ¹:

• Stationary tools need about 350 CFM to collect the visible chips and sawdust a broom would get, and about 1000 CFM for good fine-dust collection.
• To keep chips moving, air in vertical duct should run about 4000 ft/min (3800 at least) and about 3000 ft/min in horizontal runs (2800 at least).
• A typical 1.5 hp collector rated at 1100 CFM really moves about 785 CFM through 6-inch duct, 550 through 5-inch, and 350 through 4-inch.

Check the speeds with Q = V × A. A 4-inch duct’s area is π(2/12)² ≈ 0.087 ft², so 350 CFM in a 4-inch drop is 350 ÷ 0.087 ≈ 4000 ft/min: just enough. The 6-inch main at 785 CFM is 785 ÷ 0.196 ≈ 4000 ft/min too. Now leave all five gates open. The main still can’t carry much more than about 800 CFM, which splits five ways to about 160 CFM per drop, or 160 ÷ 0.087 ≈ 1800 ft/min. That’s far below the speed that keeps chips airborne in a vertical drop, so they settle and the planer’s hose clogs. The baseline log’s “gates left open” count and its clogs are the same problem.

So: one gate open per running machine, the rest closed. And one more constraint from the same source: thin duct can collapse if the collector starts with no gate open ¹. The system must never do that.

Write requirements that can fail a test

A useful requirement names one thing to measure, the conditions, a limit, and how it’ll be checked. “The system shall be fast and reliable” can’t fail anything. Each requirement gets an ID so tests can point back to it:

Keep four kinds of statement apart: a need (fewer clogs and less fine dust), a constraint (listed devices for line voltage), an assumption (a gate slide takes about 12 N to move; test it), and a design decision (servos or linear actuators). Mixing them lets a favorite solution pose as a requirement. NIST’s work on robotic manufacturing puts the same weight on stating performance so that someone else can measure it ².

Generate three concepts

Always generate more than one, including a simple one:

• A. Manual gates with a remote switch: a key-fob remote for the collector. Cheapest; still depends on remembering the gate.
• B. A listed current-sensing switch: a bought switch that starts the collector whenever one tool’s outlet draws current, with manual gates. Fixes forgotten starts for that tool; gates still manual.
• C. Full automation: a current sensor on each tool circuit, an actuator and limit switch on each gate, a low-voltage controller, and a listed motor-rated relay or contactor for the collector. Fixes forgotten starts and gate errors; the most to build and maintain.

Do rough numbers only as precise as the decision needs: can a hobby actuator move the gate (next lesson), can a current sensor see an idling tool, does the collector have enough air for one drop at a time (yes, above).

Score the concepts openly

Weight the criteria, score each concept 1 to 5, and show the arithmetic:

B wins narrowly. Now the sensitivity check: move 10 points of weight from cost to learning value, and A drops to 3.0, B to 3.5, and C rises to 3.9 and wins. A 10-point shift flips the decision, so the weights are the decision. For a capstone whose purpose is to practice integrating mechanics, sensing, and control, learning value deserves the higher weight, and C is chosen, with one change that raises its safety score: buy the line-voltage half (a listed relay or contactor with a low-voltage coil) and build only the low-voltage half. Say that openly in the review rather than tuning the numbers until C wins.

Review and record the decision

Write a short decision record: concept C chosen; A and B rejected and why; the decisive evidence (the airflow check and the log); open assumptions (gate force, current-sensor signal at idle); and what would reopen the decision (for example, if the gate actuators prove unreliable in a month of use, fall back to B plus labeled gates). MIT’s project courses close each design stage the same way, with evidence and a decision instead of a rendering ³.

Gate 2

Pass when every requirement has a verification method, the chosen concept covers all of them, the line-voltage boundary is settled, and each risky unknown (gate force, sensor signal) has a small experiment planned before detailed design.