Spot the mistake in this design: built exercise #6 (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 tests a heat rule: a pulsed radar only runs cool if its power amplifier's bias is pulsed too.
Post 1 (midweek): the puzzle¶
Spot the mistake in this design. 🔎
A made-up teaching example. A 9.4 GHz marine radar transmits 10% of the time. Its power amplifier (PA) runs from 24 V, and its gate and drain supplies stay ON between pulses. Take every part figure as given. What did the student miss?
Answer on Friday.
Image (figures and the check):
| Item | Figure |
|---|---|
| Transmit duty cycle | 10% (on one tenth of the time) |
| PA Qorvo QPA2811 | drain 24 V, idle drain current 1.24 A |
| PA bias | gate and drain supplies stay ON between pulses |
Student's check: "The PA only transmits 10% of the time, so it heats only 10% of the time: 2.92 W on average." (planted mistake: the idle current between pulses is left out)
Post 2 (Friday): the answer¶
The student forgot the idle current. With its supplies left on, the PA draws 1.24 A at 24 V between pulses too, while sending out no RF at all, so all of that power turns into heat. MagicON puts the PA at 29.71 W, not 2.92 W.
2.92 W is right only if the bias is switched off between pulses ("pulsed bias"). Re-run with the bias pulsed and MagicON gives exactly that.
Lesson: pulsing the RF does not pulse the bias. Switch the PA's drain or gate off between pulses, or count its idle current as heat.
Proof (MagicON result, for the answer image)¶
MagicON design_tr_module, two runs on 2026-10-08 through the MagicON AI connector (live public
server). Both: frequencyGhz 9.4, targetOutputPowerDbm 40, transmitDutyCyclePercent 10,
application marine_radar, topology circulator_limiter; grounding caller_asserted (a made-up
brief). Only transmitBiasPulsed differs.
Bias stays on (transmitBiasPulsed: false) |
Bias pulsed (transmitBiasPulsed: true) |
|
|---|---|---|
analysis.thermalSources tx:pa (QPA2811) |
29.71 W | 2.92 W |
analysis.totalDissipationW |
33.8 W | 7.02 W |
| Disclosure | "Amplifier bias stays on between pulses (as stated), so each amplifier's idle draw is counted as heat for the time it is not transmitting, on top of its 10%-duty pulse dissipation." | "Amplifier bias is pulsed with the transmit pulse (as stated), so the amplifiers draw no idle current between pulses and their dissipation is derated to 10% duty." |
- Both runs also disclose: "The receive branch does not derate: its bias is continuous." The other parts' heat is the same in both runs; only the PA changes.
- The 24 V and 1.24 A in the image are the catalog record's
(
backend/categorized_databases/pa/6_18ghz/QPA2811.json:drain_voltage_v24,quiescent_drain_current_a1.24), the record the engine read.
The planted change: none of the figures; the student's check treats the idle time as cold. The brief says the supplies stay on, so a reader can spot it from the image alone.
Other things the result says (not used in the post)¶
- With the bias on, the driver QPL3050's idle heat is missing (no idle current on record), so the 33.8 W board total is a lower bound.
- The mixer's heat is not counted either: the local oscillator's delivered power is not in the design.
- The 12 V board cannot make the PA's 24 V rail ("UNMET RAIL" in the BOM), as in #3.
Before posting¶
-
The two runs share the circulator JCSM0800T1200S20, whose band is in doubt (issue 361). The board totals (33.8 W / 7.02 W) include its 0.1417 W, so since 2026-10-09 the post and the answer image no longer quote them; the PA figures (29.71 W / 2.92 W) that carry the lesson do not depend on it. The totals stay in the proof table above as the engine's record.
-
Checked against the datasheet: Qorvo QPA2811 Data Sheet Rev. D, February 2026, p.1: "Bias: VD = 24 V, IDQ = 1.24 A" (our copy:
data/datasheets/pa/6_18ghz/QPA2811.pdfin the main folder; that folder is not in git). The post still says "take every part figure as given". - MagicON
check_groundingon the post texts (2026-10-08): every figure traces to a result except 1.24 A, flagged because it is a datasheet value, not an engine output (see above). - Images:
spot-the-mistake-built-06-puzzle.png,spot-the-mistake-built-06-answer.png, built bydocs/scripts/make-spot-the-mistake-built-06.py. - Check posted-log.md first; log both posts after posting.