Field guide
Prepare a drawing that the next person can interpret
Check the coordinate meaning, choose the drawing content, and keep the source with your export.
Start with the information your CAD recipient needs
Use survey points to DXF when a colleague needs a drawing containing coordinate locations, rather than another delimited point table. Each accepted source record becomes one ordinary POINT entity. You can also include separate TEXT entities for the point ID, known elevation and complete description. Optional polylines follow only the source-row sequence you supply. The drawing does not infer boundaries, create surfaces or construct linework from feature codes.
Agree on the intended use before exporting. A drafter may need visible point numbers and a few feature layers, while a recipient preparing a background drawing may want only the points. Choose that content deliberately to keep the file manageable. The source-and-settings report accompanies the geometry so another person can check where the numbers came from and how the drawing was prepared.
Map PNEZD, PENZD and other point layouts
Import UTF-8 CSV or TXT, or paste the source text. Comma, semicolon and tab delimiters are supported. Confirm whether the first row is a heading, then map northing and easting. The PNEZD shortcut means point, northing, easting, elevation and description; PENZD reverses the two planar coordinate columns. Other arrangements can be mapped individually, including files with no point ID or description. XLSX workbooks are also accepted for this input: explicitly choose the worksheet, and replace formulas, dates or error cells with plain values before importing.
Do not identify a coordinate by the letter X or Y alone. Export conventions differ, especially when a field system uses X for northing. This writer always assigns CAD X to the field you confirm as easting and CAD Y to northing. A headerless CASS-style file may put the description immediately after the point number. Review the visible column sample rather than assuming it follows a five-column shortcut.
Choose known-height 3D or a deliberate 2D projection
In 3D mode every point needs a valid, known elevation. The writer does not replace missing heights with zero, interpolate them or discard incomplete records. If your work only needs plan positions, choose 2D and confirm that the exported Z coordinate is a drawing plane. Every POINT and polyline vertex then lies at Z = 0, while the original height values remain unchanged in the report.
Elevation labels in a 2D drawing still describe the known source heights. They do not suddenly become zero labels. A blank elevation produces no elevation text, and the summary counts those omitted labels. This lets you prepare a plan drawing without claiming to have measured the missing heights. Source syntax errors and invalid numeric fields must be corrected before either mode can export.
Keep drawing units and coordinate references explicit
Unit selectors describe the input; they do not scale the coordinates. Set the receiving CAD application's insertion units to match the source horizontal unit and verify a known distance after import. This R12 export records the chosen units in the report rather than promising that every receiving application will apply a modern insertion-unit header. Metres, international feet and legacy US survey feet must not be treated as interchangeable.
Three-dimensional output requires identical horizontal and elevation unit labels. Convert a mixed-unit file explicitly before using that mode. A 2D projection can retain a different original elevation unit in the report, so state that unit beside any height labels in the receiving drawing. No map projection, datum transformation or local-grid adjustment is performed here.
Understand the entities and the tested format
The output is an ASCII DXF R12 file identified as AC1009. It uses POINT, TEXT and, when requested, POLYLINE with VERTEX and SEQEND records. These are ordinary drawing entities. They are not Civil 3D COGO points, attributed blocks or a survey database, and moving a text label in CAD does not move its associated point. Import workflows that require intelligent point objects need a separate supported point-import process.
Representative exports have been reopened with the independent ezdxf reader to compare point coordinates, text values, layers and explicit vertices. That check supports the specific format and entities used here; it is not a claim that every CAD version or font combination has been tested. Try a small file in the receiving application before sending the complete project.
Use description layers without losing the original code
When grouping is enabled, each exact source description receives a point layer. Names use uppercase ASCII letters, digits, underscores and hyphens and stay within the older format's 31-character limit. Spaces, punctuation and non-ASCII description characters are replaced in the layer name. Distinct descriptions that produce the same cleaned name receive different suffixes, so they are not silently merged.
The layer CSV and JSON preserve the complete original description beside its drawing-layer name and point count. Blank descriptions also receive a defined group. Turning grouping off places all points on POINTS. Text uses ANNOTATION and paths use CONNECTIONS, making it possible to hide labels independently. A maximum of 256 description groups is supported; disable grouping or filter a highly varied file before export.
Choose readable labels and preserve coordinate precision
Text height and east and north offsets use drawing units. Consecutive enabled labels are stacked toward decreasing northing at 1.5 times the text height. Rotation, measured in degrees, applies to each text entity; it does not rotate the coordinate frame or the offset vector. Choose a nonzero height appropriate to the scale of your receiving drawing, then inspect a trial import for overlapping labels.
Elevation decimal places affect only the text label, not the POINT coordinate. IDs keep leading zeros because they are text. Empty selected labels are omitted and counted. Tabs and line breaks in otherwise supported labels become spaces for single-line TEXT, with the number of affected labels reported. The complete source strings remain available in JSON even when the visible annotation uses this single-line presentation.
Know the text encoding limits before handover
The file contains ASCII bytes, with supported BMP Unicode characters represented as DXF Unicode escapes. Chinese and other such labels still require a receiving application and font that display those characters. A DXF reader may expose the encoded text until its Unicode decoding helper is applied. Layer names stay ASCII regardless of the label language.
Selected labels containing backslashes, carets, percent control sequences, unsupported control characters or non-BMP characters are blocked. Labels longer than 255 encoded characters are also blocked rather than truncated. This limit is smaller for text that needs Unicode escapes. Disable the affected label category or prepare an edited source copy if necessary. Turning off a label does not remove the point or alter its original description.
Connect points only when the order is known
Leave the path field empty to export unconnected points. For an explicit path, enter comma-separated original starting line numbers, one path per line. Count a header as a source line. These numbers identify source records, not point IDs; this distinction also works when IDs are duplicated or contain letters. An open path needs at least two vertices, and the closed option needs at least three.
Confirm the exact sequence before generating the preview. Every referenced source line must exist. Closing a path joins its last vertex back to its first without inventing intermediate points. The writer does not sort by point number, search for neighbours or assume that a code means a legal boundary. Repeated positions and self-crossing sequences remain your explicit geometry, so inspect them in the receiving CAD application.
Review counts, locations and the actual download scope
The preview reports point, text and polyline counts separately, plus vertex and SEQEND records and the final byte size. Five points with three nonblank labels each produce five POINT and fifteen TEXT entities. Adding one three-vertex path gives twenty-one top-level entities and twenty-five entity records when its three vertices and terminating record are included. This distinction helps explain why an entity counter can exceed the original point count.
Search the point table or select a source row to inspect its original fields. The north-up plan preview shows at most the first 2,000 point positions, with the selected point added when needed. Complete paths are displayed within a separate 2,000-vertex budget. These display limits and table searches do not reduce the downloaded drawing. The plot does not simulate CAD fonts, hidden-layer settings or final annotation overlap.
Export the drawing together with its evidence
Inputs are limited to 10 MiB and 100,000 point records. The drawing limit is 50 MiB, with at most 100,000 explicitly requested path vertices. More labels increase file size. Large workloads run in a worker and can be cancelled; changing source data or drawing options invalidates the old result so it cannot be confused with newly selected settings.
Download the DXF, layer mapping CSV and source-and-settings JSON together. Processing stays in the browser; this tool does not upload point files to a conversion service. Before delivery, reopen the actual downloaded DXF, check a few known coordinates, confirm the point count and units, and inspect both a typical label and a non-ASCII label when present. Preserve the original coordinate file as the authoritative source.