Opened 2026-08-20. Until this date every number in this repository came from inside it, its own scripts, its own run sheets, and handbook values declared at the point of use. A10 records the consequence in its own words: the bank ESR bracket came from distributor listings because "the manufacturer's own data could not be obtained; that limitation is recorded rather than worked around."
That limitation lifted. This file is the survey: which live entries an external source can settle, which it can only inform, and which it cannot touch at all.
It is a map, not evidence. Anything acted on gets a banded run and a NEEDS SOURCE line
retired, the way A64 did.
A54 priced the store at 23.44-37.36 kg and named the gap as
NEEDS SOURCE rather than guessing: specific energy and ESR x C of a film or pulse capacitor.
Published pulsed-power literature gives millisecond-discharge capacitor energy densities of 1.9-2.68 J/cm³ at roughly unit density, in metallised polypropylene with extended-foil or bifilar electrodes for very low ESR and ESL.
A64, six of six: the store is ~70 g, 522x lighter, at 400 kW/kg against the 23.20 required. P86 closed and ADR-033's falsifier 1 does not fire.
A54 was correct in every calculation and wrong in its scope. It priced an EDLC because that was the only store technology in this repository. The lesson generalises, and it is why this file exists.
| Entry | What is needed | Why it is gettable |
|---|---|---|
| P26, the bank cannot source the shot on purchasable cells | Current EDLC cell data: ESR x C for modern high-power cells | A10's whole bracket came from distributor listings it could not verify against the manufacturer. If the product range has moved, the 68 mΩ ceiling may now be reachable, and the largest live defect Gen5 carries changes |
| E5, no host propellant or control authority figure exists. And A13 band 5 passed on one anyway, P94, 2026-08-20 | Published hosted-payload interface data for a spent-stage platform | A20 is parametric in exactly this one number and says so. A single published Δv budget turns a sweep into a result |
| E30 / A47, the reliability the design must hit is unmeasured | Flight reliability data for canisterised dispensers | A47 computes what r must be, 0.99326, and cannot say what r is. Published deployment counts and failure records would give the comparator a real number instead of an assumed one |
| P44, separation hardware outweighs a femtosat | ChipSat and femtosat deployer masses actually flown | The claim is a scaling argument with no flown datapoint under it |
| P45 / A25, the flywheel is at mass parity, not a saving | Reaction-wheel and flywheel-store specific energy from flown units | A25's 1.1 kg band-4 failure is owned by one unsourced number, and it says so |
| P57, the voice-coil deployer | Marked "needs institutional access" in LEGACY_STUDY_CLOSURE.md |
May be a literature question rather than an access one |
| A21 comparators, spring dispenser performance | Published dispenser specifications | A21-R already uses one published figure; the comparison would be firmer with more |
Added 2026-08-20. Everything above looks to spaceflight literature. This section looks sideways instead, at automotive and motorcycle engineering, where several of VOLLEY's live entries describe components that road vehicles have built in tens of millions of units.
The transfer is not free and the section says where it breaks. Every vehicle device below is lubricated, cyclic and atmospheric. VOLLEY is dry, in vacuum, and fires twelve times ever. That inversion cuts both ways: it makes wear irrelevant, and it makes one-shot devices far more attractive here than they are in a car.
| Entry | Vehicle technology class | What it would settle | What breaks in the transfer |
|---|---|---|---|
| P67 / P88 / P89, seal friction, unmeasured | Telescopic fork and damper rod seals, a sliding rod seal engineered against breakaway friction as its primary figure of merit, in hard-anodised aluminium bores | A61 specified 17.8 N, 4.00 % of the pressure force. A component class whose whole design brief is minimising exactly that number is the closest existing analogue to VOLLEY's requirement | Fork seals run wet. Their friction figure is an oil-film figure. VOLLEY runs dry at −35 °C, and A61's own note is that the empirical factor is "difficult to obtain unless evaluated on empirical lines" |
| P81, the backup ejector cannot clear the tube | Pyrotechnic gas generators, an automotive restraint inflator produces 0.5 to 0.9 mol of gas from 20 to 100 g of solid generant in roughly 30 ms, filling 60 L | A53 failed band 7 because a spring stores 4.5 J and clearing the tube costs 667.2 J. A device in this class releases energy of a wholly different order, and it is one-shot, which the carriage already is. A65 tests it | Total gas temperature is 1000 to 1400 K and ADR-035 set a 473 K tube ceiling. The same ceiling foreclosed steam in A63. Pyrotechnics also carry a range-safety and handling regime this repository has never priced |
| P86, the trim stage's pulse store | Traction-inverter DC-link capacitors and capacitor-discharge ignition, the same metallised-film construction, sized for millisecond discharge into an inductive load | Already closed, by A64, on pulsed-power literature. The vehicle route is the second independent path to the same answer, and it is the one built in volume | Nothing material. This row is here because it is the worked example: the store looked impossible at 23-37 kg only because an EDLC was the sole technology in the record |
| ADR-033 falsifier 3, no velocity sensor exists | Variable-reluctance wheel-speed and resolver sensing, a passive, radiation-tolerant, non-contacting measurement of a moving steel feature | A55 records that "there is still no velocity sensor in any file", and the loop has 1.4 ms to measure in. This class needs no optics and no window | A wheel sensor measures a rotating target at kHz. VOLLEY needs a linear velocity to a tolerance A44's dispersion sets, in a single pass, once |
| E30 / A47, twelve-cycle series mechanisms forfeit the manifest | Sequential gearbox shift drums and detent escapements, a single-actuator indexer moving a fixed sequence of positions under load | A47 counts the escapement among the shared elements that forfeit everything remaining. A drum indexer is the same problem solved for a component with a very large field population | Field reliability for a gearbox is a warranty figure, not a mission figure, and it is measured on units that are serviced. It cannot be read across to twelve shots with no maintenance |
| P78, the chamber-to-tube gas interface | Internal-floating-piston dampers, a gas charge separated from the working volume by a free piston with a single dynamic seal | The gas store and the bore meet somewhere, and no file draws that junction. This class is the standard answer to the same geometry | It is a damper, sized for cyclic low-rate motion. The pressure is comparable; the rate is not |
The strongest row is P81. It is the only one that could resurrect a defect A53 closed as architectural, and A47 priced that defect at +2.27 satellites delivered against +0.37 for the entire Gen5 to historical study change, six times more. That is why it was run rather than listed.
The weakest is E30. Automotive reliability data is abundant and is the wrong kind of number. A warranty rate on a serviced, lubricated, million-cycle component says nothing about twelve unserviced shots in vacuum, and quoting it would be worse than the current honest blank.
These have a literature, and the literature does not answer this machine's question.
Water propulsion is flight-proven at CubeSat scale. A water resistojet has flown on a 6U CubeSat in deep space, performing three ΔV and eleven trajectory-correction manoeuvres at 91 s specific impulse on under 14 W, with 1.2 kg of water in 2U. A water electrolysis thruster flew in 2021 and had accumulated 408 firings by 2023, and a 20 W electrolysis thruster has been demonstrated at 1.25 mN and 185 s.
None of it transfers. Every flown system is a thruster, where specific impulse is the figure of merit. A gun is sized by pressure-volume work released in milliseconds, and Isp says nothing about it.
What the literature does establish is that storing and vaporising water on a small spacecraft is flown rather than speculative, which is why PII-21 stopped steam on bore temperature and tube mass, not on feasibility.
Added 2026-08-20. VAULT.md holds twenty-one parked entries. The question asked
of each was not "is it good" but "did it stop on a number that came from inside this repository."
That is the only kind of stop an outside source can move, and it is the lesson
A64 taught at the cost of A54.
Most of the vault fails that test immediately, and it is worth saying why in one line. Nine entries stopped on 2026-08-14 because ADR-032 deleted the subsystem they improve, no mover, no stator, no bank, no track. No literature resurrects an optimisation of a part that no longer exists. PII-9 describes a different programme and PII-8 a different vehicle; a source cannot supply a host. Those are stops by type, and they stay shut.
Three entries survive the filter. One of them has a live route back into historical study.
PII-19 stopped by attribution, not refutation: A35 measured the mover it optimises at 11 % of dry mass, so it was "a careful, banded, correct optimisation of the wrong term." That verdict was about Gen5's whole drive. historical study no longer has one, but it has a 144.01 mm motor at the muzzle, and that motor has PII-19's exact problem.
ADR-033 brought magnets back to the moving part and listed the cost itself: "P34, a payload carrying a magnetometer cannot fly in this magazine. This defect returns because the magnets do", plus E35, plus a cradle that must now hold magnets in alignment.
PII-19's whole proposition is a passive plate instead of magnets on the moving body. Applied to a trim stage rather than a drive, it would not return P34, not return E35, and not ask the cradle to align a magnet set. The entry was priced against a 9.445 kg sled where its gain was 11 %. Against a stage whose stated cost is a returning defect rather than a mass, the same idea is answering a different and better question.
And PII-19 carries the measurement that may kill its own comeback. A30 measured the
transverse edge factor at 0.0253 against the 0.55 it had been sized on, a factor of 22,
because the Russell, Norsworthy factor collapses for a secondary narrower than the pole pitch.
stator.pole_pitch is 24 mm and any secondary that fits a 15.805 mm bore is narrower than that.
That is the run, and it is not written yet. It is also why this belongs in a sourcing map rather than in a decision: the deciding number is already in the vault, measured, and it points the unwelcome way.
A second PII-19 result transfers whether or not the plate does. A32 band 4 found segment ripple "intrinsic to a segmented long stator with a short secondary", and PII-19's own closing line says it "will be rediscovered by anyone who proposes one." A55's stage is a short stator crossed in 1.4 ms by a short secondary, and neither A48 nor A55 models the entry transient A32 measured.
PII-14 split its flywheel out as an independent candidate against P26, saying so in the file: "an energy store does not care what load it feeds... it is a Phase I candidate against P26 in its own right." P45 then found it at mass parity, owned by one unsourced number, and A25 band 4 failed on it.
The external check settles it and the answer is no. The question a flywheel was competing for is now historical study's pulse store, and A64 answered it at ~70 g on published pulsed-power capacitor data. A rotating machine, its bearings, its containment and its 7.15 N·m·s of stored angular momentum are not going to beat seventy grams of film capacitor, and the angular momentum is a disturbance in a machine whose shot already dumps 3.28 N·m·s.
This is the useful shape of a vault result: the entry is retired by a number, not left ajar.
PII-11 §5 argued the straightness requirement had been overstated, "a 3.3 m deployed track is a looser straightness requirement than the machine already meets", because an ironless Halbach airgap tolerates error at 1 mm of clearance and costs thrust rather than running away.
historical study's 8.0 m tube is not an airgap. It is a bore with a sliding seal in it, and a seal's job is to maintain contact, so the same argument runs backwards: the tolerance that made a deployed track plausible does not exist here. A59 already found the tube needs seven supports, and P67, P88 and P89 all rest on a friction that bore straightness and roundness directly set.
| What is needed | Why it is gettable |
|---|---|
| Straightness and roundness tolerances for long honed and skived tube in the bore class historical study uses | It is a stock manufacturing specification, published per metre of length, for exactly the sliding-seal duty A61 specified. A61 already found a 16.000 mm ISO 6432 stock bore costs 0.00 % on the seal specification, the sourcing route into this design is open and has been used once |
This does not reopen PII-11. It takes one paragraph of its arithmetic, notes that the sign flips, and points at three live entries that inherit the consequence.
P92. Reading PII-19 against ADR-033 to see whether a passive secondary fits meant reading how the present stator reaches its magnets, and it reaches them through the tube. ADR-035 made that tube aluminium four days after ADR-033 placed the stator outside it, and nothing owns the interaction.
That is the strongest argument for keeping a vault at all, and it is not the one the vault was built on. The entry did not come back. Reading it found a defect in what replaced it.
Added 2026-08-22, after a sweep of the outside separation literature was reconciled against the record. It found the largest single gap in the historical study evidence, and it is not a number, it is a whole class of model this repository does not have.
| Run | What it models |
|---|---|
| A23 | Tip-off rates against deployer classes. A comparison, not a mechanism |
| A34 | The payload rattling across its cradle clearance, Gen5 |
| A38 | The same closed forms at the historical study point, corrected to ADR-034's stroke on 2026-08-22, P102 |
All three model the release interface. None models the guide. historical study accelerates a payload along 8.0 m of bore and this repository has no contact state along it: no straightness, no roundness, no local clearance, no force-line eccentricity, no payload centre-of-mass offset, and no lateral or angular state carried through the stroke. The chain that is modelled is chamber pressure to axial force to exit velocity. The chain that decides whether the product works is contact state to lateral impulse and torque to release pose and rate, and it is absent.
"Design and Analysis of a High-Precision Separation Mechanism for On-Orbit Launch of
Micro-Spacecraft", Aerospace Science and Technology, 2026, publisher record
S1270963826018869. Located and checked 2026-08-22 from its abstract and indexing only.
Evidence level: ABSTRACT ONLY. What the published abstract states:
- Launch-dynamics models of several separation-mechanism configurations in ADAMS, to find which configuration produces large launch disturbance and why. The cylindrical configuration performs best, and a continuous-contact straight guide rail beats a gradient rail at suppressing launch disturbance.
- Contact force and moment formulas derived analytically, then parametric optimisation of guide-rail length and fit clearance, reported as an 82.76 % improvement in launch attitude accuracy at a 270 mm rail and 0.2 mm fit clearance.
- A 3U-form-factor device in 7075 aluminium, with hard-anodised guide-rail surfaces coated with PTFE to reduce friction.
- Post-launch angular velocity treated as the quantity that decides mission success.
The 270 mm and the 0.2 mm are that mechanism's answers and are not to be imported. They are quoted here so that the shape of the result is on the record, fit clearance has an optimum, not a minimum, and because a number left unstated tends to get remembered as a recommendation. VOLLEY's bore is 15.805 mm over 8.0 m and shares nothing with a 270 mm rail but the physics.
Two of those land directly on historical study. The tribological pair is the one ADR-035 and A61 chose, hard-anodised aluminium against PTFE, and the output quantity is the one
KILL_CRITERIA.mdthreat 4 is written in. This is a guided ejection of a CubeSat-class body along an anodised aluminium rail, analysed for release attitude. It is the same physical problem.
Evidence level: ABSTRACT ONLY. The full text has not been read, so this entry is a lead: its method is used to decide what VOLLEY must model, and no figure from it is cited anywhere in this repository.
| P67's framing | Friction is not separable from the geometry that produces it. B-2 measures a force; the force is a property of a bore, a finish, a clearance and a seat. The measurement stays first, this changes what it is a calibration point for, not whether to take it |
| P78 | Stroke buys gentleness, and the same stroke is more guided length for tolerance and friction to vary over. P78 already says stroke makes the worst defect worse; this says which defect |
| P102 | Names the gap in the same words and does not close it |
| What it cannot do | Supply VOLLEY's clearance. An optimum belongs to the mechanism it was optimised for, its masses, its rail length, its fit, its release. Importing one would be the mistake this file exists to prevent, and it is the same rule already applied to the vault's straightness arithmetic |
Jiaolong Zhang, Jingao Su, Chao Wang, Yiqian Sun, "Modular design and structural optimization of
CubeSat separation mechanism", Acta Astronautica 225 (2024) 758-767, DOI
10.1016/j.actaastro.2024.09.067. Evidence level: PUBLISHER METADATA VERIFIED.
| From the publisher record | |
|---|---|
| Optimised separation-mechanism mass | 2.417 kg |
| Mass reduction | 34.9 % |
| CubeSat + mechanism total | 13.417 kg |
| Mechanism mass proportion | 18 % |
| Maximum deformation | 0.123 mm |
| Ground mechanical verification | overload, sine, random vibration, shock |
| Flight | BY-03 deployed in orbit |
A21 compares VOLLEY against a spring dispenser as a class. This is one named mechanism carried the whole way, and the ladder it exposes is the useful thing:
| Rung | VOLLEY | |
|---|---|---|
| 1 | Analysed only | here |
| 2 | Prototype with separation tests | , |
| 3 | Ground environmental qualification, overload, sine, random, shock | , |
| 4 | Flown | , |
FIGURE_INDEX.md's class-D evidence has zero members and E4 says nothing here has been
built, fired or measured. The comparator is three rungs above.
Its flight is not evidence that VOLLEY works. It is evidence about how far VOLLEY's evidence has to travel, and the two must never be confused. A separation mechanism succeeding in orbit says nothing whatever about a different mechanism that has never been made.
On mass, read it carefully. 2.417 kg is a separation mechanism for one CubeSat, at 18 % of a 13.417 kg stack. VOLLEY's numerator is a deployer amortised over a manifest, 10.547 kg per 3U satellite on dry mass, 1.2145-3.0827 kg on added mass (P68). They are not the same quantity and this file does not divide one by the other. What the comparison supports is P69's finding, already recorded: mass parity with conventional separation hardware is withdrawn.
Not added to PRIOR_ART.md. That file requires the work to have been read in
full, and this entry is metadata plus abstract.
Three records, and their value is method rather than number. None of their dimensions, clearances, tolerances or coefficients may enter VOLLEY: those belong to the mechanisms they were derived for.
| Record | Evidence level | What it contributes to P103 |
|---|---|---|
Zhenyu Jin et al., "Rapid Modeling Method and Analysis of Factors Affecting the Dynamics of On-Orbit Launch Systems for Micro-Spacecraft", Aerospace 13(6) 541, 2026, DOI 10.3390/aerospace13060541 |
PUBLISHER METADATA VERIFIED | The two-level modelling structure P103 step 2 and step 3 adopt: a high-fidelity rigid, flexible model, a reduced multibody model calibrated against it, and parametric sweeps run on the reduced one. Also the factor list, thrust misalignment, thrust eccentricity and mass eccentricity as first-class inputs, not second-order corrections |
Linfei Yang et al., "Research on the Influence of Structural Parameters of Micro-Spacecraft On-Orbit Launch Separation Mechanism on Launch Accuracy", DOI 10.3233/ATDE260113 |
PUBLISHER METADATA VERIFIED | Monte Carlo over structural parameters, guide clearance, sliding length, step thickness, with guide clearance identified as the key one. The transferable finding is that guided-contact geometry has an optimum and drives release accuracy, which is why P103 lists clearance as first-order rather than a tolerance detail |
Zengqiao Tan, Haibo Yang, Xiaoyu Tao, "General method for establishing a contact force model during the in-orbit launch of micro-spacecraft", J. Phys. Conf. Ser. 3240 012008, 2026, DOI 10.1088/1742-6596/3240/1/012008 |
PUBLISHER METADATA VERIFIED | The contact model itself. Micro-spacecraft-against-barrel collision, a Lankarani, Nikravesh continuous contact force law, a finite-element collision model with mesh validity checked, and inverse identification of the contact parameters by unscented Kalman filter against the FE result. This is the answer to "where do contact stiffness and damping come from before there is hardware?" |
Together they say the same thing three ways, and it is the thing P103 exists for: for a guided ejection, the exit attitude is set by contact, clearance and eccentricity, the contact law needs parameters that are identified rather than assumed, and the practical route is a reduced model calibrated against a higher-fidelity one. None of that requires hardware, which is why P103's model is no longer waiting behind B-2, see the amendment in that entry.
What none of them supplies: VOLLEY's clearance, VOLLEY's straightness, VOLLEY's friction coefficient, or a tolerance any of them optimised. Those are configuration-specific and importing one would be the failure this file exists to prevent.
Written 2026-08-22 and superseded the same day. An earlier revision of this section reported that the Acta Astronautica pagination was 615-624, that the 2.417 kg mass, the test campaign and the BY-03 flight could not be confirmed, and that two 2026 separation papers could not be confirmed at all.
Every one of those findings is withdrawn. The pagination is 758-767. The mass, the campaign and the flight are in the publisher record. All four records above are real and their metadata is verified.
The defect was method, not carelessness. Those conclusions were drawn from indexing summaries rather than from the publisher records, and an absence in a summary was reported as an absence in the paper, including a correction to a citation that was itself unsourced, which is worse than the claim it corrected because it wore the authority of having been checked.
The rule this produces, and it is now the rule for this file: every entry states its evidence level,
FULL TEXT READ,PUBLISHER METADATA VERIFIED,ABSTRACT ONLY,SECONDARY LEAD,NEEDS SOURCE, and no numerical result is quoted below the level that supports it. P22 is the entry that first taught this project the cost of writing a literature file from abstracts. This is the second time.
Added 2026-08-26. HOST_REFERENCE_CASES.md is a method for mapping a
real, publicly documented propulsion architecture against the three host classes in
MISSION_ARCHITECTURE.md, and its first worked case is a restartable
semi-cryogenic engine platform whose thrust class, propellant family and ignition-system test
campaign are all public.
It belongs in this section rather than in the one above, and the reason is the whole point of the entry. Public company information about an engine is real evidence about an engine. E5 is not about an engine.
| Public information can establish | Public information cannot establish |
|---|---|
| The engine class and its feed architecture | The stage's actual post-primary propellant reserve |
| A nominal thrust figure, sometimes without its condition | Host control authority (P94) |
| The propellant family | Engine or stage restart qualification, as opposed to a subsystem test |
| A modularity or customization claim | The cryogenic coast limit that sets campaign duration |
| That an ignition technology exists and has been fired | The flight interface (E31) and what a provider will credit (P68) |
So no HOST_DATA entry moves, E5 stays open, and the historical study computation
count is unchanged at seventeen. What the study produced instead is a list of the specific
provider numbers that would replace each of its assumptions, which is the form E5 would actually
be closed in if anyone ever supplied them.
The sharpest thing the exercise found is a distinction, not a number. A resonance igniter that restarts repeatedly is a subsystem result. Whole-engine restart additionally needs turbomachinery, thermal cycles, valve life, propellant conditioning, chill-down and inlet conditions, and stage restart needs tank state, ullage, settling and command authority on top of that. Reading the subsystem claim as a stage capability would be the same class of error as A54's, which was correct in every calculation and wrong in its scope.
One entry opened, P113: how the orbital work divides between host and
deployer has never been computed, and it sets VOLLEY's release-velocity requirement. That is a
PROGRAMME computation and it is deliberately not allowed to touch P108.
These need hardware, and no amount of reading substitutes.
| P67, the seal friction | Searched 2026-08-20 and it is not there. The literature confirms the component class, spring-energised graphite-filled PTFE, and confirms that friction is determined empirically, with the standard relation carrying an empirical factor "difficult to obtain unless evaluated on empirical lines." There is no published coefficient that answers a specific bore, seal, surface, pressure and temperature. A61 produced a requirement, 17.8 N, and a requirement is what a bench test is measured against |
| P88, the seal cannot absorb its own friction | Same. The conduction path out of a seal is a design, not a lookup |
| P89, the seal specification | The class ranges A61 used are handbook and stay NEEDS SOURCE until measured |
| E4, nothing built, fired or measured | The whole point. No external source changes the category of this project's evidence |
P67 is the sharpest case for why this file has a third section. Six entries depend on it, it now has a number to be measured against, and the search returned exactly what a careful reader would expect: the method, not the answer.
Nothing here is evidence. A row moves out of the middle section by being run: bands declared
first, the source named at the point of use as a technology class rather than a product, and
the NEEDS SOURCE line retired in the run sheet that carried it.
No organisation, supplier or individual is named anywhere in this repository, and a published performance range for a component class is not an exception to that, it is how A39 declared its gas model and how A64 declared its capacitors.