Skip to content

Spot the mistake in this design: built exercise #5 (draft)

Drafted 8 Oct 2026 for issue #227. Nothing has been posted. A BUILT exercise under the "Built exercises" rule in marketing-block-1.md: made up on purpose, ONE planted mistake, never presented as an AI's answer. Same style as #3: a system-level claim that a MagicON T/R budget disproves. (It replaces an earlier #5 on shared switch loss, which the owner judged too simple; that one was never merged.)

Post 1 (midweek): the puzzle

Spot the mistake in this design. 🔎

A made-up teaching example. A 9.4 GHz marine radar shares ONE antenna through a T/R switch, with a limiter in front of the receiver. It must see a buoy as close as 155 m. Take every part figure as given. What did the student forget?

Answer on Friday.

RFEngineering #ElectronicsEngineering #EngineeringStudents #Radar

Image (figures and the check):

Item Figure
Transmit pulse 1 µs, 1000 pulses per second
T/R switch Finwave FW2001 settling time 200 ns
Limiter Qorvo TGL2209-SM recovery time 30 ns
Requirement see targets from 155 m

Student's check: "The receiver is blind only while the 1 µs pulse is on, so a target at 155 m is seen." (planted mistake: the switch's turn-around and the limiter's recovery are left out)

Post 2 (Friday): the answer

Spot the mistake in this design: the answer 🔎

The receiver isn't listening the moment the pulse ends. First the T/R switch has to flip back to receive and settle (200 ns), then the limiter has to recover (30 ns). Only then can an echo get through. MagicON's timing budget puts the closest visible target at 184.37 m, so the 155 m buoy is missed by 29.37 m.

Any echo from closer than that arrives while the receiver is still switching back or recovering.

Lesson: the blind range is pulse + switch-back + recovery, not the pulse alone.

RFEngineering #ElectronicsEngineering #EngineeringStudents #Radar

Proof (MagicON result, for the answer image)

MagicON design_tr_module, re-run 2026-10-12 through Claude Code's magicon MCP tool. Inputs: frequencyGhz 9.4, targetOutputPowerDbm 40, application marine_radar, topology tr_switch, prfHz 1000, pulseWidthS 1e-6, minDetectionRangeM 155; grounding caller_asserted (a made-up brief).

  • sharedParts.switch: FW2001 (Finwave Semiconductor).
  • budgets.timing.blindRange: blindRangeM 184.37, delaysS: pulseWidthS 1e-6, switchBackS 2.0e-7, limiterRecoveryS 3.0e-8; minDetectionRangeM 155, marginM −29.37, verdict fails_requirement.
  • budgets.timing: priS 0.001, dutyPct 0.1, txRxSwitchingBasis settling_0_1db_ns.
  • Datasheet cross-check (our copies): FW2001 DS-00001 Rev B p.1 "Settling Time 50 % of VCTRL to within 0.1 dB final power ns 200" (read 2026-10-12; the same row lists "Switching Time … 85" ns, which the engine does not use because a receive window needs the settled signal); TGL2209-SM p.2 "Pulse Recovery Time <30 ns" (read 2026-10-08; the engine uses the 30 ns bound).

Why the figures changed (2026-10-12). The first run, on 2026-10-08 (MagicON AI connector, live public server), picked Qorvo QPC2040 (35 ns switching) and gave 159.64 m, missed by 4.64 m. On 2026-10-09 commit ed03ff820 corrected QPC2040's arm-to-arm isolation from its datasheet, which moved the 9.4 GHz pick to FW2001. The same inputs no longer reproduce 159.64 m, so the exercise and both images were rebuilt on the 2026-10-12 result.

The planted change: none of the figures; the student's check leaves out two delays.

Before posting

  • The 1 µs pulse at 1000 per second is a made-up brief (duty 0.1%), separate from #3 and #4's 10% duty design. Not checked against datasheets: the other part figures (the engine's).
  • Images: spot-the-mistake-built-05-puzzle.png, spot-the-mistake-built-05-answer.png, built by docs/scripts/make-spot-the-mistake-built-05.py. Their titles carry no number, because draft numbers do not match the posting order (built #2 went out first, 2026-10-12).
  • Ready-to-post folder (outside the repo): Documents/MagicON-posts/puzzle-built-05-blind-range/.
  • Answer text checked 2026-10-12 with MagicON check_grounding against the re-run: grounded, 0 findings.
  • Check posted-log.md first; log both posts after posting.