Getting interface detection wrong does not cost you a little accuracy — it points the answer the wrong way: tie the discriminator to a premise that does not hold and the instrument returns a confident wrong value, silently. This guide works by elimination, and three of its six steps argue for the cheaper method first.
How to choose liquid-liquid phase separation detection: six steps
From whether this is a separation problem at all, through the velocity gap, whether a conductivity or photoelectric unit will do, what the pipework allows, and which way the interlock should fail — six steps and the scheme is settled.
Steps
-
Step 1: Confirm this is a phase-separation problem at all
The instrument tells you which of two already-separated phases is passing. Rule out three cases first. Miscible systems (ethanol / water, acetone / water) never separate — those want a concentration measurement. A stable emulsion (surfactant present, or too little settling time) is a process and equipment problem. With three phases present at once the instrument describes what passes and cannot do a mass balance.
This step comes first because the other five assume the phases really do separate.
-
Step 2: Get both velocities and check the gap
There is one threshold number: the phases must differ by at least 30 m/s at working temperature. Common organic-against-water pairs are far apart — carbon tetrachloride 572 m/s from water, dichloromethane 428, toluene 170, and the closest, m-xylene, still 155 (handbook values at 25 °C).
Those values judge sufficiency, not calibration: how much water, salt and product your phase carries is something only a measurement knows. When in doubt send 500 mL of each phase. Note the direction of temperature too: organics run -2.7 to -4.9 m/s per °C and water +2.4, so warming widens the gap and hot duties need no extra margin.
-
Step 3: Check whether conductivity or photoelectric will do
This is the step that saves money. If the aqueous phase conducts, the organic does not, and an electrode may be fitted, use a conductivity switch — nothing is cheaper or more robust. If the phases differ in colour or refractive index and the window stays clean, a photoelectric unit is equally direct.
Those premises soften on pure-water washes (aqueous too resistive), wet organics, lined pipe that is awkward to tap, pairs that are both clear and colourless (dichloromethane / water), and dark or solids-bearing liquors. Then carry on down this list.
-
Step 4: Check the pipework
Four things. Bore — DN25 / DN32 / DN50 are standard spools, others to order. Can it be kept full — free-draining runs, falls and siphon breaks make the reading meaningless. Lined or not — a lined line takes the lined spool, with windows through the liner. And the distance from the measuring point to the diverter.
That last one gets overlooked: DN32 holds about 0.80 L per metre (0.53 L lined), and that volume plus the valve stroke is the mixed quantity every batch pays. Put the measurement hard against the diverter and give the straight run to the upstream side.
-
Step 5: Settle the interlock and the failure direction
Three relays carry three states: heavy, emulsion, light. The decision that matters is what happens on emulsion and on not-decidable — hold the current valve position and alarm, then let an operator confirm. Bubbles, flashing and window film all push the reading toward not-decidable rather than toward the other phase, so conservatism in that direction actually works.
Two parameters are set on site: the confirmation window (0.3 s default, 1-2 s where the interface tends to park) and hysteresis (10 % of the gap by default). Send sound velocity to the DCS on 4-20 mA and trend it: batch-to-batch movement is an early signal that the composition has changed.
-
Step 6: Send us the process data
We need: composition of both phases and the working temperature range; line size and material, lined or not; diverter type and its distance; whether the aqueous phase carries salt or acid; area classification; batch volume and draw rate; and the product list for a multi-product line. Send it to contact@pisonics.com and a written proposal usually follows within two business days.
If the velocities are uncertain, send samples for measurement — better for both sides than copying a number out of a brochure.