For example, we are in a same situation:
| Situation | Go with |
|---|---|
| Your parts are flat or thin, tolerances in the ±5–10 µm range, and you need to measure 50+ dimensions per piece | Vision Measuring System (VMM) |
| You need true 3D data: bore depth, flatness, perpendicularity, or GD&T on machined castings | Coordinate Measuring Machine (CMM) |
| Your parts are soft, thin-walled, or easily scratched | VMM (non-contact) |
| You need to verify a single critical Z-feature on an otherwise 2D part | Consider a composite system, or use CMM for that one feature only |
Table of Contents
- What You’re Actually Comparing
- How Each One Works (And Why “Optical CMM” Is a Confusing Label)
- Side-by-Side Comparison Table
- Six Decision Gates — Answer These Before You Look at Any Quote
- Part-by-Part Fit Matrix (10 Real Cases)
- Accuracy Claims: What You Can Trust and What You Can’t
- Total Cost of Ownership — The Numbers Nobody Puts in the Brochure
- Composite Systems: When One Machine Isn’t Enough
- Your RFQ Checklist: 8 Things to Send Suppliers Before You Ask for a Quote
1. What You’re Actually Comparing
Let me start with the one sentence that would have saved me two weeks of research when I first looked into this:
A vision measuring system is a 2D/2.5D tool that excels at flat, thin parts with many features. A CMM is a true 3D tool built for volumetric geometry and tight GD&T.
That’s the whole story. Everything else is detail.
A Vision Measuring System (VMM) — sometimes called a video measuring machine or optical comparator’s bigger, smarter cousin — uses a camera, telecentric lens, and programmable LED ring light to capture edges optically. It counts pixels. No probe touches the part. You get X and Y coordinates with sub-micron repeatability, and a derived Z from autofocus or laser attachment.
A Coordinate Measuring Machine (CMM) uses a physical probe — touch-trigger, scanning, or optical — to collect 3D point clouds. It moves in X, Y, and Z on a bridge, cantilever, or gantry frame. It is the only tool in this comparison that can verify a bore’s true position relative to a datum structure, or confirm that a milled pocket is perpendicular to a reference face within 8 µm.
The mistake most buyers make: they compare them as if they compete directly. They don’t. They overlap at the edges, and that overlap is exactly where you need to be careful.
2. How Each One Works (And Why “Optical CMM” Is a Confusing Label)
2.1 Vision Measuring System — The Basics
A VMM sits a camera over a moving stage. You place the part, the software finds edges using contrast gradients, and it calculates dimensions from pixel coordinates. Modern systems add:
- Telecentric lenses (no perspective distortion, even at the edges of a large FOV)
- Programmable ring lighting (dark-field, bright-field, multi-sector — you dial in contrast per feature)
- Auto-zoom (overview at low mag, detail at high mag, on the same part, no repositioning)
- Laser attachments for Z (typically confocal or line-scan, ±1–2 µm in practice)
What it does brilliantly: measure 100+ 2D features on a flat part in one shot. O-ring ID/OD, gasket profiles, stamping outlines, PCB pad spacing, connector pin locations — these are VMM territory.
2.2 CMM — The Basics
A CMM touches the part with a probe. Each touch returns an X/Y/Z coordinate. String enough touches together and you can reconstruct any surface. Add a scanning probe and you get thousands of points per second along a curve.
What it does that nothing else does: true 3D form evaluation. Flatness of a machined face. Perpendicularity of a bore to a mounting surface. Position tolerance called out to a composite datum scheme. These are CMM-only tasks.
2.3 “Optical CMM” — What People Mean and Why It Causes Bad Purchases
Search for optical cmm and you’ll find three different things:
| What someone calls it | What it actually is | Can it measure a blind hole depth? | ISO 10360 certified? |
|---|---|---|---|
| Vision Measuring System | Camera + stage, 2D/2.5D | No | No (no applicable standard) |
| Multisensor system | VMM + touch probe + laser on one bridge | Yes | Yes (per sensor) |
| White-light scanner | Fringe projection, full-field 3D | Yes (but no access to deep small holes) | Sometimes (ISO 10360-7/8) |
The buyer’s rule: If a salesperson calls their machine an “optical CMM,” ask two questions:
- Can it measure a blind hole 3 mm deep and 1 mm in diameter?
- Is it certified to ISO 10360-2 (the CMM standard), or to -7/-8 (the optical standard)?
If the answer to #1 is no and #2 is “-7/-8,” you’re buying a VMM, not a CMM. That’s fine — if your parts are flat. It’s a very expensive mistake if they aren’t.
3. Side-by-Side Comparison Table
| Attribute | Vision Measuring System | CMM (Bridge/Traveling) |
|---|---|---|
| Dimensional domain | 2D dominant; 2.5D with laser | Full 3D |
| Contact | Non-contact (optical) | Contact (probe) standard; optical options available |
| Best part type | Flat, thin, soft, fragile, high-feature-count | Volumetric, cast, machined, molded 3D, GD&T-heavy |
| Typical repeatability (XY) | 0.5–2 µm | 0.5–1.5 µm (MPES) |
| Z capability | 2–5 µm with laser; autofocus ~3–8 µm | 0.5–1.5 µm (touch probe) |
| Throughput | 30–100+ parts/hour (automated edge detection) | 2–15 parts/hour (depends on probe path complexity) |
| Programming | Learn-mode or CAD import; many systems measure without programming | DMIS or vendor language; requires probe path planning |
| Blind holes / undercuts | Cannot access | Standard capability |
| Soft/thin parts | Ideal — no deformation | Risk of probe pressure deformation; requires low-force probes |
| Environmental need | Temperature-stable room preferred; less sensitive than CMM | 20 ± 1°C (often ±0.5°C) strict for quoted accuracy |
| Floor space / foundation | Granite base, no special foundation | Granite table, sometimes air-isolated foundation |
| Entry price (approx.) | $25k–$120k | $60k–$400k+ |
4. Six Decision Gates — Answer These Before You Look at Any Quote
This section is the reason this article exists. Don’t skip it.
Gate 1: What Dimensionality Does the Drawing Demand?
| If your drawing calls out… | Winner | Why |
|---|---|---|
| Length, width, radius, angle, hole pitch, profile tolerance in one plane | VMM | These are pure 2D. VMM is faster and often more repeatable than a CMM’s 2D probing. |
| Flatness, perpendicularity, position relative to a 3D datum tree, profile of a surface | CMM | No camera system can verify these without a touch probe or structured light — and even then, deep features are invisible to optics. |
| A mix: 90% 2D + one Z-check (e.g., coplanarity of 20 pins) | Depends — see Gate 6 | Read on. |
Red line: If your customer’s drawing has a position tolerance tied to a composite datum (A|B|C) with a 3D modifier, stop looking at VMMs. You need a CMM or a composite system. No software upgrade changes this.
Gate 2: What’s the Tightest Tolerance You Actually Have to Hold?
Don’t answer with the print’s tightest number. Answer with the one your QC manager loses sleep over.
| Tolerance band | VMM reality | CMM reality |
|---|---|---|
| ±25 µm and above | Comfortable | Comfortable |
| ±10 to ±25 µm | Sweet spot for VMM | Comfortable |
| ±5 to ±10 µm | Possible with high-end VMM, telecentric optics, tight environment | Standard |
| Below ±5 µm | Risky — verify with on-part benchmark | CMM is the right tool |
Red line: If your critical dimension is ±3 µm or tighter and it’s a 3D feature, VMM is not the answer. Don’t let a salesperson’s spec sheet tell you otherwise.
Gate 3: What’s the Part Made Of? (The Non-Contact Advantage)
This is where non-contact measurement stops being a buzzword and becomes the reason you buy a machine.
| Material | VMM fit | CMM fit | Why it matters |
|---|---|---|---|
| LCD glass, silicon wafer, optical film | Excellent | Poor — probe scratches | No contact = no damage |
| Battery electrode foil (copper/aluminum, 0.1 mm thick) | Excellent | Probe dents or tears | Even a 0.1 N probe force deforms foil |
| O-rings, gaskets, soft rubber | Excellent | Probe rolls or compresses | Non-contact reads true profile |
| Injection-molded plastic (thin-wall) | Good | Acceptable with low-force | VMM faster for 2D profiles |
| Machined aluminum/steel block | Good for 2D, limited for 3D | Excellent | CMM owns volumetric |
| Cast iron, forged steel | Marginal | Excellent | Rough surface + 3D form = CMM |
Real case: A battery-tray stamping shop tried to use a CMM for incoming inspection. The probe tip left micro-dents on the 0.15 mm aluminum tabs. Parts passed CMM but failed downstream assembly because the dents shifted contact resistance by 12 mΩ. They switched to VMM. Problem gone.
Red line: If your part deforms under 0.5 N of probe force, you are in VMM territory. Period.
Gate 4: Can the Sensor Physically See the Feature?
Optics need a line of sight. If the feature is:
- Inside a deep, narrow pocket (depth > 3× diameter)
- On the back side of a wall
- Behind a rib or boss
…a camera can’t see it. Period.
| Feature type | VMM | CMM |
|---|---|---|
| Top-surface holes, pads, edges | ✓ | ✓ |
| Side-wall features (through open access) | ✓ (with oblique light) | ✓ |
| Blind hole depth > 5 mm diameter | Laser can estimate | ✓ Touch probe |
| Internal thread pitch | ✗ | ✓ (with thread probe) |
| Undercut | ✗ | ✓ |
Red line: If more than 20% of your critical features are optically inaccessible, VMM alone will not close the inspection loop.
Gate 5: How Many Parts per Hour?
| Throughput need | VMM | CMM |
|---|---|---|
| 1–5 parts/day (first-article, PPAP) | Slow for what it is | Right tool |
| 20–50 parts/shift (in-process) | Good | Possible with automation |
| 100+ parts/shift (sorting, 100% inspection) | Built for this | Not economical without multi-probe robot cell |
Red line: If you need 100% inspection on a stamping running 200 parts/hour, a CMM will never keep up. You need a VMM or an inline optical system.
Gate 6: Part Size and Stage Travel
| Part size | Consideration |
|---|---|
| Under 200 mm in any axis | Either system works; decide on Gates 1–5 |
| 200–600 mm | VMM works if flat; large-format VMM available |
| 600 mm to 1500 mm | Large-format VMM (gantry) or large CMM |
| Over 1500 mm | CMM or laser tracker; VMM impractical |
Red line: If your part is 800 mm long and you need flatness across the whole surface, a VMM’s autofocus Z-scan will take 20 minutes and still not match a CMM’s scanning probe for form data. Know the trade-off. If you’d like more data for your reference, view our products list.
5. Part-by-Part Fit Matrix (10 Real Cases)
Published specifications vary by configuration. Always confirm with a benchmark on your own part.
| # | Part type | Typical examples | Best fit | Why | What it can’t do |
|---|---|---|---|---|---|
| 1 | Large-format flat panels | LCD glass, large PCBs, battery trays, automotive sheet metal | Large-format VMM (gantry, open-bay, 850×1050 mm travel) | Open measurement space built for oversized flat/thin workpieces; 0.6–5× zoom covers overview and fine detail in one setup | Full-field 3D warp/camber across entire surface — needs CMM or laser scanner for that |
| 2 | Small precision parts, batch | O-rings, gaskets, watch cases, radar housings, injection-molded clips | Flash-measurement VMM (fixed platform, no programming) | Place part in any orientation; system auto-identifies features and measures 100+ dimensions in one field of view. Dedicated O-ring module gives ID/OD/wall thickness in ~1 second | No blind-hole depth; no 3D form |
| 3 | PCB & double-sided stamped parts | Connectors, precision stamping, FPC, dual-layer PCB | Dual-vision flash tester (wide-field + auto-zoom) | Wide-field lens for fast large-area scan; auto-zoom for micron-level detail on pads and pins. Movable platform breaks fixed-travel limit of standard flash testers | Z-height of through-hole barrels; buried vias |
| 4 | Pin spacing & burr detection | Optical modules, precision gears, mold inserts | High-precision VMM (~1.5 µm class as published) | Intelligent edge-detection algorithm rejects burr noise for clean pin-spacing measurement | Deep Z-features; internal thread |
| 5 | Coplanarity & spring-tab height | BGA/QFP leads, IGBT modules, connector tabs | Laser line-scan + 2D vision system | Laser line generates Z point cloud for coplanarity; vision handles 2D pitch. Built for µm-level coplanarity on leaded packages | Large volumetric parts |
| 6 | Machined housings & brackets | Mill/turn aluminum housings, steel brackets | CMM | True 3D: bore position to datum, flatness of mating faces, perpendicularity | — |
| 7 | Castings & forgings | Engine block, transmission case, forged yoke | CMM (scanning probe) | Rough surface, deep features, heavy mass — CMM handles it all | Surface finish measurement (need profilometer) |
| 8 | Gears & splines | Spur gear, worm gear, spline shaft | CMM (gear module) | Involute profile, lead, pitch variation — all 3D | — |
| 9 | Medical implants (flat profiles) | Stent markers, flat implant outlines, surgical blades | VMM (high-mag) | Edge definition critical; non-contact avoids contamination | Surface texture, 3D curvature |
| 10 | Thin-walled precision tube | Hypodermic needle, catheter marker band | VMM (backlight) or CMM with low-force | OD, wall, cutoff — VMM fast; CMM for concentricity to a datum | CMM needs special low-force probe; VMM can’t do concentricity to internal datum |
Two judgment calls worth highlighting:
- Case 4 vs. Case 5 crossover: If your need is mostly 2D pin spacing with occasional height checks, a high-precision VMM alone may be sufficient and more cost-effective. If coplanarity is a hard spec (e.g., IGBT lead coplanarity tolerance), lead with the laser-vision system as primary and treat the VMM as the general-purpose companion for 2D.
- Case 1 boundary: A 1200 mm battery tray is flat and thin — perfect for large-format VMM on the 2D profile. But if the tray has formed ribs with perpendicularity callouts, you’ll still need a CMM or a composite system for those features.
6. Accuracy Claims: What You Can Trust and What You Can’t
6.1 The Standards Gap
CMMs have ISO 10360-2 (and -5 for scanning). It defines MPEE (Maximum Permissible Error of Indication for size). A quoted MPEE of 1.2 + L/350 µm means you can verify that claim with a calibrated gauge block in a lab.
VMMs have no equivalent single-number standard. Some vendors quote ISO 10360-7 or -8 (optical distance sensors), but these test the sensor, not the full system. Others quote repeatability on a glass scale — which tells you about stage precision, not measurement accuracy on your part.
What to trust on a VMM spec sheet:
- Repeatability on a calibrated reference (glass scale, precision cross-grating) — this is real
- Reproducibility across temperature band — ask for data
- Edge-detection repeatability on a specific material — ask for a demo on your part
What to question:
- “System accuracy ±2 µm” without stating FOV, magnification, or temperature range
- Comparisons to CMM MPEE numbers — different standards, not comparable
- Z-accuracy from laser without stating surface reflectivity and angle
6.2 How to Challenge a VMM Quote
When a vendor says “1.5 µm accuracy,” ask:
- At what field of view and magnification?
- On what reference standard? (Glass scale? Cross-grating? Your part?)
- Over what temperature range?
- What’s the edge-detection algorithm’s performance on my material?
If they can’t answer, the number is marketing, not engineering.
6.3 Cross-Vendor Comparison Reality
You cannot fairly compare VMMs from different vendors using spec sheets alone. The only honest method: send the same 3 parts to each vendor, get a measurement report, compare to a CMM baseline. Anything else is guesswork.
7. Total Cost of Ownership — The Numbers Nobody Puts in the Brochure
| Cost element | VMM | CMM |
|---|---|---|
| Capital | $25k–$120k (typical); large-format to $200k+ | $60k–$400k+ |
| Installation | Granite base leveled, power, done | Foundation sometimes required; temperature-controlled room |
| Environmental | 20 ± 2°C recommended | 20 ± 0.5°C often required for quoted accuracy |
| Programming labor | 2–8 hours per part family (learn mode helps) | 8–40 hours per part (probe path, clearance, DMIS) |
| Operator skill | Moderate — lighting setup is the art | High — probe selection, path planning, datum structure |
| Probe/stylus cost | None (non-contact) | $200–$2,000 per stylus setup; wear items |
| Maintenance | Lens cleaning, light calibration, stage lube | Probe recert, air filtration (if air-bearing), granite resurfacing |
| Throughput value | High — one operator runs multiple shifts | Lower — each cycle is minutes, not seconds |
| 3-year TCO estimate | Capital + ~15% annually | Capital + ~20–25% annually |
The hidden cost on CMM: Programming time. A complex 5-axis machined housing can take 3 days to program properly. If you’re running 10 part families, that’s 30 days of engineering time before you measure a single production part.
The hidden cost on VMM: Lighting setup. Every material and finish needs a different light recipe. This is skilled work, but it’s faster to learn than CMM programming.
8. Composite Systems: When One Machine Isn’t Enough
A composite (multisensor) system puts a touch probe, camera, and sometimes laser on the same bridge. It sounds like the perfect answer — and sometimes it is.
When it makes sense:
- 70%+ of features are 2D, but 2–3 critical Z-features need touch probing
- You’re measuring connectors: pin pitch (vision) + coplanarity (laser or touch)
- Part size fits the composite’s travel (most are mid-format)
When it doesn’t:
- 90% of your work is 2D — you’re paying for a CMM bridge you barely use
- 90% of your work is 3D — the vision head is a $30k paperweight
- Your parts are large-format flat — a gantry VMM is cheaper and faster
Honest rule: If you can split your workload cleanly — 2D parts on a VMM, 3D parts on a CMM — two dedicated machines beat one composite 80% of the time. Composites win when the same part needs both.
9. Your RFQ Checklist: 8 Things to Send Suppliers Before You Ask for a Quote
Don’t ask “how much is a VMM?” or “what’s your CMM price?” You’ll get a catalog thrown at you. Instead, send this package:
- Sample part (3–5 pieces if possible) — the single most useful thing you can provide
- Drawing with GD&T callouts — highlight the 3–5 critical dimensions
- Material and surface finish — including coating, plating, or anodizing
- Tolerance band for each critical feature — not just “tight,” give numbers
- Annual volume and required inspection frequency — 100%? First article? PPAP?
- Floor space available — including ceiling height, power, air supply
- Environmental constraints — temperature stability, vibration, cleanliness
- Acceptance criteria — what measurement proves the machine works? (Gauge blocks? Your part? A standard?)
Send this to three vendors. The one that asks follow-up questions about your parts — not just your budget — is the one to trust.
FAQ
Q: Can a vision measuring system replace a CMM?
A: For flat, thin parts with 2D geometry and tolerances above ±5 µm — yes, and it’ll be faster and cheaper. For anything requiring true 3D form, GD&T relative to a datum structure, or blind-feature access — no.
Q: Is “optical CMM” just a marketing term?
A: Sometimes. If it’s ISO 10360-2 certified with a touch probe, it’s a CMM with an optical sensor. If it’s camera-only with no probe, it’s a VMM — call it what it is.
Q: What about computed tomography (CT) scanning?
A: CT is a different category — full internal 3D, but slow, expensive, and resolution-limited. Use it for internal voids, assemblies, or impossible-to-probe geometries. Not a replacement for either VMM or CMM in production inspection.
Q: Can I use a VMM for non-contact measurement of shiny parts?
A: Yes, but it takes work. Programmable ring lighting, polarization filters, and sometimes a thin matte spray (if allowed) are required. Highly reflective surfaces like polished aluminum or chrome are the hardest case for any optical system.
Q: How do I verify a VMM is still accurate?
A: Daily: check repeatability on a reference standard (glass scale or precision artifact). Monthly: full system check per vendor procedure. Annually: recalibration by the vendor or a certified metrology service.
If your parts are flat, thin, soft, or high-feature-count — and your tolerances sit in the ±5–25 µm band — a vision measuring system will outperform a CMM on speed, safety, and total cost. If you need true 3D form, GD&T across multiple datums, or deep-feature access, a CMM is the only honest answer.
The most expensive mistake is buying the wrong category. Get the decision gates right (§4), then optimize within the right category.
If you need a second opinion on your specific parts? Send us your drawing and 3 sample pieces — we’ll measure them on both platforms and show you the data. No quote until you’ve seen the results.