Perfusion and flow control: rockers, pumps and pressure controllers, and what each one costs

Perfusion is the point where an organ-chip programme stops being a plasticware purchase and starts being an instrument purchase. The decision is not “should there be flow”. It is how precisely the flow has to be defined, because that single requirement moves the cost by two orders of magnitude and the tubing burden by rather more.

If you need medium exchange and some shear, a rocker on a plate does it for the price of the rocker. If you need a stated wall shear stress in dyn/cm² that you can defend in a paper, you need a controlled-flow system, a channel geometry with a published relationship, and a budget in the tens of thousands.

The four routes

Route 1 is the default for plate-format organ-chips and should be the default for anyone whose requirement is “not static”. It removes tubing, connectors and pumps — which removes the three commonest sources of contamination, bubbles and unattended overnight failure. The trade is that flow is bidirectional and cyclic: MIMETAS lets you program tilt from 0° to 25° and rocking intervals from 5 seconds to 999 minutes, which is real control over the flow regime, but it is not a constant unidirectional shear stress. MIMETAS themselves acknowledge this by offering UniFlow, described as assuring perfusion always in the same direction to improve conditions for microvasculature — an admission that direction matters for some biology.

Route 4 is what you buy when the shear stress is the experiment, typically endothelial mechanobiology.

The verified cost of a defined-shear system

This is the number that surprises people, so here it is with sources. All figures are ibidi USD list prices verified on 2026-09-01.

A single-channel defined-shear system therefore starts around $24,000 before consumables, and a four-channel one around $34,000. That is the honest entry price for the sentence “cells were exposed to 10 dyn/cm² of unidirectional laminar shear stress”.

Set that against the rocker route, where the recurring cost is the plate and the instrument is bought once, and the decision usually resolves itself. Buy the pressure system when the shear stress value is a claim you have to defend. Buy the rocker when flow is a culture condition rather than an independent variable.

Shear stress is a property of the channel, not of the pump

The most common error in this area is quoting a flow rate as though it were a dose. It is not. For a rectangular channel the wall shear stress is a function of flow rate, viscosity and the channel cross-section, and changing the channel changes the answer by a factor of fifteen without touching the pump.

ibidi publishes the coefficients for its own geometries, and the spread across one product family makes the point better than any argument. Using τ = η · k · Φ, with τ in dyn/cm², η in dyn·s/cm² and Φ in ml/min:

Three consequences worth internalising:

1. The same flow rate through a 0.2 mm and a 0.8 mm channel differs in shear stress by roughly 15-fold (512.9 versus 34.7). “1 ml/min” is not an experimental condition.
2. Changing only the bottom material changes the coefficient. The 0.4 Luer drops from 131.6 to 104.7 simply by going from polymer coverslip to glass, because the channel height differs. If you switch to a glass-bottom slide for imaging, you have changed your shear stress by 20% unless you re-derive the flow rate.
3. Viscosity is yours to supply. ibidi’s reference tables use η = 0.0072 dyn·s/cm², but the viscosity of your medium at your temperature with your serum concentration is a number you must establish. It scales the answer linearly.

Worked from ibidi’s own table for the µ-Slide I 0.4 Luer at η = 0.0072: 1 dyn/cm² needs 1.06 ml/min; 10 dyn/cm² needs 10.55 ml/min; 20 dyn/cm² needs 21.11 ml/min. Those flow rates are substantial, and they are why recirculation rather than single-pass perfusion is the norm — a single-pass system at 10 ml/min consumes 14.4 litres of medium per day.

Where the calculation stops being valid. ibidi is explicit that the shear stress calculation holds only away from the channel walls, and that observations should be made at a distance from the side walls comparable to the channel height. For a 400 µm channel, that means excluding roughly 400 µm from each wall. If your imaging field includes the wall region, some of your cells are not experiencing the shear stress you are reporting.

For a custom device, the coefficient is yours to derive

The ibidi table exists because ibidi did the numerical work for their geometries. Nobody has done it for yours. If you commission a custom channel device and intend to report shear stress, you need one of:

• A computed value for your exact cross-section, from the rectangular-channel solution the vendors use — ibidi cites Cornish (1928), Flow in a Pipe of Rectangular Cross-Section, as the underlying treatment.
• A CFD result from the fabricator, which some will produce as part of a design programme and none includes by default.
• A measured calibration relating pump setting to flow through the assembled device, which is the only route that accounts for connector and tubing resistance.

Put this in the RFQ explicitly: “supply the wall shear stress as a function of flow rate for the as-built channel, and state the assumed viscosity.” A fabricator who declines is telling you they quoted geometry only, which is fine — but then the derivation is your work and it belongs in your schedule.

Bubbles are the failure mode

Every perfused culture eventually meets a bubble, and in a microchannel a bubble is not a nuisance, it is the end of that channel’s experiment. It displaces medium, shears the monolayer at the moving interface and — if you are measuring impedance — corrupts the readout directly; the organ-chip TEER literature lists air bubbles in the channels alongside temperature and electrode position as a primary confounder.

The published mitigations, roughly in order of effectiveness:

• Degas the medium before it enters the system. Dissolved gas coming out of solution as medium warms is a principal source. This is the cheapest intervention and the one most often skipped.
• Use a bubble trap or in-line degasser. Standard on commercial perfusion systems for this reason.
• Exploit gas permeability. In a PDMS device, raising the internal pressure can force a bubble to dissolve into the bulk material. This works only for gas-permeable materials — a thermoplastic or glass chip gives you no such escape route, which is one of the under-discussed costs of the low-absorption materials recommended for pharmacology in the material guide.
• Prime carefully and never let an inlet run dry. Most bubbles are introduced at connection and reservoir-change steps rather than generated in situ.
• Design them out. Avoid abrupt expansions, dead-ends and upward steps where bubbles lodge.

The rocker route is popular substantially because it sidesteps this entire category of failure: no tubing, no connectors, no priming step, and a closed plate.

Choosing, in order

1. Is the shear stress an independent variable in the experiment? If no, use a rocker and spend the money on cells. If yes, continue.
2. Does it need to be unidirectional? Gravity levelling is inherently bidirectional. If directionality matters — as it does for microvascular models, which is why MIMETAS built UniFlow — you need either a platform that engineers around it or a pumped route.
3. How many channels in parallel? This sets the instrument cost far more than the flow specification does. One pump driving four Fluidic Units is a different purchase from four independent systems.
4. Recirculating or single-pass? Recirculation saves medium and enables high shear stress; single-pass avoids metabolite accumulation. At 10 dyn/cm² in a typical channel, recirculation is effectively compulsory.
5. What is the readout, and does the perfusion hardware obstruct it? Tubing on a microscope stage is a mechanical problem, and a stage-top incubator with a pump umbilical is a different instrument configuration from the one you tested on.

How we can help

We do not manufacture perfusion hardware and the systems named here are direct catalogue purchases. Where a sourcing intermediary earns its place is the custom case: a channel geometry that does not exist off the shelf, a device that has to interface with a pump you already own, or a request for the shear-stress derivation to be delivered with the part rather than left as homework. If you tell us the target shear stress, the cell type and the readout, we will put the same specification to several fabricators and to the platform vendors, so you can see what the custom route actually costs against the catalogue one. Send the specification.