The Qualification Gap: Method, Cell Definitions, and Limits
How the four-way comparison of in-vacuum optical qualification capability is built, what each cell value means, and what the comparison does not establish.
Constanellis Aerospace. Authored by the Constanellis engineering staff. Individual contributors are named on request to reviewers and to press under embargo.
Citation record
- Author
- Constanellis Aerospace
- Title
- The Qualification Gap: Method, Cell Definitions, and Limits
- Identifier
- CR-2026-01
- Version
- 1.0
- Date
- 3 August 2026
- Rests on
- content/gap-map.ts
- Named authors
- Authored by the Constanellis engineering staff. Individual contributors are named on request to reviewers and to press under embargo.
- How to cite
- Constanellis Aerospace. "The Qualification Gap: Method, Cell Definitions, and Limits." Constanellis Research CR-2026-01, version 1.0. 3 August 2026. https://constanellis.com/research#cr-2026-01
Constanellis publishes a four-way comparison of who can perform in-vacuum optical performance characterization for U.S. defense customers. This paper documents the method behind that comparison: how the provider classes are defined, what each cell value means, why cells that could not be verified ship marked uncertain, and where the comparison stops. It states the distinction the comparison rests on, that screening establishes whether hardware survives an environment while performance characterization establishes what the hardware does inside it. It also states what a reference-directory census of the field can and cannot support.
Scope and question
The comparison answers one question. Which providers can measure the optical performance of hardware inside a vacuum environment, and sell that measurement to a U.S. defense customer. It does not rank providers, score them, or evaluate anything outside that question.
Four provider classes are compared across five capability rows. The values live in one data module, and the published table renders that module directly. This paper explains the module. It adds no values to it.
The comparison carries a review date and was last reviewed 1 August 2026. A comparison without a date is an assertion; the date is what lets a reader decide how much of it to trust.
How the provider classes are defined
The columns are classes of provider, not organizations. Publishing them as classes keeps the comparison about capability and access rather than about any single supplier, which is the question the comparison is built to answer. Named providers sit behind each class in the source record.
- Federal labs
- Government, in house. Facilities that hold the capability inside the organization. Where the record shows a capability exists and is not sold, the cell reads in house only.
- Commercial test houses
- U.S. environmental screening. Providers whose core business is selling environmental screening to outside customers.
- Foreign commercial
- Allied nation, export-controlled path. Providers in an allied nation that sell the capability, reachable by a U.S. program only through an export-authorization path.
- Constanellis FORGE
- Colorado, build-out. The capability under construction in Colorado. Nothing in this column is operating today, and the cell values say so.
What each cell value means
Cells take a fixed vocabulary. Five values are available to the three outside columns, and one more is reserved for the Constanellis column, because a capability under construction is not the same claim as a capability sold today.
The vocabulary is fixed so a reader can separate three different statements that a plain yes or no would blur together: what a provider can do, what a provider will sell to a U.S. defense program, and what we were able to verify.
- yes
- The provider sells the capability to outside customers.
- no
- The record shows the provider does not offer the capability.
- in house only
- The capability exists inside the organization and is not sold. Capability that cannot be purchased is not capability a program can schedule.
- not available to U.S. defense customers
- The capability is sold, and the export-control path closes it for controlled hardware.
- uncertain
- The research did not establish a value at the standard this comparison requires. The cell reports the state of our knowledge, not a judgment about the provider.
- in build-out
- Constanellis column only. The capability is designed and under construction, and is not operating. The caveat sits inside the cell so that it survives a screenshot.
Why unverifiable cells ship as uncertain
A comparison published by one of the parties compared is worth exactly its least defensible cell. One flattering value makes the whole table unreadable, because a reader can no longer tell which cells were checked and which were assumed. Fairness here is not courtesy. It is the entire value of the instrument.
So a cell is filled only from the source record. Where the record does not carry the answer at the standard this comparison requires, the cell reads uncertain and the row keeps the note explaining what is missing.
Uncertain is therefore a statement about us, not about the provider. It marks work we have not finished, it names what would settle the cell, and it stays in place until someone does that work. Cells that could be resolved by inference are exactly the cells that must not be.
Calibration-chain details for the other providers are not published at the level this table requires; marked uncertain until verified.
The measurement chain: screening and characterization
Screening proves a part survives the environment. Performance characterization proves what it does inside it.
Screening answers a pass-or-fail question. The article goes into the environment, the environment runs a profile, and the article comes back working or it does not. That answer is necessary, and for most hardware it is enough.
It is also silent on the question an optical program has to answer, which is what the article measures, points at, or transmits while the environment is applied. Wavefront, throughput, alignment, and pointing are properties of the article inside the environment, and a part that survives a profile has not been asked about any of them.
Performance characterization answers the second question. It records measured values with the environment held, and it records the calibration chain behind those values. The chain is carried as its own comparison row rather than as a footnote, because a measured value without a traceable chain does not travel from one program to another: the next program cannot tell what the number is worth.
None of this is an argument against screening providers. Screening is a distinct service with a distinct answer, and the comparison records who sells it.
Screening is the commercial test houses' core business.
Access is treated the same way. A capability a U.S. defense program cannot lawfully reach is not a capability that program has, so export-control status is a row of the comparison rather than a caveat under it.
The ITAR constraint is a row, not a footnote: sending controlled hardware to a foreign provider requires export authorization that most programs will not pursue for test services.
The reference-directory census: what it shows and what it does not
The comparison ships with one piece of context drawn from the field's own reference directories of lunar and space test facilities. It is published at the level of generality below, and the directory itself is named in briefings rather than in this paper.
The field's own reference directories of lunar and space test facilities do not yet track optical metrology, calibration, or traceability as categories at all, and U.S. commercial companies are almost absent from them. Detail behind that census is available in a briefing.
What the census supports is a statement about categories. The taxonomy a lunar test directory keeps does not carry optical metrology, calibration, or traceability as headings, and U.S. commercial entries in it are rare. That is evidence about how the field files capability, which is the point worth making: a capability the field has no heading for is a capability nobody is asked to supply, and no program manager can search for.
What the census does not support is a census of capability. A directory is a curated listing with its own scope and its own submission process. Absence from it does not establish absence of a provider, says nothing about the capacity or the quality of the facilities it does list, and settles nothing for any provider outside its scope.
Read it as a map of where the field is looking, not as proof of what is not there. The comparison table carries the capability claims; the census only explains why so few people have asked the question.
Limits
The comparison is a snapshot and it carries its review date. It is wrong the moment a provider changes what it sells, and the date is published so a reader can weigh that risk instead of guessing at it.
The sources are internal market research and one field reference directory. Neither is a peer-reviewed instrument and this paper does not present either as one. Both were reviewed against the standard stated here, which is a lower bar than replication and a higher one than recollection.
The Constanellis column states intent. Nothing in that column is operating, and the column will read yes only when there is data behind it.
The comparison covers one capability question. It is not a supplier evaluation. It says nothing about price, schedule, capacity, or quality, and it should not be cited as if it did.
Findings
- F01
Distinguish survival from performance. Screening establishes that an article returns working; performance characterization establishes what the article does inside the environment.
- F02
Read the columns for access as well as capability. Capability held inside an organization, or reachable only through an export-authorization path, is not capability a U.S. defense program can schedule.
- F03
Mark a cell uncertain when the record does not carry the value. Report the gap rather than close it by inference, and name what would settle it.
- F04
Carry the calibration chain as its own row. Vacuum optical test and vacuum optical test with a traceable chain are different capabilities, and a measured value without a chain does not travel between programs.
- F05
Treat a reference directory as evidence about categories, not as a census of capability. Absence from a listing does not establish absence of a provider.
Sources and provenance
- Market research record
- Internal market research, verified 1 August 2026. Establishes provider capability, commercial availability, and export-control status for each provider class.
- Field reference directory
- A published directory of facilities for lunar surface technology readiness, accessed 1 August 2026. Establishes how the field categorizes test capability. Named in briefings, per the disclosure practice this paper states in its limits.
| Capability row | Source class |
|---|---|
| In-vacuum optical performance characterization | Market research record |
| Optical and mechanical verification under vacuum, sold commercially | Market research record |
| Environmental screening (survival) | Market research record |
| Available to U.S. defense customers (ITAR) | Market research record |
| Vacuum optical test with a calibration chain traceable to national standards | Market research record |
Each row above appears in the published table with its four cell values. This paper documents how those values were assigned; the table is where they are read.
The field reference directory backs the census context above rather than any cell in the comparison, which is why no row cites it. A source that supports the framing and a source that supports a value are recorded separately here on purpose.
Provenance and revisions
- Data module
- content/gap-map.ts. Holds the four provider classes, the five capability rows, every cell value, the per-row source key, the context footnote, and the review date. The published comparison table renders it directly, and this paper imports it rather than restating it.
- Published comparison
- The qualification gap map, at /forge/gap-map. Same data, rendered as the table this paper documents.
- v1.0 · 3 August 2026 · First publication.