SMTPA – Surface Mount Technology Product Assembly
This post is being
written in response to Mr Stephen Hawes excellent article ‘Mid-Scale Manufacturing’. It is directly related to a challenge I am facing
at work. How do I make four thousand of something, without making
twenty batches of two hundred? I have designed and built the jigs for hand assembly, trained the staff and assembled 3000 pieces so far. Will the next 4000 simply be more staff hours and more jigs? The next 10000? A giant pick and place machine will solve the problem.
Philosophy
- currently missing middle-scale automation stack, a gap is present.
- linear relationship between staff hours and units produced, until offshore threshold reached
- ROI threshold for onshore automation based on 90’s tech, SCARA arms etc, that were scaled down versions of high-volume high-cost machines, wrong base assumptions / starting point
- open toolchains now exist for quick deployment of precision component location with decent UX; OpenPNP, Klipper, the whole family of GRBL derivatives, etc
- toolchains popularity has driven hardware prices down for a decent motion platform
- open toolchains make hiring / troubleshooting easier, more transparent and accessible
- start slow when deploying internally to limit risk; ie the charger accessory for a product after a manual process has been established / has had its costs made known
Practice
- Design products as placeable assemblies
- accept and support a slower production speed per work cell, then scale in parallel when capacity increase is required
- product design to stay within community documented/debated walled garden, just as pcb design does
-
MAY
break a DRC but that had better be the
killer-feature of the product because engineering that assembly step
will involve more risk/time; stepping outside the garden
- source components on tape-and-reel or tray; a uniform and understood feed mechanism (EIA-481, 948, 960) (Thanks Stephen!)
- can then leverage existing open source software and hardware intended for pcb assembly to assemble the entire product
- only assemble with open source hardware; financing supply chain while demanding transparency and community inclusion
- maintain documentation and interoperability, ease of access and staffing for engineering/assembly techs
Rules for SMTPA Products
- all components
supplied in standard feeders, as tape or trays, this includes weird
stuff like the plastics and batteries
- Vertical assembly of components, or at least, vertical loading of assembly stations in the work cell
- certainly more to be defined, how did the SMT placement / sourcing rules evolve from zero to JLC making a last minute conference badge for $8 with only a couple hours of engineering design labour?
- cell transfer process definitions; sub assembly moving from one cell to another, from FDM printer to assembly cell, etc. Must be positioned and delivered according to spec, fiducial etched to printed part for location during assembly, required in component model.
ProblemS(olving)
- Strict focus on targeted risk-management; give the problem an ID number and tackle it with the community, the worst problems will float to the top and get the most attention; have the most positive impact when solved
- supply chain discussion required with Digikey, McMaster-Carr and others; can we get everything in SMT packaging? What will handing exceptions look like? Cost to force exceptions into packaging to comply with toolchain too high?
- not being applied to whole-product assembly yet due to risk for startups, inertia of existing solutions however poor their fit, startup’s focus on their product’s risk and aversion to taking on any manufacturing risk
- kitting / packaging will be a mess
Case Studies
- Assembly of wearable safety device charger, 4000pcs quantity, near term requirement
- Motorcycle ECU, 1000pcs quantity, no fixed dates on delivery
Closing Thoughts
Why hasn’t this been done? Why don’t we see this everywhere? I
suspect that at thousand-plus volumes, risk management has dictated
more traditional approaches to manufacturing. ‘No one ever got
fired for choosing IBM’ has been the default since the risky
product at their organization was the widget and not the assembly
technique. Please help solve this problem, head to https://openpnp.org/ and get chatting in #mid-scale-mfg (ping me, L2K20) at https://discord.gg/Hc8SjyfRDj or leave a comment here about how a board-in-a-box is weak sauce and I should think bigger!
