What accuracy depends on
- Drawing quality. Dimension labels can be rounded, out of date or inconsistent with the geometry. The app does not judge whether a drawing is correct.
- Calibration segment. A calibration error carries over to every later result. The longer and more clearly labelled the segment, the smaller the relative error.
- Placing the points. Touching the screen has limited precision, especially at low zoom. The Precision magnifier and zooming in help.
- Distortion in a scan or photo. Perspective, creased paper and a wide-angle lens introduce errors that correction does not fully remove.
- A printout or export that is not to scale. The stated scale (for example 1:100) applies only to the original sheet. See A printout that is not to scale.
- Several scales on one sheet. Calibrating the whole page on one dimension does not cover views drawn at a different scale. You calibrate each such view separately as an Area; when a measurement enters an area with another calibration, the app warns you.
An error in placing a point carries over to reality in proportion to the scale. For example, a point that is off by 1 mm on the sheet (measured at paper size) means in reality:
| Drawing scale | Error on the sheet | Error in reality |
|---|---|---|
| 1:20 | 1 mm | 20 mm |
| 1:50 | 1 mm | 50 mm |
| 1:100 | 1 mm | 100 mm |
| 1:200 | 1 mm | 200 mm |
More on this relationship: Measurement accuracy on a drawing and the reading error calculator.
How Skalnivo calculates a result
- Two-point calibration. You pick points A and B on a segment of known length and enter that length. The scale factor is k = known length / distance AB in document units.
- Measurement. The distance between two points in document units is multiplied by k. A check example: a reference segment of 400 document units corresponds to 4.00 m, so a segment of 250 units is 2.50 m.
- Document coordinates. Points are stored in document space, not in screen pixels. Zooming, panning and rotating the view do not change the result.
- Full precision. Calculations use double-precision floating-point numbers, and rounding happens only when the value is displayed. You set the number of decimal places in Profile.
- Area and perimeter. Polygon area is calculated with the shoelace formula in real-world space; concave polygons are handled, and self-intersecting ones are flagged with a warning.
- A second dimension. Using Check with a 2nd dimension, you add another known dimension. Multi-segment calibration fits a common scale (least squares), shows the deviation as a percentage, and gives the page the Needs checking status above 1%. For a stretched scan, you can calibrate the X and Y axes separately.
- Scale 1:N. When the page units are known, the app shows the Resulting scale 1:N. It applies to a print at the PDF page size; the scale in the title block may be outdated.
The app rejects zero and negative values, points placed too close together and ambiguous polygons. Other pages do not inherit a scale unless you decide so.
Results of an internal test on real drawings
The calculations were checked on 14 real PDF files (58 pages, assembly drawings in A3 and A4 sizes). For 110 measurements in 11 files, the result was compared with the dimensions labelled on the drawings:
- median deviation from the labels
- 0.22%
- 95th percentile (p95)
- 0.99%
- largest deviation
- 2.53%
The arithmetic itself matched an independent calculation to the limit of floating-point precision (relative error 0 against a test threshold of 1e-9). Deviations from the labels come from features of the drawings:
| Effect | Size of effect | Cause |
|---|---|---|
| Calculation error | 0 (to numerical precision) | None detected. |
| Labels rounded to whole mm or cm | 0.1–0.3% for segments of about 400 mm | The label does not give fractions of a millimetre. |
| Dimension ticks in the middle of gaps | about 0.8% (1200 mm against a label of 1190 mm) | Chain dimensioning convention: the label refers to the distance between faces. |
| Label inconsistent with the geometry | 1–2.5% for individual dimensions | A dimension marked "not to scale", a manual edit or ticks on a different element. |
| Scale in the title block differs from the file scale | 3 of 4 title blocks read | An A3 sheet reduced to A4 on export. |
| Several scales on one sheet | tens of percent when using the scale of another view | Detail sheets combine views at different scales. |
What this test covered. The end points of dimensions were set programmatically on the dimension ticks, and the label values were copied by hand from the drawings. The test therefore does not include the error of placing a finger on the screen and does not replace independent certification. The files come from one sector (assembly drawings), and their data is not published.
How to check a result yourself
Verifying a measurement against a second known dimension
Step 1: Find a second labelled dimension
Choose a segment with a numeric label that you did not use for calibration. Ideally it is long and at the same scale as the first.
Step 2: Measure it after calibrating
Tap Check with a 2nd dimension in the calibration window (the app calculates the deviation) or use the Distance tool. Place the ends of the segment as precisely as you can, zooming in or using the precision magnifier and step buttons.
Step 3: Work out the deviation
Subtract the labelled dimension from the measured one, divide the difference by the labelled dimension and multiply by 100.
Step 4: Judge the result
A deviation of up to about 1% is typical for label rounding and point placement. A few percent usually means a poor calibration segment, a printout at a different scale or several scales on the sheet.
Step 5: Correct the calibration
Choose Recalibrate on a longer segment or a dimension from the same view, or add a region calibration for a view at another scale, and repeat the check. You can restore the earlier calibration using the Restore previous version command.
What the app does not do
- It is not a measuring instrument or a surveying tool; results are estimates for information only.
- It does not check whether a drawing, or the dimensions written on it, are correct.
- It does not read dimension labels automatically and does not guess the scale: you confirm a known dimension yourself.
- It does not replace measuring on site, the designer or the site manager.
Frequently asked questions
How accurate are measurements in Skalnivo?
Why does the result differ from the dimension labelled on the drawing?
What should I do with a sheet that has views at several scales?
Does zoom affect the measurement result?
Is Skalnivo a measuring instrument?
Related pages
Updated: