A quotation for a microwave board is not one number; it is five lines that behave differently when the batch size changes. There is a unit price, a one-time engineering charge, a set of tooling items, an add-on schedule for impedance control and hybrid construction, and a minimum charge that governs small orders. Buyers who negotiate only the first line routinely pay 20 % more than they need to, because the four lines that can be restructured are the ones nobody asks about. This article takes a quote apart line by line, shows how each one amortises across a batch, and ranks the levers that actually move the total.
Two Cost Articles, Two Audiences
The cost of a microwave board is looked at from two directions and the two views are not interchangeable. From inside the factory, cost is material per usable area, processing, test and the reciprocal of yield — an equation that a process engineer can attack station by station. From outside, the same cost appears as a quotation: a unit price, a one-time engineering charge, tooling, add-ons, and a minimum charge. The first view explains why the number is what it is; the second explains which part of it a buyer can change.
This article is the second view. The internal equation is set out in the cost structure of high-frequency microwave PCB manufacturing, and it is worth reading first, because every fee below maps to a physical cause: a tool change, a press cycle, a coupon measurement, a yield risk. A fee that cannot be mapped that way is a fee worth questioning.
| Line in the quote | What pays for it | How it behaves with batch size | Negotiable? |
|---|---|---|---|
| Unit price | Material, processing, test, yield | Falls with batch, then flattens | Yes, but slowly; driven by the design |
| Engineering / NRE | CAM, stack review, coupon design, first article | Fixed, so per-board cost collapses with batch | Yes — once per part number, not per order |
| Tooling | Test fixture, probe adaptor, routing programme | Fixed, and reusable | Yes — ask who owns it |
| Add-ons | Impedance coupon, hybrid press cycle, special thickness | Per board or per lot | Yes — by design choices |
| Minimum charge | Setup time that a small order cannot absorb | Dominates below about 20 boards | Yes — by consolidating orders |
The Unit Price Is the Line with the Least Room
The unit price is where negotiation starts and where it delivers least. On a microwave board roughly 60–70 % of the unit cost is material and the rest is processing, test and yield, and none of those three moves because a buyer asks. A 5 % reduction in the laminate price moves the unit price by about 3 %; a design change that removes one press cycle moves it by 15 % or more.
This is why the productive conversation about unit price is a design conversation. Three questions expose the room available, and all three can be answered before the first panel is cut. Does the stack need two press cycles, or can a hydrocarbon-compatible prepreg bond the same layers in one? Is the PTFE confined to the layers that carry the field, or is it everywhere because that was simpler? Is the finish ENIG because the assembly needs it, or because it is the default? The electrical cost of getting that one wrong is set out in PCB surface treatment processes.
The answers land on the quote as concrete differences. A four-layer hybrid of FR-4 and RO4350B avoids a second press cycle and cuts the material term by roughly two thirds compared with full PTFE, for the same board outline. On a batch of 100 that is not a discount; it is a different product. The structural side of the same decision is in hybrid dielectric PCB technology.
One-Time Engineering: The Line That Shrinks Fastest
NRE covers the work that happens once per part number: CAM engineering, stack review, impedance model check, coupon design, first-article measurement and the press trial. It is a fixed number, and because it is fixed, its contribution per board follows a simple inverse: divide by the batch.
| Batch size | NRE per board, if NRE equals 30 board-units | Share of the total for a board priced at 1.0 |
|---|---|---|
| 10 | 3.0 | about 75 % — the order is mostly engineering |
| 50 | 0.6 | about 38 % |
| 100 | 0.3 | about 23 % |
| 500 | 0.06 | about 6 % |
| 2,000 | 0.015 | about 1.5 % |
Two practical consequences. First, small orders are dominated by NRE, which is why a 10-piece prototype of a microwave multilayer can cost several times the unit price of the same board at 500 pieces — not because anyone is overcharging, but because the engineering is real. Second, NRE should be paid once per part number, not once per order: a follow-on order of the same revision with the same stack should not attract a second engineering charge, and if it does, the question to ask is what changed.
There is a legitimate exception. If the stack changes, the coupon changes, and the model has to be re-checked; if the press cycle changes, the first article has to be repeated. Those are new engineering, and they should be listed as such rather than buried in the unit price. The reverse is also true: an NRE line that is charged but not itemised is worth challenging, because the work it claims to pay for — stack review and coupon design — is exactly the work that prevents a 20 % impedance error later.
Tooling: Ask Who Owns It
Tooling on a microwave board is usually a test fixture or a probe adaptor rather than a mechanical die. A bare-board test fixture with the right pitch and probe geometry is a real cost, and it is reusable for the life of the part number. Two questions decide whether it is a cost or an investment: who holds the fixture between orders, and whether a second source can use it.
If the fixture is held by the fabricator and the buyer changes supplier, the tooling is lost and paid for twice. If the buyer owns it, the fixture travels. This is worth putting in the purchase order explicitly, because it is the one line in the quote whose value does not depend on the design, the batch size or the material.
Where a probe adaptor is required for impedance measurement, ask what it measures. A fixture that measures a coupon at a fixed frequency is cheaper than one that supports a full S-parameter sweep, and for most acceptance purposes the coupon value is what matters. Paying for a measurement capability that the drawing does not require is a common and invisible fee, and it is the fixture side of the calibration question discussed in the production windows article.
Add-Ons: The Four That Appear on Microwave Quotes
Add-on fees are where a microwave quote differs most from an FR-4 quote, and where the largest avoidable spend sits, because every one of them is a consequence of a design decision.
| Add-on | Typical form | Physical cause | How to remove or reduce it |
|---|---|---|---|
| Impedance coupon and measurement | Per lot, plus a per-point charge for VNA work | Coupon area plus instrument time | Define one coupon per layer, not per net; specify the frequency band |
| Hybrid press cycle | Per lot | A second press cycle at a different temperature | One bonding system per stack; PTFE only where the field is |
| Special thickness | Per sheet, with a minimum quantity | A laminate that is not a stock item | Stay inside the stocked thickness set; adjust W/h instead |
| Tight impedance window | Percentage uplift on unit price | Slower press cycle, more coupon points, more scrap | Ask what ±10 % costs versus ±5 % for this stack |
The tight-window line deserves a comment because it is the most misunderstood. A ±5 % impedance window is not a matter of measuring more carefully; it requires a tighter dielectric thickness, a slower press cycle and a narrower etch window, all of which reduce throughput. On a 0.254 mm stack, the thickness tolerance alone consumes about ±6 %, so a ±5 % specification is only realistic with a selected laminate or a measured press — the thickness term is quantified in common impedance designs and layer stacks. Asking for it because the design has margin is the most common way to pay for tolerance that nobody uses.
The Minimum Charge and the Small-Order Reality
Below a certain quantity, a quote stops being a unit price and becomes a setup charge with some boards attached. The reason is that the fixed work — CAM, tooling, press set-up, first article — does not shrink, and a press cycle costs the same whether it carries two panels or twenty.
The practical threshold on a high-frequency multilayer is around 20 boards for a small outline and lower for a large one, because the constraint is panel count rather than board count. Two consequences follow for a buyer. Consolidating two part numbers into one order can move both above the threshold, which is usually worth more than a percentage discount. And a prototype order should be priced as engineering plus boards, not as boards alone; a quote that hides the engineering in an apparently low unit price at quantity 10 is not cheaper, it is just harder to compare.
Expedite Fees Have a Structure, Not Just a Percentage
Fast delivery is sold as a percentage uplift, but the cost behind it is not a percentage: it is a dedicated press slot, a priority position in drilling, and engineering attention diverted from production lots. That is why a 24-hour sample service exists at all — it is a capability built around a small number of boards, not a discount scheme applied to a large order.
| Lead time class | What it consumes | Typical uplift | When it is worth paying |
|---|---|---|---|
| 24–48 h samples | Dedicated slots, hand-carried through each station | Highest | Design verification before tooling release |
| 3–5 days | Priority in the queue, standard process | Moderate | A build that must be in assembly this week |
| Standard | Normal scheduling | None | Anything that can wait; most production |
The cheaper alternative to an expedite fee is almost always a smaller first order on a standard lead time, used to validate the stack and the coupon, followed by the production batch. That sequence buys the same information for less money, because it moves the risk to a point where the board is cheap rather than to a point where the schedule is expensive.
How Yield Risk Is Priced Into a Quote
A quote contains a yield assumption whether or not it is written down, and the assumption is visible in the price of the first order. A fabricator quoting a stack he has built before prices at a mature yield; one quoting a stack with an unusual bonding film or a 6:1 aspect ratio prices the risk of learning it.
The buyer’s lever is to reduce the learning. Three things do that: use a stack the fabricator has in his process library, keep the aspect ratio inside 6:1 on PTFE, and avoid a bonding system that forces a press cycle the plant does not run routinely. Each of these reduces the yield risk that has to be priced in, and none of them changes the electrical performance if they are applied at the stack drawing rather than after layout.
The other side of the same coin is what happens when a lot fails. A price that assumed 90 % yield and delivered 78 % is a loss for the fabricator, and it reappears as either a surcharge on the next order or a reluctance to quote the stack again. The same exposure runs through the copper line of the quote, as the copper price guide describes. Discussing the expected yield openly at quotation is not a concession; it is the mechanism by which the second order is priced sensibly. The defects that drive that number are catalogued in common scrap causes.
Normalising Three Quotes Onto One Board
Comparing quotes without normalising them is the most common purchasing error on high-frequency work, because two suppliers can be 40 % apart and both be correct for what they quoted. The way to compare is to force every quote onto the same specification and the same quantity, then convert the fixed items to a per-board figure.
| Item | Supplier A | Supplier B | Normalised at 500 boards |
|---|---|---|---|
| Unit price | 1.00 | 0.86 | 0.86 — B is cheaper on the board itself |
| Engineering (fixed) | 0.00 (in unit price) | 25 board-units | +0.05 for B |
| Tooling (fixed) | 10 board-units | 0.00 | +0.02 for A |
| Coupon measurement | included | 2 board-units per lot | +0.004 for B |
| Effective cost per board | 1.02 | 0.914 | B is 10 % lower, not 14 % |
The exercise usually changes the ranking and always changes the size of the gap. It also exposes the assumptions: a supplier who quotes “impedance included” may be planning one coupon per panel rather than one per layer, and a supplier who quotes the lowest unit price may be assuming a ±10 % window when the drawing says ±5 %. Getting those two assumptions written into the quote is worth more than the discount being negotiated.
The Levers, Ranked
| Lever | Effect on total | Cost of using it | When to use it |
|---|---|---|---|
| Re-specify the stack | 15–30 % | Design time, one model re-check | Before tooling; the single largest lever |
| Consolidate part numbers | 5–12 % | None, if the schedules can move | Small or medium batches below the threshold |
| Relax a tolerance nobody uses | 3–8 % | Engineering review of the real margin | When the drawing says ±5 % but the system tolerates ±10 % |
| Match the finish to the link | 2–5 % | None, if the assembly allows OSP | Microwave boards with many launches |
| Batch-size negotiation | 3–6 % | Inventory and cash | Stable demand for a released revision |
The ordering is the same as in the internal view, for the same reason: the levers that act on the physical build are larger than the levers that act on the commercial relationship. A buyer who re-specifies the stack once gets more than a buyer who negotiates every order for a year, and the first conversation is much shorter.
What Should Be in the Purchase Order
Most fee disputes come from what was not written down. Seven items belong in the order, and each one closes a specific argument.
1. The stack revision, so that a changed laminate is a change order rather than a surprise. 2. The impedance target per layer with the coupon geometry and the test method, so acceptance is a measurement rather than an opinion. 3. The acceptance standard — IPC-A-600G and IPC-6015 with the class stated — and which one governs if they disagree. 4. Tooling ownership, so that a change of supplier does not require paying for the fixture twice. 5. What is included in the unit price — electrical test, coupon measurement, number of coupon points. 6. The NRE scope, so a follow-on order of the same revision is not charged again. 7. The date basis, meaning whether the quoted lead time starts at purchase order or at receipt of working data.
None of these seven is unusual, and all seven are routinely omitted. The one that causes the most trouble in practice is the third: a board that meets the electrical requirement but fails a workmanship criterion is a board that gets questioned, and the answer depends entirely on which document the order said would govern.
Conclusion
The cost structure of a microwave board as seen from a purchase order is five lines, and only one of them is the price of a board. Engineering is fixed and collapses with batch size; tooling is fixed and should be owned, not rented; add-ons are the shadow of design decisions; the minimum charge is the fixed work that a small order cannot absorb. The unit price itself is the least negotiable line, because 60–70 % of it is material and the rest is the process and yield that the design already fixed.
The ranked levers follow from that. Re-specifying the stack beats consolidating orders, consolidating beats relaxing tolerances, and all of them beat negotiating the unit price. And the comparison of three quotes is only meaningful once the fixed items are converted to a per-board figure, because a 14 % gap in unit price can be a 10 % gap in cost — or no gap at all, once the engineering and the coupon scope are aligned.
XCE PCB quotes high-frequency and microwave work with the five lines itemised: 2 to 72 layers, PTFE, hydrocarbon and ceramic laminates, 600×1200 mm panels, impedance ±10 %, samples in as fast as 24 hours, inspected to IPC-A-600G and IPC-6015. The process side of the same numbers — which station or which tolerance the fee is paying for — is in the manufacturing cost structure, and the design decisions that precede both are in the twelve production decisions.
Frequently Asked Questions
Why is a 10-piece prototype so much more expensive per board than a 500-piece order?
Which line of a microwave PCB quote is most worth negotiating?
What does a ±5 % impedance window cost compared with ±10 %?
Should tooling be owned by the buyer?
How should three quotes for the same board be compared?
Does a fast lead time always cost a percentage premium?
Want a Quote Broken into Its Five Lines?
Send us your Gerber files and requirements for a manufacturability review and a tailored quotation.
sales1@xcepcb.com
+86 180 3341 4657