Design Is Only Half the Story: How 14 Engineering Functions Interlock to Take an Optical Module from Schematic to Volume Production
An 800G optical module leaves the factory as a palm-sized metal box. Lay out the engineering organization of an optical module maker, though, and you find fourteen different kinds of engineers standing between nothing and that box: seven on the design side, seven on manufacturing and quality. Exactly half and half.
This piece is not about contracts, quotes, or stock prices. It does one thing: take apart the engineering chain from schematic to volume shipment, look at what each function receives, what it produces, and who it argues with, and find the four interfaces where design and manufacturing truly meet.
Full Table of Contents
1 | Why the design-to-production chain is worth taking apart
2 | Six gates from schematic to volume production
3 | Design side: seven kinds of engineers competing for the same budget
4 | The moment of design freeze: design review and design verification
5 | Manufacturing and quality: where the most expensive defects happen
6 | Design and manufacturing really meet at only four interfaces
7 | One problem, three solutions, a 10x cost difference
8 | Cross-functional trade-offs: someone pays for every number
9 | Why the gap between Taiwanese and Chinese makers is not on the design side
10 | Conclusion | Unlock with a subscription
The design side is only half the headcount, yet it sets the difficulty for the other half. Every tolerance, every watt of power, every margin that design hands over ultimately turns into yield and cycle time on the production line.
The most expensive defects happen at the optical coupling station, because that is where parts are scrapped after expensive chips have already gone in. The same 5% defect rate costs an order of magnitude more in gross margin at the coupling station than at incoming inspection.
Design and manufacturing really meet at only four interfaces: tolerances, the power closed loop, the tuning interface, and test thresholds. How these four lines are negotiated determines how fast a factory ramps yield.
1. Why the Design-to-Production Chain Is Worth Taking Apart
For the past few years, conversations about optical communications have centered almost entirely on "speed jumps" and "who won the orders." 800G volume, the 1.6T ramp, whether to follow CPO: every headline has been about technology roadmaps.
But technology roadmaps are essentially public. MSA specs, IEEE standards, and OIF electrical interfaces are all in public documents, and any well-funded company can buy the same DSP, the same EML, the same connector. What really separates companies is the ability to assemble these public building blocks into a product that ships steadily and continuously. That is engineering process capability, not a technical secret.
That is why, among companies that all announced 1.6T, some ship reliably within a year while others are still stuck at samples three years later.
(If the spec codes themselves are still unfamiliar, start with What Does OSFP 800G SR8 Actually Mean? Five Axes of Optical Module Naming.) Or try our earlier optical module spec lookup tool (in Chinese)
That wraps up the key points of this piece.
STT's full analysis — how the four real design-manufacturing interfaces are negotiated, why three solutions to the same problem differ 10x in cost, who actually pays for each spec number, and why the gap between Taiwanese and Chinese makers can't be closed by poaching a few design engineers — is in the paid section.
Subscribe to STT's paid section and read the full analysis on vocus (in Chinese): https://vocus.cc/article/6a816bfafd897800017257ff
Related Reading
[CPO Deep Dive 1/6] Have Pluggable Optical Modules Hit a Bottleneck? Understand This Limit Before You Understand CPO: where pluggable modules top out, picking up directly from the power and thermal trade-offs discussed here.
Why Is Silicon Photonics So Hard to Design? The Answer Lies in a Box Called the PDK: how design work changes when optical design moves from discrete components to silicon photonics.
[CPO Deep Dive 4/6] Fiber-to-Chip: The Least Glamorous Step in CPO That Holds Everyone Back: takes the coupling challenge from Section 5 one level further into the CPO generation.
Comments