Compensation is about overlap not brightness

Compensation is a subtraction of spectral overlap, built from single-stain controls, and it leaves brightness decisions to the voltages you already set. The decision this page supports is whether a double-positive cloud survived because two fluorophores are both present, or because one fluorophore's emission still leaks into the neighbour after a bad correction. How a pulse becomes an event is described in how a flow cytometer reads a cell. Imaging the same dyes, without a matrix, is the subject of fluorescence microscopy without the myths.

Compare instrument classes in the scientific instruments catalogue. The panel often sits in a molecular biology pathway. Optical layout and detector count belong in a quote request.

Emission spectra and filter windows

A fluorophore absorbs in a band and emits across a longer band. The emission is a curve, not a single wavelength. A detector sits behind a filter that passes one slice of that curve. The slice is chosen so that most of the intended dye's emission lands there. The tails of other dyes still enter. That fraction, measured in the secondary detector relative to the primary detector, is spillover.

The laser decides who gets excited. The filter set decides who is seen. A dye with a long red tail will spill into a redder detector even when its colour name sounds private. Two dyes that share an excitation laser are collected at the same time, so the spill is present on every event. Sequential acquisition on a microscope can avoid some of that simultaneity. A cytometer running one laser line into several detectors cannot.

Spillover is therefore a property of the dye spectrum, the laser, and the filters, expressed at the gains you used. It is not a property of how biologically important the marker feels. A dump channel stained with a bright fluorophore can spill more trouble into a dim channel than a carefully chosen dim marker does.

What the matrix subtracts

You run each fluorophore alone. In the primary detector the positive population sits high. In each other detector you measure how far that same population sits above its own negative. The ratio of those distances is the spillover coefficient into that detector. The software builds a matrix. When an event arrives with signal in several detectors, the matrix removes the estimated contribution of each primary dye from the other channels.

The control has to be at least as bright as the brightest cells in the experiment. A dim single stain underestimates the slope, and a bright sample is then under-corrected. The residual looks like a diagonal streak and manufactures false positives. A control brighter than the sample is safer than a control that merely "looks positive".

The matrix does not decide the voltage. Voltage and gain place the negative population out of the noise and the positive population on scale. If you raise voltage until a dim marker looks dramatic, you also raise the spill into that detector, and you may saturate the bright dyes. Saturation ruins compensation because the true primary signal is clipped: the software no longer knows how bright the event was, so it cannot subtract the right amount.

Spreading is what remains

Compensation aligns the medians. It does not remove variance. Photons arrive with noise, and a bright primary dye contributes extra uncertainty to every channel it spills into. After a correct matrix, the negative population in the spill channel is wider than it was in an unstained tube. That widening is spillover spreading. It is why a gate copied from an unstained plot sits inside the spread once the rest of the panel is present.

A fluorescence-minus-one tube, which contains every stain except the one you are gating, shows the spread you must clear. Compensation does not replace that tube. Unstained cells show autofluorescence, which is a different baseline. Autofluorescence that differs between samples is not removed by a matrix built on a single cell type.

Display scales matter. Linear plots squash the decades where fluorescence lives. Log plots hide negatives. Biexponential and logicle displays show zero and the negative spread without piling every dim event on the axis. Use them when you judge whether a matrix worked. A population diving sharply below zero, while its median is far from the unstained cells, is a sign of over-correction, a mismatched control, or a tandem that has come apart.

Reagent classes that break the assumption

The calculation assumes the spectrum in the control is the spectrum in the sample. Tandem fluorophores, in which one protein dye passes energy to a second dye, can uncouple when they see light, heat or age. The donor then emits in its own channel and the spillover matrix from last month no longer matches. Protect tandems from light and follow the storage on the vial. Recalculate when the lot changes.

Antibody-capture beads and cells can both carry a single stain. Beads are often brighter and more uniform, which helps the slope. Some tandems report a slightly different spectrum on a bead than on a cell. If the corrected cell plot still shows a diagonal that the bead matrix will not flatten, acquire a cell single stain at least as bright as the sample and rebuild. Mixed stains in one tube are not single stains. A panel tube cannot calculate its own matrix.

When the corrected plot still looks wrong

A diagonal streak after compensation means the slope is wrong or the spectrum changed. Check that the single stain was truly single, that it was on scale, and that it was at least as bright as the sample. Check that you did not apply a matrix calculated under different voltages. Hand-editing coefficients until the plot looks square hides the failing control. Find the stain that misbehaves by applying one single stain at a time to the matrix.

If every channel requires extreme coefficients, a filter may be seated wrong or a detector assigned to the wrong dye. Stop and confirm the optical configuration. Compensation cannot invent a filter you do not have.

If the medians line up but the positive gate still fills with events from a fluorescence-minus-one tube, you are looking at spreading, not at residual median spill. Move the gate or redesign the panel so that a bright, highly overlapping dye does not empty its noise into the dim question you care about. That redesign is panel design. It is not a brightness slider.

Safety and research limits

Compensation is maths applied to a file. The tube you acquired may still be a biosafety issue. Institutional rules, read against the WHO Laboratory biosafety manual and the CDC BMBL, decide whether the sample may be run. This explainer does not. A corrected plot is not a clinical result. Reporting practice for cytometry experiments is discussed in the community around the International Society for Advancement of Cytometry. Filter physics for the imaging case is introduced in the MicroscopyU fluorescence section. Neither source chooses your matrix.

What to send with an enquiry

State how many fluorescence detectors you need, which laser lines must excite the dyes you already know you will use, and whether the software exports the spillover matrix with the FCS file. Ask for the filter specification in writing. Use the scientific instruments catalogue and the quote request. Ask whether a quotation is possible. A detector count without filter bands is not enough to judge overlap.