Why our two compressors overheat, how a venturi with no moving parts makes vacuum, how a foam gripper mechanically notices a hole and seals it, and why every mobile picking robot abandoned compressed air. Drag the 3D views to orbit. Numbers come from the 18k-word field guide (all sources cited there).
Cutaway of a venturi ejector. Supply air chokes sonic in the throat and screams across an open chamber; ambient air from the vacuum port gets dragged into the jet by momentum transfer and thrown out the exhaust. No moving parts — but the nozzle is choked, so air consumption scales with absolute supply pressure while vacuum stops improving past the design point. Feeding it tank pressure is pure waste.
drive air (burned) entrained suction air (what fights leaks) — both leave through the exhaust.
Single-stage venturis return ~0.7 l/min of suction per l/min of air. Multi-stage (Piab COAX, Schmalz SBPL-HF) reuse the jet: up to 3.6:1 — but only at the shallow vacuum cardboard uses. That's the entire trick.
A foam bar gripper is a vacuum plenum with 18 cells. Cells the box doesn't cover are giant leaks. In each cell sits a small ball: a covered cell flows gently and the ball hangs open; an uncovered cell flows hard, drag slams the ball onto its seat, and the cell seals itself — purely mechanical hole detection (Schmalz SVK, Coval MVG, AIRBEST TXM "-V"). Toggle the valves off and watch the plenum die. Try the four vacuum sources — same gripper, wildly different wall power.
Cell state: open + covered (feeding vacuum to the box) ball latched — sealed itself open + leaking. Particles show air rushing into leaking cells.
Push porosity to "torn" and even covered cells flow hard enough to slam their own valves shut — the real failure mode that makes VMECA sell two valve-threshold classes. Blow-off at release back-flushes dust out of the balls.
A vacuum source is a curve, not a number: free flow at 0 kPa falling to zero at its max vacuum. Your box is a leak line; the system settles where they cross. Deep-vacuum sources collapse on leaky cardboard; the blower barely notices. Drag the leak slider and watch each working point (dots).
One 230 V / 16 A group. Each bar: vacuum source + the robot (~1,000 W). Today's compressor pair doesn't fit — that's the klixon story: 120 motor starts/hour on 40 L tanks, thermal trips, dead air mid-shift.
Numbers: 2× oil-free compressors 5,200 W (today) · belt-drive oil-lube compressor 2,600 W · Becker/Busch-class vane or claw pump 1,500 W · 2× 48 V BLDC blowers 750 W (the Boston Dynamics / Pickle architecture).