Where this spatter came from ORIGIN
The plan position is exact. The height is a ceiling, because every drop fell while it flew, and no back-projection that ignores the fall can put the source anywhere but too high.
The scene
Both accuracies feed the bands below, so understating them is how a page comes to promise a precision it has not got. A stain read to a tenth of a millimetre is careful work with a scale in the photograph; half a millimetre is a quick measurement in the field, and it costs more than most people expect.
Walls
A wall is the vertical plane standing on the line between two corners, given as plan coordinates. Walk from the first corner to the second and the room is on your left. The heading printed beside each wall is the way a drop must have been going to strike it; if it points into the room rather than out of it, the corners are the wrong way round and the button puts it right. The floor is always there and needs no defining.
Stains
Seven numbers each: a name, which surface it is on, two coordinates, the width and the length of the ellipse in millimetres, and the direction of its long axis in degrees. On the floor the two coordinates are x and y and the direction is measured round from the x axis. On a wall they are the distance along from the first corner and the height above the floor, and the direction is measured from “along the wall” towards “up”, so a drop on its way down reads as a negative angle. In both cases the direction is the way the drop was going, which is the tail end of the stain pointing away from you.
What it comes to
The scene in plan: each stain as a dot with its long axis run backwards, the walls as heavy lines, the shaded box the spread of positions your stated reading accuracy allows, and the cross where they converge. The cross is a real answer. The height printed above it is a ceiling.
Stain by stain
The last column is the bound each stain gives on its own, and how far that bound moves when everything it depends on moves to the ends of its stated accuracy. Three separate things push it. A nearly round stain has an impact angle that is barely knowable, because the arcsin runs away as the width approaches the length. A steeply descending drop has a bound the tangent stretches out of usefulness. And the distance out is measured from the convergence, so the spread of the convergence moves every bound at once, which on a floor stain is the only way the long-axis reading reaches the height at all, since turning a floor stain’s axis changes neither its position nor the angle it was descending at. A stain can suffer from one of the three and not the others, which is why this is computed by moving the readings rather than read off a formula.
The working
Questions people ask about ORIGIN
Why will it not just tell me the height?
Because the height is not in the measurements. The identity is h = zs + D tan β − gT²/2, where T is how long the drop was in the air, and nothing about a dried stain records how hard the drop was thrown. The textbook method is the T = 0 case of that identity, which is to say it assumes the drop did not fall while it flew. It did. So the textbook number is always too high, never too low, and the honest thing left over is the bound: the source was at or below it. This page gives you that and declines to dress it up as an estimate.
How much too high is the ordinary method?
As much as the drop fell during its flight, which is gT²/2 and grows with the square of the time. Half a second of flight is 1.2 m of overshoot. On the worked example here (nine stains generated forwards from a source at 1.40 m by integrating real drops with gravity and air resistance), the middle of the per-stain heights is 2.21 m and the mean is 2.81 m, against a truth of 1.40 m. The least bound is 1.51 m. That is the whole argument for taking the smallest rather than the average, in one line of arithmetic.
Why is the smallest better than the average?
Because every one of these numbers is a ceiling, and the average of a set of ceilings is never a better ceiling than the lowest of them. If one stain tells you the source was at or below 1.5 m and another tells you at or below 5.9 m, the second has told you nothing the first had not already told you better, and averaging the two produces 3.7 m, which is worse than either stain on its own was capable of. This is not a subtlety about estimators. It follows from what the numbers are.
Is the plan position biased in the same way?
No, and this is the part worth relying on. Gravity pulls downwards and air resistance pulls straight back along the direction of travel, so neither of them can turn a drop sideways. The horizontal direction a drop was going when it struck is exactly the horizontal direction it left on, whatever it did vertically in between. Simulating flights and comparing the two reproduces them to thirteen decimal places, for stains on walls as readily as on floors. Where in the room is a real answer; how high is a ceiling.
What about a drop that was still rising when it hit the wall?
It reads as a negative descent angle and the bound correctly puts the source below the stain, which is right: a drop still climbing at the wall was launched from lower down. No special case is needed: the same identity handles it, because the sign takes care of itself. The worked example contains one, struck dead square at 90° while travelling horizontally, and its bound is simply its own height.
Is a nearly round stain a ninety-degree impact?
It means the measurement has run out, which is not the same thing. The impact angle is arcsin of width over length, and the derivative of arcsin blows up as the ratio approaches one, so a stain within a few per cent of round has an angle that is barely knowable at all. On the worked example a stain measured to a tenth of a millimetre carries ±2.4° where it is well elongated and ±22° where it is nearly round. The table shows that figure for every stain rather than printing an angle as though all angles were equally well known.
Does the width-over-length rule actually hold?
It is exact for a sphere meeting a plane: the ellipse it cuts is the diameter wide and the diameter over sin α long, so sin α is width over length and nothing else enters. Real drops are not spheres at the instant they land, and published experimental work finds the ratio drifting from the ideal below roughly twenty degrees, where the stain grows a tail and the decision about where the ellipse ends becomes a judgement. That is a limit of the measurement rather than of the arithmetic, and it argues for reading the elongated stains carefully rather than for using a different formula.
Could you not bound the height from below as well?
Not from geometry, and the obvious attempt fails in a way worth stating. Capping the launch speed at something plausible does rule out the highest candidates, but worked without air resistance the same test also wrongly excludes true heights, found here by building a scene from a known source and watching it exclude that source. So this page offers no lower bound. A single stain is consistent with a source anywhere from the floor up to its own limit, and a range with an honest top and no bottom is what the measurement contains.
How many stains does it need?
Two is the arithmetic minimum and it is worth nothing, because two lines always cross somewhere and a pair can never disagree with each other. What matters more than the count is the spread: stains reached from different sides pin a position, stains all pointing along one line do not, and the page scores that spread and refuses to print a coordinate when it is near zero rather than inventing a point along the line.
Can this go in a report?
The working pane is written to be pasted into one, and it states the bound as a bound. Whether area-of-origin evidence belongs in a particular case is a question for the jurisdiction and for whoever is instructing you, and this page takes no view on it beyond refusing to overstate what the measurements support. Everything is computed on this device; nothing about the scene is sent anywhere, and the page works with the network switched off.
Doing this often? OBSCURA Case keeps the analysis with the rest of the evidence, in the chronology and the case file. Everything is US$5 a month or US$48 a year.