A liquid class that works beautifully on one instrument is a calibration for that instrument, not a universal truth about the liquid. The moment you move it to a different deck, a different pump, or a different lab across the country, you are making a bet that the new hardware behaves closely enough to the old that the numbers still hold. Sometimes they do. Often they need a nudge. The mistake is to assume either extreme, to copy the parameters blindly, or to throw them away and start from scratch. There is a disciplined middle path.
Transferring a class well is worth the effort because the alternative is a lab where every instrument has its own private dialect and no result from one can be trusted on another. This is how you carry a calibration across a boundary and come out with evidence that it survived the trip.
What actually changes when the hardware changes
Before copying anything, it helps to know which parts of a class are portable and which are hardware-bound. The intent behind the class travels well. The exact numbers that encode that intent may not.
- The intent travels: which liquid, roughly which volume range, which tip family, and which dispense mode. That is the identity of the class and it does not change because you moved rooms.
- The fluidic path may not: pump size, tubing length, and the volume of system liquid behind the tip all shape how a plunger movement becomes a delivered volume. A different pump can turn a perfect flow rate into one that pulls air.
- The correction curve almost never travels unchanged: it cancels a specific instrument's steady over- or under-delivery, so it is the most hardware-bound thing in the class and the first thing to re-measure.
- The tip really matters: the same nominal tip from a different lot or vendor can shift low-volume behavior enough to break an otherwise faithful transfer.
A transfer protocol that produces evidence
Treat the transfer as a small validation, not a copy-paste. The goal is not just to make it work but to be able to show that it works.
Start from the source class, not from zero
Bring over the parameters as a starting hypothesis. They encode real knowledge about the liquid, and even if the destination needs adjustment, they put you far closer than a blank form. Keep the source class intact and identifiable, so you can always compare the two later and explain any difference.
Verify gravimetrically before you trust it
Run the transferred class on the destination instrument and weigh the result. A balance and the liquid's density turn delivered mass into delivered volume, and that measurement is the whole point: it tells you whether the copied parameters actually reproduce on the new hardware. Test across the volume range, not just at one point, because instruments often agree at mid-range and diverge at the extremes.
Adjust the fluidic parameters, then re-measure the curve
If the verification shows drift, resist the urge to change everything. Adjust the parameters most sensitive to the fluidic path first, the flow rates and air handling, and change one at a time so you can attribute the improvement. Once the raw behavior is close, measure a fresh correction curve on the destination instrument. Never carry the source curve across; it is calibrated to a machine you are no longer using.
Multi-site transfer adds a human layer
Moving a class to another lab adds problems that are not about physics. The receiving site has different water, different ambient temperature, different operators, and different habits. Record the conditions the class was built under, especially temperature, because viscosity and density shift with it and a discrepancy across sites often turns out to be a few degrees rather than a fault in the method. Agree on the tip down to the lot where you can, and agree in advance on what counts as a passing verification, so the two sites are measuring against the same bar.
A liquid class does not transfer when you copy its numbers. It transfers when you copy its numbers, re-measure on the new instrument, and can show the delivered volume still lands where it should.