Working notes

Engineering log

A running record of what I'm working on, testing, changing, and figuring out.

Notes: 14
Projects: 08

Retrospective notes14

Written from project notes after the fact, about decisions whose date the documentation does not record.

Retrospective noteArraySignal

One low reading is not a fault

The first version reported every interval that fell below expectation, and almost none of them meant anything. Fixing it meant reasoning over time rather than over a number.

Time-SeriesAnomaly Detection

Retrospective noteIonospheric Plasma Irregularity Payload — Rev A

The probe is part of the instrument, not just a connector

I drew the electronics for a plasma instrument and left the part that touches the plasma for later, as though it were packaging.

InstrumentationDesign Review

Retrospective noteDwellMetry

More fitting cannot recover information the sensors never measured

The thermal model reproduced the data well and had not learned what I thought it had. The limit is structural, not numerical.

PhysicsIdentifiability

Retrospective noteCubeSat Solar X-Ray Flare Payload — Rev A

The flare detector was missing from the flare-detector schematic

I drew a schematic for a solar-flare payload, reviewed it against the mission functions, and found that the actual X-ray measurement path was not there.

Design ReviewInstrumentation

Retrospective noteCubeSat VLF Lightning Wave Receiver — Rev A

The spacecraft could become part of the measurement

The VLF receiver had to coexist with switching and digital electronics that could radiate or couple energy into the same band I wanted to measure. That made spacecraft-generated interference part of the instrument design rather than an integration detail.

EMIInstrumentation

Retrospective noteCubeSat Gamma-Ray Burst Timing Payload — Rev A

One pulse is two measurements

A single detector pulse has to answer two different questions, and I had been designing as though it answered one.

MeasurementDesign Review

Retrospective noteCubeSat Exoplanet Transit Photometer — Rev A

A transit-like dip can come from the instrument

I had been designing a detector sensitive enough to see a one per cent change. The review pointed out how many things other than a planet produce one.

InstrumentationMeasurement

Retrospective noteCubeSat Storm-Time Radiation Belt Mapper — Rev A

Four detectors are not four science channels

I drew four detector devices and assumed I had four measurements. Counting symbols is not the same as tracing paths.

InstrumentationDesign Review

Retrospective noteArraySignal

Teaching the system to say it does not know

The most useful thing the classifier does is decline to answer. Getting there meant separating how bad a condition is from how sure the evidence makes me.

Fault ClassificationUncertainty

Retrospective noteIonospheric Plasma Irregularity Payload — Rev A

Protection diodes can be noisier than the signal they protect

The clamps guarding the front end sit on the one node in the design where leakage is indistinguishable from science.

Analog DesignInstrumentation

Retrospective noteDwellMetry

An anomaly is not a diagnosis

Detecting that behaviour changed and explaining why are different problems, and merging them produces confident causes the evidence does not support.

Fault ReasoningUncertainty

Retrospective noteCubeSat Solar X-Ray Flare Payload — Rev A

Visible light is a pointing check, not flare evidence

The VEML7700 is the visible-light sensor in the Rev A schematic, which made it tempting to treat its response as flare evidence while the actual X-ray detector path was still missing.

SensorsRequirements

Retrospective noteCubeSat VLF Lightning Wave Receiver — Rev A

Anti-alias filtering is part of the measurement

I had the filter down as cleanup between the amplifier and the converter. It is closer to being the thing that defines what the instrument measures.

Signal ProcessingAnalog Design

Retrospective noteCubeSat Gamma-Ray Burst Timing Payload — Rev A

A GNSS timestamp is not the event time

I had been treating GNSS synchronisation as the whole timing solution. It fixes the origin of the time axis and says nothing about the delay between a photon and the record of it.

TimingGNSS