@silo0623 Is exactly right, that comes from the sled not being rigid. Basically we haven’t found a 3D printing technology that can make big things like the sled well so we’re stuck waiting for the injection molded version which will be glass fiber reinforced poly-carbonate which is super strong.
Here’s the earlier update that talks about it:
Side note…did you know that YouTube adds those sections to the bottom of the video automatically? I had no idea! It even did a pretty good job.
Edit: fixed the video to start at the relevant section
Hey, I know on a previous update you mentioned getting your own CAD software up and running, I currently use Easel as I’m not making anything too intricate and it’s perfect for what I need but it has 0 android support. Is this something you hope to have with Maslow software at all please? Obviously I know I couldn’t run ground controll from my tablet but I just want to be able to create small designs on the go without having to load up my pc. Thanks in advance and I CANT WAIT to get my Maslow 4. Super excited, great job.
We’re working on the CAD/CAM problem from a couple different angles, but everything is browser based so it will work on your Android tablet. For simple things like a quick design or changing the material thickness of an existing design a tablet is perfect. For more complicated things like designing something from scratch you will probably be better off with a mouse and keyboard.
We’re currently 100% focused on delivering to our Kickstarter backers so we’re sold out at the moment, but we do plan to reopen sales in the not too distant future. If you sign up for the email waitlist on our website here Newsletter Signup — Maslow we’ll send you an email with a link as soon as we open sales again.
I read many posts on accuracy, precision and calibration, while most focus on maslow 2.
For maslow 4, got the basic statement that “it is self calibrated” and “reference points which would allow better calibration are difficult to implement since such is economical available for small distance measurements, but extremely expensive for a 8”x12” setup.
Now, I’m an engineer and that should be a challenge. Since, if we improve the calibration it will have a huge impact on both accuracy and precision.
In my first session to search for something which could be used on the 8”x12” setup as reference. I came across “Shaper Tape” which is usually used in conjunction with the Shaper Origin. The tape sells for 22$ on Amazon, is highly accurate and provides a length of 45meter/150feet.
If we had two adjusted cameras (as the Shaper Origin), we could go for a similar approach, but such would be pretty complex and presumably expensive. If Maslow 4 would use 4 optical sensors which are placed around the router plate, a calibration run could move the router clockwise and counterclockwise one time along the tape. The basic pattern of the tape provides means of a distance measurement.
Since two of the sensors would always be used on each side in parallel the tilt of the router plate can be also measured.
Each maslow setup would need to place the tape one time with high accuracy in rectangle form on the work surface. But this should be doable.
Certainly the more optical sensors, more information could be sensed (like for instance the dot-code of the tape). But this could be a step postponed to maslow 5
What do you think of this proposal for a calibration function?
what is the precision of the shaper tape? especially over larger distances? can
the tape stretch as you are applying it?
remember, the shaper isn’t depending on the tape being in a particular place, or
even it being continuous (you can have lots of short sections), it’s depending
on the camera having predictable distortion as it’s reading the tape and
adjusting accordingly.
The maslow 4 has high precision measurements of the 4 belts (And with 4 belts
it’s overconstrained, 3 belts is theoretically enough to position it, this
allows it to lock it’s position with 3 belts and measure the 4th, repeating
until it can solve for it’s position)
With the original maslow, we did have someone try for an optical calibration,
but the problem they ran into was getting something to detect that was actually
accurate enough.
Shapertools does not state the accuracy of the tape itself. They only state that:
“Product Overview
Origin uses ShaperTape fiducial markers to digitally augment your Workspace and accurately track position while you are working. ShaperTape’s uniquely encoded patterns are printed using specialized production equipment designed to ensure extremely high dimensional accuracy which results in optimum tool performance. ShaperTape is also specifically engineered to be easy to apply and remove, with carefully selected adhesives that won’t leave residue on your material. Each roll of ST1-150 ShaperTape is 45 m in length. The amount of tape required for a given project depends on the scale of the project you’re working on. For reference, 45 m provides ample coverage for approximately one and a half standard 1m x 2m sheets of plywood.”
On the other hand, they state that the Shaper Origin is accurate to 0.01 mm and that the band cannot be printed by the user themselves as this would not be accurate enough.
So the shaper origin system is highly sensitive if it comes to the scale of the tape. The direction of the tape seems to be irrelevant. If the tape would be stretchable, it would also not work with the shaper origin tool.
I hope this answers your questions sufficiently. From my perspective, the tape offers an adhesive scale for under $5 per Maslow system. So in my opinion it might be worth a try.
I might have missed it so forgive me if I did. Has there been any update on the tolerances achieved with the machine. I remember the kickstarter mentioning that they hadn’t yet hit ±0.5mm. Has this been achieved now?
This video is still the most recent time that I did a careful analysis
@RomanG is currently working on the calibration process so we’ll probably do an update on it soon-ish. We’re working on making it more user friendly (ie you push one button and it does everything). Once we have it working fully hands off (right now you have to do a couple steps in the right order and it’s possible to mess it up) we’ll get back to refining the precision and then we’ll do another test and update.
a redundancy in the measurement can not be used for calibration, since you don’t know which of the measurements you can trust. You can detect defects which makes the CNC more reliable what is actually a good thing. But for calibration you need to compare the measurement with a known distance you can reliably detect.
That is where something like the Shaper Tape can help. It is not so important how accurate the Tape is placed on the board since along the tape you can detect the exact pattern of the tape and use it to calibrate this axis. Than you repeat this on the other axis.
As long the Tape is approximately placed in 90 degree on the board and almost straight.
Is there for instance a useable I2C interface which could be used for an optical detector on the maslow controller? Btw. you are right that this should not be the highest priority, since a calibration system can be added later as a addon.
With the maslow 4, you have 4 distance measurements that you can trust
(the length of each belt). what you don’t have is any angle information, and
only minimal information on the belt tension.
we know there are a few pins available, we don’t know if they make a usable I2C
interface (you could look at the firmware souce code and analyse it to try and
figure this out)
You are correct that repeated measurements in the same location don’t help with calibration, but by taking multiple measurements in different locations we can sort of triangulate the locations of the anchor points very precisely.
There is a full serial connection exposed on the board so it would be possible to add something like a tape reader if we wanted to, but it’s actually quite a bit more complicated than that.
Let’s say that we can read the tape and get a very accurate measurements for distance moved and we tell the machine to move 100mm and we read the tape and see that we moved 100.1mm…what we can then do with that information isn’t clear. We can stretch all of our movements by 99.9%, but that will only be correct in the one location where we did the test. The equations are very non-linear so the behavior in one location is not cleanly connected to the behavior in another location. What we really need to know is the X, Y coordinates of the anchor points. Knowing that we moved 100.1mm instead of 100mm is a hard thing to convert into anchor point X,Y coordinates.
It’s a deceptively tricky problem…I could be wrong, but I’m pretty confident in our current calibration system.
Ok, something very fundamental.
A universal scale factor for x and y is certainly not usable. Ideally you require for every single possible position on the board a perfect compensation factor which is applied on the measured values from your belt sensors after you transform them to x/y.
For every single x/y it is not feasible (and reasonable) to calculate the compensation values. Instead you better use a matrix which stores compensation values for points on a grid.
May be every 5cm to 5cm. For the points which are not on the grid, you can calculate the compensation by interpolation, or you just use the nearest grid point.
Ideally you could measure on every single maslow system every single grid point, but this would be very time consuming and also expensive. Instead you better measure the full grid only once on a reference setup.
On every Maslow system you measure only the grid points around the frame. The compensation values of the full matrix can then be calculated.
It is actually simpler than it sounds.
except for the fact that you need an accurate grid. It’s not enough to just get
a grid around the edges, you need the grid across the entire workspace.
The errors that you get from a cable robot like the maslow are curves with lots
of different parameters (so they aren’t simple curves to calculate)
Yes, if you can get an accurate grid across the entire workspace, you can then
interpolate between them for good accuracy everywhere, but the problem is having
an accurate grid over that distance to start with.
look through the forum archives for optical calibration and you can find the
efforts and the problems that we ran into trying to get an accurate grid to
start with.
It would take a lot of shaper tape, and a camera with a large field of view to
try and cover the entire 4x8’ workspace.
but don’t take my word for it, try it. you can use a old-syle maslow or just a
couple motor/encoders with string holding the camera frame
