The calibration instructions in the gSender documentation (place a piece of tape on the rail, move the carriage a fixed amount and measure to see if it moved the correct amount) seemed rather inaccurate to me. I used a 60deg v-bit, ran the router, made a tiny mark, used the gSender calibration function to move the router 100mm, jogged the router to make another small mark. The first time, I measured the distance between marks with a steel ruler. The best I can interpolate is the nearest 0.5mm. There is also the problem of parallax due to the thickness of the ruler. The second time, I measured with calipers. This is more accurate, but not perfect because I am trying to line up both ends of the caliper with the exact centers of the small marks. I think this is good enough.
I cut a 1” square with an IDC 1/8” compression end mill, measured it, went back to the tool data base, changed the diameter of the bit from .125 to .120, recalculated and reran the cut. Now the square measures 1.008 in the x axis and 1.0005 in the y axis. I know the IDC bit is not .005 undersized, but combined with the +/-.004 (actually +/-.01mm) accuracy published by Scienci, the bit could only be .001 undersized, which is quite possible.
My question is how do other people calibrate the machine to assure getting the best accuracy?
I am still waiting for my first mill so all my experience with calibration is based on 3D printers. What I would do, after having done the basic calibration like you did, I would cut a much larger square (any error would be more pronounced), measure that square with calipers and fine tune the calibration to give me a more-or-less accurately sized square after re-running things. I would NOT change the size of the endmill unless I had a VERY accurate reading on the actual endmill diameter. The manufacturer of the endmill has much more sophisticated measuring equipment and it is highly unlikely that the endmill is not the advertised size (unless you resharpened it).
You would also need to account for any possible error in your measurements and know both the accuracy and rounding errors of your measuring tools. This also assumes you have at minimum a caliper made by a reputable manufacturer (such as Mitouyo (sp?)) or better yet, a caliper that has a calibration certificate. Most people that I know use cheap Chinese calipers which are not accurate enough for what you are trying to do.
You mention the y axis reading 1.0005 in … that’s half a thou. Unless you have a caliper that measures to a tenth of a thou, that half thou is a wild guess on the part of the calipers. Likelihood is that a) the specs say the resolution is a thou ± something and that the last digit is also plus minus one step … ie the reading would be between 1.000 and 1.001.
Were the calipers at reference temperature? Holding calipers in your hand will change the instrument’s temperature and alter the reading.
When you take the reading, measurement technique can alter the results.
All this is to say that unless you are trained in proper measuring techniques and have high quality measuring instruments, you should really be skeptical about what you are reading. This also does not take into account the nature of wood and how accurately it can be cut.
You are chasing an accuracy that is not achievable with the equipment you are using or the materials you are using.
Edit: The fact that Scienci is asking you to calibrate visually like you describe should tell you something about the expected accuracy you can achieve.
I also have a 3d printer, a Prusa MK4. I did the 1” square test and changed the mill diameter based on other posts (not here) which said it is common practice because all bits are .001 or .002 undersized. I have no way of accurately measuring the bit diameter, it is hard to do because of the flutes. Before I retired, I had access to an optical comparator which would have worked fine, also calipers with traceable calibration, but now, I use what I can afford.
I included the 1.0005 only to show it measures very close to 1.0. I realize it probably isn’t that exactly. I am a retired engineer and spent some time in QC, so I am trained in proper measurement techniques, but again, I no longer have access to that kind of equipment. That said, you have to use what you have available and make the most accurate measurements you can, or else just give up. Scienci Labs says you should be able to get to a tolerance on +/-.004”, and I think that should be possible with the tools I have.
Good to hear that you understand the intricacies of measuring stuff!
I had a look-see about the spec’s of a Longmill and I was unable to find anything (it was a short look).
Does the spec say anything about repeatability? Where did you find this info - I am waiting for my Altmill and would be interested to see what the specs are for it.
Given that the quoted spec is ± 0.004", I would say it would be reasonable to expect an actual accuracy and repeatability of 0.01" which I think would be perfectly acceptable for a device cutting wood. Of course that is just my take on the general situation.
To be fair, I run my machine uncalibrated and use it to measure the dimensions of whatever I have strapped to the bed. The numbers g-sender indicates as being x-y-z, is what I use in v-carve.
Signs don’t need accurate dimensions, the design only needs to stay within the boundaries of the material on the machine.
It works good enough for my purpouse and I have not had a moment that I felt I needed calibration to work more accurate.
For more irregular stock I use the camera function in lightburn, measure the max boundarie with the mill as a calibration to home in the camera and design somewhat conservative onto the projected image.
Any deviation the machine has, goes away if your measuring tool is also the machine.
I used a v-bit like you did and measured between the marks. I would make the marks farther apart when doing the calibration. That way any measuring errors are a smaller percentage of the measurement. My mill is a LongMill 30x30 and I probably used 5 or 6 hundred mm for the legs of triangle to try and be as accurate as I could with what I have available to me.
A tape measure is thin so you don’t get the parallax. I use a metric tape measure and put the 10 or 20mm mark at the start of the measurement to try and avoid errors from the end of the tape measure. Just remember to subtract where you started from the result.
Thanks Michael, the only reason I didn’t go bigger was that I wanted the accuracy of the calipers, mine max out at 200mm, so I could go that big. I do have a metric tape measure so I may try that also. My machine is a 30 x 30 also.
After looking at the gSender docs I may have been thinking about the wrong test as there are tests for accurate distance on each axis and a test for squareness. You mentioned marking the rails and that’s for steps/mm for each axis and I was thinking of the squaring one.
On the tests for accurate movement distance I checked the measurement and as long as it was very close I assumed the difference was probably a measurement error on my part. The reason I did that is because I don’t think that axis travel would change based on assembly. If a motor has a known number of steps/revolution and the pitch of the lead screw is known then it’s just some simple math to get to the steps/mm and Sienci already did that math and set the defaults.
On the other hand the squareness test is subject to how accurately you assembled and mounted the machine so I was more concerned with that test.
If I remember right, it was a couple years ago, on the distance test gSender told me to go to 203steps/mm and the default was 200 and figured that was probably in the error range of my limited measuring capability given the simple tools I have. Then on the squareness test it told me to move a Y rail by 0.5mm and I never changed it because unless you mounted the machine on separate strips of material, slotted and bolted, it’s pretty hard to move a rail by 0.5mm.
Just wanted to explain my thought process about the two types of tests. Good luck! I hope you enjoy your machine, I love mine.
EDIT: I forgot to mention that you should make sure the anti-backlash blocks are adjusted properly. If there is play there it can lead to wrong measurements because if you moved the machine left to the start point and then move right the lead screw has to take up that slack before the carriage actually moves.
I had just about the identical experience. Moving one rail .5mm is pretty much impossible, and when I cut out a 12” square of plywood as a test, it was as square as I could measure with a framing square. When I ran the calibration test, I did change the steps/mm and I think I did improve accuracy by doing so. Thanks again for your input.
But the test they did for that was measuring the deflection of a 1/4" gauge pin in the collet under force while stationary. Basically just the rigidity and stiffness of the machine. Nothing to do with the ability of the motors and drive screws to place the bit in space accurately with repeatability.
So I would be careful interpreting that to mean you can tell it to cut 1.000 and expect it to be 1.000 +/- 0.004 multiple times.
For perspective: A 5 degree F increase in temperature will lengthen four feet of aluminum by 0.003".
Then we can get in to the discussion of climb vs conventional milling and how that would influence backlash and the cut. Cutting a square out of a board (so material on both sides of the bit) versus just trimming away to hit dimension.
I’m all for chasing the dragon on accuracy! It has me deep into metrology and thinking of moving from a CNC router to a vertical mill and getting more into machining metal.
I enjoy the quest for accuracy and precision. I am quickly learning it’s much more the user than the machine.
Good luck! Just trying to help provide some perspective so you don’t make yourself totally nuts with this.
I used my laser to make marks and measured the same side of the opposite lines to get my accuracy. Also I used large distances in order to adjust my stepping. So far I’m really stoked about my accuracy. I’ve made some very tight tolerances in a couple of projects and all is well.
The laser works great for this, however the z axis is something entirely different. For that I jerryrigged a set of calipers to measure.
I just measured mine today for first time…Got it assembled yesterday and surfaced it…I used 300 mm in my test after finding that my calipers were faulty( cheap digital)…kept reporting 5 mm extra when measured against a metric tape. What is it they say about the man who blames his tools???lol…Anyway short story long measuring with the tape found I was within 0.59 but not going to movea rail.5mm…
In addition to the point that @CharlieMike made about aluminum, the woodworkers among us will attest to the fact that wood moves. It moves owing to temperature, it moves owing to humidity and it moves owing to the stresses that we relieve when we carve it. It moves much more than the aluminum that Charlie mentions.
Even our MDF spoilboards can move over the course of a day or a long carve. In high humidity regions, this can really have an effect on the accuracy of detailed 3D carves.
I am not out to argue about the level of accuracy of Sienci machines. I’m merely trying to put their specifications into some sort of real world context. In the end, we all choose to accept a degree of precision that suits our projects. That will vary considerably depending on just what those projects are.
