Embroidery format

From EduTech Wiki
Jump to: navigation, search
Machine embroidery
Module - entry page
Embroidery format
to finalize beginner
See also

1 Introduction

There are several kinds of Embroidery file formats, and each contains different abilities and features. Some formats have different versions with increased features. Usually these try to preserve backwards compatibility because of the cost of the hardware in question.

This page attempts to provide a short overview on embroidery formats. For technical details, see the specialized articles (menu to the right). Any technical details in the article are for understanding why the formats are as they are.

2 The Intended Functions

A lot of embroidery confusion is because there are diverse sets of hardware, and software and it is an unfathomable muddle. We can however properly classify the problem. The most significant differences between embroidery files is the divide between intending to run a machine, and those intended to run the programs that create the files. It is like the difference between a complex layered photoshop file with undo history etc, and a jpeg. While you could call them both "image files" it's a massive distinction. We will divide them up to avoid this confusion.

3 Machine Embroidery Files

Often the actual controllers within the machines are similar and the encoding scheme for the embroidery files turns out to be quite similar. They are intended to control a stepper or servo motor, and a x-y plotter and a needle head. They don't do much beyond that. The stitch portions of files encode for three things: a control command, a dx, and dy. This encoding is also heavily influenced by the actual physical requirement of the machine. The machines must stitch, unless the needle bar is blocked, which is the difference between a stitch and a jump. When enough of these are in a row, machines will often force a trim. It's possible with some machines to block the needle bar and move very far from the last stitch, others will force a trim. There are some cases like fringe where this is needed or puffy applique. Given this requirement, the commands used within the machine code is often similar, sometimes identical when you get away from the header and into the actual stitches within the file.

The header information is often of a specific length and contains a lot of display data for the machines. These are quite often specifically intended for exactly that display module. But, it also means that some formats like .DST, .DSB, .DSZ all have the same header information even though the stitch data is entirely different. These different encoding schemes result in entirely new file types while not doing anything new.

To help matters, or rather make them worse, some companies like Wilcom will make new file types that just contain the stitch data. For example .t01 is just .dst encoded stitches without a header. So is .tap. Literally these formats are identical and interchangeable, and you could create them by deleting the first 512 bytes from a dst file. Wilcom also has .t?? with other numbers that refer to other encodings.

Different machines will have their own format specifically because the format consists of commands intended only for the particular machine in question. These are specific ways that that particular hardware interacts. This is akin to having a different filetype for each executable for each different type of processor. Which is entirely what these files are intended as, executables.

3.1 Interchange Formats

One consequence of the need to have simple machine readable formats is that when you cut away the headers, the formats can be used as interchange formats. Embroidery machines do only a few things and those need to be easily encode in ways that the hardware can understand and execute. This means these formats are pretty easy to parse in binary and consequently access the stitch data therein. The ubiquity of .dst files is largely due to the age of the format but also because many formats are derived from them, and the early hardware was from the company that produced them.

3.2 Machine Encoding

The encoding of the format for the computer varies from company to company and software maker to software maker, but often share a common history. For example Pulse developed the early software for Brother called PG1, it's also telling that the PEC blocks within Brother software just like in .dst files have a 512 block header starting with a label "LA: (NAME)" In 1982, Pulse Microsystems also developed the first PC software Stitchworks. The structure of the files is likely due to Pulse as well as Tajima's dominance in the industry in the 1990s.

3.2.1 Command Encoding Schemes

Code Length Company Associated Structure bytes Encoding
3 Pulse, Tajima, Eltac 3 byte swizzled bits, punchcard yyyyxxxx yyyyxxxx ccyyxx11 Triplet code has control bits on the final byte. Last 2 bits of that byte are always set.
2 (4) Melco signed(x,y), x=0x80 control xxxxxxxx yyyyyyyy Series of move locations with control bits being triggered by 0x80 (-127). 0x01 stop, 0x04 jump, 0x10 end)
3 Barudan unsigned(x,y) control byte xxxxxxxx yyyyyyyy 1cccccccc The control byte provides direction(x,y) aka the sign.Last bit controls jump vs. stitch.
3 Foxtron control byte, unsigned(y,x) 1cccccccc yyyyyyyy xxxxxxxx This is very similar to the previous but differs in the endness of the triplet code.
2 (3?, 4) Brother signed 7 bit (x,y) cxxxxxxx cyyyyyyy if the high bit is set, it becomes a 12 bit signed number. And the top 4 bits in that nibble (the first triggering the event), are control flags.
2 (4) Compucon, Singer signed(x,y), x == 0x7F or 0x7D xxxxxxxx yyyyyyyy The 0x7F triggers a control event, whereas 0x7D triggers long form. Read x and y as 2 sInt16.
2 KSM unsigned(x,y) control byte xxxxxxxx yyyyyyyy 1cccccccc Controls the restart of stitching after a stop. Before restart of stitching, commands are unstitched but identical to stitches.

The DST (Pulse, Tajima, Eltac) punchcard swizzle is the oddest and oldest of these. I cannot help imagine the hardware of the reader. The high bits control Y and low-bits control X. One could imaging different mechanism reading different parts of the card and controlling the x or y part of the plotter. Reading from bottom to top, you would get control bits, and the biggest movement, depending on whether there's a hole there and in the direction of the hole. With the next byte having a third of that impact, and a third more each byte you go up. It would also explain things like why in the third byte it always ends 11 ??????11. Namely, if you turned the physical card around it would be in the position of the control bits. 00000011 is the standard no-op card for that but 11000000 would cause c1 and c2 to set and therefore STOP. It might be nice to look at some precomputered punch-card of this encoding. But, it would detect a backwards card as stop, with those set bits. Welcome to the 21st century!

DST Encoding
BYTE 7 6 5 4 3 2 1 0
1 y+1 y-1 y+9 y-9 x-9 x+9 x-1 x+1
2 y+3 y-3 y+27 y-27 x-27 x+27 x-3 x+3
3 c0 c1 y+81 y-81 x-81 x+81 set set

3.3 Machine Embroidery Files

Embroidery file formats
.ext Manufacturer(s) Structure
.10o Toyota unsigned(x,y,control) encoded stitch, a separate .00o contain color
.100 Toyota 4 byte encoding. 2 bytes of control bytes, x,y with 3rd, 4th signed x, y locations.
.bro Bits & Volts 256 byte header. x == 0x80 control encoding, 2 stop, 3 jump
.csd Singer, POEM brand-specific
.dat Barudan 256 byte header, unsigned(x,y,control)
.dsb Data Stitch Barudan 512 byte DST header but stitches are unsigned(x,y,control) style
.dst Data Stitch Tajima (See detailed article) 512 byte header, DST encoded direct commands.
.dsz Data Stitch ZSK_USA 512 byte DST header, big-endian unsigned(x,y,control) style encoding, with specified 4 bit needle
.emd Elna 48 (0x30) byte header, x == 0x80 controlled encoding: 0x2A stop/color_change, 0x80 Jump, end 0xFD.
.exp Melco, Bernina (high-end models) Expanded Melco (See detailed article) X == 0x80 controlled encoding. Color change, 0x01 is followed directly by 0x00 0x00 and a jump 0x80 0x04 uses the following the X, Y position instructed, but must be repeated each new command.
.e?? Eltac 256 bytes of header, Triplet coded in DST encoding.
.f?? Forton 256 byte header, unsigned(x,y,control) style encoding
.gt Golden Thread 512 bytes of header, unsigned(x,y,control) style encoding
.hus Husqvarna Viking Compressed bytes. Using a small table arc compression. The belief is that it's from a defunct compression library ArchiveLib by a defunct company GreenLeaf Software using mode "AL_GREENLEAF_LEVEL_4" which should be backed with something akin to zlib.
.inb Inbro 8192 byte header, unsigned(x,y,control) style encoding
.jef Janome (See specific article) stitch + color, Header information, magic-number thread lookup, and 0x80 triggered control events.
.ksm Pfaff 512 byte header. unsigned(x,y,control) triplets. However, it doesn't force that encoding on all jumps. Rather after triggering a color change to a specific needle, it simply gives normal encoded x, y location (indistinguishable from stitches) until control bits of 0x17, 0x18, or 0x19 triggers sewing again.
.max Pfaff
.mit Mitsubishi
.new Ameco
.ofm Melco Compound Binary format with some stitches in there.
.pcd Pfaff these actually have a weird encoding scheme. Using absolute positioning rather than relative positioning.
.pcm Pfaff stitch
.pcq Pfaff stitch
.pcs Pfaff stitch
.pec Brother, Babylock (see detailed article) colors, stitch, 1 bit graphics, header of #PEC0001, pec block, magic number colors, graphics which are displayed on the machines, contain high-bit long-form + control triggering.
.pen Brother - Disney Graphics files and Encrypted Stitches. Apparently this was a thing.
.phb, Brother Header, bunch of info and a PEC block.
.phc, Brother Header, bunch of info and a PEC block.
.sew Janome, Elna, Kenmore magic number thread lookups. Signed x, y with 0x80 triggered control events.
.shv Husqvarna Viking stitch, Big old giant 1 bit graphic, of varying size. Magic number colors, x==0x80 controlled events. Predefined length for stitching events before color switching.
.sst Sunstar
.t01 Wilcom contains DST stitches with no header.
.t09 Wilcom contain Pfaff data.
.tap Happy contains DST stitches with no header.
.vip Pfaff (older), Husqvarna stitch, Compressed stitches. Like Hus it uses Arc table compression.
.vp3 Pfaff (newer) (See detailed article) unlike most encoding schemes Pfaff files are stored in designs, blocks, stitch blocks allowing multiple blocks and designs within the file structure (see article on specifics) stitches are encoded in several places with seek values to the next relevant set of data and x==0x80 triggered encoding with few commands, namely 0x02 which means long form (it may just mean trim, and jump, as Pfaff's software seems to do that) and 0x01 which ends long form. And 0x03 that is seen at the end of the final block.
.u?? Barudan (See detailed article) Barudan calls this FDR. 512 byte header. big-endian unsigned(x,y,control) encoded stitches in control.
.xxx Singer, Compucon stitch
.zsk ZSK USA

The x?? usually refers to file types with progressing numbers so .u00 .u01, etc, where it starts with u but then has some numbers that change by version.

3.3.1 Other Related Formats

Related file formats
.ext Manufacturer(s) Structure
.pmv Brother (see detailed article) Stitch format, 5 and 6 bit hybrid relative formatting. 5 bits of absolute location in presser foot, 6 bits of forward and back along sew path.

4 High Level Embroidery Files

At the other extreme there are vector files that serve to create embroidery files by having all by storing all the data needed to create an embroidery. This means having the vector shapes and fill types, and offsets and motifs, and which order these are located the start and stop location. And then generating a bunch of line segments through a variety of algorithms. These are then set into the machine readable formats for the machines to read and follow the set commands.

4.1 Hybrid Machine/Vector Files

Many of these types have both high and low level to run on machines. This is because they can encode this data without ill effect. For example PES contain a pointer to a PEC block that is entirely intended for brother embroidery machines to use. These blocks also exist in PEC files that simply say #PEC0001 and then have the PEC block, or within PES files or PHB and PHC files which also contain a PEC block. EMB and ART contain internal stitch data in parts of their file for the embroidery machines to jump-through and read.

4.2 Digitizing Embroidery Formats

Embroidery file formats
extension Manufacturer(s) Structure
.art Bernina Compound File Binary Format, of a series of files. Different files within the format contain the summary information, the Design Information, contents (the compressed stitch data, zlib 4 bytes in), the Design Icon, a bitmap of the what the design should look like.
.emb Wilcom vectors, icon, colors, stitch. this is a full fledged vector format stored very similarly to .ART and clearly share a code base. Several elements stored via a byte replacement cypher in a zlib compressed stream of data after a file size.
.pes Brother, Babylock (see detailed article), vectors, colors, (PEC File), contains several different layers of information.
.thr ThreadWorks full vector embroidery format for threadworks software.

5 Misc Other Embroidery Data

  • .INF, contains only color information like a thread chart.

Embroidery files are used both for stitching and editing. They need to be read by the machine doing the embroidery to process the series of commands. Since most machine embroidery is rendered from shapes and fills applied to those shapes, saving only data needed to stitch would be lossy. So many formats have a hybrid of this and store easy to read stitch data and higher level objects sometimes protected with encryption and compressed (.hus, .art, .emb). With the higher level the embroidery program can reproduce the lower level stitch commands. For most programs that read this data, they often have their own higher level objects and can read only the stitch data from other formats. When they also write these formats, they very often produce the minimum acceptable version of the file that will not crash the program reading the file. So converting from Wilcom's emb to PES will produce a PES with only stitches even if the Wilcom had access to the higher level objects and the saved version of .pes also those forms available.

6 Information that may be found

  1. Stitch Information.
    1. Direct commands go dx/dy, add stitch, go dx/dy, trim, change threads, stop.
    2. Explicit location of the points for the segment list.
    3. Stitchblocks unbroken lists of stitches in a particular color.
  2. Vector Information
    1. Shape Data, Rectangle, Circle, Path etc.
    2. How these shapes should be filled. For example:
      1. Type of fill being used
      2. Angle of the fill
      3. Angle-path of the fill
      4. Start and stop location within the shape.
      5. Pattern for the needle impacts.
      6. Randomization of edge.
  3. Font Information
    1. Text and font, how it should be applied.
  4. Design information.
    1. Design name.
    2. Design author.
    3. Design comments.
    4. Design keywords.
    5. Design copyright.
    6. Design category.
    7. Number of Stitches.
    8. Number of jumps.
    9. Size of embroidery.
    10. Start Location.
  5. Hoop Information.
    1. Specific custom hoop information.
    2. Distance design is from edge of hoop.
  6. Thread Information.
    1. Color data from a preselected list.
    2. Custom color data for thread.
    3. Thread metadata, manufacturer, pantone approximate, etc.
    4. Thread weight
  7. 2D Bitmap information, simulated view of the sewout.
    1. Bitmap representation for project. EMB contain a full color icon.
    2. Bitmap representation for each color. PEC contains 1 bit graphics.
  8. Control information for the typical editor of that format.
    1. Color of background.
    2. Scaling information.

7 Other Lists

8 Acknowledgements

Due to reorganization - i.e. the breakup of the Computerized embroidery page - names of original contributors, in particular Tatarize, do not appear in the history of this page.