Maslow Home Maslow Community Garden

Draw-Wire Calibration/Positioning System



That’s true, I only meant to point out that with one of Bee’s boards as built today you could begin work.


unless we can make the sled lighter or the motors more powerful, we can’t step
up the diameter without loosing the ability to cut along the top of the machine
(unless we change the machine dimenstions, raising the top beam…)

David Lang


look at power poles in your area, you will see that the electrical wires sag,
even though they aren’t chain :slight_smile:

any flexible thing that has actual weight per unit length and is suspended by
the ends is going to sag.

Shelves also sag under their own weight, but they resist bending so they have a
different way of calculating the sag. But flexible things (chains, wire, rope,
etc) all sag pretty much the same way.

David Lang


If the new device were to be configured into something like a tape measure
that would recoil the wire, and have sufficient tension to keep it straight,
that device could be placed anywhere. For instance, placing one each on the
lower left and lower right corners of the rig would still give you the ability
to create triangular positioning and know exactly where sled is located.

it still needs to be immune from the sled rotating, so it either needs to
connect to the center, or it needs a ring/linkage to allow it to be logically
attached to the center, but not physically interacting with the bit.

I’m only trying to point out that I don’t think the new sensors have to be
located at the sprockets location.

They don’t, but where they do live needs to be very accurately known (and
putting them below things may get sawdust into them)

David Lang


My sled is only 22lbs (probably less now that I removed part of the router and 3d printed a Zaxis). I thought it was mentioned that max weight would be 39lbs


22 pounds is about normal. We know that 30ish pounds is enough to make it so you
can’t cut at the top of the workpiece.

David Lang


Not to argue with you, but to expand my line of thinking…

Here I am going to cheat and show some pics from Goliath website. In the first pic you can see the little post on the top of the router, in the centerline of the bit…The wires ends have a magnet on them, and snugly fit to that post by the user at each set up.

Edit: After looking closely at it the post is not directly over the bit. Interesting.

In this second pic shows their “sensor” that they brag about. But only to illustrate that it can have a height to match any height that we would need to get above the z axis stuff on our router, and still hit the center post with a wire from the sensor.

If we had such a feedback system it could be easily retrofitted to all existing machines in the field. We would need a “top for the router”, and a sensor thing to mount somewhere. I give you that the lower corners have more dust than the top corners, but we could try to make the sensor dust tight.

Also, knowing its location would need to be part of the “calibration” of our machine. (But in their write ups they say you can put the sensors anywhere on the edges of the cutboard.)


Not as I read the manual. The ‘Align’ button sets the registers to zero at the present position on the incremental/absolute units. Using a quick-connect attachment like @pillageTHENburn suggests could make zeroing less of a bother that the present sequence, but too bad it would be necessary.


@dlang . PM me the specs for those driver chips - Plz


Just in case you haven’t seen this video of setting up the positions and sensors for the Goliath here is a link to it.


It’s interesting that the Goliath draw-wire issues from a hole in the top of the sensor and the sensor tower is tilted towards the center of the work area. Also, the draw-wire attaches to a point centered between the wheels, but the spindle is offset from that point.


Yep, must be a lot of trig in those programs.


I do have more detail on Goliath, but I’ve probably upset enough folks with what I’ve done already. If you want it I can post it or send it to you.


Do post it, there are some interesting things there.
For instance, the video shows them moving a tower to each of four corners of a workarea which seems to be the sheet stock to be cut, and the software defines the rectangle from those four measurements. They must be relying on that sheet to have all four edges squared true, yes? The tower seems to be shaped to aim at either 90 or 45 degrees from the edge; I wonder if there is some angle measurement of the draw-wire going on inside it.


Well here it is.

Goliath Sensors Q&A

What are the position sensors for?

Goliath CNC comes to you with a patent pending sensor system, which triangulates the position of the robot. The system ensures the target accuracy of 0.1 mm (0.004 inch) which: 1) Allows the robot to calculate its position, so that the Goliath can correct the trajectory in case of loss of traction of the wheels. 2) Define the work area of Goliath.

How the sensors communicate with the robot?

Thanks to 2.4GHz radio frequency communication, the sensors continuously update the position of Goliath relative to the work area, 500 times a second.

Do I always need the sensors to run the robot?

Yes, Goliath CNC must be always connected to its sensors to work.

How long are the sensors wires?

The length of sensor’s wire is 4 meters (13,1 ft).

How Goliath avoid slipping? How can it move and carve at the same time?

This is the magic of Goliath! We spent months looking for the right balance between the torque of the motors, the power of the spindle, wheel’s dimension and material and so on. Goliath works thanks to the correct balance of these elements. If something is goes wrong, and wheels slip, the positioning systems will correct the route.

What is the repeatability of Goliath CNC?

The target repeatability is 0.1 mm (0.004 inch). This is thanks to our patented positioning sensors that can also correct the route in case of slipping or nodes in the wood.

How big is Goliath CNC working area?

The sensor system gives Goliath CNC a work area of 3 x 2.5 m (9.8 x 8.5 ft). Since Goliath and the sensors are portable, the work area can be easily extended, making it virtually unlimited. This makes Goliath the DIY fabrication tool with the widest working area.

Does Goliath have a minimum work area?

Yes, due to the new way of working directly on the work surface Goliath needs a minimum work area. We suggest as minimum work area 800 x 800 mm (2.62 x 2.62 ft). To help you in reaching the minimum area, Goliath comes with two Panel Extensions.

What thickness can Goliath cut?

Goliath can cut a maximum of 35 mm (1.3 in). This is dependent on the material of choice and the kind of mill bit used.

What kind of materials can Goliath cut?

Goliath can cut a wide range of materials including woods (plywood, MDF, bamboo, solid wood), plastic sheets to thin layers of aluminum.

Is it possible to switch the milling tool during the same job?

Goliath CNC allows you to use multiple CNC mill bits, by having a semi-automatic tool change procedure. This process will make Goliath stop, approach to one edge of a panel, and then it will wait for you to change the mill bit. After, the robot will calibrate again the z-axis and start again working with the new mill bit.

What is Goliath’s feedrate?

The average feedrate is 1000mm/min (40 inch/min), of course it depends on the milling paramenters (type and dimension of the bit, depth of patch, etc).

What is a “panel extension”?

It’s a support surface that helps Goliath on small surfaces. It will make the work surface larger so that Goliath has no problem functioning.

How does Goliath handle traveling over parts it’s already cut?

The custom omnidirectional wheels, thanks to the three module design, ensure that two rollers are always in contact with the panel, different from the traditional omni-wheels that only have a roller in contact at a time. In so doing, Goliath movement isn’t affected by grooves or parts that it has already cut, as at least one roller is always in contact with the panel.

How does Goliath ensure it stays level?

Goliath CNC works on flat surfaces. If the panel is slightly sloping or curved it does not affect the work execution. Moreover, we saw in our experience and tests that the weight of the machine helps to maintain the panel flat, at least under Goliath. The software CAM provided with Goliath will generate a machine path that will prevent wheels from falling in the pockets in such way Goliath will always be leveled as the surface below.

What happens when Goliath gets to the edge of the work piece? Do the wheels just fall off at the edge?

Goliath will know where the edges are since in the setup phase the sensors detect the size of the panel. Then the CAM program generates the path for Goliath so that it doesn’t fall off the edges of the panel.

How does Goliath avoid driving over holes or pockets?

Goliath carves out pockets or drills large holes after it has cut all the profiles. Moreover, it carves out pocket beginning from one corner on the top and then moves backwards (see the pocketing GIF on the page) to avoid falling into what it has carved.

Is it possible to 3D carve with Goliath?

The main purpose of Goliath CNC is cutting 2.5D cuts, 3D carving it’s a feature we like a lot, especially looking at the possibilities to make particular surface finishing. We will test and develop. The necessary condition for doing an engraving is that there must be a place in the work area, sized as a footprint of Goliath, with no pockets inside it. The algorithm will manage the pocketing in a way that the wheels will always be on “safe” ground and using the pocket-free area as the last position of the process. If there is no pocket-free area the user must add an extra panel so that Goliath will end the pocketing process over that panel. The entire path will be calculated by our CAM software, which will also alert the user if an extra panel is needed.

How do I get my designs to Goliath?

Goliath CNC is provided with wireless connection systems. The user can design or download designs from online libraries, then send the file and supervise the work progress by computer, smartphone or tablet thanks to internet connection.

What kind of design files does Goliath work with?

Goliath CNC works with the standard format of CAD drawing (dwg, dxf, svg, ai). A custom developed software transforms these files in G-code telling the robot the path to follow.

Which operating systems will be compatible with Goliath?

After the Kickstarter campaign we asked to the backers information about the device they would like to connect to Goliath CNC. The results were quite heterogeneous, and for this reason we decided to develop a web application to serve as interaction between you and Goliath CNC, so the first version of the CAM app will run in the web browser of your devices. The web application will be a cross-platform solution, which will allow you to use multiple devices with different operative systems. In the future we would like to develop also the client-based versions to make Goliath CNC work also without internet connection.

Why I can’t use my CAM software?

We do not suggest using a G-code generated from other CAM software because it will not take into account the wheels position to create the path of the robot. This may cause the machine falling into pockets or moves on the already carved surface. For this reason, we are developing our own CAM software.


“target accuracy” of 0.1mm, it will be interesting to see what they actually

David Lang


Here is a thought. We need about 110" (my guess) of wire extension and retraction from a sensor. If it were a 5" drum, grooved for the wire tracks, 7 revolutions of level wound wire produces about 110" right on. One issue is that the wire moves the width of seven tracks.

Here is an idea to keep it in the same spot.


Lets say that the “pitch” of the wire wrap and the pitch of the threaded rod are the same. Assume the rod does not rotate, and the “spool” turns on the rod. As the wire is pulled off of the spool, the spool rotates, and moves left or right on the rod. keeping the position of the wire leaving the spool constant. The spool moves up and down the shaft, not the wire exit point.

Now this idea needs:

  1. A spring to keep the wire taught for its travel on and off the spool. This spring has 7 revolutions of travel, must be able to be pre-tensioned in the housing, and strong enough to minimize/eliminate the dreaded “sag” of the wire.

  2. The encoder needs attached to measure the rotation between the spool and the other shaft, or the housing.

  3. A housing.

If the wire were attached to the “bit center” (such as a pin on the top of the router), we would know the distance from the top of the spool to the center of the bit.

Need more information to know where the bit is…

Next feature needed is that the wire has to point to the center of the router (bit center) as the sled moves around the extents of the 4 x 8 sheet.

Imagine (#1) that the “housing” for the spool were mounted on a single axis swivel that could rotate, allowing the housing to swing as needed to keep the wire pointed directly toward the bit center. (Some sort of shaft with a bearing to reduce any friction in the rotation?)

Imagine (#2) that another encoder was on that shaft. This would give us the angle to the current position of the bit from “horizontal”.

Now with the distance and angle we would know exactly where the bit is relative to the pivot point of the housing.

Don’t need two of these set ups, just one.

The distance is the hypotenuse, and with known angle, the sin and cos give us x and y.

My wild idea today.


how would you rotate the thing to keep it pointed in the right direction?

today we calculate what angle the chain leaves the sprocket so that we can take that into account, it’s a bit ugly, but it works. I would suggest using the same thing.

trying to measure the angle that the line feeds out with sufficient accuracy is a very hard task, I don’t think you can do it (looking at the math, at a distance of 3m you need to detect 0.5mm or less in cross-line detection, this is an angle of < sin-1(0.0002) or an angle change of ~0.01 degrees (~36k positions/rev). In practice, the line will flex more than that and not turn the sensor. measuring angles is much harder than measuring distance (which is hard enough)

with a 5" radius you have ~400mm per revolution instead of the current 64mm per revolution
currently we have 8000 positions/rev for an accuracy of 0.008mm in chain length
with a AMT21 multi-turn encoder, we would have 4000 or 16000 positions/rev. 4000 would be an accuracy of 0.1mm, 16000 would be an accuracy of 0.025mm

that is less accurate than we currently get, but I think it’s accurate enough.

If we have lightweight line, 7x line diameter (or even 14x to allow for a nice gap between each wrap) is not going to be very much, (2mm line would result in just over 1" of height). The Router is going to move it’s end of the line more than that. This angle should be ~2 degrees at the top corener and <1 degree at the far corner if setup properly, which results in a measurement error of ~0.5mm from the angle in the Z direction

then you have the question of sag/stretch of the measurement line.

also, how much tension will your 7-turn spring apply at the full distance? the tension on the chain is only a little over 3 pounds, so if you end up with a pound or two of tension on the measurement line, that can cause problems.


Storing the line on the same shaft as the one that the encoder monitors brings complications, why not pass the line one time around a sheave on the encoder shaft then off to some storage arrangement like weights and pulleys (benefit is uniform tension) or a tape-measure spring (benefit small space)? The encoder doesn’t introduce any force on the shaft it monitors, and with a good bearing on that shaft and a moderate tension on the line, the encoder should accurately track the movement of the line. A 1" working diameter sheave would give ~.02mm resolution on the 12-bit encoder.
Your idea about a follower on the line turning a second encoder to measure the angle is very interesting. As you point out, that measurement need only come from one side, the other side can be calculated (to reduce the cost - these encoders are $50 a whack…)


Storing the line on the same shaft as the one that the encoder monitors brings
complications, why not pass the line one time around a sheave on the encoder
shaft then off to some storage arrangement like weights and pulleys (benefit
is uniform tension) or a tape-measure spring (benefit small space)? The
encoder doesn’t introduce any force on the shaft it monitors, and with a good
bearing on that shaft and a moderate tension on the line, the encoder should
accurately track the movement of the line. A 1" working diameter sheave would
give ~.02mm resolution on the 12-bit encoder.

the advantage of storing it on the same shaft is that you can anchor the line to
the shaft, no possibility of it slipping.

you could have another line on the same shaft with a weight on it (the other
shaft can even be a smaller diameter so you don’t need some other pully)

Your idea about a follower on the line turning a second encoder to measure the
angle is very interesting. As you point out, that measurement need only come
from one side, the other side can be calculated (to reduce the cost - these
encoders are $50 a whack…)

two encoders, one on each line, or two encoders on one line doesn’t make a big
difference in the cost.


I guess I missed the part about not needing two lines. As you pointed out, when the line gets long the angle gets too small for these encoders.