New SO3 - Fun, Frustration & Mistakes

I have now settled down after building my SO3 and attaching the JTech laser. I believe the machine is fettled as best as I am able to get it at this early stage and I had a go with trying out the various facilities. This is on the basis that unless the machine is used for actual projects it will constantly be used for running tests and for assessing this and that, with no real work being produced. I also thought it would be useful to write about my experiences and possibly make other people laugh as well attempt to educate and induct me into the dark art of CNC machinery.

So apropos of nothing more than I thought that some people may find this screed useful/funny… onwards. After assembling the 4.2W JTech laser, I was delighted to find it working as intended. I have been using LightBurn which is lovely software. anyone with an understanding of image editing or designing software will be familiar with its design language. What has impressed me is just how rigid the magnetic mount is and I opted for a central position on the Carbide 3D router mount. I reasoned thus, it would be faster to visualise what was happening than to keep adding a specific amount of offset to account for a one side mounting.

I had wanted to run a test image of a .svg file of a Guilloché pattern. The image was burned into a piece of scrap MDF at 150 x 150mm image size. My recall is that it took slightly more than an hour to create and herein lies a part of the problem, my speed setting was 1200mm/min at 100% power. I noticed that the pattern was developing as it was supposed to but the machine was making all sorts of strange random noise as it tracked back and forth.

It turns out that I was given Grbl to much to do because @WillAdams had passed my files to and @chamnit and Sonny wrote down his thoughts here:

The image in question is this one:

When viewed at 100% magnification, you can see that the straight lines forming the pattern are actually pairs of lines sitting very closely together. I am delighted to note that the laser burn was clean and clear with no trace of the wobbly laser impressions we have seen in some images, discussed by people who are using the JTech magnetic mount. I will also mention that I installed the high resolution lens into the laser pathway and this has contributed to the image sharpness.

I am not sure what the counsel of perfection would be here but reducing the traversal speed to maybe 200mm/mm and reducing the power output to possibly 50% would possibly bring a cleaner result. At the extreme edges of the image magnified to 100%, there are signs of scorching or smoke staining (not that very much smoke was evident during this particular burn)

In any event, I will try again with a much slower, less powerful setting and possibly I will scan my file left to right instead of following each component part individually. This may help with the random noise generation. Another possible hidden benefit of scanning when burning a complex file is the length of the lines will give the machine software a chance to keep up with the Grbl instructions à la the guidance from @chamnit.

As an aside… LightBurn has fine control of velocity of the laser head. The overscan setting permits the laser to be turned on and off at will, when the head is either accelerating or slowing down. If it is turning this setting prevents over burning of specific areas where the laser head had to slow and change direction.

The CNC aspect of my SO3 was not used up until the point I could not delay it any longer. All of the movements of the Z/X carriage and the X rail had suggested that everything was ready to run. The levelling of the bed and the wasteboard was accurate and with the belts and V Wheels having proven themselves with the laser head attached, I dreamed up a very small project that would have been quite difficult with just drill bits and a chisels.

For this project, I had wanted to tidy up my plugs for the various bit of machinery I have in my workspace. The leads were often laying over a bench or they got mixed up with each other and it was a minor annoyance to keep sifting through tangled leads every time I wanted to use a machine. I determined to make a plug board that would hold each plug neatly until I wanted to use it.

I found a technical drawing for the standard British three pin plug (BS1363) and I noticed that some of the values for constructing a socket were missing. The image I found was this one:

Off to the shed I go, to collect up my well used and much loved digital caliper. I then found a plug top from which I could measure all of the dimensions which I needed to know. BS1363 plug tops look like the images which follow:

Front elevation:

Side elevation:


Next step was to draw the sockets at their precise dimension and spacing and I used Affinity Designer for that task. It is a very good substitute for Adobe Illustrator and it can produce a variety of vector image files. After I had completed the simple design work, I chose the .svg file format to save the file into and it loaded directly into Carbide Create Pro.

I had to scale the image to the size of my work piece. This is where my technique could do with some improvement. I have no idea how to scale my image for Carbide Create Pro and I wondered if there was a formula that could be applied. I used the image at the size which I had created it in Affinity Designer.

The only reason that I can see that it worked is that in Affinity Designer, my page size was specified as the workpiece size. That way, when importing the design into Carbide Create Pro, I specified the workpiece size and all of the scaling was correct. My intuition tells me that this may not always work as well as it did. The scaling of designs (where it is not exact) can have a very significant effect on the carefully worked out measurements. I saved the work as an .nc file.

I attached the workpiece (by means of four gator clamps) and my clamps were screwed into my 10mm inserts in my 3/4 inch MDF spoilboard. I marked the centre of the workpiece with a pencil and was ready to roll.

I selected a Trend two flute straight bladed cutter. It had a diameter cut of 3mm with a 12mm length and it was produced with 1/4 inch shank. It was also capable of plunge cutting because the blades just crossed the centre line. I inserted the cutter just as far as the index line and tightened the cutter. I set the router to 18,000 rpm and attached the shopvac to the Sweepy.

I left the magnetic cover off so that I could zero the cutter. It was done with a series of rapid movements and then touching off a thin piece of paper using the smallest steps possible. I used the centre of the workpiece as my zero point and set that in the Carbide Create software at set up.

I loaded the .nc file into Carbide Motion. I left the magnetic cover off of Sweepy so that I could zero the cutter. It was done with a series of rapid movements and then touching off a thin piece of paper. I used the centre of the workpiece as my zero point and had already set that in the Carbide Create software at set up. I re-attached the magnetic Sweepy cover and set the router going. I also set the shopvac going, after putting on some ear defenders and goggles. I pressed start and the result is shown below.

My feed rate was 1143mm/min (45 in/min) with a plunge rate of 800mm/min (31½ in/min) The wood was scrap CLS pine that is used to support plasterboard (drywall?) constructions for putting up walls in a house.

Finished workpiece on spoilboard:

Workpiece in situ:

Tidy at last:

So there it is: This was just a very simple project that has made my workplace easier to work within and much safer now that I wont be tripping over loose electrical leads. While this project could have been done with a mallet and chisel, there is no doubt that the CNC machine made this job much easier. It was a great project from a new user’s point of view. I was able to plan and implement a useful project which be achieved from the very start of one’s Shapeoko days.

The next project was relatively simple but I made some errors and was less than happy with the result. It was for a relative who had want me to carve a pinning tray. He occasionally works as a locksmith might work and has a need to disassemble locks. There are many springs and pins in a secure lock and these are usually placed in a pinning tray during disassembly and assembly. The requirement was for two pockets of 10mm deep and 7 thin channels of 1mm deep.

I had a nice piece of Olive wood and we selected a piece of it and roughed out a sketch of where things would be located. The wood was not a rectangle so taking some measurements was a little awkward. I also had no real understanding of whether the SO3 would be able to hold the workpiece securely, during the machining of it.

I opted to get as much of the finishing of the workpiece done before carving, so that I would have less finishing work to do when the item was complete. I sanded the piece through many grit changes from 80 grit at the start all the way to 3000 grit at the finish. I knew that I was going to use beeswax as the finishing coat and that would be applied to the silky smooth surface as well as bringing out more of the beautiful pattern that accompanies olive wood.

The design was done directly in Carbide Create Pro (as a 2D piece of work) and then it was adjusted and a final approval given. The wood was decided upon but I also roughed out a blank of cheap plywood just be sure that all was going to work well. I used two different cutters for the project. The pockets were cut by the Carbide 1/4 inch 201 series cutter. The second cutter was a 3.2mm cutter with a radius that is unsupported in Carbide Create.

My tool-path was set to use the cutter as straight two flute cutter of 3mm diameter. My test piece had the channels cut to a depth of 1mm and they looked good. My workpiece and the file was set to be centred. Finally, the pockets were cut in the test piece and all looked great. There was one issue with the test piece mock up and that was I had made two files of the two tool-paths. The reason was I could not see what the process would be for changing the tool and trying to establish where the surface of the stock was, in relation to the newly fitted tool.

The plywood mock-up test file is below:

The clamping method was to secure the unevenly shaped piece with 6 clamps to the spoilboard and the clamps were in placed one at a time in a star pattern. This allowed the first three clamps to be used as stops while the workpiece was pushed against them without marking the edges. I did not use the set screws with the clamps because I had thought they would damage the edges and I had wanted to keep the compound curves of the edges unmarked.

The final image is the olive wood piece pocketed and sanded with 3000 grit aluminium oxide sandpaper. The finish was applied as an antique pine beeswax. My relative was pleased but this piece exhibits several faults.

Actual finished workpiece:

The highlight shows my extreme clumsiness while I was not paying attention when sanding. I took away too much material. Both of the large pockets are too high. This misplacement is due to measuring the centre before beginning and assuming it would be the same for the second large pocket file. I must find a technique that permits a tool change while keeping the file in one piece.

Anyway, for good or ill, that is what I have been doing with my new machine. Helpful comments, laughter or any kind of criticism is welcomed. I have started now and will do better next time.


on the svg import and sizing

until recently the industry was chaos and the DPI of SVG files was outright a mess.

nowadays this is standardized on 96dpi and carbide create just assumes that. Some older tools export 72 or some random other resolution by default… but most have a way to export 96

another trick that can work is to put a box of a known size around your whole work, and hten in carbide create. group the whole vector into a group and then resize it to that known size

next step is 3D stuff :wink:
well took me quite a long time to get to that, first was 2 1/2D signs for me but you seem to go quick


The tool change conundrum in this… BitSetter is the answer. It seems (as I will soon have to) that getting one’s head into the process flow of zeroing/bitsetter/toolchange is important, but once learned is a real time and frustration saver

1 Like

You basically just have to

  • follow the prompts in CM
  • and always, always use the “Change Tool” button whenever you physically change the bit.

Do these two things (especially the second one) and you’ll never have to worry about the subtleties of the workflow :slight_smile:

Hello Arjan. My understanding is that vector files are completely resolution independent. DPI (dots per inch) is relevant for the output device (printer). In image publication terms, human eyes are unable to detect artifacts or dots where the image is printed at 300 dpi or more.

One important feature of print resolution is viewing distance and if you can recall seeing a billboard poster that is say… 40 x 60 inches in size, you will note that it looks entirely readable from across the street. When you get up close you see the typical four colour rosette of process printing (cyan, magenta, yellow and black (key) where each colour of dot is screened at a different angle. (see illustration below)


When you move a vector file from one software to another, the file is not changed by the transfer. The exception to this may be that one software is written in a manner that does not permit it to parse the file parameters in the correct manner. My images below show examples of vector file resolution independence.

Vector image at 100% resolution (no magnification):

Vector image at 10251.1% resolution (magnified 10251.1 times)

There is no change in quality of the image because it is a given that the independence from resolution constraints of .svg has no effect on size of the image. The correct parsing of the file should not cause any issues for whatever software accepts the published and known image file format.

There is no case for Carbide Create to require a vector file to have any particular dpi setting. Perhaps one of the Carbide Create software engineers can confirm whether or not Carbide Create does assume 96 dpi is the incoming resolution of a Vector image being imported. (It sounds highly unlikely to me) knowing that vector images have no particular resolution. The file is complete in and of itself without resolution data and dpi only pertains to output files.

The resolution of a vector file may be considered (for all practical purposes) to be infinite in both larger and smaller directions. The screen resolutions of 72dpi and 96 dpi are a hang up from the early days of desktop publishing. The Mac was preferred because 72 dpi equated exactly to the printer’s 72 points to 1 inch of printed text height regardless of font used. Designers of pre-press materials found it easier to work with real WYSIWYG rather than converting points to 96 dpi and back again. This is the major reason that a Mac was widely preferred for pre-press design.

The box of known size is exactly what I did by creating a rectangular page in Affinity Designer of precisely the dimensions of the work piece. When created with a transparent background, there is only the design to create in Affinity Designer, which has the benefit of rulers, guidelines and numerical size of object entry. When it is imported into Carbide Create, providing that one dimension is the exact Affinity page width when resizing the image, the scaling will be exactly 1.00 and the sizes will be accurate.

I look forward to trying some 3D stuff as soon as I have a few detail cutters. I took advantage of the generous Carbide Create Pro offer of one year’s use for free. :grin:

Not a CC SW engineer here, but I confirm that CC expects the imported SVG to be 96dpi, because it is a de facto standard as Arjan mentioned. Yes, vector files should not have to care about dpi, it’s a little weirdness, but that’s how it is and if you stick to exporting at “96dpi” from Affinity Designer (or any other vector drawing software) you will make your life easier

1 Like

Wow! I am shocked. My apologies are due to @fenrus then for doubting him. That has got to be the weirdest software development decision I have ever heard about.

Screen resolution has moved on since the days when a CRT was produced by Sony for the Mac (aperture grilled Trinitron screens) and was 72 dpi to match the 72 points to an inch of pre-press design and the Windows PCs could not compete because of their 96dpi screen resolution.

My laptop today has a 220 pixels per inch screen resolution. So my scaling issues are because of this expectation. All is explained. Thanks Julien. :grin: Just tried it out with some known measurements and it worked just fine. :blush:

Thanks for this Andy. I can see that the interest in CNC work is just a start to emptying my wallet. :wink:

It’s interesting to consider Vector Graphics and resolution…

DPI is a reasonably common placeholder assumption when dimensions/scale are not part of the Vector data - it is mm, imperial or light years. So a ‘works in practice’ solution is DPI.

There is also the issue of decimal places, rounding and Vector segmentation of curves. Bezier curves means that the transit of a curve between two nodes is mathematical and thus accurate, but two issues can arise - firstly, rounding/decimal places in the absolute position of the two nodes, and secondly, the degree of segmentation that software A vs software B uses to draw curves. Some software will calculate Bezier trajectories ‘real time’ and always, others will project the curve and then draw discreet segment nodes along it to save ‘real time’ processing overhead.

So not as obviously resolution dependent as Bitmaps, but Vectors do suffer from ‘resolution like’ issues in some implementations…


The big consideration is that with the 2mm pitch pulleys which have 20 teeth and the rotation of them the coordinate system has a granularity of 1/40th of a mm

A further consideration is some software will decompose to polylines while others will preserve Bézier curves and translate them into G2/G3 arcs for smoother movement — academic papers and resources on this:

G codes for the specification of Pythagorean-hodograph tool paths — Academic paper discussing how curves, arcs and circles are handled and the consequences of representing them as polygons or polylines.

Curve fitting with arc splines for NC toolpath generation — Yeung, Millan K; Walton, Desmond J, Computer Aided Design. Vol. 26, no. 11, pp. 845-849. 1994


I see… I have no problem dealing with the file in the abstract. It is just a vector file and it is the use of it that provides the dimensions. (just imagining the screen required for a design that is light years in size or worse having to specify a line drawn in parsecs) :smile:

Yes, this was my understanding of the value inherent in using a bezier curve.

OK, I had not considered internal rounding errors… I guess I had never considered it because the software has always done what I needed, when I needed it and to the standard I required it.

I have noticed the differences between softwares that are ostensibly used for vector graphic manipulation but some produce much cleaner results than others.

These days, computers have more than enough power for developers not to have to worry about CPU overhead. Wasn’t the Mars explorer ‘Curiosity’ running just an 8 bit processor?

I recall being a guest of PanAm Airlines at Heathrow during the 70s. Part of the day included seeing around their operations room. Air conditioned and full of large machines holding hard disks. I believe the whole company was running on something like 10Mb of storage. We have no need of that sort of storage management today.As a comparison with PaAm’s global facility on 10Mb, I have 10 Terrabytes of storage available to me at home and 500 GIB available on my laptop!

Thanks Andy. It was a timely reminder.

Thanks Will. I have downloaded the first paper but the second link is not working. I will have plenty of reading material. Who knew that owning a Shapeoko would wake me up. :grin: