Prototype to Production: Invar Parts for Photonics Programs
A photonics program almost always starts the same way: a designer needs one part fast, finds a shop willing to iterate quickly on a few pieces and gets a working prototype. The trouble comes later, when that same drawing needs to become 200 units a quarter with a documented process, an incoming inspection plan and, eventually, a second source. Treating that transition as “order more of the same part” skips the steps that actually make a production run reliable.
Why prototype and production are different problems
A prototype’s job is speed and learning. The shop and the designer iterate on fit, function and tolerance together, often informally, and a handful of hand-fitted parts is a success. A production run’s job is repeatability at a defined cost and schedule, which needs a frozen design, a documented process and enough inspection data to know the run is staying in control, not just that the last part measured fine. Prototype vs production Invar machining covers why the cost structure changes between the two; this guide covers the process and qualification work that has to happen alongside that cost shift.
The core mistake programs make is skipping the middle stage. Going straight from “the prototype works” to “order 200 a quarter” skips the pilot batch that would have shown whether the process behind that one good prototype actually repeats.
The three-stage ramp
| Stage | Typical quantity | What matters most | Documentation produced |
|---|---|---|---|
| Prototype | 1 to 5 units | Fit, function, design iteration speed | Working drawing, informal notes |
| Pilot | 20 to 30 units | Part-to-part repeatability, process stability | Frozen drawing, documented heat treatment cycle, inspection data across the batch |
| Production | Ongoing, scheduled volume | Consistent cost, schedule and incoming quality | Fixed process package, incoming inspection plan, second-source readiness |
The pilot stage is where most of the qualification work belongs, and it is the stage photonics programs most often skip under schedule pressure. Running it deliberately, rather than treating the first pilot batch as an early production shipment, is what turns a lucky prototype into a real process.
Building the process package during the pilot
A process package is what makes a part reproducible by a shop other than the one that happened to build the prototype well. At minimum it should include the finished drawing with every tolerance and geometric callout locked, the stress relief and any stabilization cycle either specified on the drawing or documented as the shop’s fixed practice, the inspection method and acceptance criteria and any fixturing notes the shop considers necessary for a repeatable result. Every change made to the drawing on the way from prototype to pilot should go through a change process, covered in engineering change requests for Invar parts, so the frozen package reflects what was actually qualified, not an earlier revision.
Inspection data from the pilot batch is the other half of the package. A CMM report on a single part shows that part passed. A CMM report across 25 pilot units, referenced in the same way described in Invar CMM inspection reports, shows the mean, the spread and whether the process is centered on nominal or drifting toward a limit. That data is what a production decision, and later a second-source qualification, should be based on.
Incoming inspection at production volume
Once a part moves to scheduled production volume, inspecting every unit the way a pilot batch was inspected usually stops being practical or necessary. Programs typically move to a sampling plan, incoming inspection on a defined percentage or interval of units, with full inspection reserved for the first article of each new batch of raw material or after any process change. What that sampling plan should catch is drift, not just gross defects: an Invar part’s tightest requirement is often long-term dimensional stability, and a sampling plan that only checks obvious dimensions at receipt can miss a slow shift in a process that is heading out of control.
A practical incoming inspection plan for a photonics program tracks its sample measurements over time, not just against the drawing tolerance for each batch in isolation. A batch that measures within tolerance but has drifted steadily toward one limit over the last several shipments is telling you something a pass or fail stamp on any single batch will not: a tool is wearing, a raw material heat lot has changed, or the stress relief furnace is running warmer than it used to. Catching that trend early, while every batch is still passing, is cheaper than catching it after a batch finally fails.
Planning for a second source
A second source is easiest to add when the process package from the pilot stage already exists, because that package is what a second shop qualifies against. Trying to add a second source with nothing more than the original drawing and no documented heat treatment or inspection history, described in second-source qualification, turns a routine supply chain move into a redevelopment project. Programs that plan for a second source from the pilot stage onward, even if they do not intend to use it immediately, save significant time when volume, a supply disruption or a cost negotiation eventually calls for it.
There is a real cost to qualifying a second shop before it is strictly needed: a second pilot batch, a second round of inspection data and the ongoing overhead of keeping two shops current on any drawing revision. For a low-volume or early-stage program that cost may not be worth carrying yet. The judgment call is whether the program’s schedule, its exposure to a single shop’s capacity or a customer’s own dual-source requirement makes that overhead worth paying now rather than later, and that answer is different for a company shipping ten units a year than for one ramping toward a thousand.
A related question programs often skip is what happens to the second shop’s qualification when the drawing changes after both shops are running. An engineering change that only one shop has implemented, even briefly, means the two sources are no longer building the same part, which defeats the point of having two sources in the first place. The discipline that matters here is small: every revision goes to both qualified shops at the same time, with the same effective date, and neither shop ships against an old revision once the new one is released. This sounds obvious in the abstract and is one of the most common places a two-source setup quietly breaks down in practice.
Worked example: a resonator mount ramp
This is a hypothetical planning timeline, not a real program’s schedule. A photonics company designs a resonator mount and has three hand-built prototypes made at a shop it found through an RFQ in month one, confirming fit and alignment in month two. Rather than jumping to a production order, it runs a 25-unit pilot batch at the same shop in month three, with the drawing frozen and the shop’s stress relief cycle documented for the first time. The pilot data shows a mean within 0.006 mm of nominal and a spread of ±0.005 mm, comfortably inside the drawing’s ±0.020 mm tolerance.
With that data in hand, the company commits to a production schedule of 60 units a quarter starting in month five, using the pilot’s process package as the fixed spec, and begins qualifying a second shop against the same package in month six as a deliberate risk mitigation step rather than a reaction to a problem with the first shop. By month eight the second shop’s own pilot batch confirms it can meet the same numbers, and the program now has two qualified sources before it ever needed one under pressure.
Compare that to a program that skips the pilot stage and finds out at unit 40 of a production order that the shop’s stress relief cycle was never actually fixed, and the mounts have been drifting slightly wider apart in spec over the run. The pilot stage costs a few weeks and a modest unit count. Discovering a process problem at production volume costs a great deal more, not just in scrap and rework but in the schedule slip that follows while the shop’s process gets fixed retroactively under pressure, this time with a customer’s order already late.
What to put on the RFQ at each stage
- Prototype RFQ: state that the quantity is 1 to 5 units and that the design may still change, so the shop quotes appropriately flexible fixturing
- Pilot RFQ: state the target quantity (20 to 30 units is typical), request documented process and inspection data across the batch and treat the drawing as frozen for this run
- Production RFQ: reference the process package from the pilot, state the expected ongoing volume and cadence and note whether second-source qualification is planned
Background on the alloy family and typical properties referenced throughout a process package is on the Invar specs reference page.
When you are ready for any of these stages, send an RFQ with your drawing and note which stage you are at. We match it with up to three shops suited to that stage, whether that is fast prototype turnaround, a documented pilot run or ongoing production volume. This guide is part of the optics, photonics and laser systems overview alongside the other sourcing and process guides for that vertical.
Frequently asked questions
Can the shop that built the prototype also handle production volume?
Sometimes, and it is often the easiest path if the shop has the capacity and is willing to formalize its process. The risk is assuming this without checking. A shop well suited to fast, hands-on prototype iteration is not automatically set up for repeatable batch production, so confirm capacity, fixturing plans and documentation practice before committing volume to the same shop by default.
How big should a pilot batch be before committing to full production?
Large enough to see real part-to-part spread, not just confirm the design works. Twenty to thirty units is a common range for photonics-scale parts, enough to calculate a meaningful mean and spread rather than judging from two or three pieces that may not represent the process.
What exactly goes into a frozen process package?
At minimum: the finished drawing with all tolerances and geometric callouts, the specified or documented heat treatment cycle, the inspection method and acceptance criteria and any fixturing or setup notes the shop considers part of getting a repeatable result. It should be complete enough that a second shop could plausibly reproduce the part from it.
Does the matching service manage the production ramp for us?
No. We match your RFQ with shops at each stage, prototype, pilot or production, and you manage the relationship, the process documentation and the qualification decisions directly with the shop. We are not a program manager, a quality organization or a machine shop ourselves.
Reviewed against
- ASTM F1684, Standard Specification for Iron-Nickel and Iron-Nickel-Cobalt Alloys for Low Thermal Expansion Applications
- Wikipedia, Photonics
Values in this guide are typical handbook figures for planning. The mill certificate for your heat lot and the shop's quote for your geometry govern.
Need a quote for this part?
Send the drawing. We match you with up to 3 shops that machine Invar and reply within 24 business hours.
Get matched with an Invar shop