ECSS Test Centre — Infrastructure and Work Environment (space-systems/ecss/q2007-environment)
Use when the task is the facility step of ECSS-Q-ST-20-07C clauses 5.5.1 and
5.5.2 -- showing that the buildings, services and media a test centre offers
are suitable for the tests booked into them, and that the environment was
actually held inside its band while the test ran.
Domain quick reference
- Suitability and control are two questions, not one. Suitability asks
whether the facility can hold the environment at all; control asks whether
it did. A centre can be perfectly suitable and still have lost the band
for six minutes at 02:00, and a single pass/fail hides exactly that.
- Suitability is band containment. The facility declares a capability band
per parameter, the test declares a required band, and the capability has
to cover the requirement at both ends. The amount by which it does is the
containment margin, and the margin is the operator's room before the
facility is running against its own limit.
- A facility whose capability exactly equals the requirement is suitable and
has zero margin. Those are different statements. Report both, because the
second is what predicts the excursion.
- The margin is reported at each end separately. A capability that is
generous at the top and short at the bottom is a real and common case --
a chamber that can go hot more easily than cold -- and one aggregate
number hides which end to fix.
- Power quality arrives as a nominal value plus a tolerance in percent, not
as a band, so the band is derived: a 230 V supply held to 10 percent is
compared as 207 V to 253 V. Derive it once and compare against the derived
limits; comparing a reading against "230 V plus or minus 10 percent" in
prose is how two engineers reach two answers.
- Media -- a purge gas dew point, a chilled-water temperature, a cleanroom
overpressure -- are ordinary parameters with their own bands. The failure
mode is identical: drifting out of band with nobody watching.
- Monitoring cadence is graded on its own. A parameter sampled every fifteen
minutes against a ten-minute requirement can hold a clean record through
an excursion nobody saw, so a slow interval is a finding even when every
recorded reading is inside the band.
- Excursions carry a side and a magnitude, and the longest unbroken run
carries a duration. A brief touch of the limit and a half-hour outside it
are not the same event and should not aggregate into one count.
- Every band comparison carries a small relative slack, so a reading landing
exactly on a derived limit reads as inside on any platform. The slack
absorbs representation error in a multiplied limit; it does not loosen
the requirement.
Workflow
- List the parameters in scope. For each, state the required band the test
asks for and the capability band the facility declares. Derive a band
from nominal plus tolerance for anything specified that way.
- Refuse the specification when a parameter name is unrecognised, a band is
not a two-limit pair, a limit is not numeric, a lower limit sits above
its upper limit, or a parameter appears twice.
- Test containment per parameter and record the margin at each end. A
negative margin at either end is the suitability finding, and the end is
named.
- Walk the monitored series against the required band. Record each
excursion with its index, side and magnitude; take the worst band
utilisation across the series as the single control figure.
- Multiply the longest unbroken run of out-of-band readings by the sample
interval to get the excursion duration.
- Compare the sample interval against the required interval and flag a slow
cadence separately from any excursion it might have hidden.
- Roll the facility up into two independent verdicts: suitable, and
controlled. Report the parameters failing each.
Pitfalls
- Collapsing suitability and control into one verdict. The remedies are
different -- one is a facility change, the other is an operations change.
- Reading a zero containment margin as a comfortable pass. It is the
narrowest possible pass and it predicts the first excursion.
- Comparing a reading against a percentage tolerance in prose instead of
against the derived absolute limits.
- Counting excursions and never timing them, so a single touch of the limit
and a sustained half-hour outside it look alike in the report.
- Accepting a clean record from a parameter sampled too slowly to see an
excursion. The clean record is evidence about the sampling, not the
environment.
- Using a strict comparison against a limit that was produced by a
multiplication. The limit is not exactly representable and the same series
then passes on one machine and fails on another.
Behavior contract (gate 3)
The band validation, tolerance-derived bands, containment and margin
arithmetic, excursion detection with side and magnitude, longest-excursion
timing, monitoring-cadence check and the two-verdict facility roll-up are
exercised by the gate 3 contract test: scripts/test_q2007_environment.py
against scripts/q2007_environment_logic.py (stdlib unittest, offline).
Run: python3 scripts/test_q2007_environment.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and
paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
1---2name: q2007-environment3description: "Evaluate test-centre infrastructure suitability and work-environment control, ECSS-Q-ST-20-07C clauses 5.5.1-5.5.2. Use when a facility has to show it can hold the environment a test asks for and then that it did: check the declared capability band contains the required band for temperature, humidity, supply voltage and frequency, purge media and coolant, report the containment margin at each end, derive power-quality bands from nominal plus percentage tolerance, count monitored excursions with side and magnitude, time the longest unbroken one, and flag sampling too slow to see it. Trigger: ecss, q-st-20-07-test-centre, q2007-environment, test-facility-capability-band-containment, test-centre-environmental-excursion, power-quality-tolerance-band, environmental-monitoring-interval-adequacy, test-centre-media-purity-control."4license: Apache-2.05compliance: STANDARDS-REF6standards:7 - id: ecss8 reference-only: true9gated: false10domain: space-systems11pack: space-systems12compatibility: "agentskills.io SKILL.md; any SKILL.md host (Claude Code, Hermes, OpenClaw)"13metadata:14 source: https://skillmd.com/skills/ashfordeou/q2007-environment15 domain: space-systems16 subdomain: ecss17 tags: [ecss, q-st-20-07-test-centre, q2007-environment, test-facility-capability-band-containment, test-centre-environmental-excursion, power-quality-tolerance-band, environmental-monitoring-interval-adequacy, test-centre-media-purity-control]18 version: 0.1.019 author: Aero Agent Skills20---2122# ECSS Test Centre — Infrastructure and Work Environment (space-systems/ecss/q2007-environment)2324Use when the task is the facility step of ECSS-Q-ST-20-07C clauses 5.5.1 and255.5.2 -- showing that the buildings, services and media a test centre offers26are suitable for the tests booked into them, and that the environment was27actually held inside its band while the test ran.2829## Domain quick reference3031- Suitability and control are two questions, not one. Suitability asks32 whether the facility can hold the environment at all; control asks whether33 it did. A centre can be perfectly suitable and still have lost the band34 for six minutes at 02:00, and a single pass/fail hides exactly that.35- Suitability is band containment. The facility declares a capability band36 per parameter, the test declares a required band, and the capability has37 to cover the requirement at both ends. The amount by which it does is the38 containment margin, and the margin is the operator's room before the39 facility is running against its own limit.40- A facility whose capability exactly equals the requirement is suitable and41 has zero margin. Those are different statements. Report both, because the42 second is what predicts the excursion.43- The margin is reported at each end separately. A capability that is44 generous at the top and short at the bottom is a real and common case --45 a chamber that can go hot more easily than cold -- and one aggregate46 number hides which end to fix.47- Power quality arrives as a nominal value plus a tolerance in percent, not48 as a band, so the band is derived: a 230 V supply held to 10 percent is49 compared as 207 V to 253 V. Derive it once and compare against the derived50 limits; comparing a reading against "230 V plus or minus 10 percent" in51 prose is how two engineers reach two answers.52- Media -- a purge gas dew point, a chilled-water temperature, a cleanroom53 overpressure -- are ordinary parameters with their own bands. The failure54 mode is identical: drifting out of band with nobody watching.55- Monitoring cadence is graded on its own. A parameter sampled every fifteen56 minutes against a ten-minute requirement can hold a clean record through57 an excursion nobody saw, so a slow interval is a finding even when every58 recorded reading is inside the band.59- Excursions carry a side and a magnitude, and the longest unbroken run60 carries a duration. A brief touch of the limit and a half-hour outside it61 are not the same event and should not aggregate into one count.62- Every band comparison carries a small relative slack, so a reading landing63 exactly on a derived limit reads as inside on any platform. The slack64 absorbs representation error in a multiplied limit; it does not loosen65 the requirement.6667## Workflow68691. List the parameters in scope. For each, state the required band the test70 asks for and the capability band the facility declares. Derive a band71 from nominal plus tolerance for anything specified that way.722. Refuse the specification when a parameter name is unrecognised, a band is73 not a two-limit pair, a limit is not numeric, a lower limit sits above74 its upper limit, or a parameter appears twice.753. Test containment per parameter and record the margin at each end. A76 negative margin at either end is the suitability finding, and the end is77 named.784. Walk the monitored series against the required band. Record each79 excursion with its index, side and magnitude; take the worst band80 utilisation across the series as the single control figure.815. Multiply the longest unbroken run of out-of-band readings by the sample82 interval to get the excursion duration.836. Compare the sample interval against the required interval and flag a slow84 cadence separately from any excursion it might have hidden.857. Roll the facility up into two independent verdicts: suitable, and86 controlled. Report the parameters failing each.8788## Pitfalls8990- Collapsing suitability and control into one verdict. The remedies are91 different -- one is a facility change, the other is an operations change.92- Reading a zero containment margin as a comfortable pass. It is the93 narrowest possible pass and it predicts the first excursion.94- Comparing a reading against a percentage tolerance in prose instead of95 against the derived absolute limits.96- Counting excursions and never timing them, so a single touch of the limit97 and a sustained half-hour outside it look alike in the report.98- Accepting a clean record from a parameter sampled too slowly to see an99 excursion. The clean record is evidence about the sampling, not the100 environment.101- Using a strict comparison against a limit that was produced by a102 multiplication. The limit is not exactly representable and the same series103 then passes on one machine and fails on another.104105## Behavior contract (gate 3)106107The band validation, tolerance-derived bands, containment and margin108arithmetic, excursion detection with side and magnitude, longest-excursion109timing, monitoring-cadence check and the two-verdict facility roll-up are110exercised by the gate 3 contract test: scripts/test_q2007_environment.py111against scripts/q2007_environment_logic.py (stdlib unittest, offline).112Run: python3 scripts/test_q2007_environment.py113114## Compliance115116- ECSS standards are freely downloadable (ESA); cite the source and117 paraphrase per standards-map.yaml.118- compliance: STANDARDS-REF, gated: false.