Publish Time: 2025-12-10 Origin: Site
In modern SMT production, Complete Guide to SPI Machines consistently proves one unbreakable rule: SPI always comes before AOI. Getting this order wrong is the single most expensive mistake a factory can make, because 55–70% of all reflow defects start at solder paste printing — long before components are placed.
Today’s PCBs routinely carry 01005 resistors, 0.3 mm pitch BGA, and multi-layer stacked packages. A solder paste deposit that is just 10 µm too low can cause an open joint after reflow, while 5 µm too much can create a bridge under a 0.4 mm QFN. These tolerances are far beyond what the human eye or traditional 2D cameras can reliably catch, which is why automated 3D inspection has become non-negotiable in modern electronics manufacturing.
Many engineers and managers inherited production lines that were built 10–15 years ago when AOI was the only automated inspection available. Those lines still work (sort of), so the natural question becomes: “If AOI already looks at the finished board, do we really need another machine earlier in the line?” Meanwhile, younger process engineers who trained on Six-Sigma and CpK watch the same printing defects repeat month after month and wonder why the factory is spending thousands on rework instead of preventing the problem at the source.
SPI (Solder Paste Inspection) is installed immediately after the stencil printer and before the first pick-and-place machine. It uses structured light or laser to create a true 3D map of every single solder paste deposit. Within seconds it measures volume (nL), height (µm), area (mm²), X/Y position, and shape for every pad on the board. If anything is out of tolerance, the board is rejected or the printer receives real-time closed-loop correction before the next board is printed.
AOI (Automated Optical Inspection) sits after the reflow oven. It takes high-resolution 2D or 3D color images of the fully assembled board. It checks for missing parts, wrong parts, reversed polarity, tombstoning, lifted leads, insufficient solder, bridges, and visible wetting problems. Because the solder has already melted, AOI can only tell you what went wrong — it cannot prevent the defect from happening in the first place.
SPI is preventive medicine: it stops bad solder paste from ever meeting a component. AOI is the autopsy: it tells you which boards are already dead or dying. One saves you money upstream, the other saves your customer from receiving bad product downstream. Both are important, but they are not interchangeable.
Many older consumer-electronics factories still run AOI-only lines because “that’s how we’ve always done it.” These lines typically produce simple double-sided boards with 0603/0402 components and 0.5 mm+ pitch. Printing is considered stable enough, rework is cheap, and management hates adding new machines. The result is acceptable for low-cost products, but defect rates quietly sit at 500–2000 ppm.
Process-focused engineers — especially in automotive, medical, and telecom — treat solder paste printing as the most critical and most variable step in the entire line. They know that once the paste is wrong, no amount of perfect placement or perfect reflow profile can save the joint. Their mantra is “measure and correct the paste before you spend money placing expensive components on it.”
Leading contract manufacturers and OEMs now treat SPI + AOI the same way they treat printer + pick-and-place: you simply do not build a serious line without both. The investment is justified by first-pass yield numbers that routinely exceed 99.5 % and rework costs that drop by 60–80 %. In these factories the debate is no longer “SPI or AOI?” but “Which SPI model gives us the fastest ROI?”
IPC-7912, iNEMI, and dozens of independent studies over the past 15 years consistently show the same breakdown: solder paste printing accounts for 55–70 % of all assembly defects, placement 10–15 %, reflow 10–15 %, and everything else the remainder. Even a perfectly tuned pick-and-place machine cannot overcome bad paste volume or offset.
Fixing a printing defect at SPI costs virtually nothing — the board is simply cleaned and reprinted. Fixing the same defect at AOI after reflow requires manual touch-up, possible component removal, X-ray verification, and re-reflow — easily 20–50× more expensive. If the defect escapes to the customer, the cost can jump to hundreds or thousands of dollars per board in warranty claims and lost reputation.
Too little paste → insufficient fillet height → open or weak joint. Too much paste → excess solder balls or bridges under fine-pitch devices. Paste offset by 50 µm → tombstoning on small chip components. Height variation → voids inside BGA balls that AOI cannot see but X-ray will later find. Every one of these failures is 100 % predictable from the 3D paste data that only SPI provides.
SPI runs before any component is placed, so it has no way of knowing whether the pick and place machine later grabbed the wrong reel or skipped a part entirely. Polarity errors on polarized capacitors or diodes are also invisible to SPI because the paste looks identical regardless of orientation.
Even with perfect paste, a nozzle can drop a part 100 µm off-pad, or uneven heating can cause tombstoning during reflow. These mechanical shock or poor vacuum can lift a lead on a QFP. SPI sees none of these because they happen long after its inspection window.
Head-in-pillow, non-wetting, dewetting, and some types of voiding only become visible after the solder has melted and cooled. AOI’s color cameras and angled lighting are specifically designed to catch these surface-level issues that SPI never gets a chance to see.
The only sequence used by world-class factories today is: stencil printer → SPI → high-speed chip shooter → flexible placer → reflow oven → AOI → (optional X-ray or ICT). This order is not arbitrary. It follows the natural defect-creation timeline: first prevent printing problems, then prevent placement problems, then verify the final result after soldering. Reversing any step dramatically increases rework and escape risk.
Modern SPI systems like the I.C.T-S510 and I.C.T-S1200 send real-time offset and volume data back to the printer (closed-loop control). The printer automatically adjusts squeegee pressure, speed, or stencil cleaning frequency on the next board. Within 3–5 boards the process typically settles to CpK > 1.67. Once printing is locked in, the pick-and-place machines receive perfect pads every time, dramatically reducing placement-related alarms downstream.
With printing already under control, AOI’s job becomes much easier and more accurate. False calls drop 60–80 % because AOI no longer has to guess whether a marginal solder joint is caused by bad paste or bad placement. AOI can now focus on true placement errors and post-reflow issues, becoming a true final gatekeeper instead of a catch-all troubleshooting station.
Double-sided consumer boards with 0603 and larger parts, pitch ≥ 0.5 mm, very stable stencil and paste, low-mix high-volume runs, and relaxed quality targets (≤ 1000 ppm) can sometimes survive with AOI alone. Rework is inexpensive, field failures are rare, and management accepts the occasional touch-up station. These lines are becoming rarer every year, but they still exist in cost-driven markets.
Automotive electronics (AEC-Q100/104), medical devices (ISO 13485), aerospace/military (IPC Class 3), 5G infrastructure, server motherboards, anything with 01005/008004 components, ≤ 0.4 mm pitch BGA, or bottom-terminated packages all require 3D SPI. Zero-defect policies and warranty costs in the thousands of dollars per board leave no room for “we’ll catch it at AOI.”
Even factories with tight capital can justify SPI first. Typical payback is 6–12 months through scrap reduction, rework labor savings, and yield improvement alone. Many customers report that adding SPI cut their AOI rework stations from three shifts to one shift and reduced customer returns by 90 %. The math is simple: preventing one bad pallet of automotive PCBs pays for the entire SPI machine.
2D SPI only measures area and can be fooled by paste height variations. True 3D SPI (phase-shift moiré or dual-laser triangulation) measures actual volume and height with ≤ 1 µm resolution. For anything smaller than 0402 or 0.5 mm pitch, 2D is obsolete and will generate excessive false rejects or misses.
Look for ≥ 2 µm height resolution, GR&R < 10 % at 6σ, and inspection time ≤ 12 seconds for a typical smartphone PCB. The I.C.T-S510 achieves 8–10 seconds per board at 1 µm resolution, while the larger I.C.T-S1200 handles 600 × 600 mm panels in under 20 seconds with the same precision.
Modern SPI must import Gerber and CAD data directly, auto-generate inspection programs in minutes, display real-time CpK charts, and send correction values back to DEK/Minami/Panasonic/GKG printers automatically. Without these features you are buying yesterday’s technology.
Choose machines with fully automatic glass-plate calibration (30-second daily routine), temperature-compensated optics, and sealed projection units. The I.C.T-S510 and I.C.T-S1200 both include these features and maintain < 1 µm repeatability year after year with minimal operator intervention.
No. AOI inspects after reflow, when the damage is already done. It cannot measure solder paste volume or height before components are placed, so it cannot prevent cold joints, bridges, or voids caused by printing errors.
For 0402 and larger components on 0.5 mm+ pitch, 2D can sometimes survive. For 0201, 01005, 0.4 mm or finer pitch BGA, only 3D SPI provides the volume and height data required by IPC-7095 and automotive standards.
Yes — typically 60–80 %. Stable printing removes the random volume variations that confuse AOI algorithms and generate phantom solder-joint defects.
Modern systems like the I.C.T-S510 inspect a typical smartphone PCB in 8–10 seconds and the I.C.T-S1200 handles large panels in < 20 seconds. These times are negligible compared with placement and reflow cycle times.
Yes. IPC-7095D (BGA) and most automotive/medical quality standards effectively mandate 3D SPI to guarantee void rates < 25 % and reliable wetting on ultra-fine-pitch devices.